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

Good morning, everyone. Thank you for attending. I'm Rob Owens with Piper Sandler, I'm the co-head of Tech Research, and I manage our security and infrastructure software practice. So thank you for attending this morning, day 2. And really happy to somewhat continue the open source theme, I guess, with Elastic here and Ken Exner who is the Chief Product Officer. So Anthony is with us hiding in the front corner if you want to [ meet ] him up later. But really a unique opportunity, I think, to sit down with Ken and understand the Elastic product set and where you guys fit in the world. And I think that it's always, from my perspective, been a very interesting toolkit that's applicable to a lot of different applications. There's a lot of best-of-breed technologies that you go after. So obviously, consolidation play.

And then you have GenAI [ overtones ] as well. So maybe we'll start there, and I think it's just a massive opportunity for the company. Why do you feel Elastic is one of the better positioned companies to benefit from GenAI?

Generative AI. Glad to be here. I'll start with one of the things I talk about internally at Elastic, when I talk about this with our teams. I think with every one of these sort of replatforming events and technology, there's usually 2 types of winners. There's the people that build sort of the foundational technologies, the picks and the shovels and then there's the people that build things on top, like solutions on top. I think at Elastic, we actually have a unique opportunity to play on both sides. So we have an opportunity to be a foundational technology with generative AI with things like vector search and ESRE. But we also have the ability to use these capabilities to power our own solutions to make sure that we can create a transformative experience for observability and security. So I'm excited about both sides. I'm assuming, though, when you're asking about this, you're talking about more on the foundational capabilities. And there, I think we benefited from a couple of things, a little bit of luck, a little bit of history, a little bit of smarts. But we started down this path about 5 years ago. We were one of the first vector database. I think someone actually said we are the original vector database. About 5 years ago, we started adding vector search capabilities to Elasticsearch. And this is because customers wanted to do image search. They wanted to do video search. They wanted to do semantic search. And these kind of things require you to store vector embeddings and query vector embeddings. And we started adding these capabilities to Elasticsearch to support those use cases. And by the time generative AI came about where people wanted to use vector databases to pass context to an LLM mode or generative AI applications, we had a fairly mature offering that not only handle vector search, but all the other things that someone might need. So if you want to connect to different data stores in your business, you have 250 connectors. If you want to run inference on data and store vector embedding, we have that. If you are trying to combine different techniques, like geospatial search or graft reversal together with vector search, you can do that. So I think the combination of these technologies made us sort of put us in a very good position where we were the clear leaders for enterprises because we had all the enterprise capabilities, RBAC and ABAC and logging -- audit logging, but we also had the best search capabilities. And when you look at what people want to do with vector databases, they're essentially trying to do a search operation. A vector database by itself is a fairly simple operation. You're storing vector embedding and querying them. But literally, what are you trying to do with that? You're usually trying to do something like image search, usually trying to do something like semantic search or you're trying to do RAG. And when people realize that, they realize that what they actually need is a search engine because when you're trying to do these kind of things, it's a search operation. You're trying to get the most relevant results to do something, to pass an answer to a customer, to pass an answer to an LLM. And you're going to want to have the most relevant results. I often say like if you're going to build a generative AI application, you want it to give you the right answer, right? It's going to give you one answer. That one answer, better be right. So like what are you going to use? Are you going to use sort of a standard vector search result? Are you going to use something that's been tuned and optimized for your particular use case? And I think that's where we're different. We have all the capabilities to optimize retrieval and get the most relevant results.

That's fantastic. And I did jump the gun, I apologize because you're probably newer to this audience. So maybe we'll have you introduce yourself a little bit. Second question. And just how long you've been in Elastic, your background?

I'm Ken Exner, Chief Product Officer at Elastic. I've been with the company for just over 2 years. I hit my 2-year anniversary just a couple of weeks ago. And before that, I was with Amazon, was with AWS for more than 16 years. I was actually there from the beginning of AWS. So I got to see AWS and it's wild ride and finally left a couple of years ago.

So what brought you to Elastic as you looked at -- you obviously knew the technology.

Well, there's a couple of things. I saw that the company had good technology. But they were playing in 3 spaces that I knew had good opportunity. They were leaders in search, and I saw search and generative AI growing. They we're sort of having the start in security, and I saw the security market was growing like crazy, but they had a really good foundation, which was they had a security analytics platform, a SIEM, that could be the sort of the foundation for building the SOC platform. And then same thing with logs and observability, again, a high-growth space, where -- they hit where the leaders in logs. And they could use this leadership in logs as a way to get into other spaces and observability. So everyone needs logs, every company needs logs, not everyone needs APM, not everyone needs the other parts of observability. But if you can land with logs, you can expand to the adjacent spaces. So I saw a good technology, a good foundation for getting into each of these spaces and the start of a leadership position. So leadership in AI and search leadership in SIEM that could be expanded into adjacent spaces and leadership in logs that could be expanded into adjacent spaces and observability.

Excellent. One more topic that's come up recently is just the change in open source licensing. And maybe you can help explain to the audience what that means relative to Elastic.

Sure. So a quick background here is Elastic was started as an open source company, has always sort of been an open source company. The ethos inside the team, we live and breathe open source. We participate in the communities. We blog about things as we're developing them. So the company thinks of itself as an open source company. But a few years back, we changed a license to use SSPL, which is the same license that MongoDB uses. It's a derivative of AGPL with an additional provision that cloud services can't be built on this. One clause, that's the only thing that differentiates it from AGPL. Because of that, it's not considered an open source license. OSI governs what can be called an open source license and what cannot. And AGPL is an open source license, SSPL is not. So we started to have to use the term free and open. So people could still download it, development still happens in open source. It is still free to use. It is still part of this developer motion that we have. But we had to call it free and open as opposed to open source. The thing that changed recently is we added AGPL, we essentially triple licensed our source code. So that we added this as a new choice for people that wanted to use AGPL and because of that, we can start calling it open source again. This matters, I think, for people that want -- like have religion about this and want to know that something is open source in the OSI definition of open source. And as we look towards the generative AI space, a lot of the generative AI space is happening in local tools, is happening -- developer or practitioner downloads something to their laptop and they start developing there and they play there, and they typically use open source tools. So this would allow us to call ourselves an open-source vector database and be competing very directly against the other open source vector databases. So it was an important change for us to make for that reason. I think if it helps us just a little bit with people that care about that, then it's worth to change.

Excellent. It's often noted your solutions come at very competitive price points. Is there something structural there? Or is this more a pricing philosophy for Elastic?

It's more structural. So I think if you look at other companies, they typically have sort of a portfolio of products that are priced independently. They have a metrics product and an APM product and a RUM product, et cetera. Same thing in security. And they price them differently. What we do is we have consolidation play, but it's a consolidation platform. We are one SKU. So if you are buying us for observability, you are buying an observability platform. If you're buying us for security, you're buying a security platform, one SKU. So in observability, you get metrics, you get APM, you get RUM, all of it as part of one product. And the reason we're able to do this is that we have a unified data store that works with all the different types of data. So it is agnostic to whether it is storing tracing data or metrics data or log data, it is one data store. And what this allows us to do is it gives us the efficiency of having 1 data store as opposed to 10 different data stores and 10 different products and 10 different -- and we can optimize it because we own the data store. So like if you look at what we've done in the observability space, in the past year, we got something called TSDB. It's a time series database optimization for metrics data that allows us to optimize and lower the cost of storage of metrics. We can do that because we own data store. And we are doing something similar right now with LogsDB, it allows us to further optimize storage of logs information. We own the data store, and it's a unified data store across all these different products. So structurally, that's very different than what others do. Some of the other vendors also build on top of our -- so we have the advantage of being both the solution and the data store.

So when you talk about coming to Elastic, obviously, it's a fantastic solution, right, that has extensible opportunities. We think about the same on NextGen Security as we think about observability, but you've got incumbents in those markets that, that's all they do. What's Elastic's challenge to really move into leadership? And you've seen success, but I think you're known as the search company still, right? And you can dominate that angle of it, we all talk about GenAI relative to your search opportunity. But maybe start to weave in the other 2 vertical opportunities. And what it takes for you to really move into a leadership position?

So we are growing quite quickly. According to IDC, we are the fourth largest SIEM and the fastest growing. So the ones ahead of us are the incumbents, the first generation SIEM. So we're like the fastest growing and we're the largest of the new generation, and then same thing in observability, we're growing quickly. I think there's a couple of things that play to our advantage. One, if you look across both observability and security, we've had a couple of generations of products. There was the first generation of SIEMs that happened about 20 years ago, and they've been being replaced by a second generation of SIEMs currently. The difference is essentially the first generation of SIEMs are about collecting data, putting it in one place. The second generation is about doing automatic detections and automated response. Same thing in observability. We've gone through a couple of generations. There was the first generation of monitoring tools to point solutions and then there was the cloud-based observability platform, which is what we have now. I think both of these spaces are ripe for disruption by generative AI. And when I talk about this, I know everyone's kind of tired of hearing AI washing, every one is talking about AI. You go to a conference and everyone's introducing some chatbot or something that -- we have AI too. But there's something different about observability and security. And that's that -- if you think about what these practitioners do, there's a ton of pattern matching that's involved in their day-to-day work and their specialized knowledge that they accrue over time. And the combination of these 2 things make it incredibly ripe for disruption because you can take the specialized knowledge that this person builds up over time and you can teach that to an LLM. You can take the pattern matching and you can have a machine do that. Machines are better at pattern matching than people. So I do think that both spaces will be fundamentally disrupted by generative AI. And it's going to happen in the workflows. So when people first started introducing AI into applications, they did this as a thing on the side, but what you're going to start to see and what you're starting to see with Elastic products is the workflows themselves change. We recently launched something called Attack Discovery, our security product. And what Attack Discovery does is it automates the work that a security analyst does to figure out alert triage. So a security analyst, just a quick background. They'll spend their entire day looking at alerts. They get fired on detection rules. So they'll literally have hundreds of these that they have to sift through. And what they're trying to do is they're trying to figure out which ones are false positives, which ones are real, which ones are part of the coordinated attack. What we've done is we take all these different alerts, these hundreds of alerts, and we feed them into an LLM, and we automatically map the attack path. So automatically figure out these ones are false positives, these ones are actually real, these ones are part of a coordinated attack, and here's what you should do about it. And we're able to essentially take hours and hours of work and automate it for a security analyst. And we showed this -- we launched this at RSA, and we have people like almost crying because they were like you've eliminated like 10 hours of tedious work for me every day. So I think that's where generative AI has the probability to change both the security and observability products is when you start to automate the work that practitioners do and make every practitioner an expert practitioner.

And it's always a bit dangerous to ask the product guy a marketing question, but is there something that Elastic needs to do in your view to kind of unlock those opportunities? I mean obviously, the product set has those capabilities and we talk about consolidation quite a bit, it makes sense, but just to get that leg up because it's a very strong product set. And obviously, from an ROI perspective, offers tremendous benefits.

I think the hard part for us is when people see it, they get it. Like what I just described, like...

Do you think those buying centers are coming together, too, if I start to look at security and observability? Or are they still somewhat disparate [ math ] that remains a challenge?

I think they're still somewhat disparate. There is opportunities in top-down selling to appeal to consolidation. People are looking to save money. But they're also wanting to invest in platforms that are going to be future proof. You're not going to want to invest in something that's going to create even more lock in. So people are cautiously moving towards consolidation. But they want to invest in something that they know is not going to create problems down the road and lock them into something that is proprietary. So I think the store we offer businesses that were completely [indiscernible]-based, that we are completely standard space. It's a very compelling story because if you're going to consolidate, you should consolidate onto an open platform. And like the -- everyone we talk to like, no, of course, why would I want to consolidate onto something that creates even further lock in that doesn't actually help me long term. But in terms of the generative AI aspects of this and -- our challenge is showing people. I want to show every single customer because when they see it, they're bought in. So it's creating those opportunities where we can show this in action because when we show, we win. When we bake off, we win.

But to that end, I think that metrics around ESRE and I am going to use a better than ESRE title at some point, Anthony, just for you, has been very encouraging. Big question where I'm contemplating, and I love your crystal ball is just when do a lot of these GenAI applications move into production? Where do we just really begin to see the knee in the curve versus the environment we're currently in?

I think we're starting to see that. I mean we've seen encouraging start over the last couple of quarters where we're starting to see the beginning of growth for these types of workloads. If you think of -- if you step back and sort of think of history as sort of before ChatGPT and after ChatGPT. So after ChatGPT, it's less than 2 years now. So in less than 2 years ago, ChatGPT launched. And I think what happened is people spent the first 4 to 6 months just sort of coming to terms of what just happened...

We were writing limericks, let's be honest.

Writing limericks, yes. We were trying to learn how to get poetry into [indiscernible] voice. So yes, people were playing. And then it dawned on us like this is going to change everything. So then I think what happened the rest of last year is people have started figuring out what is going to be our AI strategy. Every executive I was talking to was trying to figure out what their AI strategy was. And because they were getting pressure from their Boards and everyone else. Somebody reminded me of cloud computing like around 2010, where everyone was trying to figure out what the cloud computing strategy was. Same thing was happening last year around generative AI. Everyone was trying to figure out what is our generative AI strategy. Creating new budgets, it created sort of an atmosphere of experimentation. Everyone wanted to start prototyping and picking things to start working with. And then coming into this year, I think we started seeing this aspiration turned into experiments. People started actually developing things this year and starting to take it to production. I think this grows a couple of ways. One is, those experiments are going to turn into things that get production -- turn into production and they're going to grow as they get rolled out into production, but they are also just the start. What people are doing is they're picking one thing to start with to learn and experiment. And then with that success, they're going to move on to other things. Like I look internally at what we've done at Elastic. And we started with one thing and we said, let's try doing this, and we built a generative AI application. It worked. It was great. We said, okay, let's build another one and another one and another one. And I think that what's going to happen. It's not just taking that first thing to production that's going to cause growth. It's going to be realizing the success of that and then moving on to other areas of a business and saying, okay, we automated some marketing materials. Let's automate some support, let's automate some legal work, let's automate some other areas. And that's going to create -- one success is going to turn into multiple successes, and it's going to create many opportunities within the...

Are we in a broader GenAI arms race where it's better to be first to market? Or are we at kind of this point of making the right decisions? And as you think about a build versus buy, a lot of capital coming into the space, a lot of companies, what's Elastic's view on how quick you need to be versus how careful you need to be?

I think it's a bit of both. I think everyone spent a bit of time trying to get started quickly and they usually use whatever was available. But then they ran into the reality of whatever they're developing needs to meet their enterprise needs, needs to be compliant with their InfoSec requirements, needs to be enterprise-grade. And I think this is where we often see opportunities, not just with existing customers, who are starting to pick us up because they already use us, but with customers who use something else and then realize, I actually need audit logging, I actually need ABAC, I actually need other capabilities that my InfoSec team is requiring. And the existing products, I have don't do that, or they start with something -- while they're prototyping, that doesn't scale to their needs, and they move to us because they actually scale to a level where they're actually not able to use that existing solution. They're not able to use the PG vector because it creates contention with their transactional database. So they're not able to use one of the purpose-built databases because it doesn't scale to their needs. And then they come to us.

We got time to take a couple of questions from the floor, if there are any. Go ahead, Ethan.

Yes. It's kind of an ongoing debate around what the effect generative AI will have on developer headcount given the productivity increases, right? I feel Elastic is price on a per headcount basis. But given your background, I'd love to understand your view on that debate. Do you think the developer population is going to shrink? Is it just going to grow slower in the future? How do you think this kind of changes?

I think the work of development changes. So I have obviously been following a lot about the developer productivity aspects of generative AI. I view it as like when sort of the modern IDEs came out, they had auto complete. And it was a game changer for productivity for developers because you no longer had to have -- and we used to have to have manuals of all the different function names and method names and have to constantly go back and forth. But you can suddenly have all that auto complete. You can -- not have to remember all the different variables and parameters, you can just auto complete it. It was a boon for productivity, and this was like a 10x improvement in productivity. I think the same thing is happening now where that auto complete is auto completing with entire snippets of code, entire functions that get generated. So it is the next 10x improvement there. But just as the last 10x improvement didn't cause there to be fewer developers, it just created more productive developers, more capacity, I think the same thing is going to happen here. It's going to change the nature of development to be more about design and less about coding, the coding tasks become something that you trust the LLM to do. It's more about designing and architecting. It allows everyone to sort of up-level their game and focus at that level.

I guess one final question. What excites you most in a 2- to 3-year road map about Elastic? This is all obviously moving very quickly, but you're speaking to a room and investors. And 3 years from now, they'll say, if I just would have listened to him then, but if you're looking out on the horizon, what probably gets you the most excited about this story?

I'll tie back to where I started, which is our opportunity to disrupt the observability and security space and our opportunity to be a foundational part of the modern generative AI tech stack. I think both of those are very real. And I think we're the only ones who are thinking and actually starting to show the product features that automate away the tedious tasks of observability and security. And like we're building this next generation set, we're building this next-generation observability platform. So our opportunity there is not to compete symmetrically against the current players on a feature-by-feature feature battle but to show what a next-generation observability and security platform looks like. And people, when they see it, believe it. It is clear that the industry is heading in that direction. And then the same thing for sort of the foundational part of this, the foundational building blocks that we provide. We have an opportunity to start becoming a part -- the critical part of the modern generative AI tech stack. And that's exciting to me to be -- when people go to build a generative AI application, they build it on us because we are the vector database. We are the inference service. We are the capabilities that allow them to get the most relevant results to pass to an LLM. Being that foundational part of generative AI tech stack gives us a lot of room to grow and be relevant for years to come.

All right, Ken, well, thank you very much. Thank you.

Thank you.