Good afternoon, everyone. I'm Sanjit Singh. I run the infrastructure software practice, on the Morgan Stanley software team. Super thrilled to have, once again this year—I think it's like, two or three years in a row now.
That's good.
We initiated coverage this year, but it's good to have you back to the TMT conference. We have CEO Ash Kulkarni, and we have Global VP of Finance, Eric Prengel. Eric, thank you for joining us as well.
Thank.
Awesome. Before we get started, let me go through the disclosures real quick. For important disclosures, please see the Morgan Stanley Research Disclosure website at www.morganstanley.com/researchdisclosures. If you have any questions, please reach out to your Morgan Stanley sales representative. So maybe just to level set, Ash, we're coming off a quite an impressive quarter. You know, post the Q1 sales execution issues you put up some really solid results since then. Revenue growth sustained at 17%. Your cloud business actually accelerated a little bit to 26% from 25%. What surprised you about fiscal Q3, particularly with respect to your cloud performance? And how would you characterize the breadth of strength across your core segments and particularly in your enterprise segment?
Yeah. Just in terms of the actual commitments that we saw, you know, the sales performance was very strong. We saw continued demand for our products, and it was pretty broad-based. In search, which is arguably what we are best known for, GenAI continues to be a good driver for us. In terms of commitments, we had five deals greater than $1 million in commitments that were related to GenAI, which was an uptick from the three that we saw in Q2, which was great. Search continues to do very well for us. We also saw strength in observability and security. In terms of the revenue performance, you talked about the cloud growth rate of 26%.
You know, for cloud, because it's a consumption-based model, consumption can be, at times, you know, it can fluctuate a little bit, but what we did see was, we had several of our larger customers actually consuming faster than we expected, which was a really, really positive sign, and that group was different from the group that really was ahead of their consumption expectations in Q2, which gives a sense that it's nice and broad-based, which is something that we are very happy about, but the core fundamentals of the business remained the same. It was the strength that we are seeing in GenAI. It's the fact that we are able to get more and more customers to consolidate onto our platform for observability and security.
and the fact that, you know, sales execution came back to what I'd describe as the pace that we were used to in the past.
Awesome. And then maybe, Eric, for you, as we look at the outlook for Q4, it does imply a slowdown back to the double-digit growth level. What are the specific areas of where the team is applying prudence to the Q4 outlook, and maybe the assumptions behind that?
Yeah. So as you think about, you know, the issues that we had in Q1 around commitments, that's still playing through. That's a multi-quarter issue. And so that's still gonna be a little bit of a headwind to Q4. If you think about year-over-year growth, there's some FX impact. And so we've calibrated that, and we obviously guided to the actual currency as well as the constant currency growth rate.
Interestingly, this year, there was a leap year in FY24, and so there's one less day in FY25 Q4 than there was in FY24. And so that's about a 1% headwind. And then, as Ash talked about, we saw strength in Q3 on the sequential revenue growth associated with some of our consumption revenue, and that's the cloud consumption business. And some of that strength came from some larger customers consuming at elevated levels. And so as we thought about the Q4 guide, we wanted to put in a level of prudence around that, knowing that that can fluctuate from quarter to quarter. There's also the monthly cloud business, and we've talked about that being generally a flattish business. And our expectation is that it continues to be flattish, and so that's baked into Q4.
And then, you know, one thing that we haven't talked about as much, you know, is some of the assumptions that we're making around the macro. And so there, we're baking in the appropriate levels of prudence there for a lot of moving parts that we're seeing. And so that's how we came up with our Q4 guide.
Looking at some of the themes from this year, one of the comments that you made, Ash, on last week's earnings call was execution getting back, sales execution getting back to pre-Q1 levels. Can you describe what the reasons behind the original go-to-market adjustments and fine-tuning that you were doing? What is the potential for sales productivity, not just to get back to where we were, but to actually improve?
Yeah. So, just stepping back, you know, what we did at the beginning of the fiscal year, there were some sales segmentation changes that we made. The changes that we made were not atypical of what other organizations, our size or larger, have done in the past, just as they've grown. There were three specific changes. One was, at the very top of our customer segment pyramid, we call that our strategic segment. We added more accounts into that segment, and we added more sellers into that segment. And that's a segment where, you know, each account executive has, you know, one to two accounts, but these are large, high-propensity accounts that we believe we can grow them into, you know, multi-million-dollar accounts for us. The second thing that we did was the next tranche is our enterprise segment.
In the enterprise segment, historically, we had a lot of accounts per account executive, sort of not in line with what is considered to be industry norm or industry best practice, so we normalized that because we wanted to drive the right kind of behavior. We want to have more focus in our accounts and our account executives to go deeper, and drive more meaningful deals so we can become a bigger and bigger part of an organization's infrastructure, and then the third thing that we did was all of the accounts that got freed up because we reduced the number of accounts per rep, we turned them into greenfield territories and assigned hunter profile reps to those, so those were the things that we did. The changes themselves, we had a lot of conviction. We continue to have a lot of conviction that they were the right changes.
Right.
The way we implemented them, we had a misstep in terms of the actual account transitions. That didn't go very smoothly, and that affected two things. It affected the pace at which we were building pipeline and the pace at which we were progressing pipeline. And that had an impact in Q1. Once we recognized the problem, we immediately started you know addressing it. And you know you've seen us do all the right things in Q2 and Q3. And like I said, the execution rate of execution has come back. The other thing which is important is the original rationale for the changes. We are starting to see some of those benefits in the first three quarters of FY25. We have added meaningfully more million-dollar customers than we did in all of FY24.
To me, that's a really good early sign of, you know, the kind of activity that's happening in the sales organization, the fact that our reps are actually doing what we expect and what we desire of them, not just close smaller transactions, but close more meaningful transactions. You know, as I look forward, like, I want to make sure that we are continuing to drive that same kind of behavior in the organization because as we grow and we scale to three million and beyond, it's gonna be that kind of activity that will get us there.
Yeah, and maybe just to follow up that, as, you know, we still got a big Q4 ahead of us.
Yep.
As we start to think about fiscal year 2026, Ash, what are your key strategic initiatives you plan to focus on to drive growth in fiscal year 2026? What are the two or three objectives that you really feel like you have to hit from a product go-to-market, change management perspective?
Yeah, so from a, let me first start with the, the go-to-market because that's something that, you know, we were just talking about. From a go-to-market perspective, the changes that we made, we are really happy about, right, so you know, they've, they've all settled down, as we talked about, and we are right now in the middle of our, FY26 planning, but where we sit right now, we are not anticipating any changes, but we'll make sure that, you know, as we, as we get to the start of the fiscal year, we communicate, you know, transparently about everything that, that we are doing. But at this point, we are not anticipating any, any meaningful changes, then in terms of the, the focus areas for FY26, you know, arguably the biggest, momentum that we are seeing is in the area of generative AI.
So I don't think that should be lost on anybody that what is happening now in the industry is we are seeing customers use us as a true runtime platform for retrieval augmented generation as they're building these conversational AI apps, as they are thinking about agentic workflows, or even something as simple as just building semantic search use cases. Like, we are being seen as the retrieval runtime platform. And that's a massive opportunity because we are so early in this entire, you know, evolution. You know, you take any average organization, you have hundreds of applications in an organization, if not thousands. And we are still at the phase where it's just a few handful of applications that are being sort of refactored, you know, to infuse AI capabilities, to automate through AI.
There's a lot of opportunity ahead of us in the coming years.
And we want to make sure that we are one of the platforms of choice. We continue to be in that position in the years to come. So there's gonna be investment, continued investment on the product side, and continued investment on the go-to-market side to make sure that we can capture that opportunity. Beyond that, you know, obviously, AI is helping us compete in everything we do. Features like Attack Discovery are making us more competitive in security analytics and SIEM. Capabilities like Auto Import are making us, you know, more appealing when it comes to observability and log analytics and so on. So that is part of the game plan.
We want to make sure that we leverage the fact that we have a native AI stack and that we leverage it to make our observability and security solutions better, and continue to drive in a more competitive way in FY26.
Yeah. That makes total sense and I really want to spend the balance of this conversation talking about those multiple opportunity sets and how AI is influencing them. Before we get, though, Eric, I wanted to just, visit, the margin story because there's been a ton of progress on that front. How is the team thinking about balancing the large opportunities that Ash just described about and why is now the time to drive maybe more margin expansion versus leaning in to capture that opportunity from an investment perspective?
Yeah. So let's rewind to the start of 2025, and we thought we were gonna see a modest margin expansion in 2025. We had the commitment issue in Q1, which slowed down the business a little bit. And when we saw that happen, we wanted to put a little bit of a hold on investing, just, you know, make sure that we could reconfigure the business to really make sure that we're driving the appropriate level of growth. And so we slowed down investment a little bit in Q1. In Q2 and Q3, we saw strong commitments. We saw strong consumptions. That obviously drove upside to revenue. With that, we saw a strong operating margin upside. And so as you think about our FY 2025 non-GAAP operating margin, the current guidance we have out there is for 14.7%.
Strong operating margin, a meaningful increase from the 11% that we saw in FY24. We hadn't, frankly, anticipated it increasing quite as much as it did. As we look forward to FY26, we see that the business, the go-to-market, is operating in the way that we would want it to. We see an opportunity to invest in putting more capacity on the sales team into play. Obviously, as we invest in capacity, reps are gonna have to get ramped, and then that's gonna turn into commitments, which then are later gonna turn into revenue. We see an opportunity to invest in capacity in the field. We see an opportunity to invest in some of the overlays that have a solution-specific focus. We have a GenAI-focused pre-sales team. We have a security-focused pre-sales team.
And so, to bring more of those people on to help with solution-specific selling, there's an opportunity to invest in marketing, and there's also an opportunity to continue to invest in product to maintain that leadership position that we've earned through all the product development that we've been putting out over multiple years. So, with all that opportunity, we are gonna be investing in FY26. We're still gonna be focused on balancing revenue growth with profitability. And as we said, there is gonna be an increase in the operating margin in FY26. It's just gonna be a very modest one. And so, you'll see our business grow, and you'll see our business also, you know, expand from an operating margin perspective.
Yeah. It makes sense, especially given the better execution and the opportunities ahead of you. I think that makes sense from an investment perspective. So let's dive into the AI search opportunity at Elastic. Maybe start, like, high level, right? So in software, you know, a lot of investors have been trying to figure out the timing of when this really starts to meaningfully impact different companies in the space. When it comes from your perspective and the Elastic customer base perspective, where are we in the cycle, from your vantage point in terms of customers getting those meaningful, consequential AI applications into production and that starting to impact numbers at Elastic?
Yeah. So, the first thing to understand is that we are truly very early in the AI journey. Where we are today, like, you see some of the adoption metrics that we talk about. Just in Elastic Cloud alone, we have now over 1,750 customers using us for building all kinds of GenAI applications. You know, think of that as a design win, right? Those are customers that have decided to bake us into their application. Now, not all of those applications will go into full-scale production, but, you know, the next phase of that is that journey becomes the commitments. When customers decide that they're gonna put something into production, they'll usually talk to us about, you know, what their sizing needs are, and we'll do a contract with them.
They'll make a commitment, and they commit to a certain spend, whatever that might be. And you've started to see some of that commitment, you know, show up in the fact that we now in Q3 we did five deals over $1 million, right? So that's the GenAI contribution to commitments is starting to now add up, which is a good thing. Over time, those commitments translate into revenue as those customers consume against those commitments. But I would argue that we haven't seen the inflection yet, right? Because you'll see any inflection that happens on the commitment side before it shows up on the revenue.
And the reason why I'd say that we are still very early is if you just think about, you know, like I mentioned earlier, that just the total number of applications, IT applications that organizations end up having within their business. You know, if you're a mid-sized company like Elastic, it's probably several hundreds of applications across different functions, across, you know, different parts of what IT implements and business implement. If you look at a larger organization, it might be several thousand.
If you talk about, like, really large, like banks, telcos, etc., it's often in the tens of thousands of applications across different divisions, different, you know, global, regions, etc. How many AI applications are you seeing being talked about in any organization today? It's in the dozens. You know, in the largest, maybe it's in a few hundred, but you're not talking about anything more than 1% or 2% of their overall IT landscape. If you truly believe that large language models and what you can now do with, the combination of LLM and retrieval engines like Elastic, if you can automate just about anything that involves unstructured data and human processes, you can now automate them because of the fact that these language models have the ability to do inference and reasoning. You're talking about a very large opportunity over time.
But where you are right now, we are clearly still very early, which is why a lot of the spend is still accruing to sort of the lowest levels of the stack, which is on the chips and on, you know, infrastructure compute buildout. This is a story that's been played out in the past with, you know, cloud computing and other things. Eventually, it rises to the infrastructure and then the application stack, software stack. Our goal is to make sure that we are preparing for that by being in as many places, in as many of these enterprise and mid-market accounts as we can be to become their retrieval engine of choice.
And our strength is the fact that we are experts at unstructured data, the kind of, you know, data and processes that people are trying to automate. Elastic has always been known as the expert at dealing with all of this kind of information, whether it's Word documents, it's logs, it's PDFs. Like, that's, that's where we shine. So our goal right now is to, is to become that platform of choice. And as and when that inflection hits, that's gonna be something that we benefit from naturally.
Awesome. Just wanna follow up on, like, how you sort of framed out the journey, right? There's, you know, commitments, and then that turns into revenue, and then we need to get more applications, penetrate.
Yeah.
More of the application real estate. But just going to the Kibana, what's the journey? What are customers doing before they make that commitment? It's like, is it a lot of trial and usage of, like, the community offerings? What's sort of the journey to commitment, if you will?
Yeah. So obviously, there will be a class of customers that will start with the community edition of Elasticsearch, in which case they might not be paying us anything, but they are using it in a very basic way. A lot of the AI-related features are, you know, only available in our paid editions. So when they start to use those, like, you know, they need to first come and, you know, talk to us. Either if they wanna run it on-prem, they need to purchase a license from us, or they go and just start using Elastic Cloud. Now, Elastic Cloud, we have a self-service motion. So most smaller organizations will, you know, what you think of as SMB will just naturally go directly there and start to use it on a month-by-month basis. Most enterprises typically don't permit that, right?
Most enterprises don't allow their users to just go and swipe a credit card, so those typically end up being our customers anyway, so they'll talk to us and they'll start, you know, if they have an Elastic Cloud contract. The great thing about Elastic Cloud is you're buying credits from us. You are not buying discrete licenses because, you know, we don't sell, we don't have a different AI SKU. We don't have a different security or observability SKU. We just have a platform SKU, so if you have purchased Elastic, you have the right and the ability to start using all of these features, and we see that. Naturally, customers will start to use it, and that's how, from the telemetry, we can tell that there are over 1,750 customers that are using Elastic for building all kinds of AI applications.
Because from the telemetry, we can tell that they're storing dense vectors. They're querying dense vectors. They are, you know, using, the ELSER, embedding model, etc. and that, that then becomes an opportunity for us to engage with them and say, "Hey, how can we help you?
You know, "What are you trying to accomplish? Let's, you know, we have a lot of best practices," and the presales specialist team that Eric talked about engages with these larger customers. We do workshops. We have a lot of prescriptive information on, you know, how to address specific cases around customer support or cases around e-discovery or cases around, you know, what have you. All the patterns that we have seen effectively, and what LLM choices might be appropriate for them. Because, you know, I don't believe there is one LLM to rule them all. And you've seen, we've seen different performance in terms of accuracy and relevance based on the domain and question. So those are all the things that tend to happen before a customer goes, "Okay, I've decided this is what we're gonna do.
This is my size and capacity requirements," and that's when the commitments happen.
Awesome. Let's talk about the ESRE and AI Search. I think one of the challenges for investors is that, frankly, this category is in English. There's a lot of acronyms. And I think broadly people know that there's this vector search thing. They maybe heard of RAG and a lot of companies doing RAG and vector search. And so maybe if you could sort of pinpoint the value proposition that ESRE and Elastic brings to bear when it comes to running these RAG-style applications that differentiates you from, like, you know, the guy, the operational database guys that are trying to embed vector search into their platform. Why will demand funnel to Elastic versus others?
Yeah. So first of all, you know, when you think about a vector database as just an end in and of itself, I think you missed the forest through the trees. A vector database is a functionality that solves a particular problem. But really, what you're trying to accomplish is accurate, relevant retrieval, right? That's the problem. What you're trying to do is make sure that you are allowing the large language model, you're giving the large language model just the right context to answer the particular question that might have been asked. And that could be as part of a chat-style conversational app. It could be as part of an agentic workflow. It doesn't really matter. Retrieval is what matters, and relevance is the most important thing. Vector search is one technique, but you'll often need a combination of vector search and textual search.
You will often need to re-rank the results because you might not get the ideal result set, and you might need to boost certain results to get the, the optimum overall result, which is why re-rankers have become so popular, re-ranking models. The way you create the embeddings and how you chunk the data has an enormous impact on the accuracy of the results that you get. So having the right kind of embedding model, having the right chunking strategy, all of these things become really, really important. We always had this idea in the sense that we wanna have the world's best vector database in terms of performance, in terms of scalability, etc. But we also want to think of this in terms of what is it that somebody's trying to solve? It's retrieval.
Creating that end-to-end platform that provides all of these capabilities with a view of giving the best, most accurate results at the lowest possible cost in the most efficient way. That's what we've built. That's what we continue to focus on, and the last thing I'd say is making it such that every Elasticsearch developer is able to build an AI application without having to relearn a whole new set of skills. We built our vector functionality directly into Elasticsearch in a very native way, and that's been our big strength. That's why we win.
Yeah. I think the point to me was you need a solution.
Absolutely.
These are not like, it's not just features.
Yep.
It's more than just vector search and inference. Frankly, you've seen some acquisitions in the space, in a couple of fields that sort of.
Exactly.
Points to the importance of embedding models and re-ranking capabilities, which it seems like you guys have had a head start on.
I mean, clearly it's a validation of everything that we've been doing, right? We were very early to this space. We saw the promise in this. We started working on our vector database product like over five years ago. Nobody was talking about this.
Yeah.
We built our ELSER model, our embedding model. We have had it out now for the last couple of years, our re-ranker, and you're seeing that others are trying to catch up, so you know, it's natural, but this is our core competency. We know unstructured data better than you know. That's our core strength, so we are gonna keep running with it.
Where do you see the opportunity around inference as a service? What do you think, Elastic's future opportunity to monetize more of an inference stack? Or do you see a chance for Elastic to become a customer's inference as a service vendor to pull in more of the revenue directly related to LLMs on top of some of the other use cases like RAG functionality that you're doing?
It's a definite part of our story going forward, right? So, you know, the first step that we did was we created a common inference API. We've always had this attitude that we wanna be agnostic to models. Even though we have an, our own embedding model, our own re-ranking model, we'll support other embedding models and re-ranking models. We don't really care, like, where you get it from, as long as you're running it on our platform. And then even with LLMs, we've had that same approach. So we built an inference API which abstracts away the individual complexities and differences between different LLM APIs. The next step is, this is something that we've talked about publicly at our user conferences, is we are working on our Elastic Inference Service.
And the idea there is, like, we'll first start with, you know, providing the ability to access open-source inference models through the Elastic Inference Service. And that, you know, there's Llama. We see a lot of usage of Llama. There's Mistral. There are a whole bunch of large and mid-sized models that are now available on Hugging Face that are incredibly good at dealing with different domains of data. And that's gonna be part of our story going forward. I can't tell you the exact dates, but, like, that's definitely something that we've talked about in user conferences and, you know, watch that space.
I'm looking forward to it. Maybe let's move the conversation over to modernization a little bit, and I'll throw it up to Eric and Ash. Definitely we're interested in your perspective as well. In terms of the vectors of modernization, Eric, when you look at the features that Ash spoke about, a lot of these capabilities are in the enterprise and platinum tiers, which have a higher rate card. What's sort of the storyline in terms of the uplift you see from when customers go from standard to platinum to enterprise?
Yeah. We see a steady uplift in pricing as they go from Standard to Platinum Enterprise. It's, you know, obviously it's a meaningful percentage. What we've seen that's been really interesting is that over the last multiple quarters, we have seen more and more customers migrating from Platinum to Enterprise. They can. There's a lot of features that they can unlock with Enterprise. Some of the GenAI functionality that lies in Enterprise. And so there's a lot of capabilities that people want to be able to access at that Enterprise tier. And it's really helped us to monetize that business.
The way you think about, though, just broadly generative AI and how we can monetize it is one of the ways it gets monetized is that more people are gonna be moving towards the higher tiers so that they can utilize the feature set that is available to them. But also, they're gonna be spinning, you know, the cost faster too as they've got bigger workloads, more workloads that are running through our ecosystem. So there's just effectively more data that we're managing for them. And that's one way that we can, you know, monetize the business more as you think about Gen AI. And Ash, I don't know if there's anything else you'd add to that.
The only thing I'd maybe add is that, on our website, on the pricing page, we do clearly call out, you know, the rate cards, if you will, between our different tiers. So it's very easy to get a sense of what is the incremental uplift. But like to Eric's point, our monetization strategy is pretty simple. You know, features that bring down your overall infrastructure cost in a very meaningful way, we monetize those. Because to a customer, you know, the ROI is very obvious, right? They don't mind paying us because they're gonna reduce their overall underlying infrastructure cost. Another area is where, you know, you're talking about management and monitoring of very large clusters.
Like, that is something that, you know, customers are, you know, happy to pay for because that is a pretty significant pain point. A lot of all of our AI functionality is in monetizable tiers. And then we have solution-specific features like endpoint security. Like, that's kind of monetized. So there are four categories roughly where we have meaty features that we monetize. And, you know, we've been quite successful at getting customers to vote up tier, but also bring more workloads onto our platform. Because data, the more data comes into our platform, the better our monetization story becomes.
Maybe using that frame right now, I'll offer it up to both of you again. One of the popular questions that we get is if we take a traditional Elastic customer that's been using you for keyword BM25 search and they're thinking about expanding that with semantic search, what does that sort of do in terms of the revenue uplift for those customers that take an application that's doing traditional keyword search and now say, "We're gonna make this application do hybrid search"? What's sort of the revenue uplift on those specific type of customer use cases?
Yeah. It tends to be quite varied and it depends largely on the data. And what I mean by that is that, first of all, what I'd say is that you, it's always more consumption because the compute intensity of these algorithms that you're talking about tend to be greater than the compute intensity of just a traditional BM25 algorithm. Now, having said that, depending upon the type of data, you know, if it's just a single log, just a log stream, and you know, every log line is what you're turning into a vector, that is probably gonna be less intense than if you have entire documents or libraries of documents where, you know, depending upon how you chunk up that document can have a very meaningful difference and impact on the accuracy of your results.
So think about it this way. If you ask a question, you know, depending upon how you've chunked or broken up the document, your answer might be in multiple chunks. You might need to do multi-pass or do something that allows you to actually look across multiple segments. So those kinds of complexities have an impact on, how much compute you tend to see. And we've seen anything from, you know, 20% to 30% on the low end and, 3 to 4X on the high end. But it, it just, it's a, it's a spectrum. And that's why it's really hard to give a number. You know, your mileage will vary, but, it always tends to be more intensive just from a compute standpoint.
That makes tons of sense. Let's actually talk about observability 'cause it's actually the biggest piece of your ARR and where you're seeing momentum there as well. I'd love to get a sense, Ash, of like the scope of Elastic's ambition in observability. Does it remain focused on sort of the log analytics opportunity or do you see your observability strategy expanding into things like, you know, incident management, ITSM on call? You've seen a lot of the traditional observability players kind of expand into adjacent categories. Talk to me about the scope of Elastic's observability strategy.
Yeah. Our ambition is definitely to go beyond log analytics. The first thing I'd say is, like, log analytics is a big, big market in itself, right? Because the one thing you can be very sure of is the more applications that get built. And, you know, we've been talking about AI and all the new AI applications that are gonna get built and are being built. Every one of them, as they go into production, as they mature, are gonna need observability. So, you know, log analytics is gonna be a continuing and growing market. We wanna obviously be a leader in that space. But beyond that, you know, whether it's tracing, whether it's metrics and infrastructure monitoring, our ambitions are beyond just logs.
Having said that, like, the way I think about it is there are some natural areas where, you know, I don't know if in the near term we would wanna compete, right? And there, like ITSM, you talked about since you brought up ITSM, we have things like the ability to do remediation or the ability to take actions. But we'll often integrate very nicely with the players that tend to be strong in this space, whether it's a ServiceNow, or a PagerDuty or what have you. We have deep integrations with them. Our customers tend to use us together. And that's how we prefer to go, you know, with this sort of a partner route.
But the core observability area where you are trying to look across different signal types to understand whether something's going wrong, predict if something's gonna go wrong based on the data, and then suggest some remedial action. And if something goes wrong, then root cause it quickly. Like, that's our core sweet spot and we're gonna keep, you know, improving and increasing the breadth of our portfolio.
Awesome. With a minute left to go, for investors that are newer to the Elastic story, or maybe they visited once upon a time, it's been a couple of years, given where the business is standing today, you know, you have a $1 billion-plus business, your margins are improving. Ash, what are you most excited about, you know, from here looking forward? Like, what gets you up in the morning that really drives your enthusiasm for the Elastic opportunity from an investment perspective?
Personally, for me, it's everything to do with AI, and you know, even the reason that I came to Elastic, yeah, a little over four years ago was because I had used Elasticsearch in my prior life, and what I found was that when it came to unstructured data and any kind of real-time analysis, I had not seen anything nearly as good as Elasticsearch, like miles ahead of everything else that I had seen. When you think about what generative AI is doing, it's fundamentally changing how we are able to use automation for decision-making, analysis, and actions on unstructured information. You know, whether it's opening a new bank account, applying for a mortgage, like you know, looking through case law and making a decision, like everything.
There are so many processes, irrespective of the business that you're in, that involve human beings reading documents, reading basically unstructured information, stuff that can't be put into a SQL database and have a very precise deterministic query. That's what large language models are allowing us to automate. We are excellent at that. This has been our core strength. Now, how we ramp into that and how that grows, you know, that's gonna be, if anybody has a crystal ball, that'd be awesome. All of you are better at that than I am. But the long term is pretty obvious to me. The long term is that this is, in my mind, a very big technology shift. I started my career in the early 1990s and, you know, I was, I saw how the internet progressed. I saw how cloud computing progressed.
It always starts with the lowest layer in the stack in silicon and then compute buildout, and then it percolates up. I'm very excited by what this means to every application that we have today that I believe is gonna be reworked in the coming years.
Exciting times. We're out of time. Thank you so much for giving us an update on the Elastic story.
Thank you.
Thank you, Eric.