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

All right. Hey, everybody. My name is Koji Ikeda, I'm a software analyst here at Bank of America covering Datadog. I am absolutely thrilled to do a fireside chat here with David Obstler.

There you go.

Did I get it? Did I get it?

You got it.

All right. Got it. Thank you very much, CFO for Datadog. And so we'll just kick it off here, a very, very high level to start. Maybe for those that are not familiar with Datadog in the room, what is Datadog? What do you guys do? How do you think about the opportunity you're going after?

First of all, thank you for having us, and thanks, everybody, for coming. So what's that? It's not going to. There we go. Okay. Yes. So Datadog has an observability and security platform that is used to monitor production workloads for modern applications generally client facing. And we essentially have a platform that then has a number of pillars or components in it. The largest would be infrastructure monitoring and metrics, traces and logs, and this allows the engineers who develop and put applications in production, the DevOps world, to see what's going on in an application and monitor its effectiveness, it's uptime, the capacity needed to provision it. It's a real time. It is used across the department. It's sort of monetized based on infrastructure and data, not by seats. And it's got a uniform data structure that allows us to add lots of different data sources for that environment to see what's going on in the production environment for its applications.

Got it. Thank you. When asked the demand question, it's been very topical over the past 2 months. And so without getting too into the weeds, I understand you guys don't get into quarter update, right. So we'll stick to that. But how would you categorize the demand environment today versus 6 months ago versus a year ago when you say more or less same, better, worse? I mean how would you think -- how should we think about it?

We're a calendar company. So we reported based on the March results and we gave a little bit of a flavor into April, and we have to stick to that at this point. But I think the major factor that we've been talking about the last 2 or 3 quarters has been that when we had the beginnings of the Fed raising and some more economic challenges -- we had a number of customers, particularly in large ramped cloud natives that began to optimize their environment. And that really started happening in the beginning of last year or maybe the end of the previous year, where certain of our clients, either because of the growth of their business or because they were running it had invested substantially in their cloud environments. And we went through -- we talked about this, and optimization in those cohorts that peaked in retrospect, in the second quarter of last year, that was where you had the most intense optimization and then began to heal. And what we've been saying since then, and the biggest factor in Datadog, which is really a usage or consumption-based model is that that cohort that was weighing on our results had begun to complete optimization and grow again. And I think we said in the last quarter, that particular group actually grew. And in fact, some of our largest customers grew at higher rates than -- and more scaled customers than other cohorts. And so that was the main factor in the demand environment. That started in signs of it in Q3 last year in Q4 and accelerated in Q1. In terms of new workloads and new customers, we've had more stability across the customer base. We've had good performance, and we included that in many of our earnings releases in customers adopting data dog and consolidating on Datadog. But at the same time, we've said that we see a continuation of a cost-conscious environment. We're by no means into a bullion. So with that recovery and that better demand environment, we still have some caution in the world some caution in their investments in IT and in applications.

When you look back into the transcripts of Datadog, you just mentioned, right, their second quarter of 2023, call it, reeked optimism. You're actually kind of talking about it before that. I remember when looking at the transcripts. And so it does feel like you might have saw things happen real time, early -- and now you're seeing the recovery real time early. Is that a fair way to categorize how you see kind of the spend environment out there?

Yes. I think we tend to because we're a consumption based -- by the way, we generally sell in commits. We do -- so clients commit. It's not just whatever they use, they commit to it, and then we recognize revenues based on their consumption. So our revenues tend to be a characteristic of consumption model. One of the things in a consumption model is you can see in relative real time how your clients are using your applications, your service. And so yes, I think we have -- if you look back to the transcripts and agree with you that we've been able to comment on what we're seeing in real time. That may have a quicker feedback loop than you might have in other economic models, seat-based, et cetera. So I think that's right. And that's enabled us to tell everybody what's going on in the cohorts and also to give sort of comments on attenuation of optimization trends. Now you're always going to have in the cloud and in modern applications, you're always going to have development of applications and optimization that goes on and we said that since we've been public, but we're talking here about the intensity of it.

Yes. Yes. That makes sense. I wanted to kind of transition the conversation over to everybody's favorite topic, AI. I mean, we asked a ton of questions last year on it. We're going to ask you questions on it this year. Generative AI, 18 months into it, vary from an explosive awareness perspective. How has AI either evolved the way your end market thinks about consumption or building new applications? Do you believe it's a massive tailwind for you? Are you seeing the fruits of that yet? Or is it all kind of really still to come?

Yes. I know we all call it like a year or 18 months, but this has been -- the machine learning in applications have been going on for a lot longer than that. So Datadog in parts of its platform, it's called Watchdog, has basically always architected this to try to use intelligence and machine learning or AI to understand what's going on in the environment. So what we're talking about is sort of everybody piling on on this topic. And essentially, in the history of Datadog, replatforming and new technologies have been a fad. So any time, whether it be Containers, Kubernetes, Serverless, et cetera, that there's been new technologies and new things that need to be monitored and new impetus for the replatforming of applications, it's helped and made Datadog. Now we don't know, and I think our CEO, Oli has said this, we don't know the exact pattern or timeline, but essentially, the injection of large language models into applications over time in production environments if history repeats itself will be a friend of Datadog. And it may stimulate either more rapid application development or more rapid replatforming. Everybody wanted that to happen on Christmas morning on day 1, but it's more of a sort of a pace over time. And so we think it's a friend, but we don't know the exact way that it's going to go.

So thinking about the customers out there and the contracts and you talked about, hey, they sign up, you talked about this a lot over the years kind of for a certain amount of usage for the year they generally hit those markets particularly in higher levels. And so it does seem like a lot of the good results that we've seen from you over the past since the IPO, it is -- doesn't really have AI budgets in there yet. I mean it sounds very positive, but the contracts, the usage trends, the commitments over the next 12 months from now even longer for the bigger customers don't really incorporate that because the ability to up the consumption contract quite easily. Is that the right way to characterize?

I mean, we monetize based on the infrastructure used to run applications and the data or logs that are used to analyze that. And so we don't monetize until the applications earn production. And then our platform takes in as much of the data as we can integrate in, and that's a lot of the investment we make, to try to figure out what's going on. So we would be correlated to not to the use of AI to chat internally, to develop marketing collateral, to test and train, but we would be influenced in applications. And that has not manifested itself in very significant ways in our contracts. There's evidence in a number of places, including a lot of use of the integrations that we built. So we have a lot of customers using integration. We have a number of a cohort of customers who are the tools providers whose whole business is providing AI tools that has grown very rapidly. We said 3.5% of the ARR. So there's evidence that there's activity but not mass adoption yet that has manifested itself into a change of the workloads. We actually don't care. We sell based on sort of capacity you use. And so any way the client wants to use that, whether it be to monitor LLMs or do logging or look at network or databases would be part of our contract with the client. So I think we're early days, and we don't know how it's going to go, but there's good evidence of activity in these different directions.

So on that point of the 3.5% of ARR, you are one of the few companies out there that are calling out a percentage of ARR and for yours it's specifically digital or generative AI native.

That would be the tools provider.

Tools providers, right. So you also mentioned that it's generally inference, that's part of that 3.5% in production and inference.

That would be very similar to Datadog use in others that would be monitoring the delivery of that application to its clients. In the same way, metrics, traces and logs, so that is an activity metric based on a set of customers getting a lot of activity and growing, and it's a pretty broad set of customers where Datadog is monitoring their production environments just like we monitor all the other applications. So it's much more similar to what we do across the board. Does that make sense?

Yes. Yes. And I do know that you've stated publicly that these companies, you kind of sit through the data and say, okay, this is clearly an AI-native company, tool company, we're going to -- does it become more difficult or, I guess, visible to you in the future as more companies have AI applications and production to truly see within the business model, where the inference is coming from? Is that...

Definitely. There's no question that this is a spaceholder, and we haven't promised to give this forever. Basically, the money will be the biggest part of the money. And I can go to some other places we're investing will be when we're monitoring LLM within client applications, and it's tough to tell because we're not running those applications. We're monitoring it. So we will do our best over time to understand the effect of the injection of LLM into the applications, but we may not even know for sure. We're essentially trying to see. But I think that's where you have not just a set of tool providers but you have our customer base, our 26,000, 28,000 customers using the functionality in the platform to monitor LLMs within applications. And that really hasn't happened yet because most of these mission-critical applications have not been put in production yet with the LLMs. It's more sort of testing. So I think we've been pretty clear to say at that point, when you have inferencing and production, we may see that manifests itself mainly in our workloads, but it hasn't happened in a material way yet.

Yes. Yes. No, that makes sense. I do want to press you on this a little bit. It does seem like there's one, a positive, you're able to call out some sort of percentage for generative AI, but also somewhat of a predicament because it becomes less visible in the future. And so because you have this data point out there, I mean, it does feel like that the community is becoming accustomed to getting it, and it does feel like it may be removed at some point or just talked about less frequently. So how do you anticipate in the future, either quantitatively or qualitatively describing your, I guess, participation to generative AI.

Yes, we hope to follow that everything we do is don't forget, monitoring reservability of what our clients do. So we hope that we get more data about what our clients are doing and either qualitatively or qualitatively can express that, that may manifest itself in more hosts, more GPU host, more -- it could also we have a product in public beta, which is LLM reservability. And that is meant to basically have a module, which allows our clients to look at the functionality and -- of the LLM models within an application, we may get signals from that that we can express. Right now, it's in beta, right? So that would be like we have with other product lines. If you go back to our most recent script, when we talked about the amount of ARR that we got from newer products and we gave a number of examples, we talked about cloud costs, we talked about database, we talked about a number of other products. It may be that as it develops further, we're able to have metrics like that in terms of modules of our platform that we can deliver and particularly when we have those in GA for our customers.

So when I think back to DASH last year, one of the more exciting announcements that we thought was your integration with NVIDIA down to -- I think it was the chipset, you kind of go straight in there. Is that something -- I mean that definitely feels like it's an AI indication? Would that be -- I'm not saying give a metric today, but would you be able to have that type of visibility with that type?

The metric we gave was over 2,000 customers are using our -- those types of integrations. So we gave -- if you look back to the script, we've been -- it's a difficult thing because we are -- and I think we maybe are doing it more than other companies, we're trying to give flavor without having that specific -- this is the use of this module. And that integration was -- that comment on integration, both last time and this time was really about what you just said which is clients using our integrations, our deep integrations and it's an activity monitor. So we haven't figured out exactly what metrics we're going to use, but that's another type of metric that we've chosen to put out there to give evidence of traction.

Okay. Okay. I'm going to ask you one question here. Not on AI, a different question. And then I wanted to open it up to the audience to see if there's any questions out there. We do have someone with a microphone, please raise your hand, and we'll get the microphone over to you for the webcast. And -- Okay. So next question before I open it up. Big deal activity, right? Every quarter, you guys talk about some pretty big deals, 6-figure deals, 7-figure deals, 8-figure deals. I mean how have these deals come to be. I know you guys are land and expand, but is the shape of the deal flow changing? I mean are customers coming in bigger? Are they expanding larger? Is there a certain type of customer whether it's enterprise, mid-market or certain verticals that you're seeing better demand out there to drive a big dealer?

Yes. I would say it's still land and expand. It would be finding use cases and finding workloads. Landing, we land in the vast majority of cases, they're landing with not just infrastructure, but 2 of the pillars. That's the majority. And we -- so most of those deals have been expansion deals where we are either covering certain of the pillars or certain business units. I would say they tend to reflect, as we've talked about over time, more consolidation as our products have gotten stronger and not only the platform, but the individual products have gotten to parity or beyond. We've had a flow over a number of years of consolidation onto the platform. So we give some examples of that. And so that may be something that's developed as we develop the products. We also have sometimes -- it's not the vast majority, we will land bigger because they're doing a rip and replace, and they're adopting us. At the same time, they're consolidating. And so you see some of those come in, but that's not the vast majority. So I would say the motion hasn't changed in that we do tend to land bottoms up. We do tend to develop use cases and then expand either in more pieces of the platform or more business units over time. It really -- whether you call it an enterprise or mid-market or an SMB, the clue is what are they doing in the development of their digital businesses. and what decisions they're making to consolidate. And so we have examples in that of very large traditional enterprises and cloud natives. I would say we've been clear over time that traditional enterprises in some very traditional industries, automotive, manufacturing, things like that, have been behind on their journey. And when you see a lot of those examples, and we give the industry, you see that those are examples of those traditional industries catching up and those are larger companies, so they tend to evolve to larger deals faster than smaller companies.

Got it. Any questions from the audience, please raise your hand, and we'll walk the microphone over to you. Questions?

So 2 questions. One, first, from an AI perspective. Do you get any feedback as to whether or not activity is picking up in terms of the tools that your customers are using? Or is it sort of stabilized at this point with people just doing proof of concepts. And then the second is there's a very large theme over the last couple of weeks that the software market has matured significantly, it's now slow growth. And if you're seeing signs of improvement, are you seeing signs from your traditional installed base that's actually coming back or are you seeing potentially market share gains so that it's not necessarily a sign that the overall market is coming back, but that you're doing better?

Yes. So the first one, I think we said that we believe a lot of the initial work is being done in training, internal applications, sandbox and gave some examples of use of integrations, which means that there's activity, but it's too soon to comment on activity in this quarter versus that quarter, it's still fairly early on. In terms of what's driving the business, the business has basically been driven through a combination of expansion of use of the platform and an additional pillars or modules. And the pillars and modules have been for quite some time been market share gains. And we know that because if you look and follow the public research or you look at the kind of ARR expansion that we've been getting every quarter, it's been larger than our competitors. And that's been correlated to both the popularity of the platform, but also, as I mentioned, the investment we've been making in different pieces of it. So I think that those have been driving the business for quite some time. We've also had, I would say, the thing that changed is we had maybe overexpansion in the bubble and then optimization, which weighed. We said we think we're back more to normal. And the new business has been relatively stable for the last -- to this whole cycle, meaning there's enough new projects and logos that it's been relatively stable in terms of what we've been getting. You could argue that it hasn't gotten to an expanded level because whereas it's been out there, we haven't had the cessation of the weight of the Fed and the economics. We don't know, but it's been relatively stable.

Maybe one more question from the audience if there is one. David, I wanted to kind of ask you a question related to that on buying patterns from your customers out there and it can be both for new or -- I know a lot of your customers when you speak with them, they always tell me, contracts a year, but we're nowhere going to run out at some point. And so there seems to be some sort of runway to remind, some sort of pattern in there whether it's 6 months, 9 months, whatever it may be. Have you noticed that runoff time change over the years, meaning was it a much shorter time call it, I don't know, 2018 to 2020. And now it's longer runway today. I mean how would you...

So we tend to sell sort of on capacity plan that tends to be conservative. A lot of these are new applications and because it's cloud and bursty. And we run our own company the same way. You generally don't buy for the peak, you buy under and then you're willing in capacity planning to do some on demand, and that really hasn't changed. And you're right. Most customers -- and we don't force it. Like we try to land at what the client wants. We're confident in expansion. So we don't force them or even incentivize them to overbuy. So we still have that motion of them staying short. When the net retention is lower, which it has been than at the peak, you would have the time on the -- to the ramping to the use of the capacity be slower. And that's purely a factor of the net retention and the growth. The net retention being the growth of the customer in using the platform this year compared to the previous period. So yes, it takes longer, but the overall motion of staying short and getting to a new commit hasn't changed.

Okay. Okay. And I guess the follow-up there would be thinking about pricing trends in the market versus the competition? Are you seeing roughly the same pricing trends out there? When I look at observability, specifically, there's been some movement, there's been some consolidation out there. What are you seeing specifically that you could comment on with the pricing trends?

Yes. I think our pricing has been very stable. Sort of optimization has been more a unit or consumption optimization. So our pricing is pretty stable. As our customers have grown, we've always priced based on discounts on volume and somewhat on term. And I think you're right. You said that if a client, and for the most part, we have very high gross retentions, so the vast majority of our clients do feel this way, know that they're going to be using Datadog long term, we have had, as we talked about in Q4 and Q1, an extension of duration, a trend towards more 3-year contracts where they do get a little more of a price break. So -- but it hasn't changed the overall weighted price because we have the next generation of customers coming in smaller and paying the higher unit price because they're not either ramped yet in terms of overall units or duration.

Okay. Okay. You are the CFO, I'm going to ask you 2 CFO-y question. One on cash flow, the other on stock-based compensation. I don't think that's -- might be coming up more and more for you from an investor debate perspective is cash taxes. You guys are profitable so how do we think about cash taxes this year, next year or into the future?

Yes. So we said that this year, we're not going to be a significant cash taxpayer. We still have either enough stock-based comp or RSUs or NOLs. We gave out an effective tax rate. And what that can allow you to do is we've had cash flow higher than operating margins. And part of that is due to the fact that we don't pay cash taxes. So if you think about just to use round numbers, the low 20s, 2021 and you think of the operating margins, you can see that I'll just use a round number, 25 in 20, that's basically 500 basis points, right? So you would have essentially when you -- if you are going to be a cash taxpayer, your ratio, all other things being equal, they may not be equal. We may have a different working flow, you're going to have somewhere between 300 and 500 basis points. We do pay some taxes. So we all -- we said all along that we don't assume we're going to pay, but it's likely that the cash flow margin and the OpEx margin all other things being equal, start to compress a little more once you become a cash taxpayer. In terms of -- your second question was?

Yes. So stock-based compensation. When I look at the model yes. Sometimes I'm like, oh, that's a lot of stock-based comp. So how do you think about stock-based compensation? Because I understand you guys definitely are R&D focused and there is a component of that to drive product differentiation. So how should we be thinking about stock comp?

Yes. I think we essentially never went crazy. So if you look at the amount of stock-based comp, we never went off the rails. We've always been more like 2% to 3% dilution in stock comp, and that's really what we think about. We understand that over time, that we need to sort of bring that down. But the most important thing is to stick within that dilution as well as to hire and retain the right people for growth as we think about it.

We're all out of time. Thanks so much. Thank you, everybody.

Thanks a lot. Thanks for having us. Thanks. Thank you.