Thanks for joining us, day 2 of Citi's TMT Conference. I'm Tyler Radke, Citi's co-head of U.S. Software. Excited to have MongoDB for the second session of the day. We have CEO, CJ Desai, and the Head of Investor Relations, Jess Lubert. Thanks for making the appearance at the conference. CJ, it's great to see you. I thought we'd just kick things off. It's been almost a year since you've taken over the helm of MongoDB. Just give us some of your observations about the business, and what are your top priorities for the year ahead?
Absolutely. First, great to be here. Thank you for hosting us. We had a print last week on Q2. We felt that overall revenue acceleration and first time hitting 30% total revenue growth in a long time was just fantastic. As I have been here for a year, there are 2 or 3 things that I'm seeing very clearly. Number 1 is, I wanted to make sure once I joined the company, where we have the biggest potential, right? Where we have the biggest potential and the why behind the potential. What I saw in last 10 months or so is that number 1, in enterprises, we are definitely becoming a standard for modern databases, whether these are financial services institution, insurance companies that have a lot of unstructured data, and they want to modernize and move it to multi-cloud estate.
We also saw over last year or so that public sector demand is there. It's early, but public sector demand is there in large enterprise. As I traveled in Tyler, Europe and Asia Pacific, India was very interesting to see lot of new digital native companies, building on MongoDB like a Zomato and others. In Europe, Sovereign Cloud, us having that flexibility that you can run anywhere, so you can self-manage MongoDB, was very prevalent in the U.K., France, and few other places. So that's in the enterprise environment. The second learning I had was that many AI native companies are building on MongoDB, which was a new thing to me after I joined.
When I spoke to folks at Mercor or whether you look at ElevenLabs, and there are many others like that we shared even last week, like Harvey, and others. These are AI native companies, and once they started hitting decent scale, we were the only answer. So that was very encouraging to see, even though some of them did not start with MongoDB, they hit scale limits on Postgres or whatever first-party database from Hyperscaler they were using and then switched to MongoDB. Now MongoDB is just a foundational operational data layer that they don't even think about. So that's my second bucket. The third bucket that we shared last week is seeing early traction, and early is an important word, early traction with the labs, as in specifically Frontier Labs.
Why are they using MongoDB? What does that look like? Because they have many products. All these labs now have many products. What are specific use cases they use us for varies by lab, by lab. Again, fairly early. We do not have history with them for multi years. That was very encouraging to see that one of the smartest teams at one of these labs said they have looked at every alternative. They thought they should build something on open source and decided MongoDB was the answer. That just gave me lot of conviction that this is a great technology for AI workloads. That is how I would classify it in three buckets. That is the learning I have. Yeah.
Oh, that is great. I remember when Dave made the handoff to you a little under a year ago. Obviously, you've had an impressive background at Cloudflare and ServiceNow, kind of calling on the C-level of the Fortune 500. I think that's not something MongoDB necessarily had in its DNA, kind of that top-down selling. How have those conversations been going? How would you sort of characterize where we are in potentially some of those larger expansion opportunities? Maybe some of them were using MongoDB departmentally. Where are we at in that adoption?
Yeah. That is absolutely a focus where I wanted to leverage the relationships I had and the customer focus I had in my prior chapters of career, and calling on to the CTOs, the CIOs. Those are typically the two line of business CIO at a bank or an insurance firm is it's going very well. So one is they appreciate having that relationship to say, "Okay, Tyler Radke, my biggest thing is that some of the large enterprises who are spending, whether it's seven-digit ARR or eight digits ARR, didn't even know that they were spending that much, and that they were using MongoDB for some of their mission-critical workloads." So once you establish that relationship, and what I noticed is, while MongoDB, like you said, was workload by workload and our sales team did a fantastic job with the product because developers loved the product.
Now you come I don't want to say top-down, but at a C-suite level, the biggest learning I had was that AI decisions were being made top-down with the enterprise architect, chief AI officer, chief data officer, pick a title or somebody like that. That's when you go in and say, "Hey, do you know we have vector search integrated, and text search. We have now Voyage AI embeddings." Almost all of those conversations, I do approximately 10 to 12 conversations a week.
They're like, "Oh, we did not know this." Right? We are now gaining traction to say I'll just give an example. Large biotech company, I meet the chief data officer in South San Francisco. Next thing he's like, "Talk to my enterprise architect." Fast-forward 4 weeks later, "Let's do a POC on our AI workload to see why we need real-time operational data via MongoDB." So that has started, so that's one. The second thing we are discovering when I talk to the C-suite, CTOs, CIOs, the modernization that they need to do to get AI ready in enterprises, is a real thing, and how can we help them? So that's where also we are a little early, but I'm like, "Hey, we can help you modernize from a particular relational database that's end-of-lifing or whatever." They're like, "Oh, I didn't know that.
How fast can you do that?" My answer is that we believe that our teams can reduce it from years to months or months to weeks, your modernization effort on the workload. We have a set of tools and products for that. They're like, "Okay, let us get started there." These are the things that make me really, really encouraged about the market size that MongoDB has and where we can play.
Okay. Going back to some of your points you referenced earlier around the AI natives and AI labs, I think it's sort of an interesting observation that maybe these companies didn't necessarily start on MongoDB, right? Now you're actually seeing more traction, maybe a couple of years into the AI adoption cycle. What are the use cases that you're seeing the labs, I guess, use you for? I know some of the recent announcements, like some of the wins that you announced in the most recent quarter. There was perhaps a win to kind of distance themselves from one of their larger partners that they use on the database side. Can you just talk about the use cases and kind of why you're winning and how you see the potential in those large labs?
I would say, first, the strengths of the platform is the reason we are winning, whether it's AI native or whether it's a Frontier Lab. I'll tell you what specifically three things that always stand out to me when I speak to an engineer who makes a decision to use us for a particular workload. Document model and JSON and document model just make it super easy because that's the language of AI.
And us being JSON-native is definitely an advantage. That not being rigid relation. I worked for Oracle for eight years, and in Oracle, the DBA has a lot of power still when you want to make a change in any database, tables, queries, views, things like that. MongoDB is very agile. So the document model combined with agility of it, because your AI use cases change all the time. You are trying to move really, really fast. If you look at these labs, they are trying to ship products so fast, compared to a typical software life cycle that we have seen in the past. They are almost like at a consumer company level. So agility of MongoDB definitely stands out. Second is just scale. We can scale out. We can scale out to massive amount and not fall over because it is a scale-out architecture.
Then the third thing I would say is run anywhere. Run anywhere does help because one lab or one of the use cases, when they saw that we can run nodes across two hyperscalers, that was a big advantage to them because they were running out of capacity in one hyperscaler, and they are like, "Okay, so are you saying that we can have redundancy in this other hyperscaler? That is really meaningful to us." So they chose for one of their workloads us. So these are the three reasons, Tyler Radke, that still why, whether it is labs or AI native company, choose this document model, scale, and run anywhere. Then in terms of use cases, sometimes, if you are interacting with agent, you have lots and lots of conversations going on.
Next time you interact with agent, it is almost like they forgot that you had that conversation with that agent. So sometimes for one particular product, one of the labs is using us as a conversational memory layer. For one of the AI native companies, we are completely, what I would describe it as just the operational data layer. What they told me specifically, having search in a different, having a pipeline for search in a different database, operational data layer separate that they were using before, so they had multiple moving pieces. For that 75 million agents, this is ElevenLabs growing like crazy.
They just could not scale because they had to maintain the pipeline between the ETL the data out- put it for search, then sync it back in. Now we are the underlying operational data layer for real-time agents, and they do speech to text and text to speech, so a lot of unstructured data. So those are the kind of use cases where we stand out. There is really no competition, because even if you use, I would say our competition for these kind of AI native companies, where they are making explicit or implicit decision, is sometimes what a coding agent may tell you something, and then you just choose that database or a first-party database from a Hyperscaler, and then eventually you hit scale limits.
Right. So even where you are running into Postgres being used, it is not from a large Postgres specific vendor, right? It is just kind of either open source or from-
Or the first party from a Hyperscaler. Yeah.
I see. Okay. Going back to your results last week, as you pointed out, total revenue growth, the highest in many years, north of 30%. I think you look at bookings, customer adds, net retention, all improved quarter-over-quarter. Yet, I think investors focused on Atlas growth. It stayed consistent versus, we look at some of your peers in the space are seeing more meaningful acceleration. How would you just encourage folks to kind of dissect those results and, I guess if we're sitting here a year ago and Atlas is bending that growth curve going higher, what are the things that could make that go even faster?
Absolutely. First, what I want to touch on for our investors, and this really matters to us, this will be the second year that our revenue is accelerating. Total revenue. Revenue is accelerating, and that is meaningful to us and our operating margin, cash flow, all the numbers that we printed last week. Fundamentally, the business is very good, right? The business is very good. When you look at the overall business, it's durable growth while profitability is increasing. That is always hard to achieve, but the teams are able to do that. I'll touch on that from a total revenue common perspective. Second thing I would say is what would you need to believe when people say, "Okay, CJ Desai, your peer group of a data warehouse company or peer group of a observability company." Tyler Radke, I've done technology for a long time.
Those are very different things than an operational database first, right? Understood, but we are put in that peer group, totally fair, and they will say, "Hey, AI is accelerating XYZ," which is totally fair and a lot of credit to them. Most of the data warehouse use cases are employee facing, right? MongoDB, when a bank or an insurance company or an AI native company like ElevenLabs use us, they are almost always customer facing use cases. In customer facing use cases, as companies, enterprises build more agents for customer facing use cases, that's when you will see our growth curve starts to bend. We still feel, like I said earlier, I speak to banks, insurance firms, healthcare companies all the time, and they are still not doing customer facing agents at scale.
When they do customer facing agents at scale, you will see our consumption go up and so on, right? Compared to right now, you do a query on data warehouse using one of the AI agents, and you get better answers in natural language. Great. That's great. If the answer is not accurate, you don't have to deal with regulators and others for internal facing application. Versus for external facing, the bar is very high, including your firm we talked to and many others. So that's number one. When will agents that are true customer facings start to scale built on MongoDB? So that's one. Second, I definitely feel that our vector integrated with now auto-embeddings that we just released a few weeks ago will also be a tailwind.
We are still early in that, but many of the banks I talked to, they have a different vector database, they have different operational database. Then they are using embeddings from a Frontier Labs or somebody else. All of those being unified in Atlas is definitely going to help out when enterprises start. That's where the big TAM is. Start building agents at scale. So that's second. Third, from a growth perspective on Atlas is we just released fully managed MCP functionality for Cloud Code, Codex and Grok Build, which was on August 13, not even a month ago. We are seeing it's been only now three and a half-ish weeks. We are seeing higher traffic on Atlas. Once that we provided fully managed MCP capabilities on August 13, we launched it at GA.
We will provide additional updates on how that is going on Investor Day on September 29th. We will have six weeks of data points on what we are seeing. But I personally feel that that is something super important that we had to do. We got it done. I would argue that we are a little behind on that compared to the other startups and others who did that. But now you can work in Atlas via Cloud Code is a pretty big deal to us. Same thing with Codex, and that will be received very well by developers who want to either build, provision a cluster on Atlas and so on. It's early, but that was definitely necessary.
We had to get done, as in must get done. Pablo and the team did a great job to get it out there, including with Vercel v0, which we are behind. You had called this out with me many quarters ago. Also Cognition Devin. We are really excited about our partnership with them, technical partnership with them, and what they are trying to do to also modernize databases and so on.
Right. That's super interesting. Those two kind of big product releases you called out, the auto-embeddings, think of that as maybe more a tailwind for the enterprise customers. And then fully managed MCP server, I mean, that could even be a tailwind for the self-serve.
That is exactly right. What is encouraging, you asked me, "Hey, what is your observation in the first year?" To add a new customer, paying customer, is really hard. When we added these many customers in last few quarters- over time, they are almost all on Atlas. Probably 99% of new customer adds-
Yeah, with all of them
are Atlas, meaning in cloud. Voyage has clearly been received really well. Because you know why? When we looked at why Voyage customers are being added at such a rapid pace, most of those customer adds came via referral from Cloud Code and Codex. Which was not happening for Atlas. Right? That now acts as top of the funnel for us to potentially upsell Atlas. That new customer growth, when that also kicks in, when they start building workloads in a meaningful way, will also help Atlas.
I see. Okay. I guess as you just think about the-- We talked a lot about your view on when we see the enterprise ramping and all that. I guess, it is difficult to put time frames on it, and it varies by industry. Certainly, I would say we are seeing customer-facing use cases ramp at the AI labs. You talked about Harvey, we had Legora on stage yesterday. These companies are really inflecting this agentic labor opportunity. Customer service is another big area. We just talked to Salesforce, companies like Sierra as well. Sounds like you are partially participating in the AI lab customer-facing use case.
Where do you think the non-AI labs are in terms of that? Again, we understand a big bank like Citi is regulated, but maybe something retail, e-commerce could be a little faster than a big bank. What is- the general timing on this?
I think that's a fair question. One of the places where we are seeing higher enterprise rollout on MongoDB for customer-facing workload or product, basically at the end of the day. If you look at ElevenLabs, whether you argue that they are a lab or whether they are a product company millions of agents running on MongoDB. Tyler, exactly a year ago, they were not a customer.
Okay? They are a customer that is growing very rapidly. That is number one. Number two, where I see higher traction among the industries, so bank would be behind from customer-facing perspective, is tech companies. Tech companies, cybersecurity companies that are built on MongoDB, whether they are cybersecurity companies out of Israel or cybersecurity companies, some out of U.S.A. We are seeing them building agents on top of MongoDB and rolling that out for their customers. We are the operational data layer. So tech is a vertical where I have seen higher traction.
It is still this year is when I saw the growth with some of these cybersecurity companies that are built on MongoDB, because we do unstructured data very well, like logs and things like that. So tech is a vertical. Retail, airlines, others, I have not seen massive customer-facing agents still being rolled out. Telecommunications company, some internal facing but not external facing yet. When I look at that enterprise bucket, what I personally feel is that after tech, most likely it is going to be retail and financial services that would follow, including insurance, before consumer product goods company or manufacturing go there.
Right. Okay. That is super helpful. I guess going back to the strength in the new customer adds, which again, really impressive, I think 2,900 new customers, the highest you have ever added. Many companies out there are talking about top of funnel- yes, right? Maybe a little bit more on the application side. What are some of the things you are observing in what is driving such outsized top of funnel strength? I think you alluded to this dynamic earlier, whereas if you go on to Claude or ChatGPT, it might initially default you to recommend a Postgres. I presume you have been working hard at getting MongoDB more in this conversation. Just talk about the top of funnel, the things you are doing, and how you see that durability of the strong customer adds.
Absolutely. First bucket, I would say us getting new Voyage customers for auto-embeddings. Which are typically almost always AI use cases for the companies that use Voyage, obviously. When we see that is, I almost guarantee you, mostly driven by Claude Code or Codex recommending, please use Voyage. Right?
They say it on Anthropic's website.
On Anthropic's website, they say it, but Codex. OpenAI has their own embedding model. But if you say, "What's the best embedding model to use?" You get pointed to Voyage. I would say the cause and effect on Voyage customer adds is primarily driven by coding agents. Then it is the opportunity for Ryan and the team, our sales team, to now upsell them on Atlas. So that becomes a very meaningful top of the funnel for Voyage. Now, when we are getting Atlas customers on top of the funnel, those customers are still very human-driven because we didn't have this managed MCP functionality that we just released four weeks ago. So human will say, "I want to spin up Atlas," and then they will use Atlas in the customer adds that we continue to get, and then they scale over time.
One of the labs told me that they actually came via self-service channel, and then we noticed it, and then we put team on it to make sure that they can scale nicely and so on. With this managed MCP functionality, Tyler Radke, I'm optimistic that we will continue to see new customer adds. Now, you cannot game prompting. You just cannot with these agents. If you use certain words, of course, it will pick MongoDB in the prompt. If you don't pick certain words, it will default to a Postgres, things like that. Yes, the teams are working hard because we also want right use cases to be built on MongoDB. If it is some employee-facing apps with 1,000 users, we really don't want that workload. We want the workload that can scale a quality workload versus a quantity workload.
I see.
That's the focus we have.
Okay. You brought up go to market a bit. I did want to touch on that, just given some of the changes this year. You brought in John McMahon. What's sort of the state of go to market? I think as we look at the commentary over the last few quarters, you've kind of consistently called out strength in North America. I'm sure there's areas that are not performing as well as North America. Give us a sense on his priorities and how do you kind of see the productivity and the ramping of the reps that he's brought on?
Absolutely. I'll do that. Jesse, you want to touch on regional performance, and then I'll go into the details.
Yeah. We've continued to see strength, particularly with our largest customers in North America. That's a very encouraging sign. North America, obviously our largest region. There are some areas for opportunity for us to improve. One is U.S. federal. We're investing in FedRAMP High, and that will hopefully open up incremental opportunities. That's one of the reasons we did the Clarity Business Solutions deal. Then APAC. APAC is an area where there's an opportunity for improvement. Japan is a very large database region. It's one we're going after, still very early stages, and we definitely see opportunities to accelerate our success in Asia. That's an area that Ryan's made some changes, and we're optimistic that they'll have positive results on a go-forward basis.
Specifically what I will tell you is when I look at the NRR number, and you cover, Tyler Radke, software industry, you see that that's a very important metric on the retention rate, 122 NRR. But even when I look at NRR underneath for our self-managed as well as Atlas, both are very healthy. It's not that one is driving the other. They both are very healthy. So the blended rate is 122. In terms of Ryan and go to market, if Asia is a smaller percent of the total business, but it's a very important business because you have a lot of digital natives in India, you have digital natives in China, Korea, all these places, and most of those regions love MongoDB. But we also need to serve the large enterprises there really well. So that's some of the changes that he's making in Asia.
Okay. Got it. As we think about the migration opportunity, right? We've talked a lot about these new use cases, and clearly there's going to be way more new database use cases than existing. But migrations have been an important part of the story over time, and you've rolled out a number of initiatives there. You talked about partnership with Cognition. This MAP program that's been announced, I think over a year ago. So how do you kind of see the migration opportunity today? Is that something you're investing more resources in, or does it make more sense just to prioritize new workloads?
Being in the database industry for a long time, both as a customer and as a provider, Oracle is going to be 50 years old next year. Think about it. 50 years old company in database industry, around 60 years. We are the most modern database. The opportunity, Tyler Radke, just to answer, absolutely exists.
The change I made once I joined with Mike and the team is that we had taken this very people-centric approach on migration. Hey, we will help you migrate from Oracle to MongoDB for the right set of use cases or some relational database. What we decided that this year, as in this current year we are in fiscal year, we are going to get the product right, and we want to provide a product to our customers that you can migrate from a relational database to MongoDB faster, and the people portion is a smaller portion of that engagement.
When we announced it was very people-driven, services-driven approach. We are changing that to product-driven approach, and AI definitely helps with that, whether it's Claude or pick one, as well as partnership with folks like Devin helps. And you do that, then you go to customers and say, "Now we can migrate faster." So we are going to put fuel into that, in the next fiscal year once the product is ready. But the product, we are currently working with very large customers on very specific, what we call design partners, to make sure that our product really works so that they can migrate to Atlas.
Yeah. CJ, in sort of the closing question here, if we're sitting on the stage a year from now, which hopefully you're definitely welcome back, what are the two or three things that you hope to have accomplished by then?
I would say you touched on this on the priority. First, we have to be the best database, whether it's self-managed, whether it's Atlas, it has to be the best database that can run anywhere, and we just want to continue to focus on that. The best product always wins in the market. Number two, that I have shared, human developers love MongoDB. I want coding agents to love MongoDB. And managed MCP was the first big progress we made, but we'll continue to invest in that so that, hey, I want to do this, of course, you should build on MongoDB. That would be the second.
Then the third thing is, you will see some of the exciting product announcements on September 29th, which is 20 days from now, around what else we want to do so that we are the default real-time operational data layer for AI. I mean, those are the three things.
Yeah. Great. Well, let's wrap it there. CJ, Jess, thanks so much for coming to the conference, and appreciate everyone for joining the session.
Thank you.
Thank you. Thanks, everyone.