Welcome, everybody. Thanks for joining us. I'm Jason Ader from William Blair. I'm pleased to introduce Mike Berry, CFO of MongoDB, and Benjamin Cefalo, Chief Product Officer. Before we begin, I'm required to inform you that a complete list of research disclosures or potential conflicts of interest is available on our website at williamblair.com. We're going to go through some slides, and then we'll have some time for Q&A. Take it away, Mike.
Great. Thank you, Jason. Good morning, everybody. Good afternoon to some of you on the call. My name's Mike Berry, and this is Benjamin Cefalo. We're going to ham and egg the presentation. This goes way back. We don't typically do these anymore, give us a break. We just normally do the fireside chat. I'll talk a little bit about MongoDB. We'll try to go through these pretty quickly, we'll jump into Q&A. Safe harbor, keep it up there long enough. Okay, MongoDB is an infrastructure software company. We market databases to companies across the world. We have over 65,000 customers. This year, we guided to just short of $3 billion in revenue. You'll hear us talk about two major product lines. Atlas, which is our cloud-based service offered in all three of the public clouds, you'll hear EA, Enterprise Advanced.
That's the on-prem version. Atlas is about 75%. EA is about the rest. We play in a market that's huge. It's $100 billion, both data warehouse as well as online transaction processing, which is where we play as well, and it's growing very quickly. It is a huge market. While we have a lot of customers, we have a huge TAM to go after across all of those customers.
[Inaudible]
Be careful.
Yeah.
Don't want to fall.
Right. The document model is the foundation of our growth, right? There's a lot of talk lately about AI workloads, unstructured data, and what all the new data that's being generated is largely unstructured. We like to think of JSON as of the lingua franca of the AI generation, and the main reason why is all of this data is being modeled and best suited for a JSON document model, which we're the only database that ever built this from the ground up from the very beginning. Every question I have is always about Postgres. What do you think about Postgres? Does it scare you? Does it keep you up at night? Well, I just wanted to walk through a couple different things here, right?
First of all, we like to consider ourselves a flexible data model, unlike a relational data model that has a very rigid schema. We have a very flexible data model, right? Number two, we have a lot of native querying capabilities around security and governance. That's even more important in the AI world. One of the reasons why enterprises are very slow to adopt AI is around security and governance, especially in regulated industries. The third is about being able to run anywhere. I think we misuse this term sometimes. What run anywhere, I think, means to some people is, okay, cool, you can run it anywhere you want to, on-premise data centers, inside of different clouds, et cetera. What we actually mean here is about multi-cloud.
Something that no one else can provide, especially the hyperscalers, is the ability to extend a singular data cluster across multiple clouds or multiple regions at the exact same time, serving simultaneous workloads. That's extremely powerful, depending on where you are in the world, different regulatory regimes, data sovereignty, data residency requirements. We can spread that data and that availability of that data across multiple geographies. Lastly, one of the things where Postgres falls down a lot is on the cost of scaling. One thing about using the hack of JSONB is that it requires you to have to scale Postgres a lot faster than you normally would have to, and that gets very expensive. Natively, MongoDB offers the ability to horizontally scale at a much cheaper cost, as well as vertically scale with how we separate out the data into different shards.
That's all built into the database. The other thing to think about is what we're doing at the Atlas level when it comes to our data platform. A lot of other providers and competitors are using multiple different technologies and making either the customer stitch them all together or understand and manage each one of those technologies, or the company that's hosting this is stitching them all together and providing different interfaces to them. Ever since the very beginning of MongoDB, we wanted to focus on the developer and making the ease of the development cycle as simplistic as possible. We keep adding capabilities to our platform, but we're keeping the same developer experience.
When we added Atlas Search, when we added Vector, we're not making the developer or the application owner query all of these different things separately and bring them all back to complete their use case. It's all a singular query language, and we do all the stuff behind the scenes to keep the data in sync. There's no more ETLs. You're not doubling up on your data size. We do it in a very efficient way. At the end of the day, it's all about JSON being built from the ground up as far as the document model is concerned. Handing it back to Mike.
Thank you, Ben. Atlas, which again is about 75% of the business, has grown very quickly. We talked about it on the earnings call. The last four quarters, it has grown by 29%+ consistently from an Atlas growth perspective. Again, it started out, it was $800 million four or five years ago. It's now over $2 billion. Keep in mind, this is a consumption business, and it's our customers that deploy that data. The good part is that consumption in a given quarter, you typically won't see that growth until the next quarter. It's hard to move the needle in a quarter, which is nice. People say, "Hey, if you see consumption in a quarter, when will you see it?" Typically, in the following quarters. Again, this is across all three of the hyperscalers. Q1, we reported strong results.
There's the Atlas growth. That's 75% of the business. EA and other, that is the Enterprise Advanced, that's the on-prem version. That's both license and support. We've talked about this, if it's 20+% of the business, almost 70% of that revenue stream is support. It's what's sitting in deferred revenue coming off the balance sheet. There is a good bit of ratable recognition there. The nuance there is that we have to recognize the license as we deploy it. If somebody does a multi-year deal, we have to take all three years up front because we've already deployed the software and they own it. That causes some volatility within that number. Net new customer adds, we have a great enterprise motion, but what really Mongo started at the beginning was what we call self-service. This is product led.
We've added, call it 2,500 up to 2,700 net new customers. Those come almost entirely from self-serve, where they'll use a credit card, they'll start using our product. We talked about it last September at our Investor Day, that a good number of the customers over $100,000 in ARR started out as self-serve. What they do is they get bigger, they get bigger, they call us, because at that point, they're using self-serve, they don't get support, and then they want it. Once it becomes an enterprise workload, then that's when the sales team steps in. Operating margins, we've taken up from the low teens now up to 18% this quarter, and we've guided the full year to about 20%. I'll talk about guidance in a second. Cash conversion.
This has been a concern I know with investors and me as the CFO is, "Hey, we're generating operating profit, but where's the operating cash flow?" Last year, actually cash conversion was over 100%. We expect that to be between 80% and 100%. If you're going to earn profits, you need to bring the cash with it as well. Kudos to the whole team for really driving that. We raised guidance across the board after Q1, and I'll just hit the big numbers. Total revenue now $2.9-$2.96. Yeah, the prior guide was $2.9 at the high end, now it's $2.96. That we took the Q1 beat, rolled it, we raised Q2. The back half for us, the entire guidance range raise was in Atlas. We think EA will do better in Q2, but we didn't bump EA for the second half.
That's while we're trying to be prudent in the second half because again, folks, it is a consumption business. We do feel very good and confident about the Atlas business. We also took operating margins up. It has a very good operating model. When you bring in new revenue, it flows through operating or gross margins at about 76%, then it cascades down. We're doing a much better job of driving efficiency in the operating margins, then that also flows through to EPS. We rolled the beat, we increased across the board. These are the long-term targets that we talked about last September. We have announced we'll do an Investor Day again in September, at the end of September in conjunction with our .local in N.Y. We had three pieces of that.
Total revenue growth in the high teens, Atlas growth above 20%, operating margin of 20%+. I want to underline the plus, it's not a cap, folks. As we continue to grow, we expect to grow operating margins as well. Free cash flow conversion greater than 80%. The guide we gave for fiscal 2027 hits both revenue growth at 20% and operating margins at 20%. We are guiding for a rule of 40 this year ahead of our targets. Before we hand it over to Jason to ask Q&A. I told you we'd finish in more than 15 minutes. Hey, four things coming out of the earnings call. I just want to reiterate. We just talked about them. Hey, we had a strong Q1, fourth consecutive quarter of Atlas growth above 20%. Another strong quarter of momentum in EA.
The importance of EA is that some of our largest customers, regulated industries, financial institutions, that still want to run their database and workloads on-prem. With AI, that's starting to become a bigger thing. Hybrid cloud is a real thing in our mind. We issued strong Q2 guidance, and now we're talking about a rule of 40 for the full year ahead of our targets. We are starting to see contributions from AI, but it's still early days, and I'm sure we'll talk about that, and we are still confident and remain very confident in our ability to drive durable growth and be a rule of 40 company. Those are four things I just want you to take away from the earnings call. With that, Jason, we'll hand it to you.
All right. Thanks, Mike. Maybe start out, I'm actually going to ask every single one of my companies that's at the event here. Can you frame the case for investors for why MongoDB is a winner in AI?
Yep. A couple of reasons, in my opinion. The first thing is we have 65,000 customers globally, and we're the system of record for a lot of those critical applications that are already in use. What's happening and what we're seeing in these enterprises is that they're starting to prototype and develop different agentic workloads, whether it's agents, whether it's simple chatbots, different experiences for both internal and external. Where's the data coming from? It's coming from the data that already is stored inside of MongoDB. Number two, as more agents are deployed, and I'm sure there's been a lot of research about this, is agents need memory to be viable, right? We just announced a partnership a couple of weeks ago with LangChain, with their memory frameworks, et cetera.
If an agent and a chatbot and an MCP server and everything is already communicating with itself and with other agents via JSON, the memory is going to be stored in JSON. What better way to store that memory is inside of MongoDB. Then third, on the longer tail of the funnel with the, I would say, new AI native companies
As I said earlier, the VHS Betamax war of the best way to model data in this world is over. It's all about JSON. Even the workloads that we see that end up going to Postgres or Supabase or Neon, you pick your poison of flavor, they're not being modeled in relational schemas. They're being modeled in JSON. Right? We eventually will get that workload if it doesn't land on us from the beginning, but we're obviously focused very hard on making sure we're there from the very beginning. The data is being modeled in JSON.
Okay, great. I want to just poll the audience here because I know there's probably some people that are not that technical, but how many people know what JSON is? I'm not talking about me, Jason. JSON, the technology. How many people know what JSON the technology is? Just raise your hand if you know what JSON is. Whoa, okay.
More people than I thought.
Maybe just spend a couple of minutes talking about what JSON is, why it's differentiated.
Sure
created your product around that, and then why it's so tied so closely with AI.
Yeah, absolutely. Everyone's used an Excel spreadsheet before, right? Good. A lot of heads. That is the best analogy for what a SQL relational Postgres schema looks like. It's a bunch of rows and it's a bunch of columns. You only can fit so much data into a cell, and think of every cell as like a record. Well, that's better for certain types of workloads that I would say is very structured data models that don't change. JSON is a different way to model data that's not trying to fit all the data into a singular cell and allows you to separate out the data into a document. The flexibility is important because we don't have to abide by a schema is because data changes all the time.
I'll give you a perfect example of this, is if you were building a, let's just call it a chatbot, you're talking to one of your cable provider about getting higher internet speed or adding HBO. An application developer could never model what the consumer would be typing into that chatbot, right? It needs to be able to handle it. That flexibility, especially in the AI world, is hypercritical. You need a data model that is able to handle that uncertainty when it comes to the data input or the data output that it might be generating. Being able to model everything in a document that has organization but doesn't have a very strict structure is what makes it flexible.
That takes a special, this is more technical, but how we actually store it in bits on a hard drive or piece of memory, and we built the database from the ground up, being able to do that from the very beginning.
The Postgres and other relational database models have the ability to support JSON, but your point is it's not their kind of grounding, it's like an add-on for them?
Correct. It's what's called a plugin on top of Postgres. It's called JSONB, and so it allows you to model data in that JSON format that I was talking about, but it needs to store it somewhere. It's on a different storage engine. It actually just puts it all into a singular cell. Well, that cell is only so big, 2 KBs, which is not a ton of data. MongoDB document sizes are 16 MBs, and I know that doesn't sound that big as how big data actually is, but that's a lot of text. We have a much bigger document size that we can support. What happens in Postgres is if you have a bigger document size, it has to spill into a second cell, and you sacrifice performance on query writes and reads when you're dealing with multiple cells for the same document.
Got you. You hit some kind of performance ceiling.
Very quickly.
Yeah.
As the app scales, that's when the performance problem would happen.
What does CJ mean when he says you guys are kind of downstream of where Snowflake and Databricks play?
A couple different things with that. I think number one, OLAP has its place. I'm not saying it doesn't. That data by nature of the use case of OLAP is old. Is that good for a data analyst or internal use cases where there's nothing critical about that? Sure. Makes sense. We firmly believe, and actually Databricks and Snowflake have both validated our belief by purchasing two OLTP companies, is that OLTP is actually the higher ground for AI because to do any type of real-world data transactions, whether it's some bank is recommending a stock trade, or you want to go update your insurance provider, or you have to file a claim, that's all OLTP. You can't do that against historical archival data.
MongoDB being an OLTP database also from the very beginning, and also the fact that all of those customers are also our customers as well. Why we say it's downstream and why we think they've seen the uptick is it's been a lot of internal use cases from playing around data analysts, crunching data, doing research and everything like that. The keys to the kingdom that are actually running all of their mission-critical workloads that are generating all of the revenue for these companies is actually MongoDB.
That means that you will see more the benefits later than they've seen them so far. Is that the right way to put it?
It's not a question of if, it's a question of when.
All right. Let's just talk about the opportunity across the three different AI customer cohorts that you guys have defined, the frontier labs, the AI natives, and enterprises.
Yeah. Let's break that down. We talked about this on the last call. You have the frontier labs, and quite frankly, we're thrilled to have them as customers. They've asked us to not talk about how we're used with them, so we won't. That's one piece of it. The second piece is AI natives, and this is where companies are making their own product, that is selling an AI product that is run on Mongo, powered by MongoDB. It's a bunch of smaller companies, but nobody has really hit exit velocity. They will at some point, and that's a great use case because, and this is as Ben was talking about it, sometimes it's a lot of unstructured data.
If that database doesn't scale with the product, as they scale, they have a problem, which is why they may start out on Postgres and then they come to MongoDB. The third piece, which is where all the money is in enterprises. As we've talked about, we started to see some activity. There's a lot of POCs, there's a lot of betas. We know customers are big banks, insurance companies, consumer products, retail, starting to build agents, but they're not deploying them at scale externally because of security, governance, compliance. Nobody wants to be on The Wall Street Journal because they had an agent do stock trades that went the wrong way. That's going to take time, and that's the third piece of it.
We feel very good about what we have in that enterprise market, and we know that customers are deploying agents, but they have not deployed them at scale, nor have they come out of beta or POC. Those are the three layers we've talked about, Jason, for AI.
Got you. I think CJ has said, for at least enterprise, a kind of 12- to 18-month timeframe is a realistic kind of timing of when it starts to become more meaningful for you.
We'd like to see it faster. Look, we can try to help them, and I know we were going to talk about this, and we have customer success folks in their sales teams, but at the end of the day, they have to move at their own speed. Even, for instance, MongoDB, we have a lot of internal use cases for us to then deploy to our customers. There's a lot of things we have to go through. We hope it moves faster. We're starting to see some pickup, which is why we talked about it. Again, like Ben said, for us, it's not a matter of if, it is when.
Okay, great. For Ben, just want to ask you a couple of things on the kind of startup side. How do you guys get in the door with the startups right away versus having them fail on Postgres and then kind of come to you, number one? Number two, is related to that, how do you make the product more agent-friendly?
Good questions, actually, they're really related. If we rewind the clock five years ago, if you're an application developer, the first thing you did was saying, "Okay, what's my stack going to be?" Right. Whether you pick Node.js or some other, you pick your technology stack, you pick your hyperscaler, you pick where you're going to run it. You knew what the security model was. You wire up the network and you'd be like, "Okay, great. Now I'm going to go start developing my application." Right. Today, what happens? You go to whatever pick your poison code platform and say, "I have an idea for an app." The LLM writes back and says, "Cool.
How fast do you want to deploy?" They're like, "I don't care." They just start asking you questions about what you want the app to do, and it lands on a piece of technology and something. Why is that? The way the LLMs get trained is on publicly available internet data, and SQL's been around for 30+ years. Postgres is 20 years older than us, right? We have a lot of content catch-up to just plainly do to just get the training in the right place to where the recommendations happen more. What's actually happening is fascinating, is all that data is actually being modeled in, as I said earlier, JSON. They're not trying to model the modern application in relational. It's actually being modeled properly, because the LLMs themselves are utilizing JSON for how it stores its own training data.
It picks the better data model, picks the wrong technology. We have a lot of focus on building up that awareness, content, varying other things. There's all this new segment of AEO, making sure to build up agent awareness. There's a lot of different focus areas there to help with the awareness. Secondly, more specific on the agent friendliness, we do a lot today with MCP servers. The way our API is written, we'll continue to obviously invest in that area. Once you have decided on MongoDB, we're very agent friendly. It's about that first step of having the agent select us without the need to prompt. If you're creative with the prompt or you have a good idea that you want to go with Mongo, it's not very hard to get the LLM to select Mongo.
If you're being generic and you're at 80,000 ft.
Yeah
it just defaults to what it does.
For a lot of people, it's an afterthought right now.
Yeah. It's an afterthought.
How do you get it to be not an afterthought?
Exactly. We know that to be the case, because as soon as it becomes a real thought, whether it's application starts generating revenue, they have a security concern, they run into a performance problem, they all eventually end up on Atlas. Which is great, and we're happy about that, but I want to save them the pain upfront.
Got you. Is it fair to say that, I think CJ kind of alluded to this on the call, that there's sort of stay tuned on the product innovation side this year, that you guys are going to be releasing a bunch of new technology.
We're always innovating.
more easy.
We have a lot of ideas.
Okay. All right. A couple more things I want to hit before we wrap up. The breakout's going to be upstairs in Maher, or Maher. I don't know how to say that. The first question is just on the new sales leadership-
You had two guys that were there a long time.
Really established and really respected. How should investors think about the sales leadership change?
Sure. Cedric and Cap, as we call them, Paul, were there for many years, I think eight or nine years, both of them. They did a great job building Mongo into what we are today. They both raised their hand and said, "Hey, it's either time for me to go do something else or bring in new leadership." Both of those two exited Mongo at the end of last quarter. What was that? Oh, sorry. We broke the team up into call it pre-sales and post-sales. We brought in a new leader, Ryan, who's going to run. He's a CRO, and he is running the sales team. He came from Confluent. He's all of five weeks into the job, folks, so he's still trying to figure out where the restrooms are, to use that analogy.
Erica came in as the Chief Customer Officer, and she has not only technical support, PS, as well as the customer success team, and also partners, which is what she did very well at ServiceNow. Very importantly, before we made the announcement, we were clear going into the year, folks, all the territories, quotas, comp plans were all done for fiscal 2027. There's going to be no changes to that. Ryan knows very well the last thing you want to do is come in and change those. We don't expect any changes for fiscal 2027 as it relates to that, where all the execution is set. The great part about him is he understands the consumption business. He's done it at scale. He also is very focused on enterprise, so that's great.
Erica not only will continue to focus on the customer success, but we've also said that, hey, we can do better in the partner community in terms of SIs or other folks that we sell through and with. That'll be a big focus for her. We don't expect any issues, and we haven't baked any into the guidance because things are going well. The next level down, certainly there's always changes at that point, but we feel very good about the leadership.
Okay, great. When we think about the FY 2027 guidance framework, it looks like it bakes in a fair amount of conservatism, and it's sort of your MO, as I've worked with you for many years, not just at Mongo, also at NetApp. Can you walk us through kind of the key assumptions with the FY 2027 guidance, and if you end up doing much better, where would the upside come from?
Sure. Let's back up for a second. Understanding that the company historically has not given specific Atlas guidance. Two quarters ago, we started to give that view. When I started a year ago, that was the number one ask for investors. Give us a little view of what you think Atlas is going to grow. When we set the guidance for fiscal 2027, we talked about Q1. We thought Atlas would grow, call it around 26%. For the full year, total Atlas growth was 21%-23%, and then EA was low single digits. That was the framework that we set. Again, we felt good about the business. It is a consumption business. The thing that we have less control over is what happens in the economy. The great part about consumption is you can spool up and you can spool down.
We always want to bake that into guidance. After Q1, we grew Atlas by 29.4%. We bumped Q2 up to 26%, we raised the second half as well. In our models, the entire second-half raise was in Atlas, which goes to your other question. We feel good about the business. If we're going to beat those numbers, it will almost assuredly come in Atlas. Is there room in EA? Sure, because we don't bet on the duration. We may get a multi-year deal, if it comes in at 3 years, it boosts the licensed revenue in that quarter. Folks, that's tough to forecast because that's not our decision, that's our customers. Quite candidly, they don't even know whether it's going to be single or multi-year because their budget is the driver there. We took up operating margins across the board as well.
We feel good about setting ourselves up for success. Very importantly on Atlas, folks, we talked about, hey, when you look at the range, in Q4, we guided Atlas to grow 27%, and it grew 29.2%. Consumption was largely in line with what we expected. In Q1, 26%, and we grew 29.4%. Consumption came in better than we expected. That kind of gives you the range, and we're guiding Atlas because, hey, the consumption business is new to a lot of you folks, and that's why we started to give that incremental disclosure to show, hey, the consumption business is a little bit different, and we said it now two quarters in a row. It's a big business. Less so four years ago, where you had customer cohorts or groups of customers that can move the needle. It's less so now, which is great.
Consumption in the quarter, while it's great in a quarter, you'll largely see that hit revenue in the following.
Okay. We'll have to end it there. Thanks everybody for joining, and thanks to Mike and Ben.
Thank you.