DigitalOcean Holdings, Inc. (DOCN) Earnings Call Transcript & Summary

[Audio Gap] of Citi's TMT Conference. My name is Mark Zhang, and I help -- I'm part of the Citi software equity research team. Today, we have the pleasure of hosting DigitalOcean CEO, Paddy Srinivasan; and CFO, Matt Steinfort, all both to my left. Welcome to the conference, and thank you so much for attending, gents.

Well, thank you for hosting us, Mark.

Yes, absolutely. So I think maybe let's track a little bit. Paddy, it's been, call it, 3 years and changed since you've taken over the helm as CEO. At the time when you joined, it's -- DigitalOcean looked a lot different than what it was today. At the time, I think paper space was the big AI acquisition and the way that DigitalOcean approached AI. But since then, a lot has changed. Maybe take us through the journey of -- number one, what you saw in DigitalOcean coming into the world day 1? And how you sort of saw this as from a strategic standpoint, how you approach AI? And obviously, there's various ways to play AI and companies roll in AI. So what made you take the bet of the path that you're taking on today?

Yes. Yes. So thank you, Mark. It's a great question to get started with. So I'm coming up on 3 years, not quite there yet, but in the world of AI that it's almost an era. So when I came in, I mean, the first thing that really attracted me to DigitalOcean was I've been working on and off on developer platforms right from my early days at Microsoft for close to 30 years now. So that pedigree and the DNA of the company which was relentlessly focused on catering to the needs of developers of all sizes and shapes and different sizes of companies was really attractive. And the second thing that was amazing about digital ocean, which still -- so it never ceases to surprise me is the strength of the product to be the primary customer acquisition engine. So those were the 2 things that I knew were the strengths of the company coming in. But then one of the first things we had to do was to make a decision on what kind of company we want to be in AI. So the big bet that we took in retrospect about 2 years ago was, do we want to focus on training or inference or both? And we made a very strategic decision to say that training was more on the hardware side. And our hypothesis was inference needs a lot of software. And at that time, it was not that evident, but now it is fairly common knowledge that inference is not only a lot heavier dependency on software, but inference is also a more durable workload. It is a workload that resembles cloud patterns. It is a production workload, and it is typically consumed when you have product market fit, and it is eventually build right back to customer usage for the most part. So that is a bet that we made. And along the way, we had to make many other bets, right? So for example, we had to make the bet that software needs to evolve to serve the needs of inference in a very different way than what training was. We also made a bet that we have to nail software first and then build scale. So that is the inflection point we are in as a company we feel very confident given that we have built our inference stack, working hand-in-hand with customers. That's one of the luxuries we have that many other Neoclouds don't have is having a front row seat in core development, co-innovation with leading-edge AI natives that push the boundary of inference gives us the ability and the confidence that we are building the right thing, we are catering to the needs of the most demanding AI native customers. So we feel we are on a really good path to a great software platform now we are addressing the scale or we are building scale to support the software. A lot of Neoclouds or most of the Neoclouds, I would say went after scale for us because they were focused on training and are trying to cobble together a software platform. And the other major decision that we continue to make is I fundamentally believe that technology platforms, especially developer-oriented infrastructure platforms cannot be assembled. They have to be built from the ground up, right? Of course, you can have a couple of acquisitions here and there on small feature gaps. But fundamentally, it is really, really hard to cobble together. This is not an application stack like Salesforce. There's a reason why you don't see Azure, AWS or Google Cloud really going into -- in an acquisition spree to build their platform, right? So developer platforms, especially in AI infrastructure have to be built from the ground up, and that's the platform we call as AI native cloud. That's 5 layers, and it is beautifully integrated into a single pane of glass, provides all the control pain required to build and scale inference applications. So a lot of small decisions that have put us in a very good position of strength, I would say, because we have some of the frontier inference workloads running on our platform.

Sure. That's terrific. Thanks for that overview. Speaking of scale and where you are in the position you are today, it's sort of like this opportunity to really inflect going forward. The inferencing engine has seen just incredible traction reaching, call it, 6,000 plus logos within a matter of 6 months. Can you -- I guess I think obviously, there's tech advantage and the software that you provide on it. But can you maybe speak to what has been the driver of the success there? Working from is a product market fit, the tech differentiation, product marketing, and how -- I understand like there's many use cases here for AI and inferencing. How are you -- like AI native using basically like properly using your products today?

Yes, it's a good question. So it has been 4 months since we launched. We went into GA. It's probably slightly less than 4 months, and we have a lot more than 6,000 customers at this point. And it goes back to our strength in product-led growth, right? We -- of the 6,000 that we announced in August, maybe there are a handful that we acquired through our direct sales motion. A vast majority, like 99% of those customers came to us by the virtue of our products, lending itself to a try and then buy kind of model. So it is just organic word-of-mouth developer adoption. And the reason why the product market fit was just jumping off the page is a couple of reasons. One, there's just a lot of latent demand for inferencing, right? The whole world anyone that is writing software now is building it as AI native, agent native, it needs tokens. It's very inference hungry. So that's number one. Number 2 is -- there is -- so there's a difference between when we say we have inference and when a Neocloud says they have inference, it's like really apples and pineapples. They sound kind of similar, but they don't taste anything like each other, right? So when we say inferencing, it is managed inferencing, where we say we will give you an end point for an API and you can define what kind of tokens you want? What is your SLA, what is your throughput? What is the quality of service you're expecting? And you leave the rest to us, how to manage the infrastructure, where to deploy it, how to load balance it? All of that stuff, what kind of hardware or accelerator it is running on, that is our headache. You don't worry about it. You want certain quality of service and certain type of tokens, we will deliver that to you 24/7. When you talk about running inferencing on a new cloud infrastructure, it is basically saying, okay, I need -- I'll give you a certain class of accelerators like MI-355 or GB300. And then you take over from there. You install the drivers, you build the development pipeline. You install the models and you manage the life cycle of the whole thing. So both will work, but for you to take a piece of hardware and run inferencing at scale, it takes a significant amount of heavy lifting that I believe only a handful of companies can do at this point especially when you start talking about advanced resenting models like Kimi K3, GLM 5.3, these are near frontier open weight models. K3 has 2.8 trillion parameters, of which 200 billion are active nodes. So you have to load up the model and keep 200 billion parameters in memory and the KV cache, it's a 1 million context window. So you need to be able to load the whole thing, have KV cache for 1 million context window and manage it and manage the throughput and do the offloading of the cash from memory to disk and this back to memory, it is not easy. That's why when you go to open router, you look at -- all the inference providers, you will see dozens and dozens of them, 90% of them are serving flash models, which are very, very small models, which are great for some use cases, but not great for near frontier tasks. So when you start getting into the near frontier type of model performance, it the general industry trend, you asked about use cases more. The use cases are becoming more and more sophisticated, right? To serve those, you need real frontier reasoning models like Fable and Astra in K3 and GLM 5.3. And a vast majority of AI native workloads prefer to have managed inferencing rather than just raw metal infrastructure where they have to manage the whole thing. So when we talk about the inference demand is unbounded, what we mean is to manage infringed demand, like the demand for consuming tokens, not demand for consumer actual infrastructure. So it's very different when we talk about our inference platform. It comprises of many, many different individual building blocks, not just a token delivery service, but there are a lot of other services that make up our inference engine.

Got you. Got you. And I know you speak to the demand coming from more sophisticated workloads and more sophistic test. I know obviously, like coating is a very sort of like top of mind task that we always think about when we think about AI. But anything sort of like cutting edge and merging like task that you see AI Navis performing that could be basically getting to the reflection point? Or anything else that gives you conviction of long-term durable demand.

Yes. So coating is the obvious 1 that we are seeing that coding today versus quoting 12 months ago, completely trade, it looks like, right, from a practitioner point of view. But we are starting to see a lot of different adoption in use cases like generative media very advanced use cases and real enterprise companies, where we have a lot of AI native customers that offer generative media technology to their customers and the end customer is typically a large enterprise or digital agencies. So the way you build digital campaigns, the way you build digital assets from ads even full feature movies to anything in between, like even Hollywood production support, all of these things are transforming. And we see a significant amount of demand and pull for generative media models. We are starting to see a lot of our customers build go-to-market applications, which are agent native. So we are seeing a lot of companies starting to deliver contact center software that is agent first, right? Customer support software that is agent for us. We are starting to see gaming companies that are becoming AI centric or AI native. So we are starting to see many of these other I call them a capital A agent software. So when we talk about agents, it's like a coding task has multiple subagents that are doing autonomous tasks. That's, I call it, like small A agents. And then you have capital A agents, which are labor replacement workflows, right? Multi-day, multi-turn, advanced reasoning. So you're starting to see agents that are replacing significant amount of human labor in many walks of life. So we are starting to see an emergence of multitude of use cases, personal productivity. Like it seems like earlier this week was news from Meta a couple of weeks ago, it was [ Grokpot ]. And then in between, there's another start-up that is getting a lot of attention called Instinct. So we're starting to see personal productivity agents really exploding. So there are at least half a dozen -- repeat workloads we are seeing. When I say repeat, I mean, there are multiple customers running their workloads on us on some of these micro verticals.

Interesting, very compelling use cases that are emerging. I think just given that, call it, there's a various multitude of products and features on the inferencing engine, how are customers -- can you maybe just walk through the life cycle of -- from the beginning to sort of like at production, how are customers using inferencing engine from a feature module standpoint?

Yes. It's a great question. So what we are observing just to -- and obviously, this is just a canonical example because this is what I'm distilling from watching a lot of AI native companies. When they start, almost everyone says, okay, I need to find out serious companies. They want to find out whether they have product market fit, right? The easiest way to do that is they pick either the latest Anthropic model or an OpenAI GPT model and go after finding product market fit. So that's typically how companies started because even 6 months ago, open rate models were 6 to 9 months behind the frontier models. Today, the open weight models, the near frontier models are barely 2 to 3 months behind the open rate models. So what I'm observing is a lot of companies as they get product market fit, a couple of things happen. One is they start observing the cost of goods sold line item, which is the token consumption just starts getting out of control. So that's number one. Number two is there is a lot of awareness in the market now that if you have an open weight model, it is a lot easier for you as an AI native company and you as an enterprise that is adopting that product to have an opportunity to own your intelligence, own your rate leading to own your intelligence, right? So for these 2 reasons, I see the equation flip in terms of token consumption from closed source models to open weight models. So I think we talked about the fact that we flipped literally from 25/75 to 75/25 in favor of open weight models, and it's only accelerating, right? And this is not just us industry trends. You can look at any of the AI aggregators, you'll see similar statistics. So in terms of a typical life cycle, so they start moving to open weight models. And then one of the first things that happen is they want to take the open weight model like a Kimi K3, and there are famous case studies like Cursor, it is famously on Kimi, right? DoorDash is on Kimi. So what they typically do is they do fine-tuning to make the model more domain-specific. And then they do reinforcement learning, which is a fancy way of saying, I can fine-tune the model to be more specific to -- we are here in New York, very Wall Street specific. But then reinforcement learning makes it even more customer-specific, so I can train it on the usage patterns of Citi every night. So the next morning, the pipeline works in a way that after the reinforcement learning, they do real-time agent eval to make sure that the model passes the val test. So the performance for those use cases that it has been reinforced to learn becomes much better the next morning and then deploy that. We are also starting to see -- I mean, these are all modules that they leverage from inference providers. So the other thing they do is we are starting to see many workflows use a mixture of models. So for certain speech to text, for example, a certain class of models may be best price for performance versus reserving a heavy-duty model like K3 for expensive but high reasoning tasks, for example. So they use model routing. We introduced another feature called model synthesis, which is for the same task, you can have multiple models work on it real time and in just a few milliseconds return back the answer and our model synthesis module will stitch together the answer. And the user will not even know whether some parts of it came from K3, other parts came from a flash model, right? So we do that to preserve the cost performance equation and it is all controllable by the builder of the AI product, and then we serve the answer. And then comes -- once you start serving it, then you need to have the ability to monitor and manage the whole life cycle of a model deployment. So these are all different aspects of the token consumption through our inference engine. And then comes the agent aspect of it. So last week, we announced an agent platform, which breaks down the -- an agent is an autonomous piece of code that does work, right? So typically, these agents are short-lived, but they perform tasks over multiple hours or even multiple days, but in short births of a few milliseconds at a time. So we introduced a new cloud primitive called Sandboxes, which are compute cycles that are switched on and off in a matter of a couple of hundred milliseconds. And the agent performs a task, goes down, wakes up a game, has all the memory, does a little bit more work and then so on and so forth. And then the ability to have a swarm of agents all orchestrating and working together to accomplish that. So these are all the capabilities of our platform.

Yes, that's a very comprehensive platform from where you were 3 years ago. But can you maybe speak to the adoption rate and penetration of this sort of comprehensive platform, whether it's from model routing to agent management, what's the penetration? And what's the sort of product adoption? And what's the opportunity here going forward?

Yes. So as you go more and more away from the bare metal infrastructure, we see the adoption of the primary front door to the platform is now moving slowly, but very, very clearly moving away from, hey, can you just give me a GPU to, hey, I want tokens of this flavor becomes the front door. But once they come to our platform for tokens, then they start expanding to all of the features that I talked about. They start building agents, deploying agents. Once you deploy agents, you have to monitor them. Agents are very data hungry. So you need storage, you need databases to persist. So then it starts proliferating. In the August earnings call, Matt and I talked about the fact that our 100,000-plus AI customers are attaching core cloud at a 70% clip. So pretty much everyone that comes for AI is attaching core cloud. In another way of saying it, AI is becoming a demand gen engine for our cloud services. And that is even before we announced our agent platform. So that's only going to become more and more acute. So from a unit economics point of view, the higher up the stack you go, it ceases to be GPU economics and it starts looking more like software economics, right? So we feel like we are just starting to scratch the surface. As these AI native applications become more and more sophisticated, I think what is the origin of tokens? Or where does the token demand come from? It comes from agents. It comes from modern applications and applications need to be executed somewhere. They need to be -- they need to orchestrate. They need to be fed data. They need to persist information. So you need storage and databases. So we are starting to see that flywheel take effect where companies come for tokens, expand into agents, agents are data hungry and agents need to be orchestrated, so you need CPUs. And the more agents you have, the more token hungry they become, so they consume more tokens, and that's the flywheel we described in the last earnings.

Yes. No, absolutely. It's very encouraging to see the momentum of that flywheel. Now maybe like just a quick one on the flywheel aspect. How sticky does a customer become once they're on this flywheel? And sort of like what's the -- maybe like ARR or revenue uplift that you see from a customer that's bought into this flywheel of, hey, like I came in for AI. Now I'm buying your core cloud. What's sort of the upsell opportunity here?

You just look at the ARR per megawatt that we generate today, it's a function of the blend of services we have. We have a small amount of bare metal, right? It's -- I think at the last earnings, it was about 15% of our AI customer ARR. Most of our AI customer ARR is inference services and core cloud pull-through. And so when you think of, okay, what do you see in the industry on a total ARR per total megawatt from the Neocloud, it's still in the high single digits and maybe it's getting to like 10 or 11. They may announce deals at higher rates. That's on newer technology. But if you look at it on an embedded base, it's $9 million to $10 million -- if you look at us, we're generating in Q2, it was $22 million per megawatt. And on an incremental basis, it's 30%, 40% higher than what you're seeing out of the Neoclouds. That's all because of the higher value-added services, the inference services, the core cloud pull-through that are both stickier and higher margin. So we think that there's a tremendous amount of upside there, particularly as the inference engine is really only a couple of months old, and we're just starting to see the economics there. And the big lever that I don't think the market fully grasps yet is when you sell inference services and you sell tokens, it becomes a price and yield optimization game. It's not a, hey, did you get $4 per hour on a GPU versus $3 or did you sign a long-term contract or a short-term contract to take advantage of surge kind of capacity pricing. it's all about how many tokens can you generate from the same amount of infrastructure. And the more you can generate, the better you can deliver those tokens, the higher you can drive that price. So without adding incremental megawatts, there's still a price lever that enables you to drive up your ARR per megawatt quite a bit.