Elastic N.V. (ESTC) Earnings Call Transcript & Summary

All right. Well, good morning, everyone. I'm Rob Owens, the Co-Head of tech research and focus on our infrastructure and security software practice here. So thank you for joining us this morning. Really happy to have the folks from Elastic with us, Eric Prengel and Ken Exner, to talk a little bit about the story, kind of where we're at and where we're going, I think. So gentlemen, thank you. Thanks for coming back to Nashville.

Yes. Good to be here.

Ken and I had a near death experience getting home last year, which we were just recounting, as our plane nearly struck another plane on the runway. And we spent 9 hours in an airport together. So hopefully, this year...

We're taking off and we're starting to ascend, and then suddenly the plane hit the brakes. I thought they just ran out of runway. But apparently, we almost hit another plane. So we were on the news and it was a lot of fun, but I came back.

Walked a half mile on the tarmac to get back. So thanks for coming back, and we're not flying this year. So excellent.

Let's go back and chat a little bit about Q1. And I think in the Mongo meeting just before you, Brent pointed out a lot of the inflection that's happening within the space. And I think it happened within your guys' numbers if we look at revenue or subscription revenue. But parsing that there was a price increase which, by the way, does happen with software companies in there, I would love to just kind of understand the different components of success that you guys saw, how you're setting up for the rest of the year and maybe, Eric, a little discussion on pricing as well.

Yes. Do you want me to start with just the Q1 and we can...

Sure.

So I'd say on Q1, we are very happy with our results. We saw both strong commitments and strong consumption. And that's really the way that we measure our business, is both commitments and consumption. The total top line of the business grew 20%, which was great. We've pointed to this metric, sales-led revenue, and that's the subscription revenue excluding monthly cloud. And that portion of the business grew 22%, which was a really nice growth rate for that. We also saw a really strong operating margin. We came in just below 16%. So that was great. And that was really driven by a number of factors. The product has seen a tremendous amount of success. Gen AI is obviously a big driver of the product. And I think that having relevance is even more important. That's what we provide as you think about agentic, where it's not just giving you an answer but it's actually potentially taking action. And so the capabilities that we bring with RAG and relevancy become even more important there. In Security, we saw a lot of consolidation, and that was tremendous for us as you see different components of the security stack, be it SIEM, XDR, cloud security starting to converge. And that's really been a benefit for us. And then in Observability, we saw strong momentum as well, fueled by some of the AI stuff that we're doing in Security with the AI Assistant and other capabilities. So across the board, the product was strong. The go-to-market performed really well. We obviously had some changes a year ago, and we've seen that come back to a large degree where it's executing nicely. And so across the board, we saw strength in Q1, and we're really happy with how that performed.

Now it was a little bit of an easier comp given Q1 of last year and some of the different sales changes. But I think that we're seeing a lot of things accelerate here in the space from an infrastructure standpoint. Do you feel we're nearing that tipping point? And I'd love Ken to weigh in, in terms of where we're at and why.

With Gen AI or more broadly?

Yes.

So the whole Gen AI thing happened a couple of years ago, and I think there's been a huge amount of excitement about generative AI and everyone is waiting for those huge inflection points. And it will happen. There will be significant growth. I think if you look at the last few big changes in our industry like cloud computing and if you look at the birth of the Internet, even the most conservative -- sorry, even the most aggressive estimates tended to be conservative. So I remember when the Internet came out, people were saying, "Oh, this is going to have like $200 billion worth of impact on the industry." And everyone was like, "Oh, no, that's crazy. That's crazy." I think the most aggressive one I saw was $1 trillion. And everyone said, "That's outlandish." But I think 30 years on, we realized that it's actually had a bigger impact. It takes time to get there but it eventually gets to step functions. I was at AWS at the beginning of cloud computing. I was actually at AWS when we launched DC2. And the first 3 or 4 years, it was kind of like the best kept secret. And we're like, why has no one paid attention to this? This is -- like start-ups are using us. And it was probably like around the fourth year that people took notice and enterprises said, "We need to figure out our cloud computing strategy." And it was still another 3 or 4 years before cloud computing actually had an impact on Amazon's business overall. Today, like we know that cloud computing has sort of changed the space in very fundamental ways, but it took time to get there. And then it grows in step functions. I think with generative AI, people realize that it's going to have a huge transformative impact on all industries. And the early couple of years has been about experimentation, has been about trying to figure out what is this going to be used for, how are people going to take advantage of this. I think a lot of the early experimentation has been around assistants and chat-based experiences and copilots. It is interesting. It's good. But I think as people move towards automation, and I think this is where agentic AI is really going to play a huge impact on all industries, because it's no longer just about an assistant. It's no longer about Clippy. It's no longer -- it's about how do you automate things and sort of fundamentally change how people are doing things and drive huge productivity. I'm excited because I think agentic AI is sort of sparking people's imagination about how AI can be used to fundamentally change how we work. So I think as we move into the next couple of years, you're going to see sort of huge amounts of automation driven by agentic AI and that agentic AI is going to be driven by context engineering, and context engineering is at the heart of what we do.

So what are the parallels? And what are the proof points we should be -- you live this ground level at AWS. What are those proof points we should be looking for? Because from an investment perspective, we see cycle times compressing, right? We look at adoption of ChatGPT versus go back to previous cycles, just people using the Internet. It's crazy the momentum these things have upfront, but then there's always that expectation that carries with it and there's kind of that period of disappointment. So to that end, if you're drawing parallels with what you went through that first time and what we're seeing now, what do you think those proof points are going to be then?

I think it's going to be seeing people move from the first experimentation within the business to like broader uses across. Like one of the things we've seen is a lot of the businesses that we work with prototype and experiment in one area. And then as they develop success, it starts to move across the business. That's one. The other is the use cases start to change beyond the first couple of experimental ones. So I mentioned sort of these chat-based experiences have been sort of the canonical example of using AI. Moving beyond that to other types of use cases is what we're -- I think it will be the sort of the signal to me. I think this year, vibe coding and using AI to drive software development has kind of been the breakout use case. Over the last 12 to 18 months, it was chat. Going forward, I'm going to be curious to see how people expand into other use cases of automation. So seeing the number of use cases expand, I think, is going to be interesting.

And what is Elastic's role in all this near term, long term?

So at the heart of what we do, we're a search engine. And why that matters in this is when you're trying to build a generative AI application, you need to pass context to that generative -- you need to pass context to an LLM. It's essentially a search function. Everyone's been hearing about vector databases over the last couple of years. We've been a vector database since 2017. We've been doing this for a long time. But it's not just being a vector database. It's about supporting the entire workflow that a developer has in building these applications, from figuring out how are they going to ingest the data to how are they going to figure out how to chunk up that data, how to run inference on that data or encode it to create vectors. Then, of course, it's vector search or combining it with different techniques. We have a range of different models to assist in this process like query understanding models and reranking models to further get better relevance out of data. You can combine geospatial search together with vector search. And then ultimately passing that context to an agent or an LLM. And you can do that a couple of different ways. You can do that through prompt engineering into a context window. You can do that through MCP interfaces. You can expose things as MCP resources or tools. All of that, the entire tool chain, is what we do. So unlike a lot of vector databases that just do the they're going to store and query vectors, we support the entire workflow.

There's one thing I want to add there that I think is important that Ken kind of touched on a little bit. But as you think about this Gen AI revolution, let's call it a revolution for now. In 2023, a lot of companies started thinking about Gen AI with the advent of ChatGPT and people -- it was sort of this aha moment. A lot of companies said, "Oh, we need to do the things to get into Gen AI." And they started chasing Gen AI. Elastic was working on vector databases in 2017. Elastic had a lot of vector capabilities. When I first joined Elastic, it was before the ChatGPT thing. I remember hearing so much about vector databases, and I'd never heard about what they were before and hearing -- there was a lot that Elastic was already doing. What I would say is rather than Elastic chasing this market, call it luck or whatever you want to, or call it preparedness, which I hope it is, Elastic was working towards this market and building out capabilities that -- where this market is really coming to us versus us chasing the market. And I think that's something that's important to note about Elastic relative to a lot of other companies that are in and around this space.

Excellent. And it seems to be translating into other areas. So maybe you can touch on the Security opportunity, which seemingly is improving around your SIEM offering, even Observability as well, and why customers are looking to consolidate these areas. Because typically, it's -- can be one buying center. But I think historically, it's been multiple buying centers in an organization, different users and just different use cases. So why do you think you're seeing the consolidation and the success that you're seeing there?

So first, let me explain. So in addition to our core search business, we are in the security business. And we got into the security business because threat hunters were using us to search through all their logs and all their different unstructured data to look for sort of that needle in a haystack. So we started being the choice of all these threat hunters for searching for information and all this data that they have in their businesses. On the observability side, same thing. People were using us as a log analytics solution to search through the mass volumes of logs that they have. Over the last few years, that's been the sort of the core of our Observability and Security business, being a SIEM or security analytics platform and being a log analytics platform. But we've been expanding that to a number of different areas. So on the Security side, we've been quietly building up a bunch of EDR capabilities and cloud security capabilities so that we can not just land in SIEM but expand into EDR and expand into cloud security. Our EDR capabilities are actually some of the highest -- actually the highest rated. If you look at AV-Comparatives, they rank us the highest-rated malware detection solution on the market. Same thing on the Observability side, where we've been expanding each metrics and APM. And our motion is we land with logs. We develop a strong relationship with the business there, and we expand into these adjacent areas. So that's going very well for us. AI is playing a huge role in this as well because what we're able to do is we're able to use AI to improve the experiences for threat hunters and improve the experience for SREs and DevOps practitioners by automating away a lot of the things that they do. If you think about what a typical security analyst does, they're doing lots of pattern matching. They're doing lots of manual work. They're sifting through hundreds of alerts a day trying to figure out what's going on in their business. Is this an alert that is real? Is it a false positive? How is it related to some other alert? All of that work that they're processing can be automated away through generative AI. So we've been using AI, our own capabilities to sort of transform these experiences to make it much faster for these practitioners to accomplish their tasks. And I've been surprised at how -- we were leaders in sort of deploying AI in these two spaces, but I've been surprised that no one's been catching up. Like when we launched Attack Discovery in Security, it won Best in Show at RSA and it was a very popular set of capabilities. It was basically taking all the different alerts that a practitioner gets and it was automatically mapping an attack chain, sort of saying, these are how all these alerts are related. No one has still figured out how to do that. And I think a lot of the things that we've been able to do is because we are an AI platform. We understand how to use the context of all their data to build these types of experiences. So I think you're seeing some momentum from people starting to pick us because we figured out how to use AI to really change these experiences.

And Ash loves to say security is a data problem. Can you expand on that?

If you look at what a security analyst is doing, they're sifting through data. Like they have massive amounts of data that they have to pour through to figure out, is there a security incident? They have to find the needle in the haystack. And there's a couple of problems. One is if you're missing data or if you're not collecting all the data, you're going to be missing potential exploits. So the first thing we always say is, "Don't throw data away. Like make sure you're keeping all the data that you -- if you have all this log data, like use the log data. AI can help you process it faster." The second part is, historically, we've always used manual means for how to sift through that data. So the process of threat hunting was getting better and better tools so that you could do the analytics. AI is proving that you don't have to do that manually. You shouldn't have to do all that correlation analysis, all that threat hunting manually. Robots are better at this. Machines are better at this. So I think that's the thing for us, is that if we can deploy AI to do that threat hunting, to find that needle in the haystack and tell you how that needle in the haystack is related to this other needle in the haystack and then to help you remediate it, that's, I think, where we can not only help people move faster but remediate issues.

And maybe shifting gears a little bit, touch on the serverless opportunity. Maybe help explain it to audience to start, where you guys are at in that journey.

This is something we've been talking about recently because we recently launched a serverless cloud offering and went GA on all three major cloud providers. I guess Oracle is becoming a major cloud provider or whatever.

Apparently it is.

So I might have to adjust my talking point there. But for Google, AWS and Azure this summer, the serverless offering is essentially a fully managed version of Elastic on cloud. So if you can think of -- there's three different ways we support customers deploying Elastic. One is self-managed. So you can run it yourself, either on-prem or on cloud, wherever you want to. You manage it yourself. You're just paying for a license. The second is a hosted offering. So this is where we will host it for you on one of the cloud providers. We will provision instances. We will install the software. We'll keep it patched. But it's a shared responsibility model. So customers are responsible for scaling, they're responsible for cluster health, they're responsible for charting, all these other things. Our serverless offering is fully managed. You can think of it as SaaS. It is a SaaS offering. So it is versionless. There's no -- you don't have to think about the underlying resources. It's completely abstracted away from you. So from a user experience point of view, it's like comparing a PaaS to a SaaS, I guess, you can say, which is a fully managed experience. The other aspect to it is it's built on a cloud-native architecture, which allows us to take advantage of a lot of the efficiencies of cloud, including object storage. It's a completely stateless architecture. So we've built essentially a data lake style architecture that underpins this. And this is important because it allows us to run more efficiently. So it's going to provide better margins to our business and it allows us to control costs and lower cost for our customers as well, especially for smaller workloads where it's -- you're paying for what you use. So it's going to be cheaper for customers. And then finally I'll mention is we are also packaging it a little bit differently. So we're able to package a security offering for security professionals and package an observability offering for observability professionals and price it that way. So if you are wanting a security solution, we have a serverless security project that is priced in a way that would make sense to you. It's priced by gigabyte. You pay for gigabytes ingested and gigabytes stored and that's it. You don't have to think about how much memory or how many CPUs you need to use for your workload. You don't have to think about the hardware.

Excellent. And Eric, if I think about the last couple of years relative to Elastic, this is a story that has flirted with inflection many times. And I know investors get excited, and it's been a pretty volatile business. If we look at your recent success, especially this first quarter, are you finding more predictability within the model at this point? Are you seeing more visibility relative to those opportunities, especially with the changes in go-to-market that happened a year ago?

Yes, it's a great question. And just with scale inherently, you're going to have a little bit more predictability, a little bit easier to see what's coming. But there's also going to be things that in a consumption model can be just challenging. And you talked about at the beginning the price increase, and we never got to that. But in Q1 -- well, in Q1, at the start of Q1, we increased prices. And we've done that in the prior year, and price increases are something that software companies all use as a lever to help drive value. And so this year, we increased prices on the self-managed business as well as the cloud business. And the cloud business price increase, we saw immediately impact the business. And so we saw a lot of positivity in that because as you think about a consumption business, it's not so straightforward as just a P x Q analysis. There are different moving pieces there. There are constant optimizations that take place, and an increase in prices will certainly have people rethinking optimizations. Also, a part of the reason we increased the price is because there's a lot of value that we've added into the platform. So as you think about Logsdb, that's something that significantly reduces the amount of storage that our customers are able to use and lowers their total cost of ownership. It's similar to the searchable snapshot capabilities that we released a couple of years ago, which also reduced some of the costs. So at the same time that we were increasing prices to customers, we're also lowering their cost with functionality that we've added. There were additional optimizations that happened. And all said and done, we saw an increase in our consumption from our customers, which means that their spend with us increased, which we saw as tremendously positive in a very positive sense. But just to talk a little bit -- to bring that back to predictability, with things like that, consumption is going to inherently, I think, be a little harder to predict than some of these seat-based models that are more of a traditional SaaS, you pay for your seat. And so as we've gotten bigger, as we've had more experience with the consumption model, I think we have gotten better at predicting it. We've started to build different models in our ecosystem. We're monitoring different things to track it. But I think that it will always be a little bit more complicated to predict a consumption model than it is some of these other models.

Great. We've got time for one or two questions if there are. Go ahead, Ethan.

Yes. Just on serverless, can you talk kind of about the process the customer would go to kind of migrate existing cloud workload or use case on to serverless? Like how hard is it to do that? How quickly do you do that? And...

Yes. It would be similar to if you're moving from self-managed to cloud. So typical process is you create a snapshot of your data and then you restore from snapshot. That said, we want to make this easier. So there's no reason we can't do this very transparently. So one of the things that we're working on over the next, say, 6 months or so is making that significantly easier for customers so that it's a push button. It will happen in phases. We'll come out with some tools and we'll continue to refine it. But I want to be able to, let's say, over the next year, allow a customer to just push a button and it's just seamlessly migrated for them.

I'd love to ask about eating your own dog food or finding...

Drinking your own champagne.

Drinking your own champagne. I like how we thought the same thing.

Where are you finding efficiencies leveraging AI and some of the different tools and some of the different capabilities Elastic has?

Well, I mean, the first thing is that we started using it to change the game for Observability and Security. So like that was the first thing. And I mentioned this before, I was kind of amazed at how long it's taken others --I think they still haven't sort of caught up in some of these spaces to use AI. But internally we use it, too. Like internally, we have a sales ops team that uses this to build sales automation. We have an IT team that uses this internally. Our support team is building all kinds of support assistance using AI. So like the support space is ripe for a lot of generative AI. If you want to really augment a support person, you can do a lot of what they do through AI. So what we're realizing is that you can use AI to automate a lot of the work that they go through, like trying to figure out, has this customer called before for this issue? What's the sentiment of what they're -- are they pissed off? Knowing they're pissed off even before they get on the phone based on the previous sets of conversations. There's all the stuff you can do to bring a ton of information to the support engineer. So we're using all that to really create a great experience for people when they call support because we can look at all the past information and understand a lot about the context for this person.

Has it resulted in tangible cost savings or lower headcount in any of these functional areas that you're starting to use it?

Yes. It's a great question. I think that we're still looking at that. There are ways where we can be more efficient. And so I don't know if we necessarily have taken heads out of the business, but we might have added less heads into the business as we think about being able to be more efficient with some of the ways that we're using Gen AI to really drive productivity and efficiency. And like Ken said, it's really sales, support. Even on the engineering side, we're starting to see some of these capabilities that we're able to leverage benefit us a ton.

Great. Well, I think we'll end it there. Thank you, guys.

Thank you, Rob.

Thank you.