Good morning. Thank you all for joining us. My name is Brent Bracelin, Co-head of Tech Research here at Piper Sandler. Next session here is a fireside chat. We have Mike Berry, we have Ben Cefalo. Thank you guys both for joining us and welcome to Nashville. Thank you. Thanks for having us. Absolutely. Listen, it's hard to not start a conversation here around AI, both the AI existential threat to software as well as a potential exponential opportunity. In our view, we think AI is just software. We think SaaS is just software. We think the death of SaaS is a little overblown. Ultimately, there is something different about AI, and that's the pace of change. As we think about it, the pace of change in AI is happening very, very fast. As you think about this, what we call race to relevancy, how do you think Mongo is changing? How do you think Mongo is responding to the needs of a lot of these application companies that are trying to race to change and keep up with the pace of AI? Yeah, thanks for having us. I know we didn't read the Safe Harbor, but we'll always say go look at it on the website just to make sure. Yeah. Again, thanks for having us, and Ben's joining as well. I'll ask him to weigh in. We're super excited, obviously, about the AI wave. I'll let you talk to everybody else about whether they think it's a risk to their business. We think it's really nothing but a tailwind to ours. As we look at Mongo being prepared, we feel really good about the product, especially helping our customers as they want to move down the path on AI, be it internal or external. We've talked a lot about, hey, it's still pretty slow in terms of the external facing because of all the risk of hallucination and stuff. We'll talk about embeddings and what we're doing there. From an internal perspective, we feel like our product really helps them. It's a true document model with JSON support, which is when you talk about all the unstructured data and everything that you need to do to modernize your infrastructure and take that as a product, we feel like we're very well prepared there. The enhancements we've made in our platform around vector search and other areas are super important. Our Voyage AI acquisition, which brings really the best, we think, the best embedding model, which is the link from your private data to the LLMs, is super important as well. We feel like we're very well prepared. It has not been a big driver of our growth yet, but that doesn't mean we don't think that it will be in the future. We see a lot of our customers, especially larger ones, starting to, I would say, play with this. Most of that focus has been on customer support, code generation, internal vertical apps versus someone like yourself actually offering something to your customers where you have the risk of hallucinations and other things. We think once that gets solved, then that will really pick up. Whether it's a risk to all the other software folks, we'll let you have that conversation with them. I think Dave mentioned model embeddings quite a bit on the last transcript, and he's very passionate about a differentiation there. Ben, for you, maybe just take a step back. When you talk about model embeddings, what is it? Why is it so important in this AI-native app space? A little background on Voyage. Yeah. The importance with embeddings is what really connects your proprietary data, your operational data to the LLM. Based on how good your model or how good the embeddings are, the accuracy of the results that are going to be returned is determined. Voyage, and one of the reasons why we want to acquire them, is that they have some of the top-rated models for doing the embeddings. When you connect that together with the operational data that's already stored in MongoDB, connected then to the platform Mike was talking about with search and vector, we're now giving that developer the easiest way possible to interact with that data, generate the embeddings, and serve up an AI application all within the same API layer and the same platform that we have inside of Atlas. What's the unit economics of a Voyage? Obviously, we all know MongoDB's unit economics, Atlas, obviously a consumption-based model. When you layer in a Voyage AI, what does the unit economics model look like for embeddings? Yeah. We did the acquisition mostly because of a product perspective. We did talk about there's about 300 discrete customers. The revenue is pretty small, though. The monetization around Voyage will be three areas. One is we do offer today the—you can buy Voyage on the website in a serverless API, and we do that today. It's usage. If it's text or if it's other data, it's going to be driven off of usage. If it's images, it will be based off that. That's how that is priced. The second piece is that we will offer it through the marketplaces, and we've started to do that as well. Those two are already there. The third piece is we will integrate it with Atlas. We're still working on the product positioning and how that plays, but we do think, obviously, it's going to drive a lot of data within Atlas and compute. It'll all be usage, and that will be the monetization strategy for Voyage. Walk me through the type of data that, as you get this integrated into Atlas, it would add that you typically wouldn't have inside of a Mongo. Is this unstructured data that you're adding? Walk me through the types of data that expand the opportunity once this is fully embedded in Atlas. Yeah. What Voyage is doing is actually generating the vectors in those mathematical equations that we store in the vector side of Atlas. We already offer vector search alongside our tech search within the Atlas platform. The connection to Voyage into Atlas means that our customers do not have to go out to another model provider to generate those vectors, that it can all happen inside the same platform. So we begin connecting the operational data of their applications that are already in Atlas, coupled with the vector search, coupled now with you don't have to go anywhere else to generate your embeddings. That whole flow flows into the same platform that we already offer. A little bit of a technical discussion around AI and embeddings, and I think it's super important because it seems to be popular. It's starting out with a small number of customers, 300 customers. Are these native AI customers? Maybe walk through the applicability of a Voyage AI. Is this going to be something that might appeal to 10% of your customers? Is this something that you think ultimately all customers will have embeddings in their MongoDB Atlas deployment? Today, it's a mix of smaller customers, but there are some very large customers. Again, they being a large entity, enterprise, but still a small customer to us. We do expect it to really resonate across anyone that wants to run an LLM and use their private data. You're going to need a top-quality embeddings model. I don't think if you're going to have AI, you need it. It's not going to be an option. We think it's going to drive usage across all of that base. We'll monetize it in multiple ways. Atlas will be the first view. We're still working on, hey, how about from a self-managed perspective as well? Do you want to add to that? Yeah. Because Atlas and MongoDB serve a very large swath of customer base, we have a lot of their applications already. It's really going to be about, on an application-per-application specificity, whether it's going to also be new applications that we don't quite have yet too. I think it's applicable no matter where they are. It's more going to be about the customer, where they are on their AI journey, whether it's a customer service app or a financial transactions app, which I think will happen later. It's going to serve, I think, the entire customer base. We started out with the race to relevancy. You're clearly seeing a pace of change happen fast. You went out, found a really unique asset here in Voyage. Can you do more? Are there more interesting tech tuck-ins out there? Can you go faster? You've been in the role here three months? 100 days. Not saying you're not going fast, but as you just think about the opportunity, what's your appetite to do more and push the team and go faster here? We very much want to go faster, but we also want to make sure that we're mindful of we're driving the car down the road, and we want to stay down the road. We want to go faster in terms of internal development. Just a commercial for everybody, we will do our investor day next Wednesday. I think that the physical space, you probably can't get in unless you want to help serve the food, but it will be online. We'll send out a press release to have all that information. We're going to talk a lot about this next week. All the work that's going on from an internal perspective around adding capabilities and functionality for AI use cases is going on. Other acquisitions maybe. We're very mindful, Brent, of, hey, we feel great about the organic growth path. We don't need to go buy anybody to increase that. If there is a build versus buy, especially around, as Dave likes to call it, the scaffolding around agents and other things, there are some interesting areas. You should expect that to be a build versus buy, something similar to Voyage, where we buy a team and capabilities that then we can embed in our solutions going forward. One of the things that kind of stood out to me that we've been talking to investors about is PostgreSQL, open source, right? Open source alternatives. That became kind of a hot button issue for you guys as we thought about some acquisitions by competitors, right? You had a Databricks buy a PostgreSQL company, a Snowflake buy a PostgreSQL company. There's a narrative out there that why would you use Mongo if you can use open source? You talked about PostgreSQL migrations. Maybe double click into the PostgreSQL migration opportunity. What are you seeing these large enterprises or large software companies run into, some limitations around PostgreSQL? Maybe talk about that migration off PostgreSQL to Mongo. Are you seeing a little bit more of those in frequency? I'll start, and I'll send to pick up as well. We have talked, and we've done it on, I wasn't there, but the company has talked about it on earlier earnings calls in terms of that for a simple data model, Postgres or another, call it SQL solution, may work. Once you start to get into any type of more sophisticated data model where performance matters, that's where we start to see that breaking. We had two that we talked about on the last earnings call. One was a bank that had used PostgreSQL, but it got to be such that they couldn't run their internal systems where they weren't able to sell loans and do other things, which is a problem, when you actually have a database that's limiting your ability to sell to your customers. They transitioned their content management to MongoDB, and their performance went up, and everything worked a lot better. That was really related to it. It was very brittle, and it just simply couldn't scale. The other one we talked about was an EV, a large EV company, where they had actually done a bake-off in terms of Postgres versus Mongo. This was their voice recognition in their cars. They realized there's no way that they can serve up as much as they need, all the data that they need to generate. It was not going to run because it wasn't performant. They went with MongoDB. There have been a lot of those examples. We think that a lot of the, call it the Postgres, if you want to use the word momentum, is really related to SQL transitions, not new applications. Versus our ability, we think a lot this came up when the company's growth came down a little bit. There's obviously, hey, it's a competitive issue. We felt like a lot of that was our internal issues that we're solving, not a competitive issue. Let me hand it to Ben to talk about why people use it versus us. Yeah. I think, first of all, you can get back to universities, computer science classes. Everyone learns relational. It's very, it's in their nature. Even on new applications, it's like, I'm just going to throw this on Postgres. It's more about throwing it on SQL, right? Postgres is talked about in a broad way, but it's actually AlloyDB versus Aurora versus Cosmos DB and all the other likes of it. They're actually a little different. You can't just move from Oracle to Aurora or Oracle to AlloyDB. It also takes some migration. I think customers go through the same thought process of like, if I'm going to modernize pieces of this, should we look at it from a much broader perspective? Secondarily, when you look at the AI use cases and why Postgres is very rapidly trying to make JSON work by shoving it into a cell of a relational database, they're running into problems with how big that cell actually can be. I think it's about2 KB versus our 16 limit possibilities that we have inside of Mongo. To what Mike was saying about the data structure, with AI, especially voice recognition, as the example Mike was using, we don't know what the data that we're going to be recording is going to be. How can you then model that into something that is very strict with a schema perspective? We're seeing all of these modernizations happen or all of these questions start being asked. That's why we feel that MongoDB is the best place for AI, because we can handle all that structured, semi-structured, or completely unstructured data all in the same database and be flexible with the application of what it's going to bring. We don't know that. Customers don't even know that about their own application based on how their users are going to use them. Outside looking in, we've seen a re-acceleration at Snowflake, a re-acceleration at Mongo in the growth rate. We've seen actually Oracle missed yesterday, but they had a pretty sharp acceleration in cloud backlog. Walk me through, Mike, you did mention a lot of the slowdown you thought was internal. How much of this re-acceleration is new AI things happening helping you versus some of the things that you're doing internally to help drive a re-acceleration in the core business? I know you've only been there 100 days. Yeah. Great job in the first quarter out of the gate. Walk us through the opportunity that you have to control things and then the external opportunity for things to get better. Yeah, great. Thanks for the question. Hey, I'm lucky enough to be there. It's a team effort, truly, the whole 5,500 of us. Let's talk first about AI. We said it is not a big driver of growth today. We'll talk about this again next week. We think it will absolutely be a growth driver in the future because we see what our customers are doing. As you look at the re-acceleration, this was much more of our core business blocking and tackling. A couple of things. One is, especially from a go-to-market perspective, we have changed, and I'm going to use the word tweak, not overhaul because it wasn't an overhaul. We have changed the go-to-market to focus more on the enterprises. We deal with a lot of the Fortune 500. Our share in that is very small. The ability to do the opportunity to do cross-sell and upsell in that market, and that's where the larger workloads sit, is significant. We move some of those resources up market. We also tweaked the comp plans to say, hey, it's less about grabbing any workload because it was much more of the portfolio theory. The more workloads, at some point, they're all going to grow, but workloads grow differently. What we really focused them on is the comp plans are more focused on, hey, go drive ARR. Go focus on the bigger workloads. That's what we all want versus just grabbing everything. We're making that transition. We did a little bit of that in fiscal 2026. We'll do more of that in fiscal 2027. The other big part of that is the corollary to that is the go-to-market product-led growth that we talk a lot about, that goes way back to my SolarWinds days, right? The inside sales model, the touchless model is really working well. That's been a work in process. We'll actually have May Petrie, who's our Chief Marketing Officer, next week, talk about this because I think it's an unknown asset within Mongo, which is we're able to move up market because we're able to then scale that self-serve model. Both of those, I think, are working much better. I think that's what you've seen in the results. We think a lot of it has been our execution. We're not going to do the victory lap. It is a process. We're getting better every day, but we feel good about that process. I like to talk a lot about the art of the possible. Before I do that, packed room here. Any questions from the audience here before I shift gears to kind of art of the possible? Perfect. It's too early still. They're still eating breakfast. Let's talk about Atlas. This is a business that in eight years has scaled from less than $10 million- a $1.7 billion ARR business. As you think about the next eight years, what's possible as this business scales to $2 billion, $3 billion, $4 billion, what's the, I know it's a slightly lower gross margin, but what's the up margin potential of this Atlas business at scale? Yeah. Let's talk about Atlas and then how it translates to margins. We won't give specific numbers. As you look at the business, and this even when I joined, I think it's a huge market. The great thing about Atlas, and EA is wonderful, by the way, I love EA because it generates a bunch of profit. It's big, huge customers committing millions of dollars to Mongo, which is awesome. Atlas is the growth engine, though. The market is huge. We have a very small percentage of it. However you want to cut that $100 billion between OLAP and OLTP, that's a huge market for us to go get. We feel really good about that. As you look at the secular growth drivers, it's not only our product and our ability to grow within, I'll call it the organic play now, but you add AI. You add what we've talked about with modernizing applications. That's all net new opportunity for us. We feel that there is a huge runway for Atlas. Assuming that the gross margins, call it, stay within the mid-70%, something like that, that is a ton of profit that comes to the bottom line. Our focus internally is, and this is different from other places that I've been, we have so much money to invest. It's not about cuts. Yes, we'll do small productivity stuff. This is about investing smarter. The great part about Mongo is the foundation is already built. We have everything we need from a go-to-market. We're in every geo in the world. We have two-tier distributions. We have sales reps. We have engineers. There's incremental spend we need to do, but there's no big step function that says go invest in that, which is great because now we can invest incrementally. Go get this to drive ROI. That's really the focus. That's why I feel good about the ability to continue to drive margin expansion. The number one driver will be revenue growth, but we will grow operating expenses lower than revenue growth and still be able to invest in developer awareness, marketing, the variable sales reps that we need, and importantly, the great engineers to drive the product. Sounds like there's a lot of investments already made in the core. Those incremental margins could be pretty meaningful if you continue to have scale Atlas, even at a 75% gross margin. Yeah, the math works very easily. The great part is that, again, outside of some small things that we need to do to drive growth, and we'll do that, those will have returns, Brent, versus, hey, you need to build it and then it'll come later. That near-term view is pretty clear. We talk a lot about software companies, risk to software companies because of AI, opportunities because of AI. Let's talk about internal. One of the biggest cost components for a software company is labor. One of the big benefits of AI is labor savings, productivity savings. What are some of the tools internally that you're using AI, leaning in on AI to drive higher productivity? What tools are working? Maybe what tools have you tried that aren't working? Yeah. This is, you go back to the margin expansion. This is a huge area of opportunity. As much as we like to espouse AI with our customers, we could do better here. We're doing a lot more around the governance and the tooling. This is a little bit go slow to go fast, which is, I think, why you see a lot of enterprises. You have to do this the right way. For us, it's focused around really not from the end customer, but internally, things like, hey, Cogent is a big area. Customer support, big area of focus. Vertical applications like Harvey, where you can really do things around legal, I think, are. As a CFO, I tell you, there's no killer use case yet for AI, but there's a lot of good things around it. My big focus around there is around ML and better forecasting. You take the consumption business, it's almost a $2 billion business. We know so much about historically how our customers have behaved. What we don't know as well is take those external shocks and build that into the forecasting. That's where things like AI and ML can add value. Those are areas that we'll focus on. As you look at the productivity layer for us, Brent, it's a huge possibility today. It's not a big driver of our cost savings, but it will be in the future. Future facts. I love this David Diamond quote. He thinks a lot of leaders today focus too much on what's happened and not enough on what will happen. As you think about the art of the possible that people might be talking about a year from now, they're not talking about today, what would be some of those things? It could be a product. It could be a trend. Let's put that kind of future cap on. What do you think a year from now people are going to be talking about that they're not talking about today? This is Mike Berry, the person, not the CFO. I think AI is going to continue to dominate for the next couple of years. I think one of the interesting things around AI is, I live in the wonderful state of Texas, for instance. All of this is interesting. We don't have the power to do it, I don't think. At least personally, I view that. I hate waking up in the morning in the middle of the summer when it's beautiful and it's hot. It's like, oh, is the grid going to hold up? Yet everybody's building data centers in Texas. At some point, that's got to get solved. It will be interesting to see how that happens because all the stuff around AI cannot happen if you don't have the power to do it. I think energy and power is going to be super interesting. Just around talent, that's the other issue. This isn't about paying people $100 billion or whatever it is. Do we have the talent to do it? I think that's going to dominate for a while. Why I love being in tech, and I know Ben's been in it as long as I have, well, not as long, but in his career, is that's a great part about tech, which is we're going to wake up and it's something new every day. Do you want to add to that? Yeah. I think from the product management side, I'm not looking for fewer product managers, but I'm looking for a slightly different skill set. Can you use AI in how you think about product management? Do you do your own mockups now, or do you build a small little app that represents your product description of what you might want to go build? Being able to augment how they typically would deliver requirements or anything else to help engineering, I'm looking for skills like that myself. I think that's going to continually adapt. That person is going to have a different outlook on how they go out and look at product or how they go out and look at engineering. I 100% agree on the power aspect of it compared to some of what other countries are doing from their grids versus what we need to do internally in our grid. I think power and energy are going to be a big one. We're out of time. Thank you so much for insights. Thank you. Super helpful discussion. Thank you. Thank you, and come on Wednesday. Thank you.