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

Welcome to the next session. I'm really happy and excited to have the team from Elastic here. It's timely conference for us actually because you've delivered really kind of strong results last week.

Maybe Ashutosh, let's start with you, like, what were the highlights from your perspective?

Yes. So like I said, the results were quite strong, 17% growth, cloud grew by 31%, 13% operating margins. Just when I step back and think about what really drove the business in the quarter, I'd say there were 3 things that are worthy of calling out. So first is we continue to see momentum with the work that we've been doing with Generative AI. I talked about -- in Q1, I had called out that we had hundreds of customers that were using our Elastic Search Relevance Engine, ESRE functionality for vector search and Generative AI use cases. We saw hundreds more in Q2. Great momentum with customers, both at our conferences but in general throughout our field organization in that area. The second thing that we saw is the play that we've been running now for several quarters. On consolidating workloads onto our platform, we started to -- we continue to see good traction with that. And some of those workloads also started to ramp up. Commitments that customers have made in the past, so that also contributed quite nicely to revenue. And then the third thing that is worthy of calling out, especially in the cloud was that the optimization trends that we had seen that started about 1 year ago, a little under a year ago, they started to -- they seem to have stabilized. I talked about that in Q1, but we saw that again in Q2. And the sense that I get when I talk to customers and we're looking at the telemetry is that, customers generally seem to have reached where they want to be. So now it's about them bringing more workloads on and that's what we're focused on.

Yes. And before I kind of -- given you worked in technology for quite a long time and have seen a lot, Ash, I wanted to kind of ask more bigger picture. If you think about Generative AI and what we understand about it now already, can you maybe talk a little bit about the role of data because you guys are a company that kind of has been dealing with data kind of all your life and working with data. Like how important is data and clean data et cetera for Generative AI?

It's incredibly important. I mean if you think about the 2 things that you really need to build any kind of Generative AI application, you need the large language models. These large language models, they all have been trained on some data sets, these are inference models, right? So they're coming up with, they're inferring, what's the right next word and then the right next word and so on. So you need the LLMs. But they've all been trained on publicly available information. And if you want to use them in any business context, what you need is, some way to ground that large language model in your business context in real time. And so the second thing that you need naturally, is a platform that will let you do that. That requires search, that requires search with the right context, with the right relevance and that's where we come in. So the data itself, all your private business data, we see as customers have so much of that on Elastic, our ability to now provide that bridge is something that is very exciting for us.

And then without going too technical but like what are the tools that we think about -- that we should think about like that you are kind of providing to be able to deliver that?

So there are a few aspects to this. So the first is the vector database functionality or the vector search functionality that we provide within the platform. And we've built that core into the platform itself, like deep into the platform and there are some really nice benefits because Elastic was always -- Elastic search was always written to be highly performant and highly scalable. We have customers that have literally tens of thousands of nodes of Elastic running in a horizontally scalable manner and now our vector database can scale in exactly that same way because of the way we've built it. So it performs incredibly well. It scales incredibly well. The second big thing that you get, when you get our ESRE functionality, the aspect of Generative AI that I talked about, is everything else that you need to build a true Gen AI application, right? It's not enough to have vector search because what people are realizing now is you need different kinds of search techniques to build the most to get the most relevant answers. And there is a term that is now incredibly popular called hybrid search. And what hybrid search refers to is, you use vector search but you also use semantic search and you use text search and you rerank the answers that you get from the different techniques to get the best possible answer. And you need to do it with additional context like geolocation. So if you ask a question and you're based in Canada and if I ask the question and I'm based in the U.S., there might be a reason to get slightly different answers because if you're talking about vacation, policies and so on that might be different. Personalization needs to come into picture. Document level permissions needs to come into the picture. Your access might be different from mine. So we provide that holistic set of functionality along with our own embedding models. Like so we've done a lot of work even in building all these kinds of embedding models for machine learning jobs to create these vectors. The third thing that becomes really important is the APIs to connect to any and all large language models. So look, we believe in choice. We -- our customers keep telling us that they have tried using different large language models and they get different kinds of results, depending upon their domain and their use case. So in our platform, we try to provide the ability to connect to OpenAI, Llama 2, Cohere, Vertex, Bedrock. Like we just want to be the open platform that lets you choose whatever you want. And then do it in a platform where the fact that you already have your data in Elastic just becomes a natural advantage. So we are building all of that ease of use directly into our platform, so now you don't have to worry about learning a new set of APIs. We've built the integrations with technologies like Hugging Face and LangChain that are quickly evolving to be part of this new AI stack.

And then as we think like there's obviously, a lot of enthusiastic -- they are asking why people are excited about AI is when you start -- new start-ups coming and they kind of having apparently the all next and best tools, et cetera. If you think about vector database in your offering, like in the market with a product, et cetera, like is there a way to think about like, some of the private competitors will say like, oh, we have, like dimensionality, and we have like so much dimensionality and whatever, like if you think about that it sounds yours is more holistic because it's kind of -- it's not just a vector database, there is a lot more that needs to kind of play in there. Is there kind of a -- from your perspective, like a playing field or a tick box that you need to have to kind of deliver all of this?

When you think about what it takes to build an end-to-end application, you quickly start to realize that vector database is a feature, not a product because you have to, first and foremost, bring all your data into that platform, then you need to turn it into vectors and that requires embedding models. And you need to build all that pipeline. Then you need to create into vectors. You store those vectors in the vector database. Then you need to have the ability to query that vector database and get all the right results from it. Then you have to worry about, like I said, hybrid search because oftentimes, you're not going to get great results from vector databases. So then you need to apply other techniques, for semantic search or textual search, rerank those results, the connectivity with the LLM. There are so many aspects that if all you're using is a vector database, you're going to end up assembling something with 10 other tools and then building all these integrations yourself. And the reality is that the vector database vendors, the pure plays that emerge, there was a lot of noise in the market, which is pretty natural for any new technology. But the majority of the successes that we have had, all those customers started with some vector database. Because -- I mean we've really only been -- we've been working on our vector technology for the last 4.5 years, but you only saw us really go into market very aggressively in the last -- since June, and when we made our ESRE announcement. So most customers have been playing around with something else and recognize that there's so much more that you need and we provide it all and we have an amazing vector database itself that has, people throughout things like dimensionality and we have more dimensionality than what any embedding model needs. So it's sort of a -- it's a fake premise that you'll often hear about. But that's -- we're in a place where we're actually winning incredibly well in the market because of it.

Yes. And it's -- that you see that in the customer understanding, customer conversations as well?

Yes. People have -- look, the first 3 months were chaotic. Everybody is trying to figure out which side is up. And now there has been so much intense activity on it. We participate in public benchmarks, like we're very open, like our heritage has always been open, right? So whatever we do, we'll talk about it publicly, we'll publish benchmarks and so on. So it's become pretty clear that we have been innovating at a very fast pace. And everything that you bring to the table in terms of the breadth, that's what customers are recognizing that they need and that's what's driving our success.

And then Janesh, over to a little bit. The -- on the -- if you think about it, it does seem like AI, ESRE, et cetera, is helping you on the kind of search side of your business. And that's kind of something that with observability and security, we haven't really talked that much about the class, hope I'm not in something you guys have said the classic search side. Can you remind us like how -- like, did you break it out, like how important was search? And am I -- do you see like a revival of that part of the business right now that it's becoming very, very crucial in this new AI world?

Yes. Conceptually, Raimo, you're thinking about it the right way, right? If I think about the personas you've got the DevSecOps persona that buys observability and security solutions and you've got the Aptiv persona where they're getting -- where the technology is being used to build Gen AI applications. And we've talked a lot about the building of Gen AI applications. Even on the observability and security side, we do get used quite extensively out there in general. And with AI, what we're starting to see is that we're using our own technology to build additional capabilities like the AI assistance that we've talked about for observability and security. Security is already GA, where people use those to actually improve the capabilities of the outcomes that they're -- or to improve the outcomes that they're getting in the business. So it's actually helping quite a bit and those features are available only in the highest tiers. And so somebody wants to access those features and use those, it requires them to use the enterprise level of the subscription tier. So that's the way we see it playing out.

And then the -- obviously, we're still early in that journey, but like and -- but I've also seen you guys kind of being very active with clients. There were a lot of events from you guys. Has that started to translate into like customer conversation, pipeline building, et cetera?

Yes, significant. And if I think about our engagement with the community and we think about the ways and various reasons which we interact with them, one of those is our Elastic{ON} Tours, which we do round the world. We had an -- our Elastic{ON} AI Tour stop in San Francisco a few months ago and since then we've had sessions in Amsterdam, in Frankfurt, Bangalore, more coming up around the world. And in every one of those, we are seeing hundreds of customers show up, lots of them participating in sessions on stage talking about how they are working with us, how they are using the technology partners are there. On the sidelines of those are, sales people are having conversations with customers. This was front and center on -- for every senior IT executive and every enterprise account that we have.

Yes. And then how would that translate into -- sorry, that's the kind of the terrible financial question. How does that translate into revenue? Is that more platform usage? Or is there an extra -- do you pay extra like...

Yes. I mean our monetization model is two-fold. One is, for certain features, people will need to use higher subscription tiers. So for example, our machine learning features which you would use to then train ML models and so forth, those are available only in the platinum tier. If you think about the AI assistance that I talked about, those are available only in the enterprise tier. So depending on what features a customer wants to use, they would need to be in the appropriate tier. And the second piece is, just the compute intensity of many of these workloads that causes the meters to spin faster. And then that drives higher consumption and then that drives higher revenue for us over time.

Okay, yes, yes. Okay. But it's like -- it does feel we're still early on that journey. Are you kind of...

Yes. I think it's still relatively early. If you think about how people are building applications, as Ash said, for the last several months, people were still trying to figure out, I love that expression, which way is up. And now they are picking their -- making their choices and taking Elastic to be their partner of choice, and they'll build the applications. The applications need to be deployed. The applications need to be to scaled. In general, I think what you see when you have major shifts in the industries, people tend to underestimate how big they can be but they tend to overestimate how quickly they'll play out. And I think AI will be similar. It's going to be a significant opportunity for us but we're still in the very early innings. We're very pleased with the progress we've had so far but I do think it takes time to ramp and we are super excited about the future.

And then does it -- how does it play into the client side or self-managed versus Elastic cloud because as you said, there's a lot of compute, there's a lot of new stuff in there. Does that changed customer thinking about where they do -- where they can utilize it.

I think customers can use the Elastic search platform, both self-managed and cloud. If you think about large language models, because of the size of those models, those tend to obviously be cloud-based led by the hyperscalers. But if you think about where we come in, if a customer has got their enterprise data in a self-managed form, by the way, self-managed doesn't necessarily mean on-prem. It could be deployed in a cloud and managed by them in a cloud in a different kind of setting. They'll tend to continue to use us where they are. But over time, I do think we've got a lot of natural shifts, more workloads start in the cloud, more expansion happens in the cloud. So we have seen significant progress from a cloud perspective over the course of the past several quarters now as we continue to grow the business. And I continue to think that cloud will continue to grow faster than the rest of the business.

Yes. Okay. Ash, back to you, I didn't want to have like the whole conversation on AI. In terms of like the advancement around platform, what you're offering, like can you talk a little bit about the other work that you're doing, I'm sure you're not doing 100% R&D on AI now, yes?

No, no, no. So AI has obviously been a big part of what we have done but there's so much additional work that we've been doing in other parts of the platform. So one of the things that is very exciting for me is the new query language that we've delivered called Elasticsearch Query Language, ESQL. This is a piped query language. And just to give you some context behind that, we've always had very rich query functionality, search functionality, visualization functionality in the platform. But one of the areas that we often got feedback from customers on was, when they compared it to something like a Splunk query language where it had some nice semantics of being able to pipe the results of one stage of the query to the next stage, they found that to be quite useful. It made it very easy for developers to do very iterative kinds of query building, testing, et cetera. And they found that to be sort of a sticky feature of Splunk's platform. And as we have been doing more to consolidate workloads onto our platform, displacing incumbents, we felt that it was really important to deliver that functionality. So we've been working on that capability for some time. It's not only a new query language but a new query processor effectively that does this incredibly fast and it allows you to do piping. It allows you to do lookups. In the future, it will do even more enrichment with joins and so on. And so that piece in data, we've got tremendous feedback from customers, lots of excitement. And some of the wins that we have seen recently where customers have moved on to our platforms, displacing incumbents, have come because of some of these innovations. Another very interesting thing that I'm excited about is the work that we have been doing on serverless. We haven't been talking too much about it but it's in gated preview, and we're getting wonderful feedback from customers and we'll keep everybody updated as that goes from gated preview to open beta and then to GA. We are focused on making the platform easier, more cloud native, making it possible for you to support different use cases powered by AI. Like there's a lot that's been going on. And that's what gives us a lot of confidence.

And the -- you mentioned the new query language, which kind of closes the gap a lot to some of the -- to one of the incumbents. Do you -- in your market or in your broader market, especially observability, there has been a lot of consolidation of late or consolidation in a way players got taken out. And usually even people get taken out there's disruption in the market. Have you seen an impact already? Because there's only -- there's not many -- I remember I used to write big notes about the observability war and everyone is coming there. And now the field is getting smaller and smaller, and you're like one of the last guys standing there. Have you seen a positive impact already from that?

So we started to see increasing interest about 1 year ago. But it was not because necessarily because some of the consolidation that is now happening. But really, lot of these incumbents were notorious for pricing models that weren't customer friendly. Just the overall costs associated with operating their platforms were exorbitant. But when you have something that's sticky, customers need a really strong reason to consider displacing something. And what changed was just the quantitative tightening that we saw. And the budget is becoming tighter and people getting more worried about what they have to implement with flatter or reduced budgets. And our platform, we have built a platform over the years that was -- that scaled incredibly well, that was faster than these incumbents, was cheaper to operate at scale. But when you have something that's established for 10 years, ripping that out requires more than that. And the economic pressure started to create that incentive for a lot of customers. So we started to see more and more activity. And that's why in the last few quarters, we've been talking about the fact that customers have been making commitments to us. Moving existing -- moving workflows onto our platform, displacing incumbents. So we have started to see this now, that's why we doubled down on things like ESQL, which takes away the last remaining hurdle, if you will. In terms of the M&A activity itself, look I can't say that I've discretely seen somebody come to me and say, because of that, I want to switch over right now. But this has been a trickle that's now has been building for some time. So I feel very good about not only what this means for us in the out years but we are leaning into it.

Yes, yes. And then you talked about like people are aware of how much they're spending and that brings me to the cloud optimization part. What are you seeing there in the field? And the question -- it's kind of slightly funny to ask you guys is that, because of your business model, you were always kind of like it's cheaper the right word, like I could predict more cost -- it was always a cost of a month to go with you because it's like they way you priced it was always an advantage. Like did you see kind of the cost optimization, like as much as the other guys? And where are we on that journey?

I don't know -- I don't want to compare between us and other people necessarily but you're absolutely right that when it comes to like the value that we deliver for a price -- were very advantageous to the customers. And that's why we started to more and more successes, but it's got tighter. What I'd say is that, in terms of the optimizations that we've seen, that's kind of like stabilized at this point. So I believe that our pricing model is going continue to be an advantage for us in the future.

Yes. Okay.

I don't know if there's anything that you...

No, I think that summarizes quite nicely. Yes, go ahead.

Like on that note, like if you think about it, the one question I got like, the results for the quarter were really good. And then people were -- but if they were so good optimization over, how do I think about guidance? Kind of maybe talk a little bit about your thinking there?

Yes. The -- I'll go back to the themes that Ash mentioned at the start that we saw in the quarter, right? One is we started to see some initial contribution from Gen AI and lots of customer activity engagement, but some very early contributions in terms of revenue. We also started to see -- we've talked for some time now how our customers have been making commitments to us. And we start to see people ramp against their consumption, against the commitments that they've previously made. And the third theme was around the optimization trend and that sort of leveling off. And so as we thought about all of those and started to think about the future and how we think about guidance, the reality is that consumption can fluctuate from time to time and we've seen that certainly in the last several quarters in the industry broadly. And that's just something that we wanted to keep in mind as we build the guidance. So it's not like we've seen anything dramatically different in the business that we are signaling in any way. But we just wanted to be -- we wanted to be thoughtful and make sure that to the extent that there is some level of future fluctuation that we consider that as we build the guide. So we were just being prudent about that.

Yes, yes, yes. It's more prudent, yes. Okay. Make sense. I mean you -- at the end of the day, you have to deliver the numbers.

And consumption can fluctuate. So we just wanted to be careful.

Yes. Yes, yes, yes. Okay. Perfect. Makes sense. And then let's change over to the costs and investments a little bit. Like we had some players that started to think more about like, okay, how do I think about investments now. We are not going to be in the downturn forever. There's always like a lead time in terms of investments. How are you thinking about that the margin situation? You've delivered really, really strong margin so far this year?

Yes. No, we appreciate that. It's been a lot of hard work from a lot of people to make sure that we deliver those numbers. But we fundamentally think about the margin on a full year basis. So we're very happy with where we are. On track to deliver the numbers that we've committed to for this fiscal year. And fundamentally, the approach that we've taken is to continue to invest in the business. But to invest in areas where we see the best opportunity to continue to drive growth. And there is very natural operating leverage that is inherent in our business model. So as we continue to do that, we are making investments in the right places. I think on the Q1 call, we had talked about shifting some of our investments towards Gen AI and R&D and in marketing. And we've continued to do that as well. We are making sure that we invest in the right way in the go-to-market functions to make sure that we build enough scale and capacity going into next year. A lot in the R&D areas, in many of the areas that Ash talked about. But all of that within the overall construct of making sure that revenue grows faster than overall expense. And that will give us the operating margin expansion that we're looking for.

Yes. That's true. And then it's funny because like this time last year, you were here at the conference and we talked a little bit about like how you budget and how fundamentally you're changing a little bit how you do that, how it's more returns focused, given the higher interest rates environment, et cetera. Is that still playing out? And do you think is that like the new modus operandi, like how you do this going forward as well?

I think in general, yes. What we want to be is, be disciplined about how we're investing in the business. And we've got a lot of experience with -- in many different areas. When we hire salespeople, we've got our internal models around productivity and how we want to see those people ramp when we make investments in marketing, when we make investments in engineering. So as long as we're doing the right things and seeing the right results, if we fail it's okay but we want to fail fast and make sure that we [ cost ] correct pretty quickly so that we can then continue to shift dollars to the right places. And that's been the approach that we've adopted. It's worked quite nicely for us this year.

Yes. And then you don't have to answer, but it's more conceptually how you think about it and it's more for the industry. The #1 question I'm getting all the time is around -- okay, so we're having a revenue suffered. We came -- like everybody in the industry came down to certain level. And so now we need to think about like recovery. How do you think about it conceptually in terms of like the reacceleration, but then everyone got bigger. So how do you think about like industry growth in terms of like much big opportunity set, but everyone bigger, like how will this play out going forward for the...

I can give you my broader perspective and then I'd love for Janesh to also add to it. But when I look at the observability and security markets, I think those -- I don't see any TAM expansion. I don't see anything that's necessarily going to expand the TAM on those markets. Gen AI does not seem to be expanding the TAM. I haven't seen that it would. So to me, like the budget, at least in the conversations that I'm having with CIOs seems to suggest that budget is going to be somewhat flattish. So it's going to be more about being able to take share and that's what we're focused on in those markets. So how do we use Gen AI to -- with the AI assistant, of this thing as an example, to be a differentiator, to take share because of some of the market factors that you're talking about, like that's really where we are seeing the energy and where we're putting our energy. When you look at the search market, Gen AI is driving a TAM expansion there, how much, how big, how fast, I think that is TBD. But DocuSign use case that I mentioned on the call, like they -- in the past, they only used to do metadata search on their documents. Now they want to do full document search across images, across -- using semantic search to search for like the context and meaning of certain things and so on. Like that is a massive expansion of the use case that never existed in the past. So I fully expect that in the coming years, you're going to see surge, have that kind of a TAM expansion, we'll see how much. And so as we think about where we are focusing our energy and what's the broader picture, it's not yet clear what the rate of the TAM expansion might mean and by when. But we are being very focused on Gen AI because we are seeing that with search. And then we are looking to see how we can leverage our Gen AI leadership to take more market share and observability and security because I think that's going to be the name of the game in the coming years.

Yes. And from my perspective, I think everything Ash said about how the business moves up into the right over the long term makes sense. As we navigate that quarter-by-quarter, we will see where, how we execute a lot of our of performance in fiscal '25 will be driven by things that we do in the second half of fiscal '24. So we just want to navigate that one quarter at a time as we go. We always just want to make sure we don't get too far ahead of ourselves in terms of how we think about the future. But overall, from a long-term perspective, super excited.

Yes. Perfect. That's actually a good closing statement. Thank you guys.

Thank you.

Thanks you for being here again.

All right. Thanks for hosting us.

Thank you.