Datadog, Inc. (DDOG) Earnings Call Transcript & Summary

Karl Keirstead, covering the software space. This is actually my first time on stage to kick off this year's tech AI conference. So I want to take the occasion to thank all of you for coming. I was telling David earlier that attendance at this year's event is up 45%, 50%. I don't think, in my career, I've ever seen 1 year to the next that kind of growth. So it's phenomenal to have 1,200 of you here and 300 corporates, including Datadog. So I'm really thrilled about that. And Datadog, you may know, I don't know if you get the list of most requested companies, Yuka, you probably know this, top 10 most requested company to see in one-on-ones of 300. So you're doing something right, David. I don't know whether you're a popular kid at high school, but you are now.

Not this popular. I didn't have a rising stock price in high school.

Okay. Well, let's get started. So let's maybe start at a high level, David, and maybe I could ask you to describe your sense of the environment you're seeing. I know on the earnings call, you described the spending environment as being relatively stable. But I think for a lot of us that tried to weave a thread through all of the results across the software industry, it actually felt a little bit better to a lot of people, including me, where we didn't have a lot of blowups, the beats were a little bit higher than normal, tone upticked a little bit. So it feels like things are a little bit better, but that's not the message you provided. So could you elaborate a little bit on what you're seeing?

Yes. As we've mentioned, we -- and the time series troughed in the third quarter of last year. And since that time, things have been progressively improving. The buying environment, the return to digital projects and the spending has been solid. And I think what we meant by continuation was it continued to improve slightly over the previous period. So it's been like a smooth upward track. It's been not just a tale of sort of all segments, everything is the same. I think we did note, and maybe that's what you're referring to, is the enterprise part of our business, customers who have more than 5,000 employees, has been in the higher range of net retention and buying, and that continued in the quarter. Our SMB, we have a very broad customer base, 30,000 customers, ranging from the largest governments and enterprises down to cloud natives and smaller companies. It was solid and stable, but really it was sort of at the same level it had been in the previous quarters.

Okay. Why do you think that is, David? Why do you think what you're seeing is enterprise steady improvement, but SMB is not yet? Why that difference? And do you think there's a plausible outlook whereby sort of SMB catches up?

Yes. Yes, good question. First of all, we're sort of, I think, more towards the top end of what you might call SMB. When you think about who's buying Datadog, you have to have a cloud subscription. So it's not going to be your corner drugstore. So we tend to be towards the higher end, I think, more towards M. But I think what we saw is that sector of size and cloud nativity had bursted a bit more on an upward trajectory during the bubble, so had a little more unwinding to do. Enterprise took it more programmatically. And then it has a lot to do with the funding environment for venture capital, what entities are being funded. It looks to us like there is -- the dollars are weighted more towards AI, and we can talk about our AI group. And so the broader group still might be constrained from a capital. And of course, the change in the equation between growth and profitability caused companies that were spending very rapidly to put a little bit of a [ censurer ] added on top of it. So we think we still see in SMB a healthy environment, but still with some constraints. What we're were doing with that is we're essentially seeing markets in SMB where we really didn't have the infrastructure in place and the sales teams, take some of the emerging markets, take Brazil, we are landing our first people in India. And we're investing in the areas where we think we -- either they're catching up in terms of their maturity. There are a lot of cloud natives that are moving more towards investing in systems like Datadog. So we're working on the distribution of our SMB sales team to try to get to the highest growth markets.

So those are some positives because if, David, the, call it, the post-COVID unwind eventually ends, rates induce a little bit more spending on the part of SMB and then Datadog itself turns around and puts more distribution strength. You can imagine a scenario at some point, hard to say when, but that part of your business might pick up. Okay.

Yes, I think that's right. I think that depends on both things we control, our resources, our products, which we're doing, our sales. And also, I totally agree with you that the venture capital, the funding environment and the founding of companies at that side of things is -- has a lot of correlation to the growth of that business.

Okay. Let's -- so the one positive we just talked about is steady improvement on the enterprise side. The other, what I thought was a big positive from your last print was the growth of "the AI natives", now 6% of Datadog's ARR, up from 2.5% a year ago. That's phenomenal. Even if it's concentrated, it's phenomenal. But when -- on your print, it initially created a little bit of a mixed reaction. There was a good and a bad. It's great that you're attached to a customer segment that's growing so quickly, but you did highlight that there could be a catalyst for some of these AI natives to search for better pricing upon contract renewals or optimize their spend. So I think the investor community was a little bit mixed initially, but I think as you can see in the stock price, with time, I think they cited on the more favorable view of the exposure to AI natives. So can we talk about both sides?

Yes, definitely.

So when you were talking about optimization, David, what were you referring to? Is there sort of a contract restructuring moment coming up where if they're blowing through their comments, they would naturally want to get lower unit pricing? And then the positive would be if there's some kind of reset after that point, you would grow off that reset base. What do you think?

Definitely, that's the way our business has always run. We've shown this publicly a number of times. We showed it when we had the cloud native optimized on the back end of COVID. So that's our model, which is essentially a commitment model where clients sign up for an amount of commitment. They can use the platform fairly frictionlessly. They aren't repricing their deal every day in real time. And depending upon their ability to project their consumption, not only the consumption in total, but where it's going to be, and also execute their engineering projects to manage that, there is not a perfect timing between the price they're paying and the amount they're spending. So what we were pointing out was because of the rapid growth in this good growth vehicle for us, albeit a small part of our business, that if these customers act like any other same customers, they will over time and we'll help them price their deal to their current volume. I mean we have active discussions, and this is just not with this group all the time. And you ramp past your commitment, let's talk about your new deal. And so what that -- it's exactly what you said, for the most part, what that means is you might have less growth. In some cases, you might have a lower ARR as they digest. It depends upon the interplay between the price and the volume. And then they generally do that in conjunction with the new commitment, which is a very good thing, and we can go to the good side of this, and then we grow with them. And this happens in our 30,000 customers. It's just that because of the growth of this, we pointed out that this is a very good thing long term, but may affect the volatility of the growth rate in the short term, which we wanted to be transparent about.

Okay. And David, how do you think about the risk profile of this group? I know you haven't named names. But if those AI natives are highly concentrated among, call it, the premier AI natives, I'm talking about the OpenAI, Anthropic, versus a scenario where they're concentrated among AI start-ups that the investment community needs to ask whether they'll even be around in 2 years, those are 2 very different outcomes that have different risk profiles. Is there any way to help us evaluate the risk profile of those AI natives? Anything qualitative you can say without naming names?

Yes. I think it's both broad and concentrated. And so there's a lot of names there. If you look at the diagram of the AI infrastructure spend, everywhere from hyperscalers to models to vector databases, they're cloud-native companies, and they are delivering a product which is the kind of customer that chooses Datadog. They're born in the cloud, their whole business is delivering it. So you have a lot of customers in there. And because when you look at who are the winners in this, on the infrastructure side, you have some concentration. You have a combination of either already large companies or companies that are getting very large and a very, very long tail. I would say it's similar to -- it's similar, although maybe more concentrated to our overall business in that we have very large customers and the tail. What we've seen is the gross retention rate in our tail has been still in the low to mid-90s. So we've seen a lot of stickiness in the smaller customer base. But you will have some that don't win and eventually go out, but we'll have to see. It's early on. It's early on to see how big this is, how rapid it is, who are the winners. We may not even know who all the winners are yet.

Okay. Good. And maybe one last one on these AI natives. Can you describe what a common use case is for Datadog? Like how are these AI natives utilizing Datadog? What are they monitoring exactly?

They're monitoring their workloads and their applications, so they would be using the Observability platform, which are metric traces and logs. So they would be monitoring how the -- their applications, which they're delivering are interacting with the infrastructure, how the code is working. So it's a very similar type of logs to investigate. No, no, I would say it's a similar use case. What they're doing with it, the end users doing with it, whether it's training or inference is less of a matter, it's more a matter of this particular vendor is delivering a software solution or infrastructure solution, and we're a choice to monitor it.

Okay. Let's talk about the other way that Datadog has AI exposure. Other than selling your core product to these new, exciting, high-growth companies, you yourself have AI products in the form of things like LLM Observability that you're selling. Can you give us an update on how that product launch is going? I know it's relatively early, but anything you can share?

There are 2 other ways, and I'll go through, one, the product or SKU side; and then two is, what about our own platform?

Yes, exactly. We'll get to that.

Because that's the third. So there are 3 real areas. So in terms of how we're designing our platform to service clients, those of you that follow Datadog know that we are rapid -- we rapidly integrate with any of the data sources that are relevant to our clients' delivery of the applications. So we have integrations with these AI vendors. And one way we know about activity is, of our 30 -- roughly 30,000 customers, 10% of them or 3,000 of them are sending us data from that. Now that's part of our platform. We don't charge for integration. That's a platform feature, always been a platform feature. And so that gives us a sense of the overall activity level that's out there. Another thing is our LLM monitoring. Now what this is, is it's like we have an APM, we have a code profiling, we have data. We have all these different features that illuminate the user on what's going on in the application, and one of them is LLM monitoring. And that is the function of the LLM model within the application. And we have not released pricing on that. We essentially have it out in the market. We have -- we said hundreds of customers. I think, to interpret that, roughly, it's about 1% of our customer base. We had a couple of anchor customers or development partners speak at DASH on that. These are customers who have successfully integrated LLM in their applications. And we are monitoring that functionality as part of our overall observability. I would say this group also is quite broad. You have some cloud-native and you have some cloud-progressive efforts at large enterprises. Still very early. We want to be able to give a metric like what's the activity on this, and we'll see how things go. But I think that's where the rubber is really going to meet the road in this thing over time.

Okay. And then you alluded to it, but the other obvious way that Datadog benefits from AI is when your customers begin rolling out AI applications that require observability.

Yes. So that's what I just mentioned that -- what I missed mentioning was that, which is essentially -- so AI or the part of the large language model is part of the overall functioning of the application, including how it deals with the infrastructure, networks, database. All of our products are around monitoring pieces of this or diagnosing. So that is correlated to the rollout, as you said, of LLM in production. And then the last piece which we're getting to is the Datadog platform itself, what is Datadog doing as a piece of software in putting large language models or chat into our platform. And we are slowly but surely rolling things out. And the areas that would be most germane would be your chat or your ability to interact with the platform, but also using LLM models to help diagnose problems. We talked about the continuum out to the brave future of auto remediation. Along that path is using LLMs and AI, which we've been doing for some time, to diagnose problems quickly. And that is we're building models and putting that in. Again, like everybody, a lot of this is not in GA production. A lot of this is being tested, but we're working on it.

Well, that's the question I wanted to ask you about the broader pull along, David, and that's your observation about where enterprises are on their journey of coming out of pilots and taking model-based AI apps to 5,000, 20,000 employees because I'm actually, even in my own work, starting to hear some companies when I ask them what piece of the tech budget might get pulled along as they roll out their AI applications. I am beginning to hear some of them mentioned that my observability spend might need to ramp. So I wanted to ask where you think enterprises are on that rollout process? Are we hitting an inflection point where you and your team are beginning to hear that there's a significant ramp in rollouts?

I think we're still very early. A lot of what's been rolled out, a lot of the investment has been in infrastructure testing and training. And a lot of it, when you talk about internal employees in some of the use cases, those really aren't the Datadog use cases. So I think we're pretty early on, but the metrics I just mentioned of the use of the integrations and the LLM and the number of customers that have it rolled out indicate -- I don't know if we're in a position to call it an inflection point, but there's a lot of activity. And so I think we're -- we think people are getting serious. They're getting more mature in their testing and modulation of these models. And in certain use cases, it is in production. So we're optimistic that this is going to affect Datadog in a couple of ways. One is workloads. Two is more rapid creation of modern software and potentially accelerated transformation of legacy application architecture to new architecture, which is a friend of Datadog.

Okay. Let's switch subjects as interesting as AI is to a couple of others. One is on the pricing side, David. So in a lot of the customer conversations we have around Datadog, to put it simply, love the product, but not cheap. And so I wonder what Datadog is doing to get those customers over the hump on the pricing side? Whether there's any impetus to maybe move to a more flexible pricing model. What are the kind of things you're doing to address that concern?

Yes. I think one thing that's maybe not fully understood to explain it, but we've been selling for quite some time on -- it's a similar model to AWS, a commit model, meaning you buy credits for the platform. And you can choose, you can use all aspects of the platform as you see fit, reserve, et cetera. So I think we've gotten better and better at that, but we've been doing that for some time. We have, I would say, we sell based discounts based on volume and term. I think we've gotten better about using both of those and plotting out capacity planning. We have a whole group that helps clients understand how they should be using Datadog so that we help them if they're not optimizing it or not using the right mix to suggest there. We have transparency of the meter ticking so they can see we've always sort of -- if things spike up, we've always said, we're not going to charge you this month, let's work on this together. And then we've, I think, gotten better at some pricing models like as we consolidate, instead of charging you a full price, a full ramped load for the first 6 months as you might migrate from another APM Datadog, we're amenable if the economics are right to ramping that buyout credits and things like that. It hasn't affected the overall average price. But what it's done is remove some friction from the consolidation effort. So those are all the things I think we've learned, and we try to work with our clients. I would say, yes, we try to value sell. We also try to bring more of the point solutions onto the product. So if they have a spend with Datadog, they're actually spending it on more products with Datadog rather than other vendors.

Got it. Okay. I also wanted to ask you a little bit about competition, David. You mentioned when we were chatting earlier that you're on your way to re:Invent tonight. So specifically, I'd love to ask you about the relationship with AWS and hyperscalers because I think, one of the attractive aspects of the Datadog story is, in contrast to a number of the consumption-based data software firms that are competing very directly with the hyperscalers, I've personally never felt like Datadog is up against an enterprise-grade version of Observability from the hyperscalers. So I just wanted to get your impression of that and what the nature of the relationship is with AWS, friendly enough to be -- for you to be heading up to re:Invent. So my sense is it's more friendly than it is an enemy relationship. But can you opine a little bit?

It's definitely a partnership on many levels. I mean I think you're right, because of the desire for Switzerland to monitor, the desire to be able to be multi-cloud and integrate all this data, some of which is germane to a single hyperscaler and some not, we tend to have a community product that is more protected than what you're talking about from the competition. So I think you're right, our partnership involves technology partnership to make sure that we're integrating what they're doing, and so we can monitor what they're doing. This helps them because they want to be able to sell a cloud that can be monitored. We sell through their marketplaces significantly. So we have a significant -- we have an active relationship both on the sales side and through our partner organization in lead. We go to -- a big part of our marketing budget is partner marketing with them. So we support them and they support us with that. So I would say with all the hyperscalers, it's a very, very strong multi-dimension partnership. And it's proven to be not one of as much competition, but how can we help each other. You sell more cloud and compute.

Okay. Got it. Let's move the conversation to sales and marketing, David, because I think that's actually an interesting part of the Datadog story. And in particular, on this last call, you made a comment implying that Datadog wants to lean into the sales and marketing effort more. I think you even articulated a view that your sales headcount growth could even pick up a little bit. So why are you doing this? And how can everybody square that with comfort in the margin story?

Yes, definitely. So we have a pretty efficient go-to-market with a really strong return. So I think we've generally tried to lean in, and we've said we're trying to grow our ramp quota capacity in line, it's correlated with revenue growth. I would say what happened was we pulled back a little bit during the bursting of the bubble after COVID. We were -- I said the market environment was a little less robust. We didn't have access to travel and being in international markets as much. So I think we had lower than desired and pro rata growth. Our sales and marketing grew 4% in the year before. So what we did was we looked at -- then the question is, why are we doing this? We looked at white space, customers, geography, ways to go to market, where we can make the investment in at a very attractive return. And we think there's a lot of places we can go. So this isn't just right now. We've been doing this during the course of the whole year. But to get the engine going and get everybody in and get them ramped takes time. So Oli mentioned that for the first time in Q3, we saw more of an inflection up in terms of quota capacity growth. And we're going to continue, and that will bear, if it works, will bear fruits next year because you got to get the people ramped up and in seats. And we do it bottoms up. Some examples are -- I'll give you an example. We've been selling to India from Singapore for all these years. Now there's a lot of cloud natives and -- but there's a lot of bigger companies, there's a lot of GSIs. And we essentially went from 0 to having 30 or 40 people in India now, and I think it will at least double that. And it's not just in enterprise sales and sales engineers, it will be in CS and commercial and marketing. It's another example of becoming more local. We were also able to sell from hubs that really paid off. But as we evolve, we learn that we need to be in more local markets. So there's a number of things we're doing that we think are going to pay off.

And David, the translation of those investments to your margin outlook, does that imply that we should be thinking more flat? Or are there efficiency gains you can extract elsewhere such that there's still potential for additional margin gains despite that uptick in sales and marketing spend?

Well, what we said was we gave a 25-plus margin target at our Analyst Day in February. Right at the time, we also had said, I think right after we said, we had soared passed it. We were at 27% EBIT and 30s cash flow. And we said, wow, we've been too good operators. We've been really -- we managed this crisis well. And so we essentially think that we -- there's so many opportunities to invest both in R&D and sales and marketing. We haven't given -- there are a lot of economies to harvest. So the business itself, because of the go-to-market, et cetera, has economies. And it's really our choice of whether we want to reinvest them or even lean in a little more than above revenues. And we haven't given our guidance yet, but I think we're signaling that we're a long-term growth company. We see a lot of opportunities, and we're going to attempt to invest in this next period of time.

Makes sense. We've probably got time for 1 or 2 questions. Okay to take a few, David?

Yes, definitely.

Anybody want to ask one, feel free to raise your hand. I think we've got a mic around as well. Okay, right in the front. A small enough room that you might be able to just ask it out loud.

[indiscernible].

In what?

In India. Okay, the question was, can you describe the competitive landscape in India?

Yes. I think for observability, India has been -- it's not one where there are like in some businesses, local companies, it's the same players that have been in APM, logs and infrastructure. We, I think, got there, as we talked about, later. So there are installed bases in some APM companies that have been there a while. So we're competing very much with the APM companies, the New Relics, the Dynatraces, the Splunks, the AppDynamics, somewhat in open source with Grafana and others and the cloud-native tools. So it's not different than the rest of the world.

I'll ask another question, David. Obviously, since the Trump administration [ won ] the election, there's all kinds of buzz about the federal government getting more efficient, presumably through the use of technology. How are you viewing this as an opportunity for Datadog? And can you remind us what portion of your revenue mix actually comes from the U.S. federal government?

Yes, definitely. To remind everybody, we started a few years ago on our FedRAMPs. We're Impact 2, we're working on Impact 3. We set up our data centers and our distribution. We have a pretty low percentage of our revenues, low single digits in Fed. We underpunch our weight. It might have to -- it has to do probably when we started, but also, as you said, we're dependent upon the modernization of IT infrastructure, and the government may be a little behind. So I think that anything that does shake that up and accelerates replatforming and efficiency is going to complement Datadog. It's too early to know, but I would say we're excited that any industry that modernizes its infrastructure is good for Datadog.

Yes, I get it. Okay, good. Why don't we end -- no. [ Nadi ], time for one more? 38 seconds, but it's going to be a good one.

[indiscernible].

It's not something that we run the company on. We know that the big correlation is on workloads in the cloud. So to the extent that we all see, and we don't see the portion of their reporting that is on the cloud delivery of modern applications, it is correlated. It's just that their data is so homogenous -- so heterogeneous -- varied in stuff always that you can't really see it. But the core cloud workload correlation is in effect.

Good. Why don't we end it there? David, Yuka, thanks for coming to the UBS event. Appreciate it.

Thank you. Thanks, everybody, for listening.