All right, look at this room! All right, welcome to Investor Day 2023. I'm Shane Xie from IR. Thanks for spending the time with us today. We have a lot to cover today. Let's just start with my favorite slide. That today, during today's presentation, we will make forward-looking statements regarding our business, financials, and future prospects, which are subject to risks and uncertainties as described in our safe harbor and most recent Form 10-Q. We undertake no obligation to update these statements after today's presentation, except as required by law. We'll present non-GAAP financials measures, which should not be considered as a substitute for GAAP financials. A reconciliation between these GAAP and non-GAAP financials is included in the appendix today's presentation. With that out of the way, please join me in welcoming, the next speaker to the stage, Confluent's Co-founder and Chief Executive Officer, Jay Kreps. Thank you, everyone. I really appreciate everyone coming and spending a little time with us today, and it's my pleasure to dive into, you know, a number of different areas of the business. We're gonna kick it off with me talking a little bit about data streaming. What is it? I know some people have been with us for a while and know all about this area in great depth. Some are totally new to the story. I'll talk a little bit about that, and then we'll dive into our product vision in more depth. We'll talk about our go-to-market model. We'll get to hear from some of our partners, some of our customers, and we'll get to talk about some aspects of the business in more depth than we have. There's exciting stuff throughout the day. I'll start with just a snapshot of the business. In the last two years since we've went public, the business has grown tremendously. We're now at a $700 million run rate, 38% year-over-year growth, 130% net retention rate, on a path to breakeven by Q4. You know, a really important part of this, especially in the last two years, has been our cloud product. You know, as we went public, this was kind of a fledgling business, but had been a huge focus and investment for us in the underlying technology. We felt very confident about the trajectory that that was on, and that has really proven itself. You know, that's now on a $300 million run rate, 140% NRR, and really become a huge part of our business, and in many ways, the default when we think about how we build, how we release features, how we deliver value to customers, and very much a part of the infrastructure for data for a lot of our customers in how data flows between environments. We've done all this really in a short period of time. Confluent's still a young company. You know, we were created in, you know, end of 2014, so this has been a fast rise for us. You know, what's allowed that is really a pretty significant paradigm shift that we're part of. When we think about data and the infrastructure for data, that's an area that really grew up around storage. You know, how do we take a pile of bytes? How do we store it in one place? That was the rise of file systems and databases. It's all about: How do I look up the right little bit of data at the right time? That turned into an enormous category and market, but it doesn't solve the whole problem. The reason it doesn't is because data isn't just about storing, you know, the bits for one application. Increasingly, it's about how all the parts of a company come together, because data isn't stored in one place anymore. It's not just about static usage at a point in time. Increasingly, data is spread in an organization across hundreds or even thousands of databases, applications, SaaS systems, data warehouses, data platforms. The task for companies is to bring all this together into one, you know, set of customer experiences, one operational backbone for the company. That was a problem that was never really considered in traditional data infrastructure. Nobody thought about how data flowed. They only thought about how it was stored. You know, for a long time, this was something that was solved really with kind of duct tape. You know, it was things you would build or a hodgepodge of different partial solutions. Nobody really thought there was a general-purpose way of connecting the parts of a company, of integrating data, that was really until the rise of data streaming. This took, you know, a whole bunch of problems that people saw as separate. You know, how data moved, how it was processed as it flowed, and came up with a very general purpose platform for this. This idea of real-time data, the idea that you could have data flowing continuously, it made that mainstream. The fundamental idea here is, instead of treating data as something that's stored, you know, in a pile, you treat it as a stream. You can imagine the stream of what's selling in a retailer, all the products that are being sold all around the world. You can imagine the stream of shipments and the logistics of any operation. You can imagine the stream of trades that are occurring or customer experiences that are being delivered. This is a very natural way to think about a business. A business is a very real-time entity. Things are happening out in the world all the time. A business is about how you react and respond to that. Having the computational capability to take these streams of data, feed them out to all the systems that need them, react and respond, and compute against them in real time, that's a very fundamental foundational capability. It's, it's actually kind of hard to imagine that it was missing this whole time, but it was, in part, because it was hard to do, and the technology to do this wasn't around. People were faking parts of the problem in different data movement tools. They were faking it in different application layers and foundations. What happened in this area was, you know, we kind of went back to basics, rethought how you did this. The idea of how you would do query processing was brought into real time. The idea of how you would think about modeling data was brought into real time. This co-occurred with the rise of distributed systems and the ability to do this across a company in the large. That enables a really powerful role for this technology, this kind of big, intertangled mess that companies have of, you know, this system kind of hands knitted into that system, which is kind of duct taped to this other system, can be turned into something that is much more robust, that's much cleaner, it's much cheaper to build against and maintain, and that enables new capabilities. You can take this big mess and turn it into something where the parts are connected in a more structured way. You know, we like to think of this as being kind of like the central nervous system in, you know, a person, right? Where, you know, the different cells are not just connected via a mishmash. There actually is some structured way that the impulses of what you see, what you feel, come through, and all the parts of the body can react and respond in real time to that. A similar thing is needed in a business, right? Every application can't hand wire itself to everything else. It can't reinvent the kind of fundamental infrastructure for working with that kind of data. You need to have something that says, "Hey, a sale occurred," and the hundreds of applications that react to that in some way need to be able to tap into that and respond and process that as it occurs in that. That ultimately is what Confluent provides. That's what data streaming is about, is that substrate of the flow and real-time processing of data. This isn't just theoretical. Early on, you know, this was seen as a kind of niche, like, "Oh, this real-time stuff, you know, that's great. Maybe some of your data movement needs will be real time, or some of your processing will be real time." Increasingly, the technology proved itself out, and it became, you know, in many ways, the default, right? It became something that could be as reliable as batch processing, as high throughput as batch processing, and as well known as batch processing. As that's happened, you've seen the adoption go up. This, this graph shows the unique users for Apache Kafka, and that has been this, you know, rapid rise up into the right over the years that we've worked on this. This has become one of the most popular open source projects in the world. This transition to real-time data, you know, it's not just a theory or something that's kind of in the very early days. It's something that's happening, you know, out in the world at large, in production, in some of the most mission-critical use cases in the world. You can see that in the adoption of this technology. If you look at the Fortune 500 companies using Kafka, it's incredibly prevalent, right? This is the ones that we know of. As with any open source thing, you don't know everybody who uses it, and so it kinda rounds up to everyone, right? This is being used everywhere. you know, we've started the commercialization of that. This is the set of, you know, Confluent customers, and you would see 9 out of the 10 top 10 travel and retail companies, 10 out of the 10 top 10 banks, 8 out of the 10 insurance companies, and so on. there's a lot, you know, here for us to, you know, go and pick up. We're really kind of just starting this journey of picking up all the use cases and customers, and we're excited about that opportunity in front of us. You know, there's a ton of open source use cases left to go get, and one of the exciting things about our cloud offering has been our ability to go convert those, you know, our ability to go pick up those customers and bring them onto our platform. This has been powered by an incredibly diverse set of use cases. There, there's a set of common use cases you would see in every industry. You know, building out real-time pipelines for data, the integration between on-premise and cloud environments, you know, building a real-time 360 customer view, fighting fraud and ensuring security, you know, acting as a backbone for microservices. Those are the kind of common IT-focused use cases. Some of the most interesting things are the ones that are unique to individual industries. You know, managing inventory, helping, you know, with claim automation in insurance, personalizing recommendations in entertainment. This list goes on and on, and you can read about these broadly on the internet because, as with any open source technology, there's lots of descriptions. You can read about how companies are putting this into practice in their business. The breadth here, I think, is incredibly important because, you know, the key analogy for Data in Motion is really to this world of data at rest, right? It's really to databases. One of the key theses we've had since the, you know, formation of the company is this is a broad data platform. It's something that's gonna form the backbone for data flow in all companies. Part of the proof points for that is you want to see not just, you know, one big use case or two big use cases, which would probably mean it would be better suited to a verticalized platform than a kind of, you know, broad platform technology. What you want to see is this breadth, where, you know, every company has not just one or two, but hundreds of use cases around the technology. There, there's been a set of trends in the world that have driven this adoption. Right? There's, you know, there's a kind of natural idea that things would move to real-time. Nobody ever said that they wanted their data slower or they wanted things to be more out of date, but there's, you know, there's been trends in our industry that have helped push this forward. I'm gonna touch on a couple of these. I'll start with just the general expansion of data, software, the cloud. You know, this is a story that doesn't need a ton of support. I think everybody knows, right? There are more applications. Those applications are more data-rich. The spend on cloud environments has increased. With that, has come a whole new set of systems that have data, and that leads to this big mess. Data is spread over more places than it was before, in part because it's easier and cheaper to adopt more of this stuff, in part because companies are just spending more. That combines with this second trend, which is, you know, the expectations that customers have on digital customer experiences have gone up. This has become increasingly a default way of doing business. Customers want the digital interaction to be a first-class citizen. They want it to know everything that they've done with the organization. They want it to be up-to-date. They want it to be personalized and contextual, and all of those things put pressure on the infrastructure layers. It's really hard to serve that if what you're doing in the back end is a bunch of old mainframes and, you know, unintegrated relational databases and antique messaging systems all spread across dozens of applications. That doesn't come together into something that meets any of those requirements. All of this puts pressure on the modernization of data flow, and in order to make this possible, you have to be able to connect this diverse set of operational data systems, SaaS applications, analytical systems, custom applications, and you have to be able to do it broadly across all the environments in real time. Increasingly, companies are split across geographically diverse environments. They're split across multiple different clouds and on-premise installations, and one of the key aspects of our technology is the ability to bridge all this, the ability to show up in each of these locations and each of these environments that a customer has, and be able to bring together all the data in real time across all of this. I'll touch now on the third one, which everyone has been waiting for me to get to, which is the rise of artificial intelligence. In many ways, you know, Kafka was created at LinkedIn to help drive the, you know, richer use of data and some of the machine learning and data-driven applications there, and that's continued. This is something that is really evolving and changing with the rise of generative AI. This is something that has, you know, some differences in its architecture. It has some differences in how it's used, has some differences in the implications from traditional machine learning. I'm gonna touch a little bit on this. When I talk to investors about this, I think broadly, people kind of understand the thesis of how we would fit in. They say, "Well, okay, it's kind of very natural that AI is gonna drive more usage of, you know, compute and more usage of data." I kinda get that, you know, you guys are gonna help move the data around, that makes sense. I wanna get a little bit more into the details of how this fits with what's emerging as the modern architectures, and why and how this is different from what already had happened with more traditional machine learning modeling capabilities, where you would see, you know, Kafka and Confluent broadly used today. The key difference for generative AI, and especially these large language models, is that the models are reusable, and they're not generally built by the company that's deploying it. This is something that kind of comes pre-trained, and it's trained off of, you know, public data sets, which are kind of broadly available. This training process is very expensive. In traditional machine learning, these models are built, you know, in-house by a team of data scientists off your data. With large language models, that's not the case, right? OpenAI is running some big process which builds this. What that means is that the key question is no longer the model building. It's really about, how do I combine this, you know, reusable model with my data, with my information about my customers, what I have, right? That's not gonna be really mostly in the training process. It's actually done at prompt time, in real time. There is a bit of a shift. Whereas with traditional machine learning, a lot of the focus was on model building, it was about, "Hey, how do I get data into my data lake? And how do I, you know, train ad hoc models off that problem by problem?" With generative AI, this now moves to inference time. The focus is now: How can I take the data I have about my customer and put that into action? How can it know the specifics of what you've done about my offerings, where things are at? How can I index that and get it ready to feed into a prompt at the right time for the right customer? You know, the roles that we have played in this with traditional machine learning was largely feeding data into the data lake and then creating some of the real-time features. What's happening is, you know, that offline batch training in the data lake, it was largely going off to OpenAI, but the, you know, the feeding in of real-time data is increasing because it's easier and easier to add more context there. So you can think of an example as a kind of, you know, if I wanna add a real-time chatbot that gives information about my business, I need to bring together the information about you, the current customer asking, about my services, and kind of broad policies, and I need to combine that with this kind of general-purpose model of the world. In order to fuse that, I have to do it by actually accessing all the systems that I have that have that data you know, this is increasingly the problem that our customers are looking at: How do I actually access this big mess of systems that I've got? How do I make sure that what I would be presenting in any output is up to date with your interaction with me, with the state of the world? The answer to that can't be that you just kind of ad hoc hook into every system you have. You need to have some kind of structured indexing, and this is where Confluent comes in, is that ability to actually bring your data together, feed it into the right systems to be able to index this for AI, and be able to look it up. I'll go through some examples of how this works out in practice in some of our customers, where you can see this happening live and in production. I think that this is actually an incredibly important role to play, and one of the reasons I think that that's true, especially if I was thinking from the perspective of an investor, is that there's a lot of uncertainty around the, you know, the final state for generative AI. You know, will these models be consumed via open source, or will they be consumed as a service? I think we're not really sure, right? If they're consumed as a service, who will be the provider of that service that, you know, gets most of the business? I think the answer is we're not sure, right? When they are consumed as a service, where will the value accrue? Will it be, you know, primarily in existing SaaS platforms, which maybe add new capabilities, or will it be in, you know, disruptive startups, which build, you know, purely around AI from the ground up? I think the answer is we're not sure. What are the kind of new specialized AI technologies that are gonna be needed to help adopt this? I think the answer is we're not sure, right? There's a lot of these areas where we actually just don't know what the end-state architecture is going to be in two years, let alone five years or the timeframe that this kind of stuff plays out on. There's a few things that we do know, though. We do know that it doesn't really matter how you consume the model, it's gonna use a lot of compute, right? That's, like, one really sure bet. The other thing we know is that companies are gonna have to really get a hold of their data and be able to feed this into these models to be able to take any advantage of this and put it into practice. When you think about the beneficiaries, I think this data flow and compute are kind of very obvious ones that are kind of right there in the supply chain for making this work. You know, I think one of the nice things about our area is, if you ask people, "Okay, how are you gonna build a data architecture that gives you this kind of real-time access to data that allows you to take advantage of this?" I think Kafka and Confluent is gonna be a very solid answer to that question that most people would give you back. That is why I think this shows up so commonly in the traditional machine learning architectures around using this, and this is why I think it also is showing up now in these generative AI applications that are coming to production now. Those are kind of the three driving trends of new things. The other trend that I think is on everybody's mind is just economic uncertainty, and this is what we hear from customers. This is one we've talked a lot about in our earnings call as one of the pressures on the business. It's worth noting that this is also, you know, something which has a silver lining, where a lot of our customers are going from thinking about, hey, big internal teams that run open source data platforms to just consuming something as a service. We've seen increased kind of build versus buy analysis and real interest in the TCO message that we have around our cloud offering. We'll get into a little bit of what that enables, why that's compelling to companies that are trying to be more efficient, and why also, like, access to data and being able to drive better automated decision-making, you know, is part of how companies actually drive savings and realize efficiencies out of the software stack that they've already acquired. That's a little bit about the underlying paradigm shift and, you know, the kind of trends that have helped drive this out into the world. That kinda leads us to Confluent and, you know, the content that we're gonna show for the rest of today. You know, our mission is to help accomplish this transition. We wanna set Data in Motion. We wanna make this a major platform in every company in the world. That's what we're here to do, and we've done this, you know, in a number of companies in really incredible use cases, and so I'll talk about a few of these examples. I'll start with Expedia. Expedia has been on this journey with us for a number of years. You know, they have a very rich and broad use of streaming technology that cuts across a lot of their interactions and helps integrate all the information about the customer, travel, et cetera. They think in a kind of stream-first way. Naturally, as they were thinking about how to integrate AI, and I think they were a little bit ahead of the curve on this one, I think Sam Altman was maybe on the board. They were early in the adoption of this stuff. The natural use case for them was interaction with customers. Despite being an online booking service, they said that 58% of their customers would contact them live. I talked with Anush from the Expedia team at Current last fall, and we talked through a little bit about this AI chatbot that they'd built, and it's actually really cool. You know, what they wanted to do was improve that experience. You know, if their customers have to call them, first of all, the customer has to wait, so it's not that great. Secondly, the experience of talking to the agent is not wonderful. How can they unlock that? It turns out it's pretty hard in a UI, right? Because what the customer wants is pretty diverse, and finding it in all the ways you click through is not that easy. The chat interface is actually very natural. They had existing ways of chatting with agents, so this plugged in, and it was incredibly successful. To do this, they had to actually bring to bear the information about your travel, right? What is the flights that you have booked? You know, which of those have actually happened? Are they delayed? You know, where's your stuff? All the things that people would actually ask about that were relevant to them. That had to be available to, you know, the agent that was doing, you know, the AI chat, right? The way that they did this is by combining, you know, a general purpose model with, you know, with that specific information about the customer and about broad questions, and putting that together in real time to produce, you know, contextual, relevant answers. If you ask, you know, "Hey, I missed my flight, can I find another one?" They have to know what your flight is and what the other options are, and be able to carry out that, you know, search. This was incredibly successful for them. You know, at the time that we talked at Current, it had already handled 29 million virtual conversations. This is one of the early AI use cases to really make it in production, and it had saved over eight million hours of live agent time. You know, this is a really big deal that unlocks a whole another way of interacting with them. Another customer is Michelin, the tire company. You know, so you would think: What on earth would making tires have to do with real-time streaming data? It turns out quite a lot. You know, this is a business that has an incredibly complex supply chain and an incredibly complex path out to market. That, that whole, you know, supply chain, inventory, e-commerce, and distribution mechanism, you know, is really being rebuilt around a live streaming view. So they know where things are, where things are coming from, all the quality control end to end, they can now tap into that and use this across different parts of the business. You know, they started with open source Kafka and then moved to Confluent Cloud, really to, you know, get something that they could depend on as a mission-critical system and realize the TCO benefits of that. You know, this went from just serving that kind of inventory use case, rolled out to then serve the Michelin Guide. I actually wasn't aware that the Michelin ratings for restaurants was the same as the tire company. Turns out it is. You know, the publishing cycle for that went from something that took two days. This is actually really important for them. You know, they're trying to go from something that's a published book to something that's an online app, where you actually find out about restaurants when you're looking for a restaurant. It's really important for them to have all the kind of data-rich features that a Yelp or other, you know, online service would have, and integrate that. This now cuts across many lines of their business. Finally, Instacart. You know, this is, you know, a very natural, real-time, technology-focused business. You know, the initial use case at Instacart was around personalization across kind of mobile and web, and giving a single view of their customer. This rolled out across many parts of the business, and is being used to capture changes from their databases, is being used to actually model a real-time view of inventory. You can imagine their core business is knowing what's in stock where, across all of these different stores, having a real-time view of those products as they evolve, and being able to provide that in a way that's contextual and send shoppers to the right place, you know, for the thing that you actually have, makes smart suggestions when they don't, et cetera. They've done actually a really great blog on the underlying architecture for some of the machine learning use cases that power this. This is actually an interesting one if you're interested in how we fit in, you know, both to the kind of traditional machine learning as well as, like, newer world. They had a very kind of batch-oriented machine learning system that they were using to do a lot of the inventory modeling, a lot of the predictive suggestions, and they talked about how they moved that to Kafka and Flink, another open source technology that we're adding to our stack, to enable the move into real time. This is an incredibly foundational thing for them. You know, in order to do this well, all the predictions and suggestions they have to be up to date with, like, what's in the store. Otherwise, it just, you know, gets the wrong answer. This was something they'd done over, you know, the last year and really wrote up a very detailed technical blog about. In the rest of the section today, you know, other folks on the management team are gonna go through a bunch of aspects of the business. We'll get into a little more depth on our product and some of the differentiation it has. You know, why is it that customers are coming to this cloud service rather than just using the open source? Well, we focus on three pillars of differentiation. You know, making something which is, you know, truly cloud native, making a complete data streaming platform that brings together all the parts, that a, you know, company would need to harness streaming data, and then making this available everywhere, in all the environments that they would operate out of. I'll touch on just a little bit of each of these. I'll start with cloud native. What does that mean? You know, to a lot of people, it's kind of a buzzword. You know, like, something is cloud native, you know, is there any substance to it? You know, it turns out there is, right? When you look at the difference between something like a Snowflake and then something like, you know, Teradata or Redshift, there really is a very, you know, different way of consuming something that's built as a cloud-native service, and this is something that matters a lot to customers. It's not just, you know, an adjective that's sprinkled on top. It actually goes right to the heart of how the service is built. What serves the underlying Kafka infrastructure in Confluent is something called Kora, and this is an incredibly sophisticated system that was built to run across the three clouds and provide tens of thousands of Kafka clusters as a fully managed service in a way that's totally elastic, and, you know, that actually provides these benefits to our customers. They can use just what they need. They can not think about how you pre-provision the right amount of storage or how you scale it up or down. Really, you know, consume this as a kind of utility. This provides better elasticity, it provides better resiliency, so we offer much stronger guarantees than anybody else who offers any kind of service in this space. It provides infinite amounts of storage. Then, you know, perhaps most importantly, it's backed by, you know, a much cheaper cost of operations. You know, we have, we think, probably about a 1,000x advantage in the human cost of operating this than, you know, a company that kind of tried to do it themselves with the open source. We have a very significant advantage in the underlying use of cloud infrastructure. A lot of that comes out of the multi-tenancy, but also just how the service is built from the ground up. This adds up to, you know, a really strong advantage in TCO. You know, something that's up to about 60%, you know, depending on the size of the team and the use case that we're targeting. This is really important. This means that just coming in for just the core Kafka bits, we can offer something to customers that is just not just more feature-full, you know, not just a better product, but also cheaper. They get something that's, you know, a better deal and does more. This is a really powerful message, particularly in the environment we're in right now, where, as I said, all the same pressures, you know, on the customer experience, on integration, on doing the new digital things are there, but at the same time, there's a pressure to do it more efficiently, right? To do it while spending less, to do it with fewer people. This is something that we enable. That's a little bit about cloud native. We'll get into more depth in that area and what it enables. Next is complete. What we offer is not just Kafka, and what customers need is not just the core stream of data. They need a set of rich capabilities on top of that. They need the ability to connect into all the different systems they have and capture the real-time feed of data. They need the ability to govern that flow of data across their organization. They need the ability to process that data in real time as it occurs, and they need the ability to share it, both within their organization and between organizations. These are capabilities that we bring around this core streaming abstraction that kind of complete the streaming platform. We'll get into more details on what our platform has here, and in particular, on the importance of stream processing. You know, this is an area that we think is gonna be a big expander of our ability to capture some of the value that we create. You know, today, a lot of the application development that customers would do is kind of out in their account. It's not really monetized by us. They might produce streams of data, we would monetize the stream, the application is theirs. As, you know, the stream processing capabilities come into our service, that application itself will move into what we're selling to the customer and what they're paying us for. It will drive up the consumption that they have with us. We're doing this with a technology called Flink. This is a, you know, very popular open source technology used for stream processing. We think that this is really important. When you think by analogy to data at rest, if you think about the stack for stored data, well, you know, the foundation of that stack would be the file system. You have one of these on your laptop, it's storing data. Usually, the layer for integration with an application is something which combines not just storage, but also data processing, and that's the database. When we think about where was a lot of the value captured in the data at rest stack, it was in that database use case, it combined processing and storage. When we move to Data in Motion, you know, maybe the analogy to storage is that stream, and that's Kafka. When you think about the integrated processing layer, that's Flink, and that's a huge opportunity for expansion for us. You know, Data in Motion is just getting to that layer, where now it becomes incredibly easy to build applications that tap into this, that use it. This, we think, can, you know, really exponentially increase the usage of streams. You would see that in the rise of the open-source Flink project, which is newer than Kafka, hasn't been around as well, as long, but is kind of following the same up into the right trajectory in terms of adoption. This is an area where we brought in a lot of the core developers of Flink through an acquisition, and we're adding that to our platform at Kafka Summit. Just recently, we announced the first release of this out to customers in early access form, and we'll be rolling that out through public access and GA through the rest of this year. Finally, you know, the third pillar here is offering this everywhere. This is something which is, in many ways, unique to streaming, that it has to be, you know, not just where one application is, but where all your applications are, 'cause it has to, you know, plug into all the different systems that you have. This ability to span all the parts of the company and turn it into one fabric for the flow of data, this is something that's uniquely important in our area. This has been important for us when we think about the competition with cloud providers, when we think about other on-premise technologies. You know, it's not enough to serve customers in one cloud or serve them just in their data centers. You have to be able to bring it all together and actually connect it together into one product, and that is a critical thing that we offer. Those three, you know, differentiating pillars, being cloud native, complete, and everywhere, those are things we've built around now for a number of years. You know, if you looked in our, in the documents around the IPO, we would have the same three pillars. We're building around the same three things. They matter to customers. We've built significant IP around each of them, and that's one of the things that's driven our success as a product beyond the open source technology. You know, we've really built, I think, a unique customer journey and go-to-market around this, and Erica is gonna go into more depths about what we've done. This comes out of the fact that the usage of this technology is a little bit different from other data technologies. You know, a database is the back end for one application, but it's kind of an island into and of itself. Streams, fundamentally, they go between things. They connect different applications, they connect different parts of the company, and that means that there's a very natural network effect that drives adoption. This starts within a company where maybe the first application comes for the features, right? If I'm a retailer and I'm building an inventory system, I need to bring in the stream of sales and to bring up the stream of product information. The next use case probably comes for those streams. You know, it comes because this is probably the only place you can get a real-time feed of what's selling or what products you have on hand. That application brings its own streams of data, and those streams attract new applications, and so on. That spins up the usage of this technology within an organization from that first use case to something that spreads. This happens industry-wide as well. You know, when we think about why has Kafka remained so prevalent over the years as this has spread, why have all the competitors kind of stayed in this single-digit market share? You know, why has nothing else been able to catch up? You know, in part, that's because we've worked hard and stayed ahead. In part, it's because, you know, there's a strong gravitational pull. If you are a system and you're building integration into the world of streaming, you're gonna pick the thing that connects to everything else, the thing that has the broadest ecosystem, that has all the connectors, that everybody knows how to use. Increasingly, this, you know, brings in a whole set of partners that help drive us, right? The cloud service providers, other technology, ISVs, the SIs that are taking and building this out into use cases within customers. These are becoming an important force for us, and we'll talk a little bit more about this in the go-to-market area. Then structuring these efforts around the natural progression within a company. You know, our customers don't start by installing a new central nervous system. They start with one application. How do we land for that first application as easily and in as low-friction a way as possible? How can we leverage, you know, kind of product-led growth and early sign-up and experimentation with, you know, as little friction just with developers? How can we help that grow within the organization and spread, you know, go up the chain of command and turn this into something that's a broad data platform, that's used across the organization, you know, becomes a mandated part of their data architecture, that grows into that central nervous system role? You know, being able to orchestrate this and build our teams, our product capabilities, you know, our consumption model all around this customer journey in a way that we're providing value to the customer in the right way, at the right time for where they're at, that's an incredibly important part of how we take this to market. That means combining the product-led part of the journey, the open source adoption, you know, the cloud sign-ups, the experimentation of existing customers with new features, the expansion to that on their consumption plan with an enterprise sales effort that is able to take this up the chain of command, that can influence data strategy broadly as, you know, we come to that. This is a critical part of how, you know, this succeeds in companies broadly. Finally, we'll touch a little bit on the business we've built around this, and we'll have some new material in a number of these areas. We'll talk about our execution so far. We'll talk about, you know, what has helped keep that strong and consistent. We'll talk about what gives us insight into where we're heading and some of the durability of growth. We'll get into a little bit more depth on the TAM and how to look at that in different, you know, with different lenses, from kind of a product lens and other areas. We'll talk about what underlies, you know, some of the customer growth and give a little bit more detail about how to think about our RPO and what that means. A little bit of information about how we're balancing growth and profitability. We'll give a little bit more depth on the growth and profitability model that we gave at our IPO and some refinements around that. All of this will come in the sections that are coming up. With that, I'm gonna turn it over to the first session, which is Shaun and Chad, our Chief Product Officer and CTO, who are gonna talk a little bit about our product strategy and how we're innovating for the future, and how we're really taking data streaming forward. Shaun? Thank you, Jay. It's, fantastic to be here. You know, professionally, I've been in technology for a very long time. I was most recently at Salesforce-owned MuleSoft, which makes software that helps enterprises join their systems together. You know, over the past 10 or 20 years, I've had the opportunity to chat with many, many business leaders across many business domains and many geographies. A common refrain across basically all of those conversations is just how difficult enterprises find it to join their data across their various different enterprise systems to deliver new customer experiences or employee experiences, and how difficult it is they find it to embed their data and their analytics in all of their business functions to make better business decisions. It is actually much harder than it looks, and it's holding organizations back from being able to take advantage of their data, and that's the giant spaghetti mess that Jay just talked about. I'd like to take a moment and talk a little bit about why this spaghetti mess exists, what is driving it, why is it pernicious, and how can we start to solve some of these problems? You know, in general, you can think about data in the enterprise as existing broadly across two major estates: the operational estate and the analytical estate. In the operational estate, the data is stored as part of running the business, and in the analytical estate, it's stored and reported on to try and understand how the business is performing and make business decisions. These different estates have very different users, different use cases, and they grew up separately, so they fundamentally have different approaches, different architectures, and entirely different technology stacks. Let me dig in just a little bit into the mess. In the operational estate, what you're looking at is the business applications that drive all the value because they're the things that actually enable you to do your business products and services. Think about something like a CRM for customer relationship management or an ERP for supply chain and logistics. Those applications obviously have need to store their data, so they have a database that sits alongside them, an operational data store. To be clear, the data as it sits on the disk is one thing, but a lot of the value is actually driven from the business processes and user interfaces that manipulate that data and use it and sit on top as part of the business application. Now, if you went back 15 years ago, we used to have very large, very monolithic business applications that could do entire large lines of business. If you went and purchased some stuff in a store, you might find that the cashier was using a terminal that was connected to a mainframe, and the mainframe application could do everything related to retail. It might have a product catalog, it might do payments, it might do inventory, it might do almost everything related to retail. In the modern world, it just isn't like that anymore. That is just too hard. Instead, we've separated out a lot of those different business capabilities into different business systems. Now, that's okay, but obviously, any customer experience or employee experience is gonna touch many of those different systems. They're gonna have to work together to deliver anything you recognize as a business capability. How do you solve that? Well, the truth is that in the enterprise, you write a lot of code, and you join these systems one by one, painstakingly together, to deliver these capabilities. Now, that's expensive and tedious, but it's absolutely critical because every organization is trying to deliver new products, new services, new business competitiveness, et cetera. That's what's going on in your operational estate. In your analytical estate, your business leaders are trying to understand how the business is performing and be able to make decisions. They're asking questions like: How is marketing going? Are we driving sales? Is sales going well? What is sales performance like? How is the overall business posture looking? They're asking these questions, and in order to answer their questions, typically what happens is there are batch jobs that suck the data from all of those operational data stores into one enormous data warehouse and try and stitch it back together. Context-free, stitch it back together to produce a view of what is going on inside the enterprise so that they can make business decisions. What's actually happening, though, is that the environment continues to get ever more complex. In the operational estate, you can see that there are an explosion of new, different business applications that are being deployed. Think of SaaS applications, microservices, new homegrown applications. Every time you're adding a new business capability or a new mobile app or a new voice assistant skill, you're adding new systems into the operational estate, those operational applications are bringing even more operational data stores and different database technologies to the operational estate. This is getting incredibly more complex, and you still have to wire everything together to deliver decent customer experiences. That same complexity that you saw in the operational estate is getting manifested back over in the analytical estate as well, which you're still trying to bring together all of that information from all of these different systems and try and make sense of it. It's become easier than it has ever been to stand up massive cloud data warehouses, but you're still fundamentally migrating data. You're usually doing it in batch because it was built to serve the needs of reports that were looked at once a week, once a month, or once a quarter. You're doing a batch, you're trying to produce a simulacrum of the world, something about how the world actually looked back in the day. Monthly isn't good enough anymore. You're looking at weekly, daily, hourly, for example, when the data is incorrect or it's not up to date or it doesn't exactly mimic the real world, it's hard to use it to make any significant business decisions. This is a very complex landscape that exists in the enterprise everywhere. You just heard me describe some of those problems. I also described how important they are in the enterprise because every organization needs to continue to innovate. What actually happened was the industry built entire technology stacks inside each of these estates to try and make the problem just a little bit easier. In the operational estate, you're looking at things like message-oriented middleware, application integration and governance, and change data capture. In the analytical estate, you're looking at tools like data pipelines, data processing, and data governance tools. Now, these fragmented tool sets, they kind of make it a little bit easier, but fundamentally, it's still a one-by-one game, solving it piece by piece by piece. What's worse is that you solve every problem at least twice, once in the operational estate and once in the analytical estate. It's unbelievably tedious and painstaking to try and make it work, worse, what typically happens is, when you move data from the operational estate into the analytical estate, you haven't brought all of its context. It's lossy. It's not a perfectly great representation of what truly has happened, even when you're ready to take action back in the operational estate, you can't easily do that. It's kind of a one-way door from one estate into the other estate. It's really messy, yet the world just marches on and is not waiting for this mess to get solved. Consider your average customer experiences these days. When I'm shopping on a mobile app for a specific item, I wanna know: Is the item available? When will it be delivered if I order it right now? I want you to give me personalized offers, personalized coupons, personalized recommendations. It all has to just work. Those are my customer expectations. When I actually go ahead, and I order the item, a whole bunch of things have to happen, again, all in real time. You have to process my payment. Sure, that's obvious. You also have to run fraud checks. You have to reserve inventory, the item in the warehouse. You have to reserve space for the item to be shipped on the next truck roll that's going to my ZIP code and going to my house. Even beyond that, it goes all the way down into your supply chain. If this item is going to run out soon, I need you to order more stock from the supplier so that the next customer is not disappointed, and I don't lose a sale because I don't have any more of that item ready to sell to you. A whole bunch of predictions and actions and reactions happen when any of these interactions happen. What you really need is you need all of your operational systems to work together as one, and you need all of your analytical systems to have the exact same ground truth view of the world, all at once. Now, when you try and brute force this problem by connecting systems one by one by one, that is what is creating this giant spaghetti mess, and that is how we got here. Putting it really simply, these two estates, the operational estate and the analytical estate, just cannot exist separately anymore. It is just impossible to make this really scale. That's what Confluent is solving from the ground up in a whole new way by delivering a central nervous system. What we're doing is we're making it easy for your operational systems to work together, transparently, trivially, and to enable that exact same view to be shared with the analytical systems so that everybody has the exact same view of the world. With Confluent, high-quality, consistent, up-to-date data is shared naturally all the way through all of your estates, everywhere in inside your enterprise. Now, this actually has massive benefits for everybody. For the CTO, this is a whole new architecture that enables them to build new applications that deliver business services, business capabilities, and drive business automation faster than they ever could before. Think about getting more from your development efforts than you ever could. It also benefits your COOs and your CROs because they get unparalleled business visibility into everything that's happening in every one of the business systems, all in real time. Beautiful, incredible business-level visibility. We provide a better way to put data to work. We package data out as products. We're changing the conversation from, "What is my data? Where is it? How do I access it? How do I query it? Is this even up to date? Is this even accurate?" To, "What is my data, and how can I put my data to work?" That's the whole point. In doing so, we're rethinking and reimagining how data is spread across organizations and how teams, applications, and systems work together to deliver business outcomes. We liberate data from silos, we make it all interoperable, and by doing that, we unlock endless different use cases across all of the businesses where we're delivering a central nervous system. You may be wondering, how does this relate to the various different technologies I just talked you through? We have been doing this for years and years and years, and it has not been solving this problem. If we keep doing more of it is unlikely it is going to continue to solve this problem. Confluent is reimagining how this problem is solved from the ground up with a central nervous system. In the operational plane, we make integrations between applications, making applications work together to deliver new products, new services, business automations, much easier than it has ever been before. In doing so, we can replace a lot of the need for message-oriented middleware, application integration, and change data capture. Over in the analytical estate, we make it easy to gain visibility into everything that's happening in your business, all of your most important data, all in real time. By doing that, we replace a lot of the need for these old batch, ETL, ELT, data pipelines, data processing pipelines that exist in the enterprise. I want to be really clear about one thing: We are not another technology joining this hodgepodge of technologies. We are fundamentally solving this problem from the ground up, and we are unifying these two estates. We are not in neither the operational estate or the analytical estate, we are in both, and that's what enables us to empower all of these different use cases I have just been talking to you about. It's that which is what's driving the $60 billion addressable market for Confluent. We are breaking down data barriers in the enterprise. There's no need for more fragmented, duplicative technology and processing layers across different use cases. There's no need for siloed processing systems across the analytical and operational estate. We are purpose-built to set Data in Motion and to solve the problems I just talked you through. Looking over the horizon, we're going to take pieces of this pie as we deliver on this central nervous system, and we're gonna unlock significant growth opportunities. Now, these are really important problems, and these are problems I care very deeply about. Bridging the divide between operational and analytical, decentralizing data strategy, and democratizing innovation are passion areas for me as an individual, and they make me very proud to be able to join Confluent. I have spent many years of my life working on this problem in my prior roles, and it's very clear to me that doing more of what we have already done, doing more of these old approaches to this problem, can never solve this problem at scale. It is impossible. I'd encourage you to consider that when you look at this $60 billion TAM, that doesn't even include various things. It doesn't include the opportunity ahead of us as we monetize more of the open source Kafka ecosystem and bring them into our managed streaming platform, where they can get a lower TCO and a better cloud-native offering, and replace their infrastructure and personnel spend. You also need to consider the incredible rise of more and more data in the enterprise, more and more real-time data in the enterprise, driven by customer expectations and AI, which I know all of us are looking at right now. That's kind of what we're doing. Let me dive in just a little bit into how we are doing it. Obviously, it all starts with streaming. With streams, data is not static, it's not passive, it's in motion, it's constantly moving, constantly evolving, constantly being processed and shared everywhere across your organization to power your business. We make it easy to connect to the static and historical data and bring in data from wherever it sits, anywhere across the organization, as streams into Confluent. We take those streams, and we process them, enrich them, evolve them, improve them, and we end up with streams that actually are carrying the most important data at every one of the enterprises we're working with, the lifeblood of the data as it flows through those enterprises, and we take that data, and we govern it. We make it safe, we make it reusable, we make it known, discoverable, and so that other people can safely consume it and use it. We enable that data to be shared as data products elsewhere inside the organization through a catalog where everybody can find, manipulate, and understand all of the data that's going on inside the enterprise. That data, those real-time data assets, can flow like water into the operational applications I've spent so long talking to you about and into the analytical applications that I've spent so long talking to you about. The same source of truth into any type of application, build it all from the same ground truth of source of truth. Our product investments, you heard Jay talk about them, they primarily are across those five dimensions I just talked about streaming, connecting, governing, processing, and sharing. I'd like to dig in a little bit more into our product innovations in each of those different areas, and for that, I'd like to welcome up my friend and colleague, Chief Technology Officer at Confluent, Chad Verbowski. Thank you, Shaun. You know, I'm excited to be at Confluent, where I can bring together my experiences from creating a bunch of cloud services like Azure SQL Data Warehouse and Google BigQuery, along with what I learned from my time in Microsoft Research and Bing. It's taught me that there are five key foundational areas that we need to innovate in to be able to build a data streaming platform for the enterprise. Today, I'm gonna walk us through some of those unique things that Confluent is doing to bring each of those five areas together. The first area, of course, is streaming, and it's powered by our cloud-native, complete, and everywhere Kafka technology. Second, you need to easily connect this streaming core to all of your existing SaaS, database, data warehouse, and any other system in your enterprise that's producing or consuming data to enable real-time interactions across your entire data ecosystem. Third, you want to enable everyone in your enterprise to be able to work with and process data in real time. You're gonna be mixing, filtering, transforming these raw inputs and reusing those outputs and combining them to create even more valuable insights. Fourth, this rich data set is then shared, reused, and perhaps even further processed by other departments or even partners to your company. Finally, for all of this to be safe and compliant, we ensure that the data, all of these operations, and the interactions are securely governed end to end. Confluent Cloud covers all of these, the complete stream, connect, processing, sharing, and governing journey to make it easy for organizations everywhere to unlock all of their data. Now, let's talk first about streaming, where we've reimagined data streaming to be secure, resilient, scalable, cost-effective, and of course, it has to be everywhere that your data exists. Now, as Jay mentioned, Apache Kafka has become the foundational technology for data streaming. 75% of the Fortune 500 companies and nearly all of the major players in every segment are using Kafka. This is a huge win for Confluent because it means that most enterprises are already excited about using streaming, and therefore, it's very easy for them to adopt and consume our products. Now, anything great for your business is, of course, gonna grow. It's gonna expand with new use cases, more teams are gonna adopt it's gonna be deployed in even more environments, then these products are gonna get mission-critical. As the usage of this grows, so do the expectations of these systems. You're gonna need to scale them, and you're going to need to provide them with some major operational guarantees, and that's gonna take more engineering resources to run. Operating open source Kafka is hard. You've got to deal with sizing the infrastructure, load spikes, expanding your clusters. You have to manage the costs. You have to provide it with high availability and disaster recovery, and of course, you've got to secure it. You've got to manage the upgrades, not just to the Kafka system, but to that underlying infrastructure as well. This is why it's such a huge opportunity for a fully managed, cloud-native Kafka service that handles all of these complexities. This enables companies to innovate faster, lowers their TCO, and it eliminates all of these operations so companies can focus on their core business innovation. Being truly cloud-native is not as simple as just taking your open-source software and running it on a bunch of machines or cloud instances. To be truly cloud-native, what you need to do is provide something that is very cost-effective, it has to be efficient, it has to be secure, and it needs to serve literally thousands of customers on that same set of infrastructure, not just one. To be a multi-tenant, scalable solution with these isolation and optimizations that you need, you need to work very closely with every cloud provider. Not every cloud is created the same. They have their own ways of building networking, processing, storage, security. If you're gonna optimize your system to have maximum performance and cost optimizations, you need to tune what you're building to every one of those environments. This proprietary innovation that we have at Confluent enables us to have differentiated cost efficiencies and reliability, and differentiated ease of use, with capabilities like elasticity for automatically sizing what resources are needing for your instances, pay-as-you-go pricing, and of course, eliminating this operational burden. Confluent Cloud is the easy button for using Kafka. This is why we built Kora. It's a new engine, 100% compatible with Apache Kafka, and it's a completely reengineered system from the ground up. We've deployed it across 85 regions in the world, and it runs on all three major cloud providers. It replicates data across availability zones, so it can provide strong availability guarantees, and again, eliminates that operational burden. We've done that work with the cloud providers, with close relationships between our engineering teams to really maximize that elasticity, resilience, and scalability with over 5 million engineering hours put into this core system. In 2023 alone, with Kora, we've already delivered a 4x improvement over open source at Apache Kafka performance, we've also delivered improvements to our data and workload balancing, along with massive improvements in our network performance and cost optimization when we're driving very big data loads. Kora powers more than 10,000 clusters on thousands of customers running their critical workloads, saving them time and money, of course, infrastructure and operational management costs. On a number of dimensions, Confluent Cloud is at least 10 times better than any self-managed open source system or even some of those semi-managed systems, where people are just taking open source Kafka and running it in the cloud. With our fully managed cloud services, our customers no longer need to worry about any pre-provisioning infrastructure. They can elastically expand from 0 all the way up to gigabytes per second to meet the demands of their business with just a few clicks and without any manual intervention. With automated load balancing and continuous monitoring, We provide guaranteed reliability of 4 nines, and our customers can be confident that with their mission-critical use cases are being powered by highly reliable and resilient platform. We've eliminated limits on storages, so customers contain all of their current data and all of their historic data as well for as long as they want, eliminating the burden of balancing the need for longer data retention with rising infrastructure cost. This means that customers can use Confluent as the source of truth for all of their data across their entire ecosystem. These innovations also do double duty here. They provide us with a dramatic advantage in delivering our cloud service. For instance, we have less than five Kafka engineers for running these tens of thousands of production systems. This gives us a cost structure for operations that we believe is over 1,000 times better than our competition. Now, with our deeply differentiated stack, we remove the complexity and the operational burden of this do-it-yourself Kafka, and we're driving compelling savings and faster ROI for our customers. An independent total economic impact research by Forrester highlights how Confluent has helped customers save millions of dollars in TCO, with a payback period of fewer than six months and an ROI of more than 250%. Now, in our mission to build a complete data streaming platform, we need to connect all of the systems that you have in your enterprise to make it easy to ingest and egress data continuously between Confluent and any system or application, whether that system is on-premise or in the cloud. While capturing dynamic data is foundational to our data streaming strategy, we also provide the seamless experience for connecting to any non-streaming source. We minimize the effort to integrate and ultimately unlock that data so that you can enhance your existing real-time use cases. Our connect strategy really has 3 core capabilities. Number 1, it reduces the TCO of integration with our prebuilt and existing managed connector offerings. Number 2, it enables custom connectors so that anybody can build unique connections to any system they have in their enterprise. Number 3, we wanna ensure that every customer can use their tool of choice, so we have native integrations with any of the major native applications that they're using. Let's dive into each of these 3 areas. In the streaming world, we have the best ecosystem of zero code connectors built out across the various data estates, operational systems, analytical systems, and SaaS platforms. You don't have to know Kafka to get data into and out of our platform. We have over 120 pre-built connectors and over 70 fully managed connectors for our customers to use, and an ecosystem of integrations that our community continues to build to connect their data sources to and through Confluent. This means our customers can quickly and easily connect their systems with Confluent Cloud using a few clicks or lines of code. We take care of all the heavy lifting so our customers can focus on unlocking the value from their real-time data estate. Now, this can save them typically 1-2 engineering years of work or investments they might have in proprietary tools that can be prohibitively expensive. Today, our customers are running thousands of connectors across our Confluent Cloud and Confluent Platform products. When customers begin to use these connectors, this of course drives more data into our Kafka Platform, but that in turn leads to more processing. This processing is powering a flywheel to increase consumption of Confluent resources, and this highlights the networking effect of real-time data. This flywheel spins even faster as customers push higher volumes of data to power their generative AI workloads. One thing I'm quite excited about is what we call custom connectors. It's a new capability we launched last month at Kafka Summit London, and custom connectors gives our customers added flexibility to simplify integrations with any possible sources that they have. Let's say a customer or some advanced user has a specific requirement to integrate their particular application or custom source of data. Now with a few lines of code, they can create this. They upload it to our Confluent Cloud. We take care of everything to do with running it and making sure that it's operationally sound. They don't have to worry about the infrastructure details of managing this connector because we take care of it for them, saving them time and the cost of data integration. Now, in addition to enabling our customers to bring their own connectors, we also wanna make it easy for them to use their tools of choice. We never want them to leave the UI of the tool they're working in, so that they can seamlessly connect with any of the data that we provide in real time. We're working very closely with our technology partner ecosystems to build Connect with Confluent, so they can build these native integrations into their own platforms. We're launching this next month, and we have many partners in our ecosystem already lined up for the launch. Continuing on our mission to build this complete data streaming platform, we need to be able to easily derive value from these real-time data systems. We do this by providing powerful processing capabilities to enable filtering, joining, analyzing, and reusing these results to generate actionable insights in real time. I want to take a moment to signify the importance of stream processing. Any data or event has, by itself, some intrinsic value, and traditional messaging systems or even simple point-to-point API connectors, they can capitalize on, but they're missing a huge opportunity. That is that data is more valuable when it is combined with other data. If you combine one event with data from another, you can get much deeper insights and way more meaningful actions. For example, let's say you're making a purchase on your favorite retail site. If that retailer is able to join that purchase information with your customer information to know the location you are in, then they can provide much more meaningful shipping options. If you take that a step further, if they can further analyze that information about what's being shipped to your location, they can, of course, bundle many of these together and ship them to you in a single package. This is an example of just where you can take one real-time event, join it with some of these offline systems, and produce even more value. This ability to share, reuse, and enhance processing results in the same data reuse over and over again. This is the power of stream processing, and it's one of the key value propositions of our data streaming platform. While Kafka Streams has dominated in building standalone applications, there has been a whole family of standalone frameworks that have emerged to manage the lifecycle of application development. Because there are ecosystems around streaming, this is an opportunity for us to pull in additional things into our platform. In the ecosystem of stream processing frameworks, there has been one technology that has come to the forefront in this space, and it has become an emerging de facto standard, and that, of course, is Flink. Flink, as Jay pointed out, is one of the most popular open source projects ever. We can see that it's growing at a similar rate to what Kafka was about four years ago. In fact, we're seeing a high level of overlap between companies that were early adopters of Kafka now becoming adopters of Flink. Many of these innovative companies are use these technologies together to support mission-critical workloads in their enterprise. This means that just as Kafka is the de facto standard for streaming, Flink is the de facto choice for stream processing. For example, the AI recommendation system powering TikTok, called Monolith, is built on Kafka and Flink, so they can respond in real time to changes in user preferences. There's a lot of synergy between Kafka, Confluent, and Flink, but Flink doesn't have a cloud-native offering to make it easy for customers to adopt. Customers are spending a lot of their time and engineering resources on operating, maintaining, monitoring, and scaling Flink infrastructure. It naturally makes sense for Confluent to deeply integrate Flink into the best-of-class cloud-native Kafka offering that we have. It allows us to reuse our expertise in building this cloud-native offering, and it also allows us to reuse some of those native, cloud-native primitives to provide a Flink-native service as well. Immerok is a company founded by Apache Flink community members, and six months ago, Confluent announced our acquisition of Immerok so that we could have the preeminent Flink leaders at Confluent as well. This really complements our existing stream processing offerings, which extends it with rich capabilities for pro code and low code developers or even data analysts. All product development and data engineering can be built from this one data estate. We will provide a fully managed cloud-native Flink service on all three of the major cloud providers in every location that we have our Kafka service. Our customers won't have to worry about infrastructure details or pre-provisioning of any infrastructure. They will only pay for what they use. Today, Kafka Streams is a library that runs in the client infrastructure, and it isn't monetized by us today except for the data it produces. In the future, we will bring that processing into Flink, so we can monetize the processing as well as the data that's being created. We're integrating all of our stream processing capabilities to provide an end-to-end data streaming platform that makes it operationally easy to access all features needed to build streaming applications. Often, the most valuable data to work with is also the most sensitive, and this is why it is foundational for our data streaming platform to guarantee that all the data and processing is end-to-end governed. We ensure the observability, the compliance, and the confidentiality of data that's always on the move. The explosion of data and its commoditization has given rise to new challenges for enterprises, like how to foster collaboration between producers and consumers with the right checks and balances in place. How do we free up and unlock data to innovate and push businesses forward without losing control over that data? How do we make sense of data and ecosystems as they grow and become more complex? Organizations balance this data access between innovating faster, onboarding these use cases with security that they need to protect and meet enterprise regulatory and compliance needs. Solving this is critical. In fact, a lot of companies have had to build this themselves because there really was no commercial offering that's provided governance for Data in Motion. Many organizations already use Kafka and trust it for their most important data. They unlock this critical data for reuse in any applications across business boundaries, and it's really critical for accelerating their agility in the enterprise. It's only possible if organizations can do this safely. Confluent Stream Governance Suite empowers distributed teams of engineers to have this balance, this balance between moving fast and staying secure. We provide three things as part of this offering. The first one is our Stream Catalog, which enables people to increase their collaboration and productivity with self-service data discovery that lets teams classify and organize the data and find the streams as they need them. The second part is Stream Quality, which enables customers to deliver trusted, high-quality data streams to drive the business with data integrity. Third is data lineage, which enables customers to understand the complex data and uncover insights and value with their interactive end-to-end maps of data streams. Stream Governance is key to fostering collaboration and knowledge sharing necessary to become a data-centric business while remaining compliant with ever-evolving regulatory needs. One area that Confluent really dominates in and is enabling operational workloads, mission-critical workloads that run the core businesses in our customers, and it's handling things like social security numbers or credit card numbers. We make sure that our customers are confident that only the right people with the right access can work with this data. Our innovations protect customer data confidentiality. They help maintain contracts. They enable early detection of any problems. They can assess the quality, and they provide the fidelity and shape of the data, making it easy to discover inconsistencies and anomalies quickly. This empowers developers to use all of the data that's at their fingertips more confidently and quickly. This creates a network effect. A well-governed set of data is more reason to use the Confluent data streaming platform as an interchange for all of the data in your enterprise. The final piece of our data streaming platform is to enable data sharing. Sharing enables self-service and up-to-the-date, trustworthy data and analytics effortlessly for everyone in an enterprise or even to your partners. Developers spend a lot of time discovering trustworthy data to use for their applications. We make it easy to do this discovery, consumption, and sharing so that developers can spend less time looking for their data instead of just consuming the right data at the right time, in the right format. This means they're better equipped to use data to power their real-time applications without worrying where data is or how to access it. Customers can easily integrate data into real-time systems. Ensure that it's secure and safe for sharing. First, we're building a developer portal to enable easy discoverability and access to ready-to-use data. Using this portal, a developer can search for a data stream, click a button to request access to it, and this request gets sent to a central operator, where it can be approved and they'll ultimately be granted access to use it. In addition, we're the only company to support any combination of on-premise hybrid cloud and multi-cloud deployments for your data streaming architecture. Our customers can connect all of their systems, all of their data and applications across their distributed architecture and unlock valuable data in these siloed environments and stream that data in real time wherever it needs to go. No native-hosted cloud offering can deliver this, and if they ever choose to migrate from one provider to another, we can help our customers maintain optionality and eliminate lock-in with seamless data mobility across clouds. Lastly, we make it easy to share data in your organization. We also deliver that same simple real-time experience between organizations because we use common. The use case is actually quite common. There's actually no organization that exists in isolation. For example, inventory companies or delivery companies or even financial trading companies, they're constantly exchanging real-time data internally and with their external partners. They're often using very primitive methods. They're sharing with things like file uploads or email attachments or one-off APIs, that means that the data is stale, and every new source requires yet another IT project to get things wired up. When data flows across organizations in real time, both parties have the same view of what's happening and what is possible. Stream sharing provides the easiest, most secure sharing across organizations in real time with just a few clicks. Now I'd like to invite back Shaun. With some of the aspect of our technology, which will unlock the Data streaming era and data streaming across the organization everywhere. You all would know that it's not just about technology, it's also about how that technology is brought to bear in every enterprise, and there's a lot to do to truly unlock that potential. One of the biggest limiting factors of the way data is used in most enterprises is actually the people. It's not just the people that have access, it's the fact that most of the most important data that's stored in streams are available in the enterprise is often only accessible to developers. I love developers. I was a developer. I am a developer, but I recognize that there are actually only 20 or 20 million or so developers in the world. 20 million developers. There are many times that data consumers in the enterprise. If the data is just available for use and leveraging by those developers, it can't ever reach its potential. That's why we've implemented technology like Stream Designer. Stream Designer, which you can see here on the right, is a completely no-code technology that enables a non-developer, a data consumer, to access all the streams that exist, understand what is in them, use them, filter them, evolve them, improve them, process them, and to produce your own new streams that go out the other side, and those streams can then be consumed by that person, but by other people inside the enterprise, not just other data consumers in the analytical estate, but even by developers to build new applications to deliver new business services as well. It's a true unification of the way data is used in the enterprise and democratizes data. Of course, if we're going to unlock the data for use by everybody everywhere, we need to do a couple of things. Firstly, the data needs to be governed. You heard Chad talk extensively about governance and making certain that the data is reliable. We also need to be able to deliver a cloud platform that meets the world's highest standards for security, governance, compliance, and reliability. We have been working on that for a very long time. We have the industry-leading controls, monitoring, compliance, certifications, and we're continuously reaching higher to offer the world's highest standards to our customers. You know, this has actually been a journey that we've been on for a long time. Apache Kafka was invented about 12 years ago by Jay and the other founders of Confluent as part of LinkedIn, and Confluent has been around for nine years. Over that time, streaming has gone mainstream. You don't have to explain to anybody what Kafka is anymore. You don't explain to people what streams are. Streams are effectively ubiquitous. In all of the largest businesses, in all of the most advanced startups all over the world, streaming is a thing, and they are all leveraging it to deliver brand-new, incredible experiences. Our journey continues, and we have an opportunity to take a lot of that open source Kafka streaming market, and to bring it into our cloud data platform, where we can offer a lower total cost of ownership, and as Jay mentioned, yet offer more features, more capabilities. It's a better product that costs less. That's our opportunity to convert more of the Kafka ecosystem into our cloud. It's not just that. Even that is just the journey beginning. You've heard us talk a lot about streaming, putting Data in Motion, about it being a way to convert data From being static into flowing through your organization. To truly achieve that, we have to deliver a complete data streaming platform, not just streams, a data streaming platform. We have to be able to stream, connect, govern, process, and share. That's the data streaming platform that we're building. It's that that unlocks the $60 billion plus market opportunity that I talked to you about earlier today. With that, I believe we're gonna break into a quick Q&A session. Welcome back up, Jay and Chad. That's right. We're gonna be doing a 15-minute Q&A section, focusing on products. We've got a mic here, and Faheem has a mic on the other side. Take a minute here, set things up on stage. Go ahead, click once more. Oh! Yep. Hey, guys. Over here, Brad Zelnick, Deutsche Bank. Thanks so much for today and for the presentations. Just a simple question: can you dive in a little bit on the investments required to bring Flink up to the standard of where you are with your core products today, and just how you see the evolution? Like, when we look back in five years, what will be the relative scale of Flink relative to the rest of the business? Thanks. Yeah, do you wanna take that, Chad? Yeah. One of the things I'm most proud of in my career, actually, I've actually never seen this anywhere else, is we were able to decide we were doing Flink. We bought Immerok, and we announced our first product in less than six months. I've been just super proud of the ability for us to take a company of, you know, 20, 30 people, merge that with a company of, you know, equal size, and get those teams to work together. We really were able to take something that was designed to be run more or less on machines, non-cloud native, and we actually have a multi-cloud offering. I think it speaks to the innovations that we've made in Kora, that we were able to take this new set of people, new set of technology, and really deliver that within five months to make it profitable. I'm very excited about what we can do. I think Flink brings one more piece to Confluent Cloud that we didn't have before, which is hosting third-party code. A lot of our work is really around making sure that we can do that super securely and super scalably, so that we can basically run and power any of the streaming applications that are out there. I would just add the three pillars of differentiation I talked about, you know, being truly cloud native, like really building from the ground up to be a cloud service, you know, being a complete platform and being everywhere. You know, when we thought about it, that really applies very directly to Flink as well, right? There's an ability to really build a processing layer as a service. I mean, we've seen this with the cloud data warehouses, as Chad would know much better than me. You know, the ability to integrate this with Kafka, with our governance, truly make it one product that all fits together, and then offer it to customers, you know, multi-cloud across all the environments they run in. You know, very much the same kind of three pillars of value, and I think that's helpful for us, just, you know, both in terms of guiding our investments and then also just kind of simplifying the message to customers. You don't wanna have different modules with different value propositions. You wanna have just a very simple description of, you know, why you're better, why you're the thing that they want. When we think about where are we investing, I think those three are a pretty good guide for us. Obviously, a lot of details of, hey, what does it mean to be cloud native? What is the integration into the rest of the platform, et cetera? You know, those are kind of the right buckets. Sanjit Singh, Morgan Stanley. Jay, I wanted to revisit the topic around AI and this sort of next wave of Wave 2 AI applications that are going to get built, and where you sort of see the logical flow. As I sort of imagine it, we imagine a customer or stakeholder user asking a question in natural language, and we have to surface that response that could span both structured data as well as unstructured data. I guess the question is like, what is Confluent's role in providing those answers, and what% of applications do you think are going to need to be real-time going forward? Is there any difference between the real-time requirement for this next generation app versus the stuff that we've built the past couple of years? It's a great question. There is certainly a difference from the kind of more traditional machine learning architectures and the newer AI architectures. I would say that that difference tends to be less about the model building, which to some extent is kind of done for you, or at least mostly done for you by the kind of foundation model builder. Our role is very much, you know, the connectors that tap into the different applications, bring together data, kind of munge it into the right format, help it be indexed for surfacing in these applications. Some of those details, I think, are still being filled in around us, right? Will the model that you're integrating with be like an open source model that you kind of fine-tune yourself, or is it gonna be something that you consume as a service? I don't know. To some extent, it doesn't change that architectural pattern or our role. You know, will you be indexing and storing data in more traditional systems or into some kind of vector database? Will there be other aspects of integration into LLMs that have yet to emerge? I actually don't know the distribution of that, but it doesn't really matter to the end-state architecture. The, the key problem is the data that's needed for these models is locked up in this mess of old systems. It was not meant to serve any kind of real-time, you know, query requirements, especially broadcast out across a bunch of systems. That's the kind of key problem that has to be solved to, you know, really unlock those use cases. That's kind of a very obvious application. I think there's other integrations beyond that, right? Chad mentioned kind of the usage at TikTok, which is, you know, not just bringing together data in real time, but actually training the model, you know, for you based on responses as they come. That's much more sophisticated. I think that is. You know, that kind of real-time learning probably is an end state for a lot of these things, but it's not the starting state for most companies. Most companies are just trying to figure out how to, you know, combine some version of their data with some version of intelligence and get that into a usable product right now. I think in these early iterations, we're gonna see, you know, a lot of that kind of integration problem for data, and we'll see a lot of, you know, applying AI to an incoming stream to augment it or improve it, or classify data in different ways. I think those are the use cases we're hearing from customers, and I think are likely this first wave. You know, beyond that, I think, the world opens up a lot more. We're still talking about, you know, language models, but there's a lot more that can be done on different data sets and applications in many other domains, That's where it gets harder and harder to predict where this could go. I mean, I think it'll be interesting, and I think there'll be a role for us in a lot of those areas, but that's where it gets beyond kind of what people are doing today with what the technology does today. Hi, Dave. Mark Parr from KeyBanc. Good to see you. I'm sorry, the kind of a riff on Sanjit's question, you said that obviously more compute, more data is a, the, it's, you know, a feature of having to do real-time GenAI. when you're talking about the fact that you need to interact at prompt time and at inference time- Yeah. I'm wondering what kind of feedback you're getting already, whether it's from commercial software application developers who are trying to develop applications and want that kind of data integration at prompt time, or whether it's enterprises that are seeking to build their own applications. So what are you hearing back from customers so far in terms of their needs for real time in order to hit at prompt time? The motivation for real time, you know, I think is pretty clear when you consider the concrete use cases, right? You know, that would be the case whenever the data that would be the right answer to your question needs to be kind of up-to-date and accurate. The inclination people have is they think, well, what will happen is we will take, you know, these models, and we'll train them on the enterprise data, and that will be the way we answer questions. I think that that's probably not the most obvious direction for this to go. It could be, right? That was kind of the TikTok example Chad gave. The challenge is basically a fewfold, right? One, that enterprise data is private. You know, two, it has to be up to the moment, not, like, a year old or even a month old, or even a, you know, a week old, right? Like, when you're asking Expedia about your flight, you don't wanna know the state of your flight, like, a week ago. You wanna know, like, right now. In fact, when people interact with a company, it's about something right now. the, you know. I think the majority of use cases will have some aspect that needs to be in sync with what you've done with the company recently, what you just bought, your interaction on the web, your interaction in person. I think a lot of these kind of copilot-type or agent-type use cases will have some element of that needing to be in sync with the world, when that's not the case, customers are just confused. They don't understand why they're getting something that's, you know, out of sync, wrong, hallucinated, not right. I think, you know, if you think about how this is working in practice, it's actually not that difficult to understand, right? You're kind of materializing a lot of data about the customer and what they're doing in their interactions and other things, and you're, you know, effectively augmenting the prompt that you're sending out to the language model and saying, "Hey, It's not just: Where is my flight, right? It's: Where is this customer's flight, given all this information about the flights they were supposed to take and the status of those flights and alternative routes and blah, blah. That enables the kind of complete and rich and correct answer off of the up-to-date status of the world. Hello, Jason Nader with William Blair. First, just a clarification on the $9 billion in the TAM in the database market. Can you talk about what you're addressing in the database market? Secondly, can you talk about the differences between Flink and Spark ecosystems? Yeah. I'm assuming that's where you're gonna run up a little bit against Databricks. Can you talk about that relationship as well? Yeah. When we think about where we're kind of taking spend from, as you said, a portion of that is out of the database market, and that happens in 2 different ways. One usage for databases is, hey, you're trying to do some data processing. You're effectively loading it into a database, running some processing, and shipping out the result. Often, it's some kind of batch job that would happen, you know, at the end of the day or periodically throughout the day. That's the kind of thing that's just inevitably going to real time, that's moving to streaming, right? For the things where there's a lot of pressure to make it faster, it'll happen sooner. For things where there's less pressure, it will happen over time as it becomes cheaper and more convenient to do it that way. Over time, I think, you know, this kind of batch processing is really kind of an odd artifact of mainframe computing that just doesn't really make sense with how the world works or what the needs of applications are. It's, it's literally just been held back by aspects of technology that made that more convenient, that are, you know, entirely going away, right? There's no inherent limitation of streaming that means it has to be more expensive or that it won't get the right result, or there's no inherent limitation on that that says, "Oh, there's gonna be this category of things that will always be batch forever." It's really just a question of, you know, convenience, you know, driving down cost, building the ecosystem, you know, to be able to make that the most obvious route. I think, you know, we'll see that conversion happen over time. I think that there are a blend of other database use cases once you take that kind of obvious batch processing stuff, that move beyond just kind of the, you know, the processing and into some of the materialization and storage of data. I think we'll see some of those, you know, move as well. To your second point on, you know, what's the difference between Flink and Spark? You know, there's some overlap. You know, Spark was kind of a batch computing framework that added some streaming capabilities. You know, Flink was more directly targeted as a pure streaming framework that added batch processing capabilities. So there's some analogy between the two. When you look at the user base, I would say that where Flink has really, you know, grown and thrived has been these kind of application, you know, run the business use cases, and where Spark has really grown and thrived has been the kind of data scientists and data science workloads. Obviously, there is some boundary between those two countries. You know, there's and there can be some border dispute there, but by and large, you know, we don't see the data warehouses or, you know, data lakes or lake houses as competitors, and we certainly don't kind of battle deal by deal with them in any way. You know, right now there's a very kind of rich and exciting competition in that analytics world. The use cases targeting streaming tend to come to us, and often data flows out of that streaming estate into all these things, into the data warehouse, into the lake house, into the new generative AI systems. You know, there's a bunch in that area that we're helping to feed. You know, by and large, these are partnerships, and we've won Partner of the Year awards from Databricks, from Google around the analytics area, for that reason, because it is, you know, primarily cooperative rather than competitive. All right. Thanks, everyone. This is the end of our first product-focused FAQ. We have more time later in the program with the rest of the management team as well. Right now, we'll take a short break, and we'll come back in five minutes. Thank yo. Hello, everyone. We'll resume our program in one minute. Please be seated. We're about to get started. All right, welcome back, everyone. Right now, I would like to introduce our next speaker to the stage, Erica Schultz, our President of Field Operations. Good afternoon! It's great to be here. As Shane said, for those of you that I haven't met, I'm Erica Schultz. I'm the President of Field Operations. I'm going to talk to you about the power of our go-to-market model. Our go-to-market model has evolved a lot in my nearly four years at the company. We're excited to get into it. First, let me start with the massive market opportunity that we're going after, that we're trying to capture. You've heard a lot about this today. Let me share a couple of more comments. As you heard Jay talk about, Sean, and Chad, traditional data architecture is based around this fundamental assumption of data at rest. This leads to this giant mess of silos and point-to-point connections and batch processes. As we know, batch processing has been around for decades. It's become a habit. It's very ingrained in a lot of organizations out there. Oftentimes, we find that even though no one really wants to use batch, it's there, and that's the default. It's still everywhere. To meet the demands of real-time applications and business processes that many companies have, many of our customers have, more than 150,000 organizations have turned to Kafka for data streaming. Confluent, as we know, counts about 4,700 customers and counting, 4,690. Not only do we have a lot more Kafka usage to go in the market to, you know, within many organizations, but even in the organizations that we do count as customers and even many of those using Kafka, the usage is still in very early innings. We see the use of data streaming and the use of Confluent in very early innings in the market, and even within many of our current customers today, most of our current customers today, in terms of the potential for data streaming's use across the enterprise. We are the proven leader for Data in Motion. We are fortunate to count the customers from many different industries, a diverse set of industries, as our customers. This was on full display at our Kafka Summit London event, which we held last month. In London, we had about 1,500 attendees, another 2,300 online. Just fantastic representation from so many different industries, primarily from Europe, given the location of the event, but a really fantastic opportunity to connect with customers. I will tell you, a lot of the product innovations that you heard about today, and in particular, Flink, were some of the most interesting to this customer group. A lot of conversations and sessions on Flink and other topics. You know, one of the customers that in the U.K., Sainsbury's, the U.K.'s second-largest grocery retailer, is a great example of kind of the adoption of Confluent. They started with Kafka, and what they really needed to do was they needed to access data that was stored in legacy systems, in on-premise systems, and as they were developing more cloud-based, customer-facing applications, they needed to access data that was locked away in those legacy systems, and their developers didn't wanna touch it. They adopted Confluent, and they've been able to really break down these data silos, give their developers access to this really valuable data, and been able to deliver inventory and supply chain management applications that are really game-changing for them. Both some customer-facing applications and also some new business process applications. It's a great story. I love the quote here from Sainsbury's: "In retail, everything is becoming data-driven. It's no longer just a facet of IT. It's quickly become essential to the entire business" Kafka has become a company-wide nervous system for us, Confluent has really taken it to the next level. This brings me to how we segment our TAM across three primary segments: the Fortune 500, the rest of the large enterprise market, what we call commercial and mid-market, which includes many digital native companies, startups, and even emerging markets for us around the globe. I wanna spend a few minutes talking about how we segment the broader market. First off, let's talk about large enterprises, and we're definitely in very early innings with our opportunity in large enterprises. We think about the Fortune 500, we think about enterprises around the globe, as defined by last 12 months revenue, over $1 billion. When we met last time at Current, we talked a little bit about this bottoms-up modeling for TAM, so I won't go through it in detail again here. We definitely see that the potential ARR in the Fortune 500 is $7 -$8 million per customer, and significant ARR across the enterprise market as well. We resource those markets with a slightly more higher touch coverage model. There's, you know, I'm gonna talk about a customer example in a minute, because what we see is once we get into a company with that first use case, and they realize the power of data streaming and a data streaming platform and democratizing data, we get into that flywheel effect of bringing more and more applications and data sources onto the platform. A little more insight into our penetration in the enterprise segment. Let me start bottoms up on this slide. Today, we count 1,075 customers over 100K ARR. I wanna point out that more than 50% of this customer base is using Confluent Cloud. That it's been a material source of growth for us with our large customers as well. Our 100K+ customers have grown 34% year-over-year. If we look at the next tier, we have 135 customers who spend more than $1 million a year with us. That cohort has grown 53% year-over-year, and within that cohort, we have nine customers who spend $5 million or more, and three customers who spend $10 million or more. In those top tiers, those numbers have really grown in the last couple of years. A couple of years ago, we had three customers who spent more than $5 million a year with us, and none who spent more than $10 million. I think this is a great demonstration of the potential in these very large enterprises, and some smaller companies who are just making really big bets on data streaming architectures, and I'll talk about a couple of those. That gives you a little bit of a panorama of our current footprint within large enterprises. I wanna talk through a specific customer example of how Confluent and data streaming can grow over time. This is one of the leading global online job sites, and this graphic here, you've seen it before, this is kind of our typical customer journey. As with many of our customers, this company started with open source Kafka, and this was back in 2017, in their initial steps in their journey to an event-driven architecture. As Kafka and streaming grew, they realized that managing Kafka on their own was untenable. We started with a small deal in 2020 for a very specific use case. It was a competitive deal. We won it, due to our cloud-native benefits and the completeness of our platform. By the next year, that initial commitment quickly grew to $1 million as the customer expanded to more lines of business, and they grew to more than 600 users and 1,500 applications once they got the data streaming platform up and running. All of this was on Confluent Cloud. Today, we have become critical to really being a unifying real-time data layer across the organization, and they're running many use cases with Confluent. Use cases like employer campaigns, sponsored jobs, targeted searches that help candidate matching for employers, as well as optimize the job seeker experience. What's interesting as well, we know that all companies, especially those that are very heavily invested in the cloud, this is an environment where there's a lot of pressure to optimize cloud spend. In this environment, their commitment to Confluent, their spend with Confluent, their consumption continues to grow significantly. We're up to a trajectory of more than $10 million today, which is larger than their spend on MongoDB, Snowflake, or Databricks. As you can see, what started as a very small land has grown quite significantly, more than 100 times over the last several years, and we've become a very strategic partner to them in the data space. Hopefully, this is a valuable illustration. Here's a quote from the customer: "Confluent's here to stay. Confluent helped us become cloud-first by making it easy to migrate and scale our Kafka deployment globally across multiple clouds. All of our Kafka use cases now run seamlessly and efficiently on Confluent Cloud. Really proud to count these guys as a customer, and we're not done yet. We see future growth. I'm gonna shift from talking about the large enterprise segments to what we call commercial and mid-market. As I mentioned earlier, this segment counts digital native companies, startups, and a number of emerging markets for us, and this is a really exciting segment for us. You know, we generally define the companies in this segment as having revenues less than $1 billion in the last 12 months, a little bit less than that, smaller than that in some of our international markets. We really focus on the startups and digital native companies within that segment. Many of these companies, as we can imagine, are born in the cloud, and many of these companies are starting with Confluent with a data streaming architecture, which is really exciting. We have really strong product market fit in this market because these companies, being young, they're kind of starting with the mindset of adopting managed services, data streaming is core, and they don't want to put resources on self-managing, you know, open source. The product market fit is really strong, really great value prop. We're excited about this market, not only because of the volume of companies in this market, but the really high propensity for us, even within relatively small companies. We've seen incredibly strong performance with our go-to-market engine in the commercial space. A few stats specific to this commercial segment over the last three years: We've seen a 7x increase in customer count. We've seen a 13x increase in customers spending over $100K with us. I referenced on a prior slide, we've about 1,075 customers who spend more than $100K with us. More than 20% of those are in the commercial market. These are, you know, companies, relatively small companies, but material spend on Confluent. This engine has helped us achieve global reach. We're in now more than 100 countries, excuse me, in large part due to the power of this commercial engine. Our product-led pay-as-you-go model is very robust in this market. Over 90% of our customers start via a self-serve sign-up in this market and often start spending and consuming, you know, a few thousand dollars in their first few months. As they get up and running and successful, their consumption accelerates, and we see, you know, from a few thousand dollars to over $100K in ARR in less than five quarters on average, which again, as you can see, is happening at pretty significant scale. We're really excited about, you know, what's been accomplished in this market and the high propensity for us across this market, as well as it's been an engine for us to kind of perfect this product-led, bottoms-up motion that we can then bring into the enterprise to really complement with that enterprise sales motion. I wanna talk about a specific customer example from this market. Metronome is a customer of ours. This is an early-stage company, a hyper-growth startup, for those of you not familiar with the company, but they offer a real-time, usage-based billing platform, so a lot of demand in the market for their offering. It is core to the business model for so many startups today, who are depending on real-time, usage-based billing. The billing platform is built on Kafka to compute usage and invoices in real time for millions of their end users. Given their size, this is a great example of what I mentioned before. Metronome quickly realized this is not where they wanted to spend significant engineering resource, managing open-source Kafka. It was costly, and it diverted valuable engineering talent away from their core mission, to low-level infrastructure management. With Confluent Cloud, they were able to reallocate at least 60% of that engineering time to delivering new product innovation instead, which is great. In just a couple of years, their consumption has increased more than 10 times, growing with the demand for usage-based billing. They count OpenAI as one of their top customers, and their recent 6-figure expansion with Confluent puts Confluent as the vendor with the second-most spend after AWS in their environment. They're making a pretty big bet with us. They're very early in their company journey, we see a bright future for our partnership with them. Worth mentioning that, again, just to round out kind of our opportunity in the startup market, we have a Confluent for Startups program, where we have almost over 100 participants, which we're really excited about. We see this as a really great opportunity for us. Let me spend a few minutes on our customer growth go-to-market model. This is how we describe our unique go-to-market model. You've heard me talk about this before, but let me just reanchor us in what differentiates our customer growth go-to-market model here at Confluent. We are product-led, consumption-oriented, and purpose-built for the Data in Motion journey. You know, one of the beauties of this go-to-market model is that as we bring Flink to market, this is the ideal go-to-market model for Flink as well. It maps to the customer journey, to the personas, and so, we see a lot of synergy as we bring Flink and other offerings to the market as we move up the stack. First, product-led. Jay shared a similar slide earlier, and we've talked about this before, but this product-led motion and the fact that product-led and enterprise sales motions are complementary is really fundamental to our go-to-market motion. Our product-led motion revolves around that self-service, pay-as-you-go, frictionless adoption, targeted at the practitioner personas. We might wanna make it so easy for developers and architects to get their hands on our products and get familiar, tinker around, and ultimately start in, you know, maybe dev environments with Confluent Cloud. We layer on an orchestrated set of what we might think of as traditional enterprise sales motions. Getting to the economic buyer, working with InfoSec to get through security reviews, doing some joint planning to make sure we have aligned definitions of success with that first use case, and a clear vision for the power of data streaming across the enterprise. Maybe we do some TCO modeling. That's where we bring. Of course, we work on a contract. That's where we bring in some of those traditional enterprise sales motions and roles in the sales organization. Meanwhile, our technical stakeholders in the account can get right into the product, and that experience can be, it can be as self-service as is appropriate. Of course, we'll wanna partner with them to do really robust technical evaluation, test scale, and things. That's a little bit how these two pieces work together. The second pillar of our customer growth go-to-market is consumption-oriented, and this is really powerful, our cloud consumption model. We, you know, given the mission criticality of the use cases that we power across both revenue-generating applications as well as data-driven business operations, we've seen our product and go-to-market model drive really exciting network effects in our accounts. We've talked about this, and this consumption model where, you know, we have usage-based billing for our customers. It's just a really powerful and frictionless way for them to consume more use cases and scale their usage with us. It also comes into play, it enables us to move up the stack seamlessly with Flink and other offerings, so this has been incredibly powerful for us. I think that, you know, as we see new use cases come on, where maybe the usage, the future usage isn't known yet by the customer, a lot of Gen AI-type use cases and that sort of thing, this is a really beautiful model because customers can kinda get started without having to size an environment and, you know, plan for volumes. We can kinda get into it a little bit together before we look at those elements. Our consumption model's been really powerful for us. The third pillar of our go-to-market model is purpose-built for the data in motion journey. We've talked about this slide, but I just want to highlight that our Data in Motion journey, we think of it, you know, generally in kind of five stages, where early on, there's early interest and developers getting engaged with the product. Maybe there's an early production application, and then it moves into production, or we move, you know, the data streaming architecture to other more mission-critical applications across disparate LOBs. Then, ideally, the network effects start to be envisioned, and customers start to, in this stage four, really think about a business-wide integrated streaming strategy. This is where we partner with customers, as you can see here on the slide, across more topics like people, process, technology, and we can help them build a data streaming center of excellence. We might, you know, bring in partners to help companies with the transformation associated with really making streaming a core capability enterprise-wide, with that ultimate destination being Confluent as a central nervous system. There's a great quote here from Bankers Healthcare Group talking about making Confluent the backbone of the company, and it really helping them integrate critical apps and data systems together. You know, this is a company that's been partnering with Confluent for a few years, and, you know, their deployment spans a number of modern cloud-native services, both, you know, in the cloud and also on-prem, databases and batch ETL, but we've really become the backbone of that environment. A great example. The power of this customer growth go-to-market model really is manifest in the growth that we're able to see within our customers. As we talked about across those 4,690 customers that we have, we have customers in many different stages, but the vast majority are in relatively early stages, and we can see the power and the potential of the growth when we look at the customers who've grown with us over the last several years. We know that even these customers who are on the larger side have plenty of growth to go. Three customers here that I'll talk about. The first is... All three of these started on our pay-as-you-go offering before moving to a commit contract. The first one is a generative AI research technology company which uses Confluent Cloud to feed real-time data into their billing applications, so that they're leveraging our service to minimize the risk of lost revenue. Their consumption has increased 19 times in the last three quarters, primarily due to the recent rise of Gen AI. We're obviously still in really early innings with this customer. Next, we have an online delivery service who uses Confluent for use cases like real-time inventory management and fraud detection. Here we have a 12 times growth rate, driven by the increased demand for online delivery, as well as the customer incorporating Confluent Cloud into more parts of the business, more use cases on their way to a central nervous system. Finally, we have an online gaming platform. This company uses us to power use cases like player analytics for millions of users, as well as a community safety use case to detect and prevent in real time if players are using harmful language on the platform. Here we've seen a 7x growth rate over the last couple of years. Just a couple of examples of how our customers are growing with us as they're incorporating more and more use cases. All right, I wanna spend a minute here on TCO. You've heard us talk a lot about our TCO advantage at various junctures, but this is really important, and especially in this macro environment. Obviously, a really hot topic for companies. There. You know, most companies are not looking for incremental spend, but rather, how can your solution help me achieve lower total cost of ownership than what I'm doing today? We believe that Confluent offers a fundamentally better, faster, and cheaper solution for our customers for data streaming compared to any alternative, the most common of which is self-managing open source Kafka in-house. Our TCO advantage comes from savings across two primary sources of spend when self-managing Kafka. The first is the underlying infrastructure required, and the second is the engineers and operators, operating talent, responsible for configuring, deploying, and managing Kafka. We've built a deeply differentiated stack that drives compelling savings in both of these areas. Let me just give a couple of examples of how we save companies money, save our customers money across both of these cost categories. First, on the infrastructure side, multi-tenancy. Confluent Cloud's multi-tenancy allows us to pool our thousands of customers on shared infrastructure to drive higher utilization with a serverless experience, and then we get to pass these savings on to our customers. I'll also highlight elasticity. Our intelligent tiering of data between memory, local storage, and object storage helps manage the cost of stored data and allows instant scalability, enabling higher utilization of the compute resources. There's some very concrete items that help us drive lower TCO on the infrastructure side. And on the development and operations side, first, our infrastructure improvements actually do double duty here. The improvements I outlined drive vastly higher utilization, and so we manage an order of magnitude less infrastructure than our customers otherwise would. That's a win. The big difference in our operations is that it's done by software, not people, that our customers have to staff. We're able to automatically, for example, detect and remediate the kinds of rare problems that become common at scale, and we have real-time monitoring and checks for every aspect of the integrity of the system. These capabilities provide Confluent Cloud with a dramatic advantage in the cost of human management. Some very concrete areas where we help customers drive costs down. We work with customers to do customized TCO assessments, sometimes in the sales cycle, sometimes once customers are into deployment, we'll work with them to go back and quantify the TCO. We have a business value team that helps us, where we need a high-touch method to do this. We've done this with hundreds of customers, and across those hundreds of customers, we see an average savings of 45%, 75% of these customized assessments that we've done have demonstrated 25% or more than 25% in savings, for a total across this pool of customers that we've partnered with, of $300 million in total savings. We're quantifying this very directly. As you heard earlier, we've also leveraged a third-party study we muster through the Total Economic Impact study with us to quantify very significant savings across the customers they spoke with as well. This is very real. I wanna talk about one specific customer example here. This one happens to be a premier global nutrition company, where we were able to drive $7.1 million in quantified savings with them. This customer looked to Kafka and data streaming to modernize its data infrastructure from a myriad of legacy ETL tools, and batch-oriented technologies as they modernized to an event-driven architecture. We were chosen as the best fit for this modernization initiative, not only due to our platform capabilities, but also because of this quantified TCO. This was really important to them. By switching to Confluent Cloud, this customer identified $2.1 million in savings and infrastructure, primarily related, in this case, to storage, which is included and fully managed by us. By leveraging our cloud-native capabilities, customer identified another $2.1 million reduction in engineering costs because they'd were able to drive a 60% reduction in engineering time dedicated to this environment due to our automated operations. In this case, the customer was also paying for paying a third party for Kafka expertise to augment their internal staff, so they were able to move away from that and free up some costs. Finally, with Confluent Cloud, customer was able to get to production 14 months faster on use cases such as customer experience and supply chain optimization, and this contributed to $2.2 million in accelerated operating income. Very quantifiable. This is a great story, and we're pleased to be able to do this work with many, many customers. I think that the power of this TCO offering and the fact that we've been able to quantify it with so many customers, really enables us to capture far more of the open source Kafka market. Because oftentimes, you know, folks will make analogies from on-prem open source models to the cloud, when in reality, the fully managed cloud offering offers a very different value prop, as we've talked about. Traditional on-prem open source business models typically offer, you know, a premium product, but it's better features for more money. As a result, the companies that are monetizing that open source are only able to capture a smaller fraction of the market. In contrast with a fully managed cloud offering, we're not just replacing free software, but we're also replacing expensive infrastructure and people costs, and so we can help customers drive costs down. As the mindset shift is happening in the market, more and more companies are seeing this opportunity. Fully managed cloud services is an opportunity to drive costs down. This is a great win for us, and so we think it's gonna enable us to capture more of the open source Kafka market, and we're definitely seeing this across our customer base. All right, let me spend a couple of minutes on our partner ecosystem, and following my presentation, we'll actually have one of our partners from AWS join me on stage for a fireside chat. Let me provide the overall frame of Confluent's partner ecosystem. We view the partner ecosystem as absolutely essential for a data streaming's data streaming platform's role as a central nervous system. Partners obviously play a critical role in our go-to-market strategy. They expand our reach, increase our efficiency, but it's really strategic for us as well. In order to become a central nervous system in an organization, we need to be integrated into our customer's ecosystem. It means that we have to really think of ourselves as an ecosystem company and make it very easy for our partners to send and receive data from us. We build the infrastructure, the integrations, and the connectivity once, and then every joint customer can leverage Confluent to build their business outcomes about across both our products and our partners' products and services. We've talked about the network effects that are very present in our market. These network effects really drive partner engagement. I would highlight that what's been exciting about our partner ecosystem is that the wins for our partners are so clear when partnering with Confluent, in addition to the traditional, you know, benefits to us, from partnering with folks. You know, CSPs, for example, benefit as they can capture valuable data that's locked up in legacy data sources. We can get that into cloud environments and unlock more cloud use cases for our joint customers. That's been a really big win for customers and for the CSPs. ISVs benefit because we take care of the ingestion for them, allowing them to focus on value-add services versus data acquisition. The SIs have a big opportunity because they can help customers really get to that business outcome and where there's a business transformation layer to take care of. There's a lot of a lot of interest in the SI market. This becomes a continual flywheel and an ever-expanding network of the ecosystem and mutual customers who are benefiting from our highly differentiated ecosystem and go-to-market. There's really three types of partners. I think these are fairly self-evident, but we generally bucket our partners across CSPs, ISVs, and SIs. We think about, you know, kind of the key pillars of how we partner with each of them a little bit differently. CSPs, first and foremost, it really is about joint innovation and integrations with their CSP native services. We spend a lot of time with them kind of doing roadmap sharing and planning and making sure that we have seamless integrations for our joint customers across various services. The CSPs are also important, you know, for lack of a better term, kind of distributors and provide us reach to market. Their marketplaces are very heavily utilized by a lot of our joint customers. We do a lot of transactions through the CSP marketplaces, which helps customers burn down some of their commits to the CSPs. Of course, the CSP sellers are highly incentivized to partner with ISVs for transactions through the marketplace. As I mentioned, a lot of joint innovation. On the ISV front, integrations with ISV services are key. The network effects that drive mutual consumption, we see a lot of just very clear, kind of accelerated consumption on both sides when we partner and make integration easy for our customers. Based on that, we're going to market with some joint sales plays with a number of our top ISVs. Then SIs, as I mentioned, are really helping our customers drive bigger transformation, get to business outcomes. SIs are interested in building services businesses, you know, based on these modern technologies, and we're interested in leveraging SIs to, you know, help our customers drive utilization and consumption faster. There's really a win-win. All of these pieces add up to increased leverage in our go-to-market model, and that's really what it's all about. I wanna just summarize the levers of growth that we see for Confluent. Each of these areas are opportunities for us to continue to drive best-in-class top-line growth. We see a large and growing TAM for Confluent, which we've talked about from a couple of different perspectives. The secular shift to cloud is definitely a tailwind for us. We see that in our cloud results. It's part of every conversation with customers, whether they're digital native customers, where the entire conversation is about cloud, or whether it's more traditional enterprises, we are core to their shift to cloud and part of that journey. We will continue to penetrate and expand in high-propensity segments, from digital native to financial services to public sector, especially as we deliver FedRAMP certification in the future, but we're very focused on our highest propensity segments and how can we, Tier 1 markets internationally, and how can we penetrate and expand in those. Growing and harnessing our partner ecosystem, we believe we're in early innings there. There's a lot of opportunity for growth and leverage in the different partner types that I talked about. Continued international expansion. Today, about 40% of our revenue is from outside the US. We are very focused on the highest propensity opportunities first, very focused on Tier 1 markets for Confluent, and putting a disproportionate amount of resources there, while leveraging efficient go-to-market engines to be a little more opportunistic in more markets around the world, so that we have a very deliberate and intentional and efficient growth strategy. Finally, we see large cross-sell and upsell opportunities within our account base. Many use cases. Jay talked about the proliferation of business use cases in every industry, and with every new capability that we bring to market, like Flink, and every new demand driver in the market, like generative AI, those count on data streaming, and really help us identify more opportunities, and grow the opportunity for Confluent. These are the levers of growth that we see. I'll wrap on this, that of course, top-line growth is critical, but so is leverage in the model. As we're building out our growth strategies, at the same time, we are very committed to driving increased efficiency overall for the company, but starting with our go-to-market model. We're very pleased with this trajectory, these results, and this trajectory of sales and marketing spend as a percent of revenue for the company. We're on a really solid path, and coupled with, you know, continuing to put up best-in-class NRR, 130%+ for eight consecutive quarters, and getting efficiency from continued cloud adoption and more than 50% of our cloud customers having started on PAYGO, with an increasing percentage quarter after quarter. That's a little bit about our go-to-market model and how we're balancing growth and efficiency. With that, I'd like to transition to a fireside chat. I'd like to welcome to the stage Matt Yanchyshyn from AWS. Come on up and join me, Matt, and we'll talk a little bit about our partnership. Thank you. Thanks for having me. Excellent. Welcome. Thank you. Thank you for joining us. It's great to have you. It's great to be here. I'm going to introduce you here. I'm gonna read your bio, and then we're gonna get into some questions. Hopefully, I've got this right. You can correct me if I miss anything. For those of you that don't know Matt is the General Manager for AWS Marketplace and Partner Engineering. It's a new role for him at AWS, but he's been with AWS for, I think, 11 or 12 years. Quite a while. He's managed a number of different service lines. He's been in the digital media and technology industry for more than 25 years. Of course, if he's been at AWS for about 12, he's been there since relatively early days. You've led a number of different AWS service teams and technical sales organizations, including AWS Control Tower, Service Catalog, Well-Architected, and now all engineering for AWS Partner Organization, which is pretty vast. Before joining AWS, you spent a decade at the Associated Press, where you helped build field offices all over the world and led modernization efforts for the AP's digital media distribution systems. You hold multiple AWS patents. You're an Amazon bar raiser with more than 650 interviews under your belt. You've had the privilege of leading great teams, you've launched dozens of new AWS programs and features and products. Wow! Thank you for joining us. Yeah. Thanks for having me. Yeah. Great. A few questions for us. The first one, you know, this is obviously a hot topic, generative AI, and specifically the importance of augmenting large language models with streaming data to make enterprise applications even more useful. Would love to get your take on that and how you see gen AI evolving, real-time gen AI, evolving in the near term, and what that might mean for enterprises. Great. I mean, first of all, it's evolving quickly, way faster than I think anyone could have predicted a year ago. You know, it's no secret we like to say data is gravity, and the more data you have, the better your models are. I think specific to AWS strategy and what we're able to do with Confluent, our strategy is pretty simple. We have a strong partner strategy when it comes to generative AI. When you look at SageMaker Bedrock, which is our upcoming gen AI service, three out of the four foundational models that we're launching with are actually third-party models, so Stability AI, AI21, Anthropic, with their very cool Claude foundational model. I think what sets AWS apart when it comes to gen AI, and again, how it ties back to what you're doing, is we're all about, yes, foundational models are amazing, but they're even better and they take on better relevancy when you train them and customize them with your own data. Mm-hmm. Right. Importantly, if you're going to customize them with your own data, you don't want that data leaking. Privacy and augmenting our core foundational models with data you bring in yourself is really sums up our strategy. Then, you know, customer choice. Not all models are made the same. Models are really, really expensive to train, but you can't just rely on one or two. You need a selection. We have our own Titan model, but if you want to use. You know, I was just using Stable Diffusion to generate some amazing images for my kids. My kid just turned 11, so I made him a picture this morning. You know, that picture, what you can generate, or the text, or the documentation gets better with the private data that you incorporate into the model. The more data you have, the more secure it is, the better your generative AI results are going to be. Yeah, that's great. You're a long-time AWS veteran, and you've recently stepped into this new role, looking after all of partner engineering. We've been fortunate that, you know, among the demands of being new in your role, you've, you know, you've invested in Confluent early on, which has been great, and I know that joint innovation is one of the most important pillars of our partnership. You've been spending some time with Sean and Chad, which we appreciate. Would love to just, you know, get your perspective on, you know, why we're a priority for you and kinda how you view the partnership. Yeah, I mean, well, I don't say this lightly: I love you guys. I mean, when I was just looking at the slides, I was, you know, acting as a customer or as an investor in the audience. One thing that stood out, and actually the last slide you had, was the 50%, your go-to-market plan, and the 50% of the pay-as-you-go opportunities converting into. That really resonates with actually, like, my and our current strategy for marketplace. It's no secret that we're an enterprise procurement vehicle for software and data and professional services, but the overwhelming majority of those enterprise-focused private offers start with self-service and start with that discovery phase. When I think of Confluent, you're really at the vanguard of that self-service, sort of, re-revolution. I say re-revolution, 'cause when cloud marketplaces launched, you know, a long time ago, about 10 years ago, almost as long as I've been at AWS, it was a self-service vehicle at the beginning. Very quickly, we pivoted into private offers, enterprise motions, and now we're kind of returning. We're realizing that, hey, if we can drive up really high conversion rates and really bring in organic, you know, joint opportunities, that can feed the private offer flywheel, like you just said, and we can all be successful. Your strategy is actually highly aligned with our current marketplace strategy. Not to mention that your top partners, like Databricks and Mongo, I mean, those are also my top partners. Yeah. The fact that Confluent, in particular, works really well in these tri-party, multi-party solutions for customers, really aligns well with solution-based selling as well, which is all we're doing these days. Strategically, we're just a, you know, two highly aligned companies, and you're willing to lean in with us. You know, you're co-investing in a lot of these new innovations and features we're coming out with, and not every partner is willing to kind of put themselves out there and you are. You know, innovate and take risks together, and that's awesome. Yeah. Are there any specific areas of joint innovation between the two companies that you're most excited about? Yeah, I mentioned sort of self-service. Specifically, we're really interested in sort of reimagining the sort of end-to-end journey of the different personas who are now buying through cloud marketplaces. It's really evolved, you know, it's amazing. We started with the developer, but now there's procurement offices, even increasingly, lawyers and governance teams who are using Marketplace. You know, like, really co-innovating on features that target those different personas. Again, investing in some of our self-service features together, but also the post-procurement journey. You know, I won't get into too much details in such a huge room, but some of the work we're actually currently doing on what we can do after that purchase motion. In terms of configuring, you know, getting things to work well together after the procurement motion, and that's, I know, an area that you're really interested in. You know, how can we get people using the software faster. Certainly, an area where we want to help. That is probably the area I'm most excited for when it comes to co-innovation. That's great. We're very aligned there. I know we're working on some exciting stuff. Yeah. I think it would be worth spending a minute on, because not everyone's familiar with how this works, why the AWS field organization is interested in partnering with Confluent or ISVs in general. Maybe talk, you know, a little bit about the incentives that are out there for the AWS field organization to partner with ISVs, and then why Confluent is interesting to AWS sellers and solution architects. I mean, well, you just touched on a couple. I mean, there's the obvious stuff like, you know, commit, drawdown, enterprise, discount plans, et cetera. I think in AWS, really in the last couple of years, our field has really woken up to the fact that when you sell with partners, those deals close faster, they're bigger deals, and it's kind of a win-win. You know, I always say, "Don't believe me." You just talked about a Forrester total economic study. We commissioned our own, and it showed that. It was from Q1 2022, and it showed that deals that are transacted through Marketplace, AWS Marketplace specifically, are 40% faster and 80% bigger. Mm. When you say that, and you couple it with, "And by the way, those 80% bigger, 40% faster deals can help you draw down the customer commit, and, you know, hit your quota," their eyes light up. You know, the field, you know, once they started to see this third party, this analyst information, they wanted to double down with the top partners, the ones that are actually, you know, driving this. When I look at... I think we had, what, 300% year-over-year growth in shared customer wins last year. If you look at the conversion, I don't think I'm allowed to share that stat, but the high conversion rates that you're seeing on Marketplace, and again, that pay-as-you-go to private offer and sort of long term, they're gonna go with the partners that are seeing success, so they can double down on what Forrester's telling them. That's why they're excited about Confluent. Yeah, that's great. I think, it is worth pointing out that even in these areas where you have, you know, you have a breadth of native services and also your sellers are very highly incentivized to sell partner solutions, which is great. Yeah, I mean, they don't differentiate. They wanna do what's best for the customer. Yeah. I know that sounds almost cliche for coming, but it's true. AWS, like, they drill it into our heads every day: "Do what is best for the customer." You know, data's king these days. It's particularly king with generative AI and what's coming out these days. Yeah. You bring more data into the platform, you work well with top partners, you're closing a lot of deals and converting opportunities. It's a win-win. It's easy for them. Yeah. That's great, and we have so many great stories together. Yeah, which is fantastic. We're about, we signed a Strategic Collaboration Agreement with AWS about a year and a half ago. We've been partners for a number of years before that. We really kind of upped our game together a year and a half ago. Can you talk a little bit about the specific benefits of that collaboration agreement, and maybe some of the specific programs and results that have strengthened our partnership? You know, first of all, strategically, I work for Ruba Borno, who leads our global partner organization, and she's big on SCAs. She's big on Strategic Collaboration Agreements, but not with anyone. She wants to sort of have a select group of, again, successful, highly motivated partners to do these with. In terms of the benefits, we're increasingly tying our top incentives, for example, like, the ability to draw down customer spend. You know, some of our biggest incentives. Like, we have a very successful migration program we call our MAP program. Some of the top incentives we're tying directly to these strategic agreements. Mm-hmm. When you enter into a, you know, a long-term Strategic Collaboration Agreement with AWS, we're not just gonna double down on co-sell, we're also gonna insert you into things like our AWS ISV Accelerate Program, or, you know, the bigger deal part of our Migration Acceleration Program. It kind of opens the door to a next level of co-sell, where we do it programmatically, much more globally. Also from an engineering perspective, I'm looking for partners who are committed, honestly. Yeah. If I'm gonna, you know, take the time with my engineering team to sit down and co-build features, I wanna do that with partners who are all in, honestly. It's one thing to say, "Yeah, we're built on AWS." It's another to say, "Yeah, we're built on AWS, we're all in AWS, and by the way, we signed this multi-year collaborative agreement." I'm gonna take the time to sort of invest my product leaders, my engineers with those partners. Yeah. That's great. It's been great for us because we feel like, you know, we know we've gotten access to a lot of those top programs that are good for customers, good for your sellers, and so it just gets us kind of in more, you know, in front of more opportunities, which is fantastic. Yeah, it's been very positive. Any other comments about what you're hearing from customers, I know you spend a lot of time with customers, just about the uptake of data streaming? Yeah. I mean, it's no secret that it's hard. You know, I think just speaking as someone myself, like, I have to. Like you said, we have a large portfolio of services, and even just the service metrics, let alone the business metrics, and, you know, I have a constellation of dependencies, and I have to track everything's working. You know, I myself understand and appreciate the need to have, you know, first of all, high speed, reliable, but very flexible data collection mechanisms that they need to scale in ways that they didn't 10 years ago. Like, I'm collecting, you know, hundreds of millions, if not billions of pieces of information every year just to run my own services. Customers, you know, I'm built on AWS services just like any customer. Customers know this too, and they want solutions that can scale. That's not a differentiating part of the business. Like, I don't make money off of my analytics backend, like I sell on Marketplace. Right. You know, customers have really sort of wisened up to that. Also with intelligence, you know, just for us, for example, we're looking at how we can improve, partner discovery through generative AI, through our own product, and that discovery only gets good with data. If that data is super messy, can't be easily combined, is impossible to find, you won't make progress. I think we're starting to see that tipping point, where it's not just about, you know, "Hey, is my latency good?" Or what, you know, these operational metrics. It's: "Hey, how can I take these metrics that were just part of my operational rigor and turn them into business value? Yeah. To do that, you need to have a ton of data, and now the compute and the technology you have with things like generative AI actually allow you to turn that into meaningful data. We've really kind of flipped the bit in the last couple of years, so it becomes ever more important to use solutions like yours. Yeah. Great. With that, I just wanna thank you again, Matt, for the partnership, importantly. Yeah, my pleasure. For joining us here today. Thank you. Thank you. Great. Thanks. Thank you. All right, I'm gonna wrap there. Thank you for the time, and I'd like to invite our CFO, Steffan Tomlinson, to take the stage next. Thank you. Tomlinson. I'm the CFO. You've heard today a lot about the secular drivers that are really driving Data in Motion, data streaming, stream processing, I'm really excited today to walk you through how we're leaning into this opportunity to really be driving efficient growth. The starting point of this journey is our operational and financial execution. In a very choppy environment over the last year and a half, we've been building a track record of delivering on our commitments even in the face of these choppy waters that are out there. We've been able to drive very durable growth. The product market fit that we have, coupled with our go-to-market execution and this move from batch processing to streaming, has really been driving high rates of NRR, compounding, and durable growth. We've been delivering all this growth while really managing and balancing profitability. We've been very much focused on working within a framework. We had a target model that we outlined at the time of our IPO. That's guided our journey on this, and we've been able to deliver very high rates of growth, greater than 30%, while delivering lots of operating margin improvement. In fact, in the last three quarters, we've driven over 10 points of annual operating margin improvement for each of the last three quarters. I want to take a few minutes to look behind the details of these, and this starts with the strong execution. The momentum since our IPO, we went public about two years ago. Our total revenue run rate has now reached approximately $700 million, and that's doubled. Our cloud revenue run rate is approximately $300 million, and that's grown more than 5x. We've delivered 18 points of operating margin improvement during this time period, and it's also driven by best-in-class NRR and gross retention metrics. When you look at our progress on both the top line and our path to profitability, we've achieved roughly $700 million in revenue, and we expect to be profitable in Q4 of this year. The company is nine years old, and that puts us in the top quartile of performance relative to peer group. On multiple dimensions, is differentiated from the alternative offerings. The TCO and ROI characteristics that you've heard about earlier today are compelling to our customers. We're able to serve our customers' needs wherever they are, on-prem, in the cloud, in multi-cloud environments. This creates a big competitive moat for us relative to most solutions that are out there. This differentiation has driven a 70% CAGR, and in the early years, most people know that the growth was primarily driven by Confluent Platform, our on-prem offering. Over the last several years, growth has really been driven by our Confluent Cloud offering, and that became a catalyst for future growth. Cloud now accounts for 42% of revenues, and we expect cloud to account for about 48%-50% of revenues in Q4 of this year. All of these characteristics have led to strong customer commitments. RPO at the end of 2022 was $741 million, and that grew at a 70% CAGR. All of this RPO performance has really helped us with visibility and predictability. One of the key measures that we track is CRPO, and CRPO coverage to revenue is tracking at or above 60%, and the CRPO CAGR over this time period is 69%. Again, these dynamics have created visibility and predictability, and it's really helped us deliver on our quarters ever since we've been a public company. While our hybrid model of selling both Confluent Platform and Confluent Cloud is important, I'll tell you, cloud is leading the way. It's powering the top line. Customers are developing new applications with a cloud-based premise first, and they're moving existing workloads to the cloud. We are a cloud-first company today, and we are 10x better than a do-it-yourself Kafka approach, and our cloud product now has best-in-class NRR. More than half of our customers that have over $100,000 in ARR, are using Confluent Cloud either on a standalone basis or in conjunction with Confluent Platform. Another driver of this growth and strong, consistent execution has been our diversified customer base. We have 4,690 customers. The customer count has grown 10x since 2018, and we are in over 100 countries around the world. We serve nearly every industry that's out there, so it's a highly diversified customer base and an industry base. Now, let's turn to the durability of our growth and the three core drivers that underpin it. First, it starts with TAM. There's a large and growing $60 billion total addressable market today. Our differentiated customer growth go-to-market model, which is purpose-built for Data in Motion with best-in-class NRR, has led to this compounding durable growth. Our cloud consumption model is powering growth of... as our customers are running mission-critical workloads across all of their environments, whether on-prem, in the cloud, or multi-cloud. I want to take a moment to double-click into each. The $60 billion TAM, it's a new category that we've created. Earlier today, Sean and Jay went through the TAM construction. I would really like to focus on a couple of things. The first is, with the addition of Flink, this enables us to capture our TAM even faster than we had the opportunity before. That's an accelerant to us. The explosion in data volumes that we've seen, and the explosion in applications that's happening across the entire enterprise landscape, and what will be coming with generative AI, are catalysts for future TAM expansion. Over the next three years, we anticipate our TAM to be increasing at a 19% CAGR to about $100 billion. The net of it is, this is a big market, and there's a lot of opportunity for growth in front of us. The TAM is also supported by the significant opportunity that we have for converting open source, do-it-yourself Kafka customers to Confluent Cloud. There are over 150,000 do-it-yourself Kafka users that are out there today, and we have 4,690 customers. Cloud, our Confluent Cloud offering is far less expensive than the alternative offering. If you think about what we deliver, there is less personnel that's needed to run our Confluent Cloud offering than a do-it-yourself Kafka implementation. There's less bolt-on software, there's less infrastructure costs, and the overall TCO and ROI is very compelling. In a nutshell, it's faster, better, and cheaper. The technology has broad adoption across all of the industries that we play in. You saw these stats earlier today. Over 75% of the Fortune 500 run Kafka, and we have about 36% of those customers today. There's a big opportunity for us to grow in this segment. Now, our product market fit and go-to-market model is a powerful combination that's led to strong growth in large. We have 1,075 customers that have spent over $100K in ARR, and they account for more than 85% of revenue, and that's been consistent ever since we've been public. This cohort has grown 6x since 2018. Now, moving up the stack, we have $135+ million customers t hat has grown 12x since 2018, and we have a growing number of $5 million and $10 million ARR customers. When we evaluate our customer base, and we look at the opportunity to grow and expand, in many of these customers, we're still in the early innings of our ability to satisfy their needs for real-time data in motion. We have a lot of room for, and to capture more wallet share. The second core driver of our differentiated customer growth go-to-market model, which powers our land expand and consumption engine, is these NRR in gross retention metrics. For 8 quarters in a row, we've delivered net retention rates above 130% and gross retention rates above 90%. In particular, our cloud business is a standout performer. We're above 140%, the highest NRR that we see is for the newer cohorts. You've heard this earlier today, we have very strong network effects within the organizations that we serve. This network effect has really translated into, as more applications and data stores get connected into our platform, it begets more applications and data stores need to be connected. We've seen that manifest itself in a 3.5x+ ARR expansion after four years. We have a powerful cloud consumption model. Erica presented this cloud consumption model a little bit earlier, but I wanted to share a few additional points. The reason why it's unique and powerful, it's because it's a revenue-generating model due to the network effects I just described, and we're running mission-critical workloads. These workloads often come pre-optimized, and over the last 12 months, in a tougher business environment, our consumption model has actually held up very well. We've been showing compounding growth in our consumption model, and it's due to the characteristics I just talked about. Our consumption model, I would just say, has held up better than maybe some other consumption models that are out there, and that's due to the nature of workloads and where we sit in the technology stack, and the mission criticality of the workloads that we're running. All of this has led to compounding growth. Our top 20 customers for cloud span 11 unique industries. Approximately half of them started on pay-as-you-go, and about 40% of them are hybrid customers. In this cohort of customers, it's taken 6 quarters for them to get to $1 million of ARR. Very, very quick. The average ARR across this cohort base is $3 million, and that's up from $0.6 million just a couple of years ago. A total ARR multiple is 17x. This demonstrates the power of our customer growth go-to-market model and the land expand and consumption characteristics that we have. Increasingly, as the enterprise continues to move workloads to the cloud, our install base of Confluent Platform customers are adopting Confluent Cloud. This perhaps is maybe one of the areas that isn't as well understood by some folks who are looking at Confluent from an investor standpoint. Looking at the cohort of these hybrid customers who run both Confluent Platform and Confluent Cloud and spend over $100,000 in ARR, the customer count has grown 70x. There are a few attributes to highlight in this cohort. The majority of these hybrid customers started with Confluent Platform. There's an inherent cross-sell and upsell opportunity that spans both Confluent Platform and Confluent Cloud customers. This cohort, in particular, has a high ASP and a higher cloud expansion rate than the average customer that's out there. Doing another double click on our cloud business, to put a finer point on things around the power of our consumption model, here are two examples of marquee customers that started their journey with us and what the profile of their expansion is. The first is a large international bank that started with about $100K in ARR and grew to 34x- $4.6 million. The next is a multinational manufacturer that started on cloud as pay-as-you-go with $50K and expanded over 192x - $4.3 million. The progress that we're making in these customers puts us on a path to be the central nervous system as they deploy Confluent more broadly across their environments. I'd like to take a few minutes to look at how we have managed growth and profitability. We've been committed to driving efficient growth, and we shared a framework and a target model at the time of our IPO that we're gonna be updating today. First, I'm gonna start with our path to date. Starting with the top line, even in this difficult macro environment, in FY 2022 and FY 2023, we've been able to navigate the environment pretty well, and our growth has been resilient. We continue to expect that we're gonna hit $760 -$765 million in this fiscal year. While there are about three weeks left to go in the quarter, and like every quarter, we're gonna have work to do in these last three weeks to land the quarter. The consumption patterns that we called out on our earnings call are holding up in line with our expectations, and we're currently tracking to our Q2 guide. The growth drivers, we previously covered these, section, so I'm not gonna cover each of them, but I do wanna just reemphasize that there are multiple growth drivers to this business that we can cover, you know, during our cocktail hour as well. Turning to the unit economics of the business, we've made great progress on gross margin, and this has been a really, like, high point for the company. We've done this despite the revenue mix shifting to cloud over the last three years. We're actually updating our gross margin target for FY 2023. Originally, it was 70%-72%, and we now expect gross margin to be in the range of 72%-73% for the year. The drivers behind this, we've continued to have very healthy gross margins for Confluent Platform, but the work we've done around our Confluent Cloud offering has been very robust. We've achieved scale in the cloud hosting costs. The new product and future releases that we've been coming out with have increased the value that we're delivering to customers. There's been a lot of work and effort going into engineering and infrastructure engineering to get more optimizations, and we're also seeing an increasing mix of multi-tenancy for cloud, and that it comes at a much higher gross margin. We also understand this work is never done, and this is something that we're laser-focused on, and we're gonna continue to work to drive efficiencies. Turning to OpEx and operating margin, we've been disciplined in terms of where we're making investments. We've been controlling the rate and pace of hiring to ensure we're driving leverage and while also delivering high revenue growth. The investments in product and sales capacity have been key drivers, and the pay-as-you-go onboarding of customers has been a key way to drive low-cost onboarding of customers. We've also seen great network effects that have driven expansion, like I talked about earlier, and we've been focusing on improving sales productivity. These factors have contributed to a very healthy pace of operating margin improvement as we continue to target achieving break-even in Q4 of 2023. Looking at our path to date and how we've managed growth and profitability, our investment decisions that were powered by this historical model that we had at the time of the IPO, were also influencing how we allocated capital, and it's been guided by this growth and profitability framework. The key attributes to call out in the old model, it was predicated on growth rate grain being greater than 30% in a growing cloud mix of revenue and a high net retention rate and gross retention rate. We're always focused on improving the unit economics and proactively managing expenses, and we actually accelerated our path to profitability by a full year, given everything that had been going on both in the market and then how we looked at the power of our own financial model, we pulled profitability forward. Now, I wanna take a minute to talk about the future and what the path forward looks like. As we look at managing growth and profitability over the midterm, we're looking to resource the company for a 30%, approximately 30% annual revenue growth rate. We're continuing to proactively manage expenses and the rate and pace of hiring. We continue to expect to achieve operating margin breakeven in Q4 2023, and once we've become profitable in FY 2024, we intend to remain profitable for, you know, on a full year basis, with about 0%-1% operating margin for the full year. That's gonna be another year where we're gonna be delivering outsized operating margin expansion of 13-14 points on a year-over-year basis. Over the midterm, we're looking at expanding operating margin to roughly 5%-10% while delivering top-line growth. Some of the key assumptions that underpin this, that macro does not deteriorate materially, and that Flink and FedRAMP continue to perform according to plan. Now, the shape of our business is changing over time, and it's changing to that we're basically becoming predominantly more cloud. We expect approximately 80% cloud revenue mix long term, and with this change comes some benefits of the compounding growth that we've talked about with a high NRR and gross retention rate. Factoring these dynamics into the financial model, we're updating the growth and profitability framework that we had at the time of IPO. In this new model, we're increasing our midterm and long-term target ranges for gross margin. We're modestly improving the long-term target margin for sales and marketing as a percentage of revenue. We're increasing both the midterm and long-term operating margin targets, and now expect to deliver over 25% operating margin in the long term. It's important to note that the top-line growth will continue to be a determining factor around rate and pace of operating margin expansion over time, because we are continually calibrating the opportunity that's in front of us and the amount of operating margin that we're gonna be delivering. We wanna make sure we're taking as much wallet share as possible while continuing to build a profitable business. There's one more item I'd like to cover. This is how we're managing dilution. Starting with stock-based compensation expense, most of you know it's a lagging indicator due to the ratable recognition of expense. A lot of this that we've been carrying for the last year or so has been impacted by the pre-IPO options and the significant amount of hiring that we did over the last couple of years. In the second half of 2023, the pre-IPO option expense will end. We should start to see some meaningful improvement for SBC as a percentage of revenue. Over the midterm, we're targeting SBC to be approximately 25%. We're gonna be driving that down to the mid-teens. Now, what we're doing to control future SBC is focusing on managing dilution today. We're committed to driving net dilution down in FY 2023, targeting 3%-4%. In the midterm, we're targeting 3%, and in the long term, we're driving that down to less than 2%. That wraps up my presentation. Now I'd like to introduce Rohan Sivaram, SVP of Finance, to cover financial-related modeling topics. Rohan? Thank you, Steffan. Good afternoon, everyone. I'm Rohan Sivaram, Senior Vice President of Finance here at Confluent. The purpose of today's modeling session, in my mind, is to help you get a better understanding of our business model and some of our key financial metrics. As I was thinking through the materials for this session, we've incorporated a lot of the feedback that you've shared with us, so thank you for that. With that, let's just jump right in and talk about the power of our hybrid business model. This is more of a setup before I jump in. As you've heard from most of the presenters today, Confluent needs to be wherever our data and applications reside. Be it in the cloud, we need to be there with Confluent Cloud, or be it on-prem, we need to be there with Confluent Platform. Starting with Confluent Cloud, in the last reported quarter, Confluent Cloud was 50% of the business, as you know, it is our fully managed cloud-native offering that we have. Confluent Platform was 50% of our business in our last reported quarter, that's our self-managed software offering. What is not here in the slide is our hybrid set of customers. We have a growing cohort of hybrid set of customers that Steffan touched on earlier, if you look at the 100K+ cohort, we have over 200 hybrid set of customers. Services is another offering that we sell along with our product sales, it is primarily sold to accelerate our consumption and obviously help our customers adopt our product better. What I'm gonna do probably over the next 10-odd minutes is to walk you through a couple of customer journeys, one for Confluent Platform and one for Confluent Cloud. As I'm doing it, I'm gonna talk about three things. First, we'll do a little bit of a setup, talk about some definitions. Second, I'll walk you through an illustrative example. Last but not least, hopefully, we'll leave you with some key takeaways for each of these. Let's get started with Confluent Cloud. Companies and prospects can start their journey with cloud by signing up for free on our website with a few hundred dollars in credits. Subsequently, they can put in their credit cards and become what we call is a pay-as-you-go customer, and pay-as-you-go customers have no contractual commitments. Alternatively, the pay-as-you-go customers or net new customers can commit to a contract with us, whereby they're committing to a level of consumption over a period of time. These commitments typically happen in either annual or multi-year form, and they show up on our financials as RPO or remaining performance obligations, or CRPO, which is current remaining performance obligations. One thing is when a customer consumes more than they commit, we recognize that incremental revenue as overage revenue. When a customer is under-consuming, the delta is recognized as expiration revenue. Finally, time to ramp is basically the time our customers take to ramp up to their committed contractual run rate. With that, let's jump into an example and talk about Data in Motion. Data in Motion Incorporated is a company that's currently using open-source Kafka, and one of the teams within the company wants to experiment with a managed service, and they decide to choose Confluent Cloud. They come to our website, quickly sign up, and through our in-product tutorials, best practices, and some of the common recipes that we have available, they quickly become a pay-as-you-go customer. What's interesting about this product-led motion is that it drives organic adoption, it is frictionless, and more importantly, it comes with very low sales touch. Let's look at some of the financials. In month one, Data in Motion Incorporated spent $1,000 with us. We recognize that $1,000 of consumption as revenue. From an RPO or CRPO perspective, there is no RPO or CRPO at this point because the customer has not committed to anything. Finally, when you think about the total customer count, Data in Motion Incorporated is now contributing to our total customer count. Moving on, the product, the Data in Motion is starting to use our product and they're finding value in it. From a experimental use case, they slowly start moving into production. Confluent offers a lot of benefits, so they decide to sign a $120,000 committed one-year contract with us to take advantage of the volume discounts we offer and to take advantage of the benefits that Confluent provides with respect to support and our expertise. At this point, at the inception of the contract, we see an RPO and CRPO balance of $120,000. In the first month of the committed contract, in this particular example, they consume $3,000, which is quickly recognized as consumption revenue for us. As we exit the first month, we are left with $117,000 of RPO. You'll see there is an inverse relationship between revenue and RPO as we move forward. We are a few months into the committed contract and the consumption keeps happening, and we reach the six-month time period. At this point, the company is spending approximately $10,000 with us, and they've actually hit our net consumption run rate that they've committed to us, which is $10,000. What's also interesting to note that at this point, they've also crossed the $100K threshold. When we report our 100K+ ARR customers, this is when they are part of our reporting. One interesting, I'd say, observation here is our time to ramp of six months is very unique in infrastructure and is actually a key differentiator of our consumption business model. From with respect to rev rec and RPO, we continue to recognize revenue based on consumption, and the RPO balance continues to get depleted. It's also important to note that Data in Motion does not stop growing at this point, which takes us to month 12 of the contract. As you can see from the chart, they're consuming at $15,000 in month 12 of the contract, which is a run rate of $180,000, higher than what they had initially committed for $120,000. Couple of observations here. Like from a setup perspective, two things that is really working well for Confluent. First, we're very well set up for net retention and gross expansion at this point, given the run rate of $180,000. Second, I also feel that we are very well set up for future margin expansion because typically a year 2 of a customer is much better from a unit economics perspective than year one. At this point, we're in month 15. We don't have any RPO and CRPO, but the customer is consuming, and they are getting value out of our product, so they decide to sign a renewal with us. As they sign a renewal, the size of the renewal deal will typically be at or above the $180,000. Although it's a illustrative example, this shows a typical journey of a customer or a cloud customer that we have. Bringing it all together, as most of the speakers said, be it from PAYGO to commit to renewal expansion, there are a few things that connect all of these. First, it's a deeply differentiated cloud-native model, which helps our customers get started, expand, and scale. Second, our customer growth go-to-market model partners with the customer at every stage of their journey and with the right product, expertise, and go-to-market strategy. And last but not least, our customer success team partners very closely to drive best-in-class time to ramp. With this as a backdrop, let's talk about a few key takeaways for Confluent Cloud. First, as Erica's touched on it, pay-as-you-go is a very important piece of what we do, and 50% of our cloud customers start on pay-as-you-go. Our time to ramp is approximately six months, which is best in class for infrastructure software. We've spoken about this concept in our earnings calls over the last few quarters. Cloud, as a percentage of new ACV, has been at or has been above 50% for five consecutive quarters. As a result, our cloud, as a percentage of total RPO, is approximately 50% in our last reported quarter. From a revenue mix perspective, we continue to expect to exit Q4 of 2023 with a mix of 48%-50% for cloud. From a revenue seasonality perspective, we expect the lowest sequential ad to be in Q1s and more pronounced increase in second half of the year. Finally, when you look at the cloud NRR, that's of course, going to be the highest NRR for the company and greater than 140%. Let me switch gears and talk a little bit about Confluent Platform. Confluent Platform is sold as a term-based subscription that's typically built annually in advance. The Confluent Platform typically has an upfront revenue recognition that happens for the license piece, which is 20% of total contract value. The remaining part of the contract, which is the other 80%, is recognized ratably over the period of the contract. Finally, for RPO and CRPO, Confluent Platform RPO, CRPO is very similar to Confluent Cloud, where we recognize based on our customers' commitments. Let's take the example of Data in Motion again. This time, they are in an industry where the data resides heavily on-prem, and they come to Confluent and sign a two-year commitment for $200,000. What happens at inception? At the inception of the contract, we recognize $40,000 as licensed revenue. That's 20% of $200,000. For the remaining, we typically recognize that over the period of the contract, which in this case is eight quarters, $160,000, that's about $20,000 per quarter we'll be recognizing. Let's fast-forward to end of year one. We continue to recognize the maintenance revenue, or PCS revenue, as we call, and we exit year one with an RPO and CRPO balance of $80,000. As you'll see here, the rev rec lines here are very interesting because on Confluent, it's like a curve, where it depends on consumption, where this is a lot more predictable and it's, you know, it's every quarter, you get the same amount. As you exit year two, you're basically depleted all your RPO and CRPO, and you've recognized the maintenance revenue over the period of the contract. At this particular time, the customer is basically in a position where they're going to sign a renewal with us with some amount of expansion, and NRR is going to drive that. Finally, I think, a few takeaways for Confluent Platform from a modeling perspective. Like I said, 20% of Confluent Platform revenue is recognized upfront, and the remaining 80% is recognized ratably over the period of the contract. Confluent Platform drives more revenue seasonality than Confluent Cloud due to the upfront revenue recognition. Confluent Platform leads to less RPO than an equivalent Confluent Cloud deal. That's, again, due to the upfront revenue recognition. Finally, for some of the key metrics, like 100K plus ARR customers, $1 + million ARR customers, or NRR for Confluent Platform is based on commit, I beg your pardon, and not consumption. Let me take us home with some key company-wide modeling points. These slides will be posted on our IR website shortly, so I don't plan to go through all of them in detail, but I'll cover a few of the seasonality-related metrics. Let's start with revenue seasonality. When you think about revenue seasonality, we typically expect second half to be more weighted than our first half, with Q1 being the seasonally weakest quarter and Q4 being the seasonally strongest quarter. When you look at non-GAAP operating margins, we expect Q1 to be the lowest. That's driven by the lower revenue seasonality you see in Q1s and large one-time expenses that we see, like company kickoff and sales kickoff, that typically happens in the first quarter. Talking about seasonality, when you look at free cash flow, we expect Q1 to have the lowest seasonality or lowest free cash flow. That's driven by our company bonus payout as well as our ESPP settlement. Hope this helps you get a better understanding of our hybrid business model, you know, you'll have this information handy on our IR website. With that, I'd like to thank all of you and invite the management team over for some Q&A. All right, we're setting things up. Takes a few seconds. Well, we have the full management team here, so if you have questions, raise your hand. Hey, Michael Turrin with Wells Fargo. Thanks very much for taking time and for all the content we provided. It's an investor session, I'll start with the target model question because Stefan just laid it out for us. Margins showing clear progression, particularly relative to where we stood at the IPO. The growth assumption tweaked a touch from above 30 to now the squiggle line, 30%. Can you just frame out, is that tied to just sales capacity being at a different point versus what you were previously expecting? Is it a change in the macro? Actually, one of the questions I field most often is how to think through Confluent's growth in a rebound recovery scenario. It's not the world we're in currently, but can you just help us think through some of the inputs on that side as well? Thanks. Well, on the top-line growth, we're resourcing the company to be growing at approximately 30%, and a lot of that is a reflection of the macro. We've been very much focused on preserving and increasing sales productivity and capacity, I should say. That's something that we're gonna be, like, looking at continuing to do, but do it judiciously. Then as far as if the economy rebounds down the road, I'm sure that would benefit lots of companies, including Confluent. The last thing I'll say is, as I think about the structural drivers of growth, we have a wonderful go-to-market organization, the product teams that are working not only on the Confluent Cloud product, but the upcoming Flink product that will be released, that should help maintain our growth rate. You know, in some scenarios that we've run, it could be an accelerant. Then we haven't really baked in anything material on the role of generative AI in the short term. We think that that's gonna play out in the enterprise over a longer period. That. That's something that we think about because we're really excited around what our customers are gonna be doing with generative AI, and we should be a beneficiary of that. I think this is gonna play out over time as opposed to something more immediate. Thank you. Yeah. Great. Derrick Wood at TD Cowen over here, you just led into what I was going to ask about in terms of new product efforts. You mentioned, Stefan, that FedRAMP and Flink are, you know, you've got certain expectations around that. If you guys could walk through kind of what the milestones are to expect, Erica, how you're thinking about the go-to-market focus once they come out, especially around Flink, but also what you're doing with FedRAMP? On the generative AI side, what you guys are potentially doing to lean into that opportunity right now? Yeah, there's a lot there, so maybe we can divide it up into chunks. I'll start with the generative AI one. You know, we laid out a little bit of the kind of architectural pattern we're seeing with customers. I think there's additional investments we can make that will help make some of that easier, but the big opportunity is definitely the flow of data to power these models and the, you know, actual application of this to streams of data that customers have. That's what we want to make easier and easier. Yeah, we have, I think, you know, very big aspirations for Flink. It's, you know, kind of logically understandable that monetizing the compute and streaming applications will be an obvious vector of growth on top of the streaming data that we already have and kind of naturally combines with what's there. The key milestone that everyone should look to and expect is the GA of our Flink SQL service this year. You know, we've communicated that that was coming. We've seen good progress against that. You know, it's in early access with customers now, but that's kind of the key thing that will unlock monetization next year. We, we do expect that will ramp like any cloud service. It's not, you know, 0 to 1,000 on day one. You know, that will roll out to customers who will, you know, bake it into their applications over time. We'll be adding capabilities to some of the other Flink APIs, so we'll open up more monetization off of that, you know, over the coming years. That's gonna be one that we think can be quite significant. You know, we think the opportunity around stream processing is as big as the opportunity around kind of core Kafka itself. It will play out, you know, over a number of years as we grow into that business. I think there were several other side parts. Do you want to remind me? Just one other on the FedRAMP focus. Yeah. Yeah, that's one for us as well. You know, for those who don't know, it's a key certification for the federal government. When we look at our public sector business, that's, you know, contributing quite substantially today. But it's contributing almost entirely with Confluent Platform. You know, a lot of the public sector U.S. for the federal government is cloud, and requires that certification. That's something we're actively working on, and as that unlocks, we should see acceleration in that part of the business. Hi, Rob Owens, from Piper Sandler. Do you wanna square that comment with something Stefan said relative to Flink? You talked about it as an accelerant, but not incremental. You held your $60 billion TAM, which is enormous given the scale of the company, but why isn't it incremental at this point, and how should we think about that opportunity? Stefan, just to nitpick a little bit, you did raise gross margin by one point, but operating margin stays the same for the year. Where does our extra point go? Thanks. I'll take the first part there. Yeah, the way, you know, the way we're modeling that in TAM is, you know, primarily in. It shows up in multiple places with stream processing, right? It shows up in some of the cannibalization of older integration tools. I think the most significant is when you look at that segment of database usage, that kind of 9% of database usage, you could look at that number in different ways, right? You could actually, if you look at the scope of batch processing, that's actually a little conservative. The most important variable is not, hey, how much of the database market, which is quite big, are you taking? It's actually the pace that you take it with. The reason we're not calling it, you know, TAM expanding is we already had that database segment in there. We already had some preexisting stream processing capabilities. This is really about getting that 9% faster, increasing our probability of really winning that segment of it, and really bringing that into the monetization of our, of our cloud. You know, you can think about it either way, but we kind of had that in scope already, and this is just a better way to get it more quickly. Do you wanna take the second point? Yeah. On the gross margin side of the house, we have a track record of at least delivering the operating margin guidance that we've delivered or beating it. I would say that from an operating margin standpoint, the kind of, like, the cushion's gotten a little bit better with the improvement in gross margins. We've been targeting roughly break even for Q4, and that's something that we will continue to do, and the improvement in gross margin for, you know, for the back half of the year is something that gives us even more confidence around the break even for Q4 2023. Hi, it's Gregg Moskowitz from Mizuho. Thanks for the presentation today. Question for Jay or Erica on TCO and ROI, y ou've done hundreds of TCO assessments for your customers, you know, but that's a pretty small fraction of your install base, and the data that you presented was so powerful, and I'm wondering how you can ramp up these assessments to drive even more transparency, even more business value. Yeah, do you wanna take that, Erica? I'm happy to take it. Yeah, I think it's an opportunity and something that we're focused on, especially in this market, is going out in a more scalable way to be able to do those types of assessments. That's actually a key strategy of ours right now. I think it's important, as we've talked about, you know, both in the sales cycle, to prove kind of the why do anything, as well as in, you know, retrospect, help customers quantify the savings that they've realized. Yes, we wanna do that across more of our base in a more programmatic and scalable way. Hey, guys, you have Mike Cikos here from Needham. The question for you is, like, I understand the pull of real time, right? If we had been batching on a monthly basis, you go down to daily, even micro-batching, it just seems like a workaround towards getting more and more towards real time. At the same time, you guys have put in place things to reduce those barriers to adoption, like custom connectors. My question is: As you see these customers going on this journey, are some of them altogether skipping going to, let's say, Kafka first, and just jumping right to a commercialized offering like Confluent, or no? Does the customer journey still require that initial pain point on managing that infrastructure yourself in that Kafka environment? Yeah, that's a great question. One of the things we believed really early on is that the natural journey in cloud starts with cloud. It doesn't start with downloading some piece of software, running it, realizing it's hard, and switching. Yeah, indeed, a lot of our customers directly just sign up for our cloud product and start using that as their very first thing. Increasingly, the scope of that product is well beyond just Kafka. We view that as a good thing. That's kind of the natural way of starting with a cloud offer. However, you know, as I think, Erica laid out and we laid out in, you know, a number of these presentations, there's a ton of open source Kafka, so that Kafka conversion motion, we also wanna be really, really good at. Even for those who are starting with our cloud product, in the back of their minds, you know, as kind of the alternative, would be, "Well, we could, you know, run it ourselves," right? That kind of TCO analysis, some of the other aspects of that still matter in terms of the differentiation. Yes, it's absolutely the case that, you know, a significant portion of our customers are just kind of starting, and their first project is on Confluent Cloud rather than a kind of open source adoption and conversion over. Hi, Howard Ma with Guggenheim Securities. Can you comment on adoption or expansion use cases, between a Confluent Cloud only? Those who start on Confluent Cloud only versus your hybrid, for those hybrid customers, do you expect them to eventually all convert to cloud? You know, if they do that, you know, what does that adoption look like? If you can. I guess one last part is, if you can comment that within the framework of the five foundational elements of, like, you know, so is it gonna be more on the process side, on, you know, on Flink, or is it more govern and, you know, or data sharing as well? Thank you. Yeah, maybe I can start with kind of what's the end state, and then the rest of us can jump in on some of those other aspects. Yeah, you know, we in the very end state, perhaps everything will be in the public cloud, like all applications. It is a very end state. When you look at how enterprise technology moves, it's not that quickly. Many of our customers still have significant mainframe installations, right? There's many generations that they've built out, and even if you looked at that graph of enterprise spend I showed earlier, the cloud was largely an addition to an existing on-premise footprint spend, which has kind of flattened off, but all the growth came in the cloud. Why is that relevant? It... For this technology, it has to be where your applications are. The critical question is, you know, when are customers are gonna not have anything on premise? The answer we're hearing from customers is, like, not anytime soon for kind of large, mainstream enterprises. We're open to it. We love our cloud business. It's extremely successful. You know, some of these are growing very aggressively and will be out of their data centers, and that's a real plan for them, and we'll serve them entirely with Confluent Cloud. Some of them will be on premise for a long time. You know, regardless, that outpost on premise is very important to actually feed some of these newer investments in the cloud product. You know, that's a barrier. You know, that's a moat for us, covering all of those. There's a lot of kind of, you know, difficulties that are not obvious in how data flows between these environments and making that something that's safe and secure and governed and done well, and it's actually a driver of growth for Confluent as well. To your end state, you know, the short answer is, it will follow the overall migration of enterprise to the cloud, which I would expect will continue to be at pace, but, you know, we'll still have some on-premise footprint for a long time. I don't know if folks want to jump in on some of the other points there. Yeah, I think just briefly, like, a hybrid model that you should think of as an enduring strength. You know, I think that there will always be data in various different outposts across the organization, including in various different places on-prem or other hard-to-reach places. It's not just about where the data is, it's about where it's flowing to, right? You need to be able to get it from where it is and make it flow to where it needs to go to, and all enterprises are complex and dispersed by business unit, by function, by geography, by office. When you look at some of the capabilities that we're offering, we are going to where the data is, you know, that hybrid deployment model you just talked to, but we're also offering these global capabilities on top: governance, processing, sharing, connecting. Organizations are looking to put their data to work wherever it is, but putting it to work means governing it, sharing it, and pushing it across all the rest of your enterprise. I think there will always be those outposts, but our global capabilities, the five dimensions you talked about, are what really adds to that ability of the data to go anywhere it needs to be, in any way it needs to be used. Great. Can you hear me? Hey. We can hear you. I must be hearing the echo behind me. Tyler Radke from Citi. Good to see everyone. Two questions. One, I was... You know, with a lot of the slides, it was pretty clear there's still a huge opportunity, right? You only have a low single-digit fraction of all the Kafka users out there. I was wondering if you could give us an update on the success of the paywall change that you made, I think it was about a year ago. Some of your peers, like MongoDB, talked about a record number of customer ads this past quarter. How are you expecting that to trend for your business based on what you can see in the top-of-funnel data? Second question, just on generative AI. There's been a lot of excitement in the investor community around vector databases and vector search, and I think at your Current conference in October, you had some customers talking about how Kafka is used to interact with vector data. I was just curious if you could comment on the vector opportunity and whether you believe there'll be new products from Confluent to address that opportunity. Yeah. Yeah, those are both great questions. On the topic of the total customer count, what's the interaction with the paywall? At that very first step of how do you spend your first, you know, few cents with Confluent, our goal is to make it as easy as possible for customers to get going. That's gone pretty well. You know, I think as we've removed that means that a lot of our open source advocacy can now directly start with our cloud product. It means that, you know, that kind of early development can go in many cases, totally unmonetized. That does, you know, have an impact on the total count, but we think probably leads to more, you know, 100K+ customers over time. We're really focused on getting those sticky customers that are gonna persist, rather than trying to monetize the early experimentation, for the obvious reason, you know, in that chart and kind of ramp up that Erica showed, that, you know, there's lots of people kicking the tires. They can kick the tires with Apache Kafka, we would rather they kick the tires with our cloud product. The monetization opportunity really comes as it kind of starts to get to production. We feel like that's gone pretty well and has really opened up the kind of direct integration of the product into the sales motion, into some of the advocacy motions and the self-service side. I do think there's an opportunity to further accelerate customer acquisition. Like, when you look at the raw numbers, another way of asking the same question is like: Hey you know, 4,000, maybe heading towards 5,000 customers. Are you happy with that number as a percentage of the total Kafka count? The answer is no. Like, we think it should be much higher, and that's absolutely what we're building towards. To get to that, it really does mean, you know, driving down the kind of amount of friction of adoption of cloud. I think it's something MongoDB has done an excellent job of. As you do that, you kind of access this broad pool of users, and that's definitely a focus for us. To the second point, yeah, the key question with AI is exactly what I said: How can I combine the kind of unique and private data of my enterprise, with this kind of general purpose model that has some understanding of the world? If you think about how humans do that, sometimes it's by memorizing a lot of facts, i.e., training, but sometimes it's by looking something up and, you know, getting the answer out of a book or a computer. Both of those strategies are gonna be pursued to some extent. The vector databases or direct lookups are kind of looking it up on the fly, and that becomes increasingly, you know, important in generative AI. If you look at what people are doing, say, with the OpenAI APIs, their newer models don't even allow fine-tuning, let alone, like, retraining with your data. It is all prompt augmentation and what's out there. They're adding some of that stuff in. Some of their older models do have it, but it is all about, how do I get this data at the right time? That can be either by a direct lookup, or it can be out of a vector database, which allows a more inexact search. For those who don't follow this stuff, effectively, a vector database allows me to take the model of language that my LLM has and apply it on my data for indexing. Which of those strategies will win out? I actually don't know, and even within the vector database world, there's, like, you know, a dozen of them, as well as a number of, you know, existing things like Elastic and PostgreSQL, et cetera, that have kind of vector capabilities. It's a little hard to predict. The data flow into vector databases is easier to predict. We're predicting Confluent as the mechanism for that flow. Yeah, whether it's an exact lookup, you know, an inexact lookup by, you know, aggregating data into a vector database, something else that we haven't thought of, you know, I think that's why I was kind of pointing out that one of the safer bets in what is a very rapidly evolving ecosystem is in the data flow layer, whereas a lot of this other stuff is very much in flux, which makes it a little bit harder to call, you know, the future state, whether that's, you know, models as a service, models trained by me, vector databases, which vector database, other lookup mechanisms. You know, regardless of which of those, kind of forks in the road you take, the fact that your data is spread across, you know, 117 out-of-date systems in the wrong format and not supporting kind of real-time load and lookups, that's a problem you're gonna solve with data flow of one sort or another, and we think the basis for that is gonna be Confluent. Sanjit Singh, Morgan Stanley again. Thank you for all the content today. Really appreciate it. I had one question for Erica and then one for you, Stefan, and potentially Rohan as well. For Erica, you talked about, you know, really doubling down on higher propensity to consume segments. I wanted to define, like, what the lower propensity ones would be, because you seem to signal an opportunity both in the large enterprise as well as in the mid-market, and so I wanted to maybe get a sense of potentially where you guys are pulling back from. Yeah. I would think of it as segments within of any company size, within enterprise or within commercial, where there's high propensity for data streaming to reduce TCO and have meaningful ROI. What are the industries that are making big bets on data? Certainly a lot of startups who are digital native, all digital native companies of all sizes. Financial services has been a very high propensity segment for us for a long time. Telco is another one. Public sector has been a great segment for us for a long time across both different branches, including Department of Defense, including different citizen agencies. It's less about the size of the company and more about what are the industries and the companies that are making big bets on data, is how I would think about it. Yeah. Great. I appreciate that clarification. For Stefan, on the pace of margin expansion of 5-10 points per year, should I think that as getting to that 5-10 points up to the target model, or is there any sort of considerations on whether some years below that or some years above that? How do I think about that in terms of the target model? The other part on the expansion business, and I think, Rohan, this was on your slides, about talking about there being a really high margin expansion segment of revenue, which was encouraging for me to see, 'cause there's been other players in the space historically, like a Cloudera, where it wasn't land and expand, it was land and mini land. It seemed like the customer had to restart its investment after every sort of use case. As you go from fraud detection to inventory management, to whatever the use case was, it seemed like it was very costly for customers to do. I was wondering to see if you could, either of you, could sort of break down the drivers of that margin expansion for the expansion business. Sure. I'll take the first one. You take the second? The midterm target model of 5%-10% operating margin is something that we're gonna get to, you know, like, over time. The rate and pace, once we get to that midterm target model, is really gonna be dependent upon our ability to grow our top-line revenue. If we can grow at elevated rates, we're going to show what I would just call very modest operating expansion on our path towards getting to our long-term model. If revenue growth rate were to slow fairly dramatically for a prolonged period of time, we would have a faster rate to that long-term operating margin target. We think that the market is big enough where it's a very large market. There's a massive do-it-yourself Kafka market out there as well. We have differentiated technology. We have this great customer growth, go-to-market motion, coupled with what we think is the best engineering and product team in the space. We want to play to win, we're also delivering this robust growth while improving margins, that rate and pace of margin expansion is really going to be determined on that top-line growth rate. Yeah, to your question, I think there are a few concepts that you need to put together to come up with the answer. I'd say, three things I'll talk about. Erica touched on her presentation, you know, 50% of cloud customers start from PAYGO. When you double-click into that, 90% of our commercial customers are essentially start from PAYGO. You couple that with our cloud net retention rates are greater than 140%. That's the three... That basically tells you that a lot of our cloud customers start small in a frictionless manner with fairly low sales touch, right? That journey over a period of time is more profitable for the company in general. Typically, as you go to year two and year three, because of the expansion side of it, the incremental dollar of expansion is less, it's less costlier than the incremental dollar of land. Those concepts, like, put together is basically where my comment came from. I mean, from a customer standpoint, of course, like, consumption that we drive, the network effects and sharing of use cases, streams of data getting picked up by other teams, that's essentially what drives the incremental consumption that you see. I imagine the cloud makes that easier, going from use case 1 to use case 2, 3, 4, 5. Yeah, I think that's absolutely. Like, if you compare us to, like, a Cloudera, I think there's a number of critical differences, but that's one of them, is, you know, the fact that, you know, you can just continue that expansion in the development team with no kind of infrastructure effort, ordering servers, setup, operationalization. A company can standardize on a technology, and it can become just one of the ingredients used in virtually every application to some extent, so that that expansion can happen over a very long period of time as applications are built and rebuilt and can be done without kind of a big setup and wait time period. Okay, we've got our last question here. Thanks, guys. Matthew Hedberg, RBC. This has been great. It seems like you guys are squarely positioned for GenAI, I guess the question is in the future, do you see a way to quantify the potential benefit that customers are, you know, attributing Confluent to GenAI? Is it API calls? Is it some form of mechanism of monetization? Secondarily, Erica, you know, when you're talking to these $10 million customers, where are they in their journey? Do you see the path to a $20 million, $30 million, $50 million ARR customer on sort of the high end of the Fortune 500 scale? I'll speak to the first half, and then maybe others can jump in as well. We, you know, As Stefan said, we haven't included anything in the TAM around GenAI, and I would say at this point, it's just really hard to put, like, a number that we felt we could stand behind around that. It doesn't mean it won't be valuable, it just means we don't know what it is. You know, I do think that there are opportunities for additional product features that will help capture that, make these use cases easier, play to the patterns. In part, we want those to kind of co-evolve with customers and with the ecosystem as that kind of settles out and we see people using it. That is one of the nice things about these kind of open, you know, popular development platforms, is that to some extent, your customers and the larger community around the technology kind of shows you what to do. You just kind of watch what they're all doing, and then eventually fold in the right thing, rather than trying to guess upfront in a very fast-moving ecosystem and doing something that's, you know, totally off-base that you end up supporting for eternity. If anybody wants to add on to that answer, please do. I just thought one thing that I would add is, like, when you look at GenAI, it's squarely in relationship to the what we've all been talking about streaming and Flink. Consider, you know, when you're chatting with a GenAI bot, it needs to know everything about you, but not just everything about you at the moment you began the conversation. It needs to know everything about you as the conversation is happening. That's the whole point. That's what you talk to an agent for. Consider an agent for travel or an agent for any other type of business process. When you, for example, are working with a bot, you're asking it about your travel, you're asking it to change your flight. You want to know if your bag's moved. You want to know if your rental car is going to be updated. You want to know all the follow-on effects of the actions that you're taking during your chat. It might be obvious that, you know, you would need a platform that does streaming to bring together the profile of you, your flights, the flights that exist, opportunity for changing flights, your baggage, your rental car, all of that stuff that has to be brought together. That might be obvious. Now, to do that, you have to have the streams, you have to use something like Flink to bring it together, to basically bring together all of those existing streams to produce a view of you. As time goes on, you still have to update that during the conversation. I think I'm just trying to illustrate that, streams themselves are one thing. Stream processing is like yin to yang of streaming. It takes these existing data streams and makes them super rich, super powerful, and then giving that to a bot is what makes an experience feel like it knows you, feel like it can actually take action in the real world and do really powerful things. Your question is, you know, how will we quantify that, monetize that? One thing you'll see is what we already see, which is that, streams are valuable. Streams drive reuse. When streams are joined and enriched, they create even more valuable streams. Kind of it's a, it's a virtuous cycle, where you get more and more consumption of those streams, and those streams become higher and higher value. Now, will we be able to directly kind of productize pieces of that? I think as time goes on, we'll be looking more about how the landscape evolves. It's worth mentioning that Flink itself has a pretty rich and popular machine learning tool set already. Very good point. Great. I'll take the second part of your question around what's the potential we see at the high end within, say, our Fortune 500 customers. You know, today I shared the example of a customer who's currently at about $10 million in consumption, and we can see in the near term, $15 million. That customer is not a Fortune 500 company, not even close, and so that's just, you know, just to put that example in context. When we met at Current last year, I talked about another customer example, currently, already in the $10+ million cohort, this is a large global bank, and we shared that we have direct line of sight to $20 million plus for that customer. I'm confident that there's potential for us beyond that, but that's kind of what we have line of sight to now. We're very comfortable with that $10 million on average, ARR potential across the Fortune 500, 'cause we do see it play out customer by customer. All right, thank you. This is the end of our Q&A section. We will now transition to a customer panel with three amazing customers. Great. Thank you. Thank you. Let's go to our next one. All right. As we wrap up with the day here, saving the best for last, which is our customers. It's always great to hear from Confluent and everything we're up to, and hopefully, you've got a lot out of that today. You know, ultimately, it's the customers who have the challenges and are solving the problems and driving value in their business. Today, we have three amazing customers here with us. I'm gonna have them do proper introductions in a second, but I just wanted to give you a backdrop of what we're looking to cover on this last session. One, we just wanna make sure you know a bit about the people in these companies, and how they're engaging with data streaming and Confluent. Many of you have asked us about the personas. You know, who are you talking to in an organization? Who are responsible for these roles? You know the obvious ones, like Kafka developers, but executives in the organization. The second is, many of you have asked us about how does this tie to the business? You can give us all the technical underpinnings, but connect this to the business value and what are the business challenges that you're trying to solve, and that's what we wanna have in this conversation here. We then wanna get into why Confluent over an open source option or a competitive option, and these folks are gonna speak freely and candidly about that. They're gonna talk about the value they're driving in their organizations to the extent that they can publicly here. I will say that they're gonna be joining us at the reception afterwards, so they're gonna be available to talk to you all one-to-one as well. Then we'll look at moving forward with data streaming in terms of what are the possibilities, what are the things they can move forward with next. With that, let's dive in. We have here Mahesh from McAfee, we have Sarvant from Penske Transportation Solutions, and Babak here from Vimeo. I will kick off with you, Mahesh. Tell us a little bit about yourself. Thank you, Stephanie. I'm Mahesh Thiagarajan. I'm currently the VP of Platform Architecture and Protection Tech at McAfee. It's all the backend side of the house. Going back maybe about 10-12 years, I've been focused on taking monolithic code bases to the cloud, breaking them up into microservices, and starting to deploy it at scale in the cloud. Walmart.com that you see today, was built by my team between 2012 and 2015. Subsequently, I had a similar role at Kroger, and after Kroger, I've moved to McAfee, and I'm doing something very similar on breaking up what was there for about 30-odd years. We have actually transitioned all the workloads to the cloud. We have rebuilt all of that as microservices, and this is the year we're migrating out of the monolith into those microservices that have been built. I wear three hats. One is the platform engineering, the other is architecture, as we actually change for all of the stuff that is going on right now. The third is all the backend protection stuff like VPN, antivirus, identity protection, and so on and so forth. Thank you. We'll come back in a moment on about McAfee, what you guys are up to these days. Sarvant, tell us about yourself. Absolutely. Happy to be here, Stephanie. I'm Vice President, Data and Emerging Digital Solutions at Penske Transportation Solutions, where I lead the creation and implementation of data management, IoT, AI/ML, generative AI, and broader IT transformation strategies. I'm an industrial engineer by training from IIT Roorkee, and I got my MBA from Vanderbilt. I've worked for companies like Tata Steel, implementing ERPs, and before Penske, I led information management at Cummins. I've been working in the data analytics and on the cutting edge of technology, throughout my career. Thank you. Awesome. Babak from Vimeo. I'm Babak, Director of Data Engineering at Vimeo. I've been working in tech for close to 15 years, nine years in data analytics, data engineering, currently, I lead the data engineering, data ops, and data platform at Vimeo. All those three teams interact with Confluent, where at different capacities. Now that you've got a little bit of backdrop on each of them as individuals, let's dive in to each of your organizations. I think most of us probably know all these names. I know in my time interacting with you guys, I learned a lot more about what each of you are up to. Maybe, Babak, talk to us about Vimeo. Most of us have seen a video and know a little bit, you're a pretty innovative tech company. Maybe tell a bit about your business priorities and what you are up to as an organization. Yeah, absolutely. Vimeo, we take pride in providing an all-in-one, end-to-end solution, a video solution for our customers. Currently, we serve a variety of customers, from creative individuals to marketers, small businesses, and enterprises as well. We have over 260 million active users, 1.7 million subscribers, and billions of video views every month. As you can imagine, we collect and consume a lot of data. Fantastic. Sarvant, I was guilty when we got on the call and talking of thinking of the yellow trucks for Penske, but I learned you're a lot more than that. Can you tell us a little bit about Penske Transportation Solutions as a whole? Absolutely. At Penske, we love our yellow trucks, but we are definitely more than that. We provide innovative transportation and logistic solutions that we believe are vital to the success of our customers. Our product lines include full-service leasing and contract maintenance. We have a truck rental service that provides both commercial and consumer truck rentals in North America. We also have a logistics business. Penske Logistics is a leader in logistics and supply chain management, and it is our wholly owned subsidiary. We operate more than 400,000 trucks and serve our customers from 1,400 plus locations. We are an information-intensive business with a lot of information getting exchanged between our customers, associates, and partners. In our industry, vehicle uptime and supply chain visibility is extremely important. Fantastic. Thank you. Mahesh? Sure. As many of you might know, McAfee is in the protection of you as a person, your devices. We started off as antivirus, but as your data and your identity and so much of your information is out there online, we provide that holistic protection of you as a person, identity, your credit, your data privacy. If you go look up how far your data is in all different places of the web, you'll be surprised. We take care of you and your identity and your family. That transformation. We're going through actually three different transformations at the same time. One is the product mix that goes from just the devices to protecting the person and account. An account could have a family, could have a household, could have a lot of members associated with it. That's one aspect of it, which is all aspects of your identity and data, apart from just the devices. The second transformation we are undergoing is, we have a technology lift and shift that we have done, we are going through this microservices transformation. We're breaking up the monolith so that it allows us to have better velocity, and that comes with all of the joys of, like, taking something and making it a cloud-native architecture. The third is also our endpoints, that is really on the devices, have gone through a significant upgrade in the last couple of years. Even if you're using it, unbeknownst to you, it's gone through a pretty big change, and you just see something, a new product show up in your hands as we've actually been able to deploy that out. A lot going on right now at McAfee in that respect. You had shared with me, and I think it's pretty cool, that you had over 100 million active endpoints right now. I mean, that's a lot of data, and maybe we bridge now into talking about data and data streaming and then why Confluent. Let's talk about those 100 million active endpoints and how you're thinking about data streaming to address. Absolutely. We have active endpoints on PCs, Macs, all the phones. We sell very significantly through our partners in the OEM, retail, and ISP world. It comes pre-installed, then you become a customer. There's a huge amount of data that's making its way back. It's both telemetry and business metrics, so we can know to actually protect you better by looking at samples or by looking at events that happen in the endpoints. That's one part of it. Mm-hmm. The main reason we chose Kafka and Confluent was the move to the microservices architecture. We are completely decoupled. We have domains that we're at 450 microservices right now. We were at three hundred microservices about three months ago. By the time we're done with the migration, would be about 1,200 microservices, and these are completely decoupled. They don't call each other unless they're in the same domain, in the domain space. Like, user is different from account, which is different from the subscription, and so on and so forth. In those cases, we actually event between the various domains and various services. Somebody would say, "There's a subscription," and the protection domain will say, "Oh, I have to protect your identity." It'll subscribe you for identity protection when the subscription domain says, like, there's a new subscriber that's shown up, things like that. That was, I think, use case one. We were already using it for the business metrics and the telemetry that was coming in. We are also using it for reporting and analytics. The microservices world, you cannot do ELT or ETL very well. We generate events, and that becomes the source for how the data is viewed from analytics, how we report into our financial systems, and so on and so forth. We also do that same thing for real-time business events. The more interesting thing and, like, it's usually the forgotten word, migration, and it's very difficult to actually migrate large amounts of data. We are using the eventing system for that as well as we actually go through with the migration. When you say eventing system, maybe, share with the group where you started, what were you using, and then your path to Confluent, and why Confluent? Sure. The microservices was a greenfield build-out, so we chose Confluent directly for the microservices build-out. It so happened that my architect came from prior experience of having worked with you, so that helped bridge the gap. We did look at. We were running open source Kafka before. We looked at why we would do this, one of the things that we chose Confluent is the manageability, the governance, the security. More so, I would rather put my engineers to look and work on protection stack than to be fully brought up to speed. Kafka is difficult to manage, taking that difficulty out of our hands and putting it into the hands of the experts, gives us the ability to rely on it and redeploy our resources into what we do best, which is protecting you as a customer. Thank you. Thank you. Sarvant, talk to us, in the same vein, the journey at Penske, in terms into streaming data and how you got started, and then how you arrived at working with Confluent. Yeah. Let me share an interesting statistics first. Truck fleets deliver approximately 80% of the freight in the U.S. The supply chains are extremely lean. Customers who are operating trucks, they have a need to know where, when, and how their trucks are operating. Similarly, those customers who are waiting for their freight to arrive, they have a need to know where and when their freight will be delivered. Without having access to real-time data, it's a huge issue. Disruption to the movement of freight can have a domino effect, whether it is by truck, railway, or ship. We all saw during the pandemic how disruption to supply chains had a domino effect and led to scheduling delays at manufacturing plants or stockouts in a local grocery store. Real-time data helps trucking companies and those who depend upon them, in the supply chain, make adjustments in real time. Perfect. Thank you. Maybe share with us your Babak, your use cases, how you're utilizing streaming today, as well as your journey, starting with open source Kafka and then moving to Confluent? Sure. I'll start with the journey of going from traditional data warehousing to real time, 'cause this is something that limited us to making decisions with a day of lag when you're only relying on daily batches of data and you're limited to midterm and long-term decisions, as opposed to being agile to pivot, to change decisions, or make decisions quickly after a product launch. I would say we have two major use cases for real-time data. One is for Vimeo as a video platform. We have a reputation in the industry as a high-quality video platform, and we want to maintain that and improve that. Part of that is the experience of the user when they are viewing a video, when they're playing a video or watching a live stream. That it's important to get that, to collect data in real time. For example, the network connection of the user or any information from the user platform or the infrastructure in general, to adapt the video quality. For example, the package sizes that are delivered to the customer to provide that seamless video experience in live stream and video playback in general. Secondly, we as a SaaS company, we are interested in understanding user behavior and to make product decisions. That is to help improve and optimize the user experience, to understand what's the user journey like, and based on that, we can optimize that user experience, and also help user with product discovery, for example, so they can get the most out of Vimeo as a software. This is... again, we are getting the higher demand from product managers, product marketing people, marketing analytics folks, that they want to understand how a campaign is performing in real time. They don't wanna wait a day when a campaign is launched. When we launch a new product, when we are running an experiment, an AB test, we want to really be able to pivot if we need to or make decisions quickly. That timely decision is really important for Vimeo as a SaaS company. Overall, with that would drive growth and also would help optimize the cost that eventually impacts Vimeo's profitability and bottom line. That's why it's important where already, it's a mission-critical thing for Vimeo to have access to real-time data and analytics but the importance is growing. Fantastic. Thank you. One of the things, as all of you are on your path, with streaming data and with Confluent, is maybe sharing with this group how it's impacting not only the customer experience, but the business value it's driving. I'll caveat by saying, you know, this is a public forum, I know that this question's within limits of what you probably can share. I think to the extent you're comfortable with the group here, just talking about the changes in your business that you're seeing, the positive changes that you're seeing in your business, the results that you're seeing in your business as a result of streaming and Confluent. I think that would be helpful. Like, what are the gains that you're getting out of your business as a result? Mahesh, why don't you kick us off? I think I'll hit the non-functional area, and then it's a business benefit in allowing my team to redeploy resources from what they were doing to support what we were managing on our own, to being able to free up that resource to be concentrating on looking at other things. That is one part of it. The second part of it is we do real-time metrics far more. We had a small number of real-time metrics, which we were doing through ELTs. We had load databases, and then we'd pull from there and then form some metrics. Now, pretty much everything that we do is through events. We've built this real-time event lake, if you will, and we can run. We use Druid for it, and we're now in the process of building out, "Oh, you ask me any question, I'll give you an answer, and I can give you a real-time answer." There's a use case for the long tail and the longer-lived analytics and the data lake, but there is also this real-time metrics that we can see right away if cart conversions are changing. We call it ATP or ATU, activation to pay. All these metrics are very important to us. It allows us to monitor a lot more and slice and dice a lot more because we are able to bring that into a real-time system. I'm not getting into numbers per se, but you can imagine if you had a few metrics and then you have a lot more, you can certainly look at all the dials and turn it differently. Thank you. Thank you. Sarvant? Absolutely. For our customers, vehicle uptime is extremely important, as I mentioned earlier. Before real-time data streaming, if there was a problem and let's say a driver is experiencing an issue, it was very difficult for us to find out where the truck was located and what was happening to it. Now, with real-time data streaming, we can predict when a truck is going to fail. In case of a failure, since not all failures are preventable, we can, real-time data streaming helps us track where the truck is located, what's happening to it, what kind of fault codes it is generating, and we can monitor the health of the vehicle remotely, and that's where the value lies. The real-time data streaming is also helping us deliver better driver experience, and it is making our technician's job easier, which is especially helpful given the technician shortage the transportation industry is facing. There is definitely, value, significant business value, being able to understand this, sensor data on a real-time basis. Yeah. Thank you. Babak, how about you talk to us a little bit about the benefits that you're deriving from it at Vimeo? Sure. As I mentioned, what we are currently getting value out of is the real-time monitoring of our campaigns, our product launches, experiments, and campaigns. If I want to go beyond that, what we want to achieve and we're looking into, is to go beyond analytics and having a human in the decision-making process, and enable real-time and automated flows. By that, I mean, streaming the data and the analytics back into our product and operational systems. That means adapting the user experience in an automated way, by providing recommendations or changing the user experience, making it customized and more personalized, and also for marketing campaigns. Whether it's an email marketing campaign or an in-product messaging one, we want that to be customized and also in real time. I'll give you an example: If a user signs up, the first actions matter, and if we wait a day to send a message to user to, for example, I don't know, upload a video or show them how to do it, how to create a video, that wouldn't be as impactful. That's why these automated flow that can be unlocked and enabled by real-time data are important. That's an area that we're looking into. Another area is real-time enrichment of data. I mean, we have our data warehouse that has that's hosting a wealth of data. Again, what we need when we're talking about the enabling operational systems, that's where we want to be able to enrich data in real time, because that will take the load away from our application, that improves the performance of our application. Also, again, the, as I mentioned, will create that impact in real time. This is where we are looking into implementing using Stream Designer and then in Flink, that would enable these opportunities. Thank you. That's great. As we start to wrap up here, I would love for you guys to share the forward path for you guys on your streaming journey. You've talked a little bit about everything from microservices to analytics, to the supply chain. I mean, how data streaming and how Confluent are helping to drive your business around those. It'd be interesting to hear any future use cases you're thinking about and your path forward. Maybe, Mahesh, let's start with you. Yeah. This year, we'd be using it extensively for just our data migration. We have an enormous user and subscription base. We have to bring them over, that would be the most immediate tactical use case. Then we would like to move our data closer to where our customers are. I'm hoping that at some point in the future, we use that to geofence the data in various locations. We do business in 46 different countries. We're in this baby step, right? First, get to the cloud, get to in one place, then get to many locations, and so on and so forth. That's the journey that we'll be very much looking forward to and using this as a methodology to get us there. Fantastic. Thank you. Sarvant? Absolutely. Transportation technology is evolving very, very fast, we will see this industry transform itself as we see more digital innovations. We will see a lot more data getting generated as more EVs, electric vehicles, join the fleet. We will see the change in the behavior in terms of as vehicles get used, as more autonomous or semi-autonomous vehicles roll out. There'll be a lot more predictive maintenance happening. We may also see what is inside the truck, that getting monitored on a real-time basis. For example, real-time monitoring of temperature of products inside a refrigerated unit. We may see that happening. The combination of artificial intelligence, IoT, real-time data processing, and computing on the edge, that is going to expedite this race to the left, and I see business and technology architectures evolving accordingly. I believe that's going to unlock a significant amount of economic value in this sector. Overall, I see a lot of very exciting things happening in the transportation technology world. Fantastic. Thank you. How about at Vimeo? What's next? What's on the horizon? Well, part of it is that the opportunity exists with providing a better video experience for our users, and that's something that we are leveraging, but there's so much more we can do in terms of monitoring infrastructure. For example, looking at the network latency, average playback time, and to identify issues. If there is an infrastructure issue, to identify and resolve it in an automated flow. That requires analytics in real time, 'cause we are talking about the milliseconds of... We can't afford minutes even, when we're talking about providing that experience to users. The other is again, related to video, is content delivery, when we have automatic content delivery network selection, for example, understanding your and reading user preferences and adjusting the video according to that. That is, again, that's something that needs to happen in real time. When we're talking about billions of views and millions of users, that requires a robust and scalable and reliable way of processing that data in the streams, yeah. If real-time insights means no buffering of videos, I'm happy. Yes. I think we all will be, very much. Live stream as well. I'd have the mention about the use case of AI in video. Currently, it is being used in creating videos. You know, with a few clicks, you can pick a template and create a professional-level video within minutes. We're talking about the future, the generative AI, and the concept of text-to-video. Instead of the way that we interact with video creators, video creator tools or video editing tools can change, where we can describe what we want, a video as a marketer, for example, or how we want to edit it. Then the tool generates it. That's again the back and forth with the data. The processing of data in real time is important because we can't afford that kind of delay in batch processing. Fantastic. Well, I wanna thank all three of you for your time today. Again, all of folks here will be with us at the reception for any one-off questions that we can ask, but I know it's a massive investment of your time to come here, be here. I'll probably state the obvious, but it's not always easy for customers to also speak at an investor conference. Many thanks to you guys, and we really appreciate the partnership. If we can give them a round of applause. Thank you. you. Thank you for having us. Okay. With that, we are wrapping up the day. A couple quick housekeeping things for all of you. The wall is going up right now, on the other side of the room here. When you're heading out, just around the corner to the right, we have the reception. We have all the panelists there. We'll have some demonstrations for you. Please stay around and join the reception. On your way to the reception, we are do have a quick survey for you. Very quick, and in return for your time on that survey, we have some wonderful jackets. Hopefully, you'll fill out that survey. Thank you. All the slide decks are now on our website, so feel free to go there and download them. Thanks to those joining us again online. Take care.