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

At our afternoon slump off is a conversation with Datadog really thrilled to have David Obstler, CFO of Datadog, on stage with me. Thank you for being here.

Thanks for having us. Appreciate it.

So let's get right into it. I want to level set and kind of set the stage with you at the highest level. Year-to-date achievements in the key milestones out of the second quarter in terms of a recap, things that I think would be worthwhile reemphasizing for the last year.

Yes. So in the second quarter, I think we -- we accelerated our top line growth like a number of software vendors, we are being complemented in the investment cycle. We're going through in AI tool companies. We have, and we've always said we have the solution that is being significantly adopted by modern software companies and we thought that like cloud natives, that would be the case with AI natives. And it has been. We said on our call that we have -- that's contributed 10% to our growth and that we have, I think, 8 of the 10 largest AI tool companies. We have more than a dozen over $1 million. We have 80 over $100,000. So in this sort of wave of software companies, we are being used for the observability. We are being used in the same ways that were being used by other companies, which is to monitor their production workloads. In terms of the rest of the business. It's really kind of a false distinction to carve out a part of the software. But we've seen similar trends, meaning we've seen growth that was similar in Q2 as in the previous couple of quarters with strong winning of logos. We gave on the call a number of logos where we've consolidated or displaced. So we continue to have that happen. We've seen a reasonably healthy SMB environment, stable to reasonably healthy, which probably is reflect -- that's excluding AI, which is reflective of maybe a more positive funding environment and growth dynamics in the industry. And we've been successful in scaling our investments, both in R&D and in go-to-market with our go-to-market investments focused on expansion of our footprint, and particularly in some international areas. One last thing. We have given milestones as to when we have on the product side, past certain revenue numbers and -- as you recall, we've given milestones on infrastructure, APM, Synthetics, RUM, logs, et cetera. And this time, we gave a milestone of crossing $100 million in security with very good growth. And at our DASH to conclude on the products, we made a number of product announcements, some of the ones that were exciting, included our Bits AI for service management, which is in private preview right now. There was a lot of enthusiasm in helping to handle cases and resolutions more quickly as well as some developments in data observability. We made an acquisition in product analytics and a number of things around the log business, whether it be the use in SIM or Frozen Logs or denominating logs in different ways for different use cases. So continue to advance a lot in product.

I know we're going to jump on a couple of those different topics that you brought up. But what I do want to peel back the onion on is the AI native customers. I think you characterized this as necessarily a false bifurcation of AI native versus a non-AI native. And I think in your defense, it's you're damned if you do, damned if you don't. If you don't have AI native customers, what are you doing? If you do have AI native customers, let's talk about the slippery slope of navigating the concentration risk, right? So with that in mind, maybe at the highest level, clearly, it has been occupying a lot of investor mind share. And so from that standpoint, are there any helpful perspectives you can share on the usage behavior, the consumption and usage patterns, the product portfolio adoption of the AI native customers relative to the cloud natives. And I'm framing the question this way is because we kind of saw the feast and famine, hot and heavy behavior from the cloud-native companies back in the COVID days, right? So if you can kind of help us give an analog on how these things can go from 0 to 60 [indiscernible] I think that would be really helpful.

Yes, it's important. We're a consumption model used to monitor production environment. So we're not paid unless there are workloads, okay? So one of the big takeaways is that a number of these companies, and you're reading about it all the time because they're publishing their revenues and their funding rounds. They are experiencing significant increase in workloads. So that's the main thing that's driving this. And we're seeing net retention that is similar to what we saw in the highest growth of cloud natives. This is a much smaller segment than that. It's still a relatively small percent of our revenues. But we're growing with them because they're being used by clients, and we're monitoring workloads. The types of products are very similar to our other clients in that we're managing production environments. So it's metrics, traces, logs, then RUM and Synthetics and all of that. I think it's confusing because they are training their own models and all of that. And we're really more production environment. So we're really -- the demand cycle is very similar to other high-growth software companies. We are seeing a pattern where they are committing with us and then they're growing past their commitments. So this is no different than any other high-growth company. We then use the same techniques. We have discount pricing for higher commitments and longer commitments. And so we're doing that. And I think it's characteristic of what the IT world is investing in because we wouldn't be paid unless the workloads were increasing.

And the feverish pace at which a lot of these AI native companies are driving their own businesses forward, you had made a point around net retention rates within the AI native vertical having similarities to the cloud-native behaviors, right? I'm wondering, is the baseline NRR behavior of the AI native cohort significantly above and beyond the company average because that is a metric that you share? Or is it pretty much close to the [indiscernible]?

No, it's above the average and it has to be with the growth of the -- I mean, don't forget most of our revenues come from existing customers. So yes, it is above the average. This is in the growth dynamics like the cloud natives, but a much smaller percentage of the business, 10%, 11%. So it's similar. And then the next question is, is there going to be an optimization. Now I do think many of people in IT and development learn some things. And so it really is dependent on whether that client is managing well their cloud use and optimizing or controlling along the way. So it's difficult for us to know that. Certainly, any time you have that type of significant growth, you could have a period of more rationalization. It could happen. We don't know for sure, but we've, I think, told everybody that could happen in this cohort.

Downsell, churn, all facets, same side of the coin, right? When these large customers who are in very consistent growth experimentation mode, the optimization risk is a faith [indiscernible], right? So when you think about some -- your largest AI native customers and in and of itself, it's a small cohort. How should we brace for a potential churn, downsell and all the way to the other end of the spectrum, a full-on in-sourcing event where they say, hey, I'm just going to -- I'm big enough. I'm going to do this in-house. How are you thinking about that level of risk to the business in terms of the gradient of that risk?

Yes. So as everyone knows, when we give guidance, we don't assume these rapid growth rates. We're heavily discounted. So I want to separate that out. So when we talk to you, we are discounting and risk assessing that. But in the business itself, we are -- full in-sourcing is quite rare, okay? So there are companies that believe in it. You can see from our gross retentions that for the most part, our risk is not about full in-sourcing. And we know that when we see very large tech companies and ones that whose policies are mainly do it themselves, they still use Datadog for mission-critical workloads. We can't be certain. So I think the question is a lot of companies use a portfolio of products. Sometimes they do things themselves, how will that evolve? Now we're trying to work, and I think we learned a lot in what happened in the cloud natives. We learned a lot about how to work with them in account management to try to advise them and get them to help themselves. That has a number of different things. If there are surges, we will tell them there's a surge, you're sending us while you shouldn't be sending us. We're not going to charge you for that burst, but turn it off, and we help them. We even go on site sometimes with them. And so we worked on our own infrastructure with things like Flex Logs, Frozen Logs in order to look at metrics with elements, all sorts of things to try to price in a segmented way versus the cost and not charge grossly the same SKU price for all use cases. It doesn't cost us as much in some. So we're doing that in some of the product announcements and infrastructure. And we have a volume and a term-based structure similar to the hyperscalers that give discounts to customers on volume. So we've been trying not just with the [indiscernible], but across the board to work with our clients to become a long-term partner, and that's some of the tactics we use.

And just a little bit of a non-sequitur. As you think about being proactive in your customers' usage of the platform, ensuring they're getting maximum ROI from more deeply using the footprint. From a sales rep and a sales organization perspective, how is a salesperson incentivized to actually tell their customer, hey, you're actually spending too much here or you could be better suited to utilize these capabilities at a certain rate, Flex Logs, Frozen Logs. So how are you kind of straddling those 2 dynamics where fundamentally the conversation is, hey, use it a little less or use it more efficiently, which is right by the customer, by the way.

So I think one is for a lot of these cloud natives or AI natives, they're not covered. They're covered by our customer success organization who is metric in a way that incentivize them to work with the client in that way. In terms of what we're doing, we are trying -- the objective is to capture more of the wallet share by pricing appropriately. For instance, we know that in certain cases like cloud SIEM, the logs are not going to have to be accessed as much in real time. Sometimes there's compliance uses, other types of uses. So what we're trying to do ultimately is expand our revenues with the client by capturing more use cases and not mispricing use cases that could cause the client to have uneconomic results. So our goal is to optimize this for ourselves over time in winning more and more of the wallet size and having more revenues. And that's what's happened in Flex Logs, where we've captured actually incremental use cases and also not had that type of situation with a client where they're using logs with retrieval that's immediate when they don't need to do that. Those are some of what we're doing.

I appreciate that nuance. Zooming out, we talked a lot about the AI native customers. What about the blue-chip classic enterprise customer that you have and what that AI opportunity looks like for them because the "core business" for you has been workload migration to the cloud. And those are the coattails that you have been very successful in writing, right? So what is that broader AI opportunity from the vantage point of a normal organization, a normal financial services firm and what have you?

Definitely a good question. I mean I think it's still -- I think the migration is still the anchor. And we believe because you need to modernize your stack in order to inject AI that there's even going to be more impetus for the traditional companies. And so we think that's the case. So that is one way we think we're going to monetize. Some metrics we're looking at, for instance, we basically use integrations to get data. And we have 4,500 customers out of our more than 30,000 sending us data from those AI integrations. So what they're doing is they may have call-outs to OpenAI, Anthropic, Perplexity, et cetera, and they're sending us data. So we're seeing signs that it is starting to enter our workloads. That's part of our platform. But still, that's a small percentage. I mean, that's a little over 10% of our total customers. But I think we are little by little seeing that the use of the LLM product, which is now in the many hundreds, still small. And I think as we go from private preview to GA in the Bits AI service management, we'll start to be able to report on numbers of customers using some of these things. So it's happening, but most of it still is in internal use or training or efficiency or call-outs through APIs to the models, but it's happening.

Staying on this topic, there has been a lot of handwringing about the whole commodification, commoditization of the software development process, the software development life cycle, i.e., if I can have 90% of code generated by an LLM, am I hiring as many people? And so the train of logic here is as follows: what is the strength, the weakness, the opportunity and threat for someone like a Datadog where maybe an organization is hiring fewer site reliability engineers or cloud engineers. So where do you fall in that paradigm?

That's a really good question. I think when you're talking about having an enterprise-grade solution, mission-critical that has to aggregate data from everywhere, organize it and then in real time, be 100% reliable for your mission-critical applications. We believe that agentic function is going to be our friend. It's going to help the platform be more valuable. And it's going to accelerate software creation and therefore, the modernization, have more of it flow through our observability. I mean we're not a consumer company. It's not like, okay, fine, if it works, it doesn't. This has to be for security, reliability, privacy, everything. And the bar to have that handled completely agentically is very, very high. I can't say never, but certainly way out in time. So we think it's going to be our friend to both enable our platform and accelerate the complexity of workloads, which will help us.

David, you and I have talked about this offline in the past, but the secular trend around cloud migration, the effects of IT and architectural modernization, those have been fine friends to your business. Do you think the AI appification of the enterprise, enterprise business processes, do you think that adds to the estate growth opportunity in the cloud? So is that more volume for you to capture? Or does that accelerate the pace? So there's a volume argument and then there is a time argument. What is your perspective on that?

Yes. Well, I think there may be some short-term disruption. There may be like figuring it all out might cause the investment cycle to be more distributed between sort of the research projects and the training projects. But long term, I think that it's going to result in more -- what drives us more market share from legacy applications, especially mission-critical to modernize applications delivered in the cloud and enhance both our workloads, the size of our workloads and the complexity of the workloads and therefore, give us more opportunities to monetize those workloads. We don't know the time frame. But if you look at what happened in containers and service list, it brings a lot of the same bells as what happened there and what the art of the possible is, is exciting for us if -- we both monitor anything that comes along. So we have to make sure we're keeping up. And we're putting that into our platform so that nobody can come along and say, I invented a better mousetrap because the mousetrap is right there in our platform. That's our strategy.