How is everybody holding up? First day, it's just a few meetings in, and we've got three more days to go. Are you ready for three more days? But before we get to the- Everybody looks tired. No, it's just that. That's just the post-lunch. I see. I think there's a lot of our clients who also have flown from different parts of the world. The jet lag's gotta be a huge factor as well. Not for the local people like me, but Jay, thank you. I think it's your third year consecutively back- Yeah -Back at Goldman Sachs- Excited, excited to be here. Yeah, and Technology Conference. Thanks for making it once again. Really, really appreciate your participation. So I think it used to be easy to start off the conversation on Confluent: "Tell us what you do, what is Confluent?" Now, most people understand that, but maybe you could share with us your vision for the company. How has the company changed over time, and where do you see it five years from now? Where do you want Confluent to be? What are your goals and aspirations? Yeah, well, in some sense, the vision hasn't changed that much. You know, even since the founding of the company, the goal has been to build a platform for real-time data in enterprises. You know, the area of, like, data warehousing has always been this awesome center of data that's very much like, end of the day, ship all the data there, but increasingly, it's about how do all the applications and systems work together in real time to run the business, and that's the role that we wanna fulfill, and, you know, so in some sense, it's a very boring answer in that like: Well, not that much has changed. You go back and read the early blog post about starting a company, and it wasn't that far off from that, but in another sense, you know, it's changed a lot. The... You know, when we think about kind of the phases for the company, the first phase of the company was around the software product, and as we were going public, you know, that was really the bulk of the business, with then a little bit of the next phase, which was cloud, and you know, we feel like, hey, we're kind of entering a third phase here, where we're adding what we would call the, you know, the rest of the data streaming platform, so it's, like, the rest of what you need to really manage real-time streaming data in a company, and for us, that's all the connectors into different systems, the real-time stream processing capabilities with an offering around open-source technology called Flink, and the ability to govern real-time data. We feel like the role we can play for our customers in enabling applications, enabling workloads, is much more significant when you have a solution end-to-end to how you capture data, how you process it, transform it, work with it in real time, how it connects up and flows across the organization, and how you can govern that in the large. You know, we've talked about this chunk of functionality a few quarters back. We said it's about 10% of revenue now, but kind of outgrowing, you know, even the cloud business, which is outgrowing the kind of business overall. We feel like that's a really strong kind of tailwind to the business, as well as an exciting kind of, you know, next step in the journey for us. That is the, you said, 10% of the business today? Yeah, the cloud rev- Okay. Yeah. What do you call this? I mean, is there a product that's branded or? Yeah. So, you know, we think this larger platform companies want, we would call it a data streaming platform. Yeah. Right? And the, you know, within the capabilities for that, Kafka is the core stream of data that would get around to the different parts of the organization. But the new things that we're adding is Flink, which is the processing layer. So you can think of it as, in database terms, people work with SQL, which is kind of the language you would use to query your data. Flink would bring that to this kind of real-time streaming. So in a database, maybe you'd have a stored set of data. You could write a query like, "Hey, how many customers do I have in California who match this criteria?" It would churn through all the data and give you the answer, and the answer would be, you know, whatever it is, forty-two. Right? In stream processing, the idea is, instead of doing that at a point in time, you know, the end of the day, and having it be out of date all, you know, right away, you run this continuously as new customers join or leave or cross thresholds. You have a running calculation that's always correct, right? And you can always look that up. And so when you think about how do not just humans interact with data, which is maybe just at a point in time when you look at the report, but how do applications interact with data, it's very continual. It's like that. And so this kind of stream processing is really coming to the fore as a key capability in the world of data and a key capability for us around these streams, where we're already kind of a leader. And in some sense, the value of any new technology becomes more evident as you build a platform and an application, set of applications, and then, like ERP was a killer application, and CRM was a killer application for the cloud. The applications tend to make sense. You, you make sense of the whole infrastructure platform, and all the capabilities of the applications drive a sense of tangibleness to it, right? And that's been the one of the aspects of the conference, right? That, that those applications make it clear as to what is it that you do, the streaming product- Yeah. -the integration- Yeah. the connections, et cetera. Yeah. Where are we in the mainstreaming of applications that are natively built for the Confluent Platform? Yeah, yeah, that's come a long way. You know, increasingly, I think many parts of the data world have to either, you know, be built to produce or, consume streams of data, and that's a big tailwind for us. You know, when we were getting started, it was very much about bolting this on to a world that really just, like, wasn't built for real time. Yes, yes. That was a bit of a hard task. Mm-hmm. ... for us. Now, this is increasingly something that I think every product is building around. And I would say there's actually a broader theme in the data infrastructure space. You know, I would say we're in a period where, y ou know, maybe we're coming out of a period in 2021 or whatever, where there's really kind of a thousand flowers blooming. And now I think you're seeing more standardization around common interfaces. So for the, you know, kind of operational databases, I would say it's really Postgres. And so everybody has to, you know, look like a Postgres service. There may be many competitors who are trying to do that, the cloud providers, other companies, but that's what companies want to have. In the analytics world, technologies like Iceberg that provide an interface to data that's common across different systems have come around. And then for real-time data, you know, the world is very much standardized on the Kafka protocol, and that's pretty much the thing. And, you know, I think that's a positive force for us. When something starts to become a known quantity that many things can build around, then you get, you know, many other technologies and partners that integrate into that. That becomes, you know, a bit of a force that drives- You use it as a platform. Yeah, that, that's exactly right. And so, I think that's happened in the streaming, you know, streaming space, and that's part of just what I think is happening overall in the world of kind of cloud data infrastructure, is this kind of standardization. I know there's an announcement after the market close, or maybe before the market open, you announced the acquisition of WarpStream. Yep. So- Yep. I think, we're fortunate to hear it from you. Yeah. This audience here. Yeah, this is genuine new news. New streaming. New news. Real time. Yeah, that's right. That's right. New events are occurring. Yeah. Yeah, so this is a company in our space that offered a streaming product with kind of a particular architecture, and we felt like it filled the gap that we had. So we had a product which was, you know, self-managed software that you can take and run in your data center, and we had a fully managed cloud product. But one of the things that came around is actually a really good implementation of something that's kind of semi-managed, right? So you can take it, and it runs in your cloud account. The data doesn't leave your cloud account, but it kind of takes streaming out to some of these workloads that would be hard to access with a fully managed cloud product, but where they don't really want to do it all themselves. And so it was kind of a nice way of, you know, fully addressing that niche in the market. And we think it's important, you know, at least for us, we want to get all the streams and kind of soak up all the usage of open source Kafka. And so when we look out at that, you know, there's still a lot of usage of the open source technology. We want to make sure we have all the kind of packages and configurations that can go and serve those customers. Particularly, we felt this would help us address some of these very large workloads around maybe observability in some of the big tech companies, where we felt like, "Hey, we've got great customers, but we're still, you know, just scratching the surface of what's there." So we're excited to add it to the portfolio of offerings. You know, we think it'll be, you know, a great addition. It doesn't change our, you know, plans over the, you know, course of the year. You know, it's still an early startup product, so it'll take some effort to kind of fully integrate and harden it, but we're really excited about adding that to, you know, our set of offerings. So you have the cloud, which is managed by you- Yep. ... on-premise, managed by the customer. This is kind of in between product? Yeah, this is, this is like a cloud product that runs in the customer's account. Okay. You know, if people are foreign to this, most products that are cloud products are kind of all in the provider's account. That allows you to do things like multi-tenancy and full management. Yeah. Most of what, you know, AWS would build kind of works that way. But there is an alternative. So, like, Databricks offers something that's in the customer's account. Mm-hmm. That's, you know, for customers that want to keep their data very close- Yeah. ... but they don't want to do it all themselves. Yeah. That's a good alternative for them. And so kind of accessing these different models allows us to make sure we can just kind of go soak up all the Kafka usage that's out there. Where are you going with this? I mean, could this lean more towards being cloud like, or more on-premise like? Yeah, it's sort of in between the two. Yeah. You know, one of the things when people look at the data space, I do think the assumption has been everything will go to cloud. Mm-hmm. But then, when you look at the actual spend of dollars, what you would see is, you know, the kind of data center spend is very flat and cloud has grown. Yep. And so in reality, the data center spend didn't really move. And, you know, we- We all thought it was gonna go down to zero. Yeah, and maybe it will, but like a lot of things in the infrastructure space, it moves slowly, you know, at least on the downturn. So, you know, what we've felt is, yeah, there are a set of customers that are, you know, set up to manage open source Kafka. Mm-hmm. We want to get them into the portfolio in the lowest friction way possible. Maybe the end state for those is something that's fully managed, but a nice stepping stone is, you know- Yeah. ... something that's halfway there. Yeah. We think that there's a significant opportunity to take that out to customers. Yeah, you're right. The on-prem, it has gravity. I remember people telling me ten years ago that Microsoft had this product they called Server Tools. Yeah, yeah. And, people would say, "It's gonna go to zero!" Yep. It's still growing, annoyingly, 2-3%. It's a big install base of massive compute. The more it keeps flatlining, the more bullish we are, because all that stuff is coming to the cloud. Yeah. Yeah, yeah. The chance just keeps getting bigger and bigger for this. Yeah, that's one of the things we felt like was like, look, there's, you know, over 150,000 organizations that are using open source Kafka. Right. We have, you know, on the order of 5,000 customers. Our goal is soak all that up, right? Mm-hmm. And so- Big unlock. Yeah, of course, we could be, you know, very purist about the, you know, architectural style we want to serve, but one of the things we've realized in our space is actually there's strength in covering all the use cases a customer has, because getting all the streams and plugging them all together, that's kind of what gives you that central nervous system across the parts of the organization, so we feel that, like, the self-managed offering strengthens the cloud offering, and we think that this will help plug some of the gaps of what we, you know, weren't covering for customers, today, and you know, we're really excited to add it to what we're doing. So does it also have the same kind of capabilities that your core products have on the cloud? Yeah. No, I mean, it's early, right? Like any startup product, it, you know, it doesn't have every bell and whistle, so it'll take some time- Yeah. ... you know, to get there, but, but nonetheless, represents a nice step in that direction. Thank you. Congratulations on the acquisition. I know the question of generative AI comes in from time to time. What are the lessons learned from Confluent versus, say, a year back or so when it came on the scene, captured our imagination? What have you learned from your customers' deployment of generative AI, and how does it inform your product strategy going forward? Yeah, yeah. I think there's been a set of really exciting use cases in this space. You know, the first part of the business to kind of really pop was actually selling to the AI companies, right? So customers like OpenAI and, you know, other companies offering solutions in that space, but along with that, and I think the bigger opportunity, is the enterprise use cases, and, you know, there, I think there's a sequence of, you know, kind of sophistication. You know, starts with kind of these internal chatbots, moves to something that can maybe interface with customers, and then, you know, companies would like to get to something that's more like an agent that's, you know, not just taking input and providing text, but taking input and providing action, and, you know, each company is somewhere on that spectrum. You know, there's some companies that are doing something end-to-end, where there's, you know, at least today, where it's relatively low risk. But, you know, today, a lot of the more conservative organizations, it would be something that's kind of internal focused. Our role in these architectures is about, you know, gathering all the data across the organization, getting it into the right form, and providing it, you know, to be combined with these language models to actually serve customers. And this is actually really important. You know, when you think about, you know, the platform for supporting this, inherently, these use cases are about what's happening right now. You know. If it's a support agent, what was it you were trying to do before you called, you know, or started chatting with them? You know, if it's something that's helping an internal part of the team, what's the actual state of the business that you're interacting with? So it demands something that's in sync with the rest of the business, and you know, we've started to see these kind of come out into production. We've talked about use cases across, you know, companies. Really, you know, from trucking companies to travel companies, a bunch of tech folks have been kind of the fastest movers on it. I think we're still just scratching the surface, you know. So for every application that's kind of out there in production and referenceable, there's probably a dozen that are kind of in the early stages. You know, I think the rate- These are applications that were built on Confluent. Yeah. Yeah, that's right. I think the rate at which those kind of come through, you know, that will be something we'll learn over the next year or so. Any interesting use case, or an application comes to mind that was built on Confluent Platform using Generative AI as the use case? Yeah, I think one of the examples that's probably easiest to understand, that was actually relatively early in the adoption, is just been the, you know, customer-facing interaction. So, you know, a good example. There's half dozen that we've kind of given publicly, but a good example is a travel company where, you know, the bookings are all online, but a lot of the time, you know, the booking isn't quite what you wanted, and then you end up calling them, right? And for whatever reason, the customers can't figure out the interface to change, make complex changes in their travel arrangements. And so it ends up being that although they're, you know, an online booking company, 60% of the customers are interacting with them, you know, live, which is not the best experience to kind of wait on hold. So what do you need to do to be able to serve them with AI and deflect some of that or reduce the amount that actually have to go through to an agent? Well, you have to have, across a pretty diverse business, the up-to-date view of what happened, right? And for them, they have a number of different properties, and travel is inherently real- time. It's like you reach out when your flight was delayed, or your luggage was lost, or your hotel was full, or whatever it was. And so they need to have a up-to-date view of things, or it's just very frustrating. And so, you know, the role for us was really provide that, you know, kind of real-time flow of data across parts of the business, get it into the right form, and, you know, be able to get it stored in a way that can be used in what's called a RAG architecture, where you're kind of. Yeah, I was about to ask. ... combining stored data with a language model at runtime. And, you know, I think that's a durable pattern. I think we're gonna see that continue to exist. The specifics of how it works may change. You know, maybe we're not using vector databases in the future, but nonetheless, like combining- I'm curious, though. You know, if you think about the vector database pattern, you know, so what we're doing now is a little bit of a hack, where we're kind of injecting data into the prompt, and that's probably not the only way that a language model can interface with stored data, right? So, you know, you could imagine that improving over time. But nonetheless, the need to gather together a bunch of data to, you know, structure for use in this type of application is very durable. It's not gonna be the case that that goes away, because by having this stored data, you can enforce all the, you know, kind of access control things that make sure customer A never sees any data from customer B, all the things you have to do in an enterprise context. And so I think that general pattern is gonna be, you know, very durable and a key part of how these technologies are put to use. And as we transition from kind of, you know, chatbots to agents, I think it becomes even a better story. In addition to bringing the kind of input data, you're also then carrying out some action, which is gonna trigger, you know, asynchronous activity in the rest of the business. That's the output. And, you know, so I think that's a role for us as well. Got it. I think you brought up agents. I'm curious to see what you make of it. Are we? Is that the next step of chatbots and having these agents?... go do things on their own, on our behalf. Yeah, yeah. Do you really buy into that vision, and how are you going to take advantage of- Yeah, yeah. I mean, I mean, so the answer, the answer is absolutely. I think the question is: what's the timeline, right? That's, that's where the biggest, you know, probably open question is. I'm close to this space in a couple different ways. I'm on the board of, Anthropic, a model company. Of course. And of course, we see these use cases on the other side, and so, you know, the- They're gonna be presenting tomorrow. Yeah, yeah, yeah. It's a, you know, phenomenal organization and really exciting work happening there. So, yeah, there's no question that informing humans to take action is, like, good, but you'd rather just complete the task where you can. And now the challenge is, it's a lot harder than it sounds. And so I think that probably the gap from, you know, the chatbot to the agents is a little harder to do than we think. So like, we think about this internally. We have something which we use, which has information about our product, which can help internal team members. We open that up externally to customers. And, you know, the state that it's currently in is it's available to some customers, but hasn't proven that it actually makes either the buying or support experience, you know, statistically better. And so they're iterating on it. It'll be in partial deployment until it really moves a measurable metric, right? So that's where Confluent is in the adoption. So then we ask ourselves: how much better would this thing have to be before we would actually, instead of just answering questions about your deployment, allow it to actually operate some of the internal software? We'd love that to happen, right? Like, we have a big portion of the engineering team that does this kind of large-scale operations. The answer is, it would be a pretty big jump in terms of confidence- Yeah. ... and quality of decision-making, et cetera, and just understandability for us to unleash it in that way, because it's a huge problem. Yeah. ... if things go down or there's any kind of issue. So getting that human out of the loop, I think it's harder than we think at times. But so I think the initial thing of just augmenting has probably got some legs on it. I think in low-risk areas, you would already see this happening, where it's, you know, where machine learning is used today, where it's, relevance-related, you know, nice-to-have operations. But I think there is a real requirement on model quality- Yeah. ... to kind of complete the, you know, close the loop. But nonetheless, that's definitely the direction, and there's a lot of progress on model quality, so it's not like that, it's not moving. So the loss of jobs is greatly exaggerated, then? I don't know. I mean, a lot of these is about the pace, right? Yeah. You know, I think once one of the exciting things and scary things about AI is like, yeah, actually, the aperture of what happens gets pretty wide pretty quickly as you get some years out, and it depends a lot on what you assume. And so, you know, it's kind of well beyond my pay grade to predict all of that. You know, for me, I just need to help plot a course for Confluent. Exactly, exactly, but more need for real-time information, more augmentation leads to demand for RAG and real-time information. Confluent should benefit from being able to feed these LLMs with real-time information. Yeah, I think that's right. I mean, you know, bottom line, from a Confluent point of view, there's definitely a set of use cases which are very data hungry- Yeah. ... require data broadly across the business that are emerging, and that's a positive- Yeah. ... you know, force for us. Yeah. The other thing I wanted to get your view was the cloud modality versus the platform. Yeah. Where are you in your journey to persuade customers that the cloud... Of course, you've made an acquisition that kind of lands- Yeah. ... in between the two modalities. Yeah. But- Yeah. ... where is the company in achieving that sense of comfort with the customers that, okay, the future is definitely- Yeah. Well, we've made a ton of progress. I mean, just, you know, you could see it in the numbers. When we went public, you know, it was whatever, 14% or something of revenue was cloud, right? And now it's, you know, the majority of the business, and that's obviously huge progress. Yeah. So, you know, obviously, things have gone well on that front. Now, I would say that's still not full unlock, you know, to really address some of the most conservative or security-conscious organizations, and not just sell to them, but really open up their usage broadly across every use case. There's still more to do, right? And so that's- What is the unlock there? You know, it's a really long, boring list of stuff. And so people ask that question, you know, in the public sector, you know, for the U.S., they just bundle it all together as FedRAMP, and that's, you know, one of the criteria. In, say, financial services, there's not really an equivalent. It's more like there's a regulatory regime by, you know, area, but each organization has their own interpretation of what that means. But nonetheless, you kind of burn down this, you know, long, fairly boring list of operational and security-related concerns. And as you do that, you can serve more organizations with less friction. And that's an area that we've made a lot of progress on over the last two years. You know, I think there's still more left to do, and that tends to open up, you know, more of that opportunity for us with cloud for the most conservative enterprises. And, you know, some of those are very big spenders on cloud, so, you know, it's important to do it. Got it. Got it. Got it. Excellent. Let's talk about the, I should call it the GTM change, but it's, i nstead of the way you incentivize your sales force, that has changed. Yeah, it was a change. Yeah. Yeah. It was a change. Yeah. Talk to us about what have you learned from the process, having implemented it for- Yeah. Coming up on three quarters now. What are the things that are going to be advantages to Confluent as you fully get the benefits of this incentive model? Yeah. Is it playing out- Yeah. ... the way you expected? Yeah, yeah. Can you- So I'll first just, like, restate what it was that we changed, and then we'll say what, you know, what's the impact, so one of the things coming into this year we wanted to change was really orient the, you know, go-to-market around consumption, so that, you know, the use cases that our customers would take to production that would drive Confluent revenue, that would be the incentive for the team, and we wanted to do this for, you know, a set of reasons. We felt, you know, first of all, we weren't landing enough new customers, and so we needed to kind of get in there early, you know, even with a smaller land, and just kind of plant more seeds. Secondly, as we were getting to scale, we didn't feel like we were driving workloads at the rate that we wanted to really get, you know, as many use cases as possible going. And third, as we were bringing these other data streaming platform components, we didn't really have a mechanism to drive them. 'Cause at the end of the day, what we were selling was, like, a general commitment to spend with Confluent, not really a particular usage. And so, you know, like, you have a new product, and despite the general commitment, there's nothing really pushing on, you know, taking that out to the world. So for all of those reasons, we wanted to make a switch to consumption. This is something a lot of the peer companies had done, you know, over the last few years and has been very positive for them. Snowflake got on it. Yeah. Yeah, that's right. This was a good year. Yeah, that's right. So we made a, you know, a change towards that. That's gone pretty well. The, you know, what we were hoping to see has pretty much played out. So we've seen higher velocity of customer lands. You know, I think Q4 was roughly on the order of maybe 40 net new customers, which we felt was, like, woefully low, given the number of Kafka users. And, you know, most recent quarter, you know, was north of 300, so a really substantial step up, and we think that can be, you know, sustained at a much higher rate. We'd love it to go even higher. And I think that's really important, is just, hey, are we getting out to all these, you know, new customers, new use cases? Are we actually capturing that? Now, obviously, those new lands, you know, they really start to contribute over time as they grow. But really just, you know, this year's kind of growth customers are last year's lands, and next year's, you know, growth customers are this year's lands, so we feel very positive about that. You know, next, we're really tracking and driving the new parts of our platform, the new workloads. You know, those have been really successful. As with any of these changes, there's, you know, a bunch of things you tune along the way, in, you know, how you're comping different things or what the different incentives are. But overall, it's been, I think, a really positive change and kind of sets us up for what we wanna do in the years ahead, so I feel really good about it. Got it. I'll check anybody with questions, just raise your hand. One of the things you'll see in our fireside chats is we will not talk, we'll not say, "I want to segue, double click." We'll try to use plain English words. Okay, no segueing into- No segueing- All right. No double clicking. What are the other things that people say that are very common, annoyingly common? The one that gets me is, "We have to action that," which I think means do it. Just, just do it. Yeah. Yeah. With the move to the consumption model, I know things... You can often interpret metrics very differently. NER, where do you like the net expansion rate for the company to go? I know it was a little bit south of what possibly you'd expected. Yeah. What are your aspirations for NER? Where would you like it to be? Yeah, yeah. Well, obviously, we'd like it to be as high as possible. Do you want a DBNRR? Sorry. Yeah. There's so many variants of these things. Yeah, yeah, yeah. So we have felt a little bit more pressure there. You know, overall gross retention has remained strong, you know, it's been consistently above 90%. But, you know, look, it's been a tighter environment for infrastructure spend, with a bit more optimization of the existing usage. And that obviously puts pressure on expansion. And so, you know, I think the good news is, you know, we feel like there's pretty strong tailwinds with these DSP components, which are kind of getting to the point where they're big enough to start to move things. They're, w e feel like we've made good progress on just landing a lot of customers that can kind of drive growth. We feel like we've made a lot of these go-to-market changes, so those are positive forces. But yeah, we would, you know, like to see that kind of, you know, at or above 125%, and, you know, it's been a bit below that. And anything as far as planning for next year is concerned, as you talk to your salespeople and talk to customers, what is top of mind? What are budgets looking like for next year? Is the election even a thing that your customers are talking about? How did this, t hat's what I'm planning to ask you. Yeah, yeah, it's a good- It's a good question. So, you know, we don't do as much, kind of, call it macroeconomic forecast. Like, what will software spend be? You know, what will the impact of this election and this jurisdiction be? All of that absolutely matters, but, like, our information about that is much weaker than, say, your information about that. So what we do a lot of forecasting of is, you know, what are your plans around data streaming? Yeah. There we feel like we have an inside track. Yeah. You know, there's, y ou know, it's early to call, you know, kind of any numbers or trends for next year. Yeah. But, you know, I would say the good news in this space is, look, you know, if you ask people, you know, data streaming, you know, what's the role this is gonna play, and how would you have answered that question a year ago? You know, I think for companies, this is becoming more central, more a part of their kind of big picture architecture that thinking about in a really serious way. And I think that's obviously a tailwind for us. Even in tighter times, where people are also optimizing, et cetera, you know, it's really important that you be part of the next generation stack that, you know, your customers are building towards. One of the companies you hosted earlier today, MongoDB, they're talking about streaming capabilities. I mean, I know that this came up about a year back or so with the announcement of the streaming product. Confluent had to come out and say that, "This is what we do, and this is what they do. It's very different." Yeah. Any changes to that observation from? Is there more overlap than perceived or actually less overlap than perceived? I think there's probably less overlap. You know, so if you think about streaming, it's kind of a paradigm shift in the data world. So every company that deals with data will have some integration with streaming. Either you're producing streams or you're consuming streams. Right. Yeah, that's, that's net good for Confluent. Like, a world where nothing works with streaming is not a positive world for us, right? The role we play is kind of at the center. We're like the distribution, you know, hub for that streaming data that kind of acts on it. That's not what Mongo is trying to do. It's not really what any of the other, you know, kind of standalone destinations are trying to do, nor are they particularly well suited to grow into that role, you know? It... And that's just the nature of the operational database is just fundamentally not the hub that is distributing data in organizations anymore. And so, yeah, I don't think that there's a lot of competitive overlap. With Mongo, there's we actually partner really well with them, work on a whole set of use cases, and have been really successful with that. So yeah, I don't see it as really at all competitive. But you will see more, you know, mention of the word streaming, you know, in the data world, and it's actually positive. You know, that means that all of these destination systems, which are stores, can now, you know, participate in that kind of exchange of real-time data, and that allows us to hook up to them in a much more meaningful way. So I think it's broadly positive. Now, there's always some, "Well, we could do this a little bit, that's right on the edge in Confluent or in, you know, that product." So there's some kind of push and pull there, but it doesn't really mean we're trying to do, y ou know, trying to play the same role in organization's architecture or kind of competing for the same thing. Got it. A play on words, mainstreaming of streaming. There you go. Final one, if there aren't any questions from our clients here. Oh, it looks like there's one. Now, can somebody get the mic over to? We can also try and repeat it if- Nadi, just you can just speak out, and I'll work. Yeah. [audio distortion] Yeah. The hyperscalers are talking about optimization starting to attenuate and no longer being as much of a headwind. It doesn't sound like you're seeing that yet. I think some of your, you know, peers in the consumption space have had sort of similar view, that they're not really seeing the headwinds attenuate yet. Can you talk a little bit about why we might be seeing a little bit of a disconnect between hyperscalers and yourselves, given that you'd normally expect some pull-through? Yeah, I mean, it's, So the question is ultimately like, "Hey, some of the hyperscalers have said, you know, there's a bit less optimization than there was previously. It's attenuating. You guys have said, 'Hey, there's still optimization.' Some of the other peer companies have said there's still optimization. So how can we square all that?" You know, I'll try to, although obviously, I can only speak to their experience to some degree. I do think that, you know, there's probably a set of easier to optimize and harder to optimize things. Probably the closer to production you are, the harder it is to optimize and make changes. The hyperscalers are kind of a pool of all kinds of things. There's, you know, analytics stuff, and there's observability things, and then there's, like, operational databases. And I think you probably feel the effect most quickly on the things that are easy to optimize, and then over time, maybe on the harder to optimize things. So I think, like, you know, one explanation would be, yeah, you know, the kind of overall, we went into a regime where there was a focus on cloud spend, maybe a MongoDB or Confluent, kind of feels that only with a bit of delay because you are actually making application changes. Whereas, you know, a data warehouse, maybe you just change data retention tomorrow, and it, you know, kind of drops suddenly or maybe some of the observability things see it that way. That would be one explanation for that kind of difference, is you are just kind of seeing the same thing, but with a little bit of a time delay on it. Overall, you know, like, when we look at our customers, if I look at kind of the digital native segment where we talked about that more, you know, it's a mixture of growth and optimization, right? The kind of, you know, balance in that, you know, in those customer cohorts. It's not like they're doing nothing but optimizing their usage, but it is, you know, kind of puts and takes off the overall revenue number. You know, that said, like, similar story to what you would have heard out of these other companies, yeah, there's a few things you can do to kind of optimize your pattern. You can't do it twice, right? So as companies kind of get into the shape that they wanna be in, in terms of their usage, you know, that's a solid foundation and probably stronger TCO to build off of. And so, you know, hopefully, that provides a little bit of color. So on that note, thank you so much for being here once again. Yeah, my pleasure. We hope to have you back- Thanks, everyone. 2025? That's- Mark the date. Yeah, all right. Wow. Thank you, everybody, for your- Take it easy. ... and your participation as well.