Okay, let's get started. Day four. A ton of familiar faces, given that I've been looking at you and you've been looking at me for four days now, and you guys should feel good that this many people are here on day four after I know, because I saw them. Many of them were up quite late at The Thirsty Camel bar last night, so heroics to you guys. I love having Snowflake here. Sridhar and Brian, Brian, by the way, congratulations on the role. Nice to have you here with different stripes.
Thank you.
Yeah, formidable team. Obviously, Snowflake just reported, so we'll have a chance, obviously, to talk with Brian about some of the aspects of the quarter. But maybe we'll start with you, Sridhar. So we had you here last year. That was a great discussion we had. When you look back on 2025, what, in your judgment, were a couple of the things that you thought went really well, and were there any things that didn't quite go as planned?
Thank you for having me. It's great to see all of you, so both last year and this year, the two main objectives for the Snowflake team and me were around accelerated product velocity and taking these products to market with a globally dispersed sales team. That's the essence of Snowflake. When people ask me, what do I worry about when it comes to AI making Snowflake better, it's getting these two functions to be better, and they're both tricky things. Product velocity, as anyone that's watched big companies put lots of effort into things knows, is a lot more than just putting people to work on problems. It's making good choices. It's having taste. Products are still magical. None of us can quite tell why a good product works the way it does, and the so-so product's kind of annoying.
I'm very pleased with just the craft and speed that has gone into products like Snowflake Intelligence. Everyone wants to create an agentic solution, so a little bit clearly of a buzz phrase right now, but our ability to create value quickly, to make that entire life cycle of creating an instance of a Snowflake Intelligence agent, and for me to be able to show it to you and say, hey, this is what it does, and for you to, as a user, to be able to relate to it and go, you know, that's great, and I'd love that, that's still magical. I think we got a number of those things right. Also, a number of companies that we acquired, Datavolo, for example, has turned into OpenFlow in Snowflake, and the team is doing exceptionally well.
We took on a broader lens of being there for the entirety of the data life cycle, number of product efforts, all within a tight Snowflake umbrella. I'd say that part has gone well. On the go-to-market side, it's been about just taking these new products, making sure that we educate our sales team on it, bring in specialist resources where it made sense, but always with an eye towards how do you drive the broader team to have more and more skills. There's only so many specialist teams any of us can afford to have. I think that's actually coming along. It's a much more quantitative and deliberate team over the past two years, which I think is very beneficial. In terms of what could have gone better, you know, I like anecdotes, so I'll tell you a little story from me growing up.
As you know, most people that have done exceptionally well in the U.S. go through this gauntlet in India called IIT entrance examination. These are terrifying exams. Hundreds of thousands of people take them, and a couple of thousand qualify. And I came in some 35th or 36th in this exam a long time ago. I told it to my dad. He said, really? Why 36th? And so that's a little bit of what can be better.
Right, I get it.
It's a time of infinite opportunity. I think the speed at which products can be developed is truly remarkable. You see people like OpenAI and Anthropic do it day in and day out. They just think differently. They're not beholden to any of the failure patterns or any of the ways of working that traditional software companies are. So we need to adapt in a very big way, both my product and engineering team, but also the go-to-market teams. I think the things that they're expected to do are going to be very, very different. Driving change at speed through thousands of people and telling them they have to earn their living differently is just really hard.
Okay. Let's talk about some of the broader trends. Sridhar, I do, and I know everyone in the audience does. When we're talking to customers and we're talking to partners, we always hear this refrain that especially as we prepare for this AI era, we've got to do a better job aggregating, synchronizing, utilizing our corporate data. It's like such a common theme.
Correct.
You can see that reflected in the shares of most data software stocks, where the stocks have done well because this group appreciates that trend. In most cases, the growth rates are accelerating. It's pretty clear something really interesting is happening. If you were to dive into a couple of the interesting demand trends that you're seeing in the last pick your period, last couple of quarters, what's starting to change a little bit that you would encourage this group to keep their eyes open to?
Yeah. I think the broader theme here with AI is something I call it the beginning of the industrialization of thought. Industrialization, first of all, plays out over many decades, if not centuries. But what is, I think, unique about this moment is that ability for these language models to plan, to be able to execute things that only a human could have done. And for the better part of the last 50 years, the age of computing, let's say, data systems have always been a little bit of a back office thing. Honestly, no one cared. You just wanted your quarterly earnings report. You didn't care that it went through 800 people with lots of people editing little spreadsheets. It was a little bit of a cottage industry.
What Snowflake Intelligence, our products like that, dramatically reveal, and on the consumer side, products like ChatGPT dramatically show is if you have the right data, you can do magic. The kind of things that you folks can get done with a deep research report, for example, I'm sure all of you know that you had analysts doing that kind of work, and it would take them a week to produce the equivalent of a well-done deep research report. But much of that technology pretty much has been trained and only uses the open web. That's the simple explanation for why is data such a big deal.
If you want the caliber of thinking, if you want the caliber of planning that wows you when it comes to these products, these AI products, whether it's a Gemini or a ChatGPT deep research or ones from Anthropic, or for your own enterprise, you need high-quality data. And I think that's the excitement. Yes, it's faster access to data, but more and more smart CEOs also realize that having this data in platforms like Snowflake, where they're readily accessible, they're readily transformable, is also the basis for transforming their business. Because you can say things that were previously done by humans passing paper or PDFs around is now more automatable. And that's why this idea of an AI-ready platform is such a big deal.
And that's the pull that we see for Snowflake demand itself, because data in Snowflake is data that's AI-ready, first to Snowflake Intelligence, but there are many other things to come. But they all build on this notion of data transformation, thought on transformation.
When I talk to customers and I ask them specifically how they're going about this, there's multiple paths.
That's right.
Where some just want to get their data into the cloud infrastructure of their choice or into platforms like Snowflake. But then I talk to others, and some in the audience joined me for a discussion with the UBS IT folks. We're trying to deploy something different that's more of a data mesh, where we're trying to keep all our data where it is, not make copies, not move it all, and utilize it at rest. So it feels like there's multiple avenues to go to modernize your data stack. Do any of those paths benefit Snowflake more than others? I'm sure the former does, but do you still benefit when customers go down the path that UBS does?
First of all, no company should attempt to do mass transformation of everything that it does in one day. This is something I explicitly tell all of our customers to never do. You're just putting too much risk. It's something I would never do. I no longer accept two-year projects from my teams without clear deliverables, honestly, every month, because I'm like, two years is too long. No one should trust any plan like that, so being incremental is very much a thing. On the other hand, there is a reason for the circular movement of computing over to the cloud. On-prem systems involve, first of all, boom and bust capital investment cycles, and increasingly, that is not where the center of attention from software engineers from great companies is.
Many of the systems that are on-prem and software are also firmly in the realm of value extraction, not value creation. There's a reason why people migrate away from those systems, because if they want to increase the amount of compute that they want to put on a problem by 5%, that helpful vendor will come and tell you you have to pay twice as much, because they're very much in that phase of how do I extract every single dollar from every single customer. While on Snowflake, you don't even have to tell me that you want to spend 5% more compute on some problem because your team found it to be interesting. There are these kinds of circular reasons for why cloud computing platforms like Snowflake are indeed preferred by lots of teams.
We also have the best tech that can act on top of the data to be able to create things like AI and agentic solutions. There's a lot that we have, but it's absolutely a heterogeneous world. And things like open formats, absolutely. We can read data from Hadoop systems if they expose it as an S3 API. The real world is messy and complicated, and we will play nice with it. But our structural advantages are also strong and will only compound from here.
Yeah. I certainly, when I'm talking to customers, sense a growing interest in migrating more of their data into the cloud, so that rings true to what you said.
That's right.
Maybe this is actually a bit of a segue to you, Brian, so I think you were quite clear that Snowflake benefited in the July quarter from a number of large migration activity. I think you narrowed it down to a lot of financial services and telco customers, and maybe the results that you put up last night didn't have that same degree, but there's just quarterly variations, so how would you describe the pacing of that migration activity where it can surge in one quarter, be more normal in the next quarter?
Yeah, absolutely. With a pure consumption model, the quarterly results have a little lumpiness and by default, so Q2 was a very, very strong quarter for migrations, but think about it, you have thousands and thousands of companies, and they aren't planning their migrations around your quarter end. It's around when they're doing their transformation internally.
When we report, we're snapping the chalk line, and depending upon where all those companies are on their migrations is what we recognize from a revenue perspective. Unlike a SaaS company that actually, once it gets built, it's daily recognition, and it doesn't really matter as much on usage. And so what we really like to point people to is the FY guidance. And we're really happy with the quarter, reported 29% year-over-year revenue growth. There was nothing in the quarter that was unexpected. We did mention on the call one thing. There was a hyperscaler outage that caused roughly $1 million to $2 million worth of headwinds. But everything else played out pretty much as expected. And then we raised our full-year guide by $51 million to reflect what we're seeing inherently in the customer behavior over all those migrations .
Sticking on this migration theme, though, Brian, when I take your 4Q January guide, and you would probably discourage me from doing so, but I'm assuming call it a two to three-point beat, you're going to land at a place where actually the product revenue growth rate re-accelerates in the fourth quarter. So are you seeing anything in the January quarter that's a little bit of the reversal of the trends you saw in October, where you're seeing some goodness, maybe a little bit more migration activity?
Yeah, absolutely. When we report earnings with a consumption model, you can imagine being a data company, we look at the data every day and have all these machine learning models and numbers of people actually doing daily forecasts. And so we take all the observed customer behavior into effect when we give our guidance. And so what you're seeing is what we've observed over the last 90 days up until when we report. And so we've seen an improvement overall in migrations over the last 90 days.
Okay.
The only color I'll add on to this call is I think we like, as humans, we like to see binary outcomes, meaning it's tempting to call something an acceleration or a deceleration. But where we have eyes on the prize is to be close to that 30% mark, which to me is a great place to be in. Obviously, we had one quarter that went a little bit more than that, another quarter that's slightly less than that. But to me, to be able to operate at that realm is great. If anything, you should be challenging Brian on what it will take him to hit 40%.
Okay. Okay, I may do that.
And that was hypothetical.
Obviously.
Yeah. But in fairness, there's not that many software companies at your scale that are going at 30%, Sridhar. So I'm with you. Let's get back to on the AI side. One subject that interests me is not so much how customers are behaving in this AI era, but, Sridhar, how you and the engineering team are incorporating AI into Snowflake's product set to actually improve your own query speed. In the same way that a number of hardware changes have occurred over the years, the chipsets are getting better, improving query speeds. How is Snowflake actually embedding AI in your core product to drive price performance improvements for your customers?
What did you mean by query speed here?
Just customers that are hosting data in Snowflake are querying it for business intelligence reporting needs. You're embedding AI in a way that perhaps they can interface more easily with their data and query it a little bit faster.
Yeah. I would break this up as two separate questions. There are a large suite of improvements that we make to performance in Snowflake. Some of them come from things like newer generation of chips that the hyperscalers or Intel, for that matter, produce. They will often involve price performance trade-offs, meaning with a new chip, you might be able to get 20% more performance, meaning queries finish faster. But on the other hand, the chip itself might cost you 10% more per unit of time. We also make a lot of software improvements that make queries just go faster. These are generally almost orthogonal to AI. And we have struggled to figure out how to roll these out in previous years. And we have had discussions with many of you about how we roll out performance improvements, so on and so forth.
But one of the geniuses in my team, they came up with this idea of a new generation of warehouse, which delivers a lot of these performance improvements, but is roughly price neutral. That's the Gen 2 warehouse. And the idea very much is that it's a win-win. We don't see any reduction in the amount of money that we make. At least that's the aspiration. It's a complicated modeling problem to price correctly. But on the other hand, our customers have a lot of the work that they do just go faster. They don't have to do anything. And that's the kind of trade-off that we are increasingly headed to, where we can carefully apportion a bunch of benefits to customers, but also have a throttle on how much do we want to take in terms of a price hit on our side.
There are second order effects that become difficult to model. If you let your customers do a whole lot of queries just much, much faster, then they often do more of them, because you can just analyze things better. You can model things better. But even without taking that into account, I'm very happy with Gen 2, because it kind of removes this question of what's the trade-off that we need to make in terms of rolling out improvements in the core platform. Now, part two was more about how are you using AI to make the act of using Snowflake, configuring Snowflake, optimizing Snowflake a whole lot better. This is actually a really exciting area for us as a whole. We have one product. It's in private preview. It's called Cortex Code.
The idea very much is it's a data agent, comes as part of Snowflake, is able to handle pretty complicated tasks for you, so much so that much to the terror of my engineers, I can write prototypes in an afternoon, because it's much more oriented towards the outcome that you want, and it helps guide you along the way. Yes, it will be used to optimize queries, but it will also be used to do things like configure a complicated connector like OpenFlow to extract data from an Oracle system and put it into Snowflake a whole lot faster. But this goes back to my point of this is an era in which product development needs to be rethought in a fundamental way. But product rollout and how people like solution engineers or services engineers use software also needs to change in a big way.
We think this is going to have a dramatic effect on things like migrations. You touched on that earlier. During the entire time that I've known Snowflake, two and a half some years, migrations have been gated by the capacity of the Snowflake team and our partner team to handle them safely. Each migration is a high-stakes exercise, because there is some critical system that is sitting behind. And the business owner is saying, you better be exactly the same before and after. It's terrifying. We think AI can be a huge accelerant in making those go faster. They all fall in the same bucket of how do you use AI to make the act of doing these complicated data jobs just a whole lot faster and safer.
That's interesting. So that could be an accelerant to that migration activity over the coming years.
100%. I think there are step changes to be made. I've consistently talked about it for the previous three quarters. I have a few pet projects that I personally pay attention to. AI-driven migrations is one of them, because the potential is just incredible.
Okay. Maybe a couple of thoughts on some other developments in the space. Sridhar, I asked you this question back in the summary. You may not remember, but I pointed out that a lot of the SaaS companies, the apps vendors that this group pays attention to are all in various ways. Salesforce might be a good example. Stepping from the roots has workflow automation, SaaS firms into the data arena. It feels like every SaaS company is attempting to become, in part, a data company as well. What are your thoughts on that transition? And is there any part of the data space that they would sort of have a right to win and beyond which might be a little bit out of their wheelhouse?
I mean, just to put a little bit of historical context into these, most SaaS vendors were operated basically transactional systems. These are systems of record. You go into Workday, if you want to file PTO, or if you want to hire someone new, somebody goes and makes an entry there. Reporting for these folks was always an afterthought. I used to run the AdWords team. We had a reporting team. But to be honest, that reporting team was a little bit of a tax on my regular team. I'd rather have them work on how to make more money, not give more stats to advertisers. That was the general attitude that all SaaS vendors had about reporting and analytics, and it's part of the reason why platforms like Snowflake that specialized in being very good at analytics even came of age, because we were very good at doing that.
And over the past many years, we have established ourselves as a place where different kinds of data can be brought together, juxtaposed, to get more of the 360 view of what's going on within an enterprise that we all crave. With AI especially, but even before that, with analytics becoming more and more prominent, people are beginning to understand that the mechanics of having deeper insights on how a system functions or how processes are functioning is an essential part of making these more efficient. There's more and more of a realization that there indeed is a closed loop around data. And AI accelerates this, because people now understand that if somebody, Snowflake, has a copy of all of the most critical data about a company, it can be the place where decisions can be made about what do you optimize, what do you do next.
Hence the many, let's call it, aspiring data clouds. One seems to come up every other month or so. Roughly in terms of our view, first of all, we don't view this as a zero-sum game. We think there's lots of value to be created. We have gone and basically done bilateral partnerships, let's see, with Salesforce, with ServiceNow, with SAP, with Workday. Several others are in the works for basically these kinds of agreements. The idea very much is by doing this, these folks are able to create products that can make money, because this data is indeed valuable. We make money as well, because with these bilateral agreements, we can take the data, juxtapose it with other data. Our customers, like you, end up getting a lot of value from using Snowflake. It's not a zero-sum game.
There'll be agentic solutions developed on top of Snowflake. There'll also be agentic solutions that are developed on top of the platforms that these folks provide. And it's a little bit of may the best product win. And we feel good about where we are, because we've been doing this for a very long time.
Sridhar, you've always had rivals, some large, like Google and Microsoft. But let me ask you about a smaller one that's hit my radar and I think others, and that's ClickHouse. So I think they're well known for low-latency analytics jobs, especially certain log events, things like that. What are your thoughts on that? And where is Snowflake on its journey to, frankly, launch features that can frankly do that?
Interactive analytics are an interesting category. As you correctly point out, one that Snowflake hasn't always paid attention to. A number of our customers, even our large ones, whenever they want to serve data from Snowflake, by that I mean put data that's in Snowflake in front of users like you with very tight latency requirements. If you're looking at a trading screen and you want to see some summarized data, you have a low tolerance for that thing taking even a second. You want that to paint immediately. It's not an area that we paid attention to, but we think it's a natural adjacency to what we are doing. We have actually introduced a product feature called interactive analytics that is focused on high-performance analytics.
Our aspiration very much, and I can speak as someone that's run load tests on these systems, is for these to be like sub-200 ms for simple queries, so that it can be deployed at scale, and the underlying Snowflake technology is sort of truly amazing, and we can support hundreds of queries per second, which can translate to millions, if not more, of users right on top of Snowflake. They involve different trade-offs from our regular Snowflake systems, but this is what we are really, really good at. There's a crack team that's working on it, and it's coming along well. I think it'll be an interesting category for us.
Yeah. I look forward to seeing that next year.
Thank you.
Brian, new set of eyes on the margin structure at Snowflake. My view is that there's actually pretty good EBIT margin potential at this firm. Maybe that's one of the things that actually attracted you to the platform. But Sridhar and team have built an at-scale $4 billion to $5 billion revenue company. Yet in my view, it's got an EBIT margin structure with room for improvement. Do you share that view? And where do you think improvement over the next several years can come from?
Yeah, absolutely. I'm a big believer in, I think you can look at the companies that I was at, Pass, GitLab, and Verisign, that you can grow, but you can do that responsibly, and so Sridhar and I are 100% aligned. One of the things that I just recently done was we sat down and gave out the annual operating plan to all the ELT and really driving accountability by using AI, getting more efficient, and just not throwing more bodies at a problem, and so this is just a ginormous market. It's a super interesting market. It's changing all the time, and so we're very, very focused on growth, but we'll do that responsibly.
Yeah. OK. We've got a minute, two, and it might be good for you to ask Brian anything on your mind related to the print. I'll give you a chance. I think we have a hand up, Malcolm? Yeah? I think you can just shout it out.
on top of that, ingest data, get the schema types, and kind of really pretty comprehensive systems just completely in the Snowflake environment. I just like your thoughts on that with regard to the opportunity.
Yeah. I would phrase this much more as data in Snowflake has gravity, and we are making it easier and easier to bring data into Snowflake, but our superpower with that data is that layer of governance and security that we put in. People that bring data into Snowflake often will set up fine-grained permissions, and we can handle that at hyperscale, tens of thousands of roles, intricate relationships between roles, modeling a complex company that has 100,000 people, and where applications become interesting is there are a whole set of folks within these enterprises that say, I want to build a slick interactive application, but I don't want to relitigate decisions about governance and who has access to what data. How can I make it super easy for it to just work as part of the Snowflake system? Streamlit was one such attempt at it.
Honestly, this was two, three years ago. We didn't do such a great job of making it performant and easy to use. Again, AI is a big game changer here. Part of our thrust with coding agents is right now, and again, I've done this. You pretty much write two, three sentences and say, I have this data set in these tables. Help me make a Streamlit to visualize the data, output the options, and you can tinker with it and so on. But we also recognize that there is now an entire ecosystem of companies that have specialized in rapid applications. All of you folks, I'm sure, know about companies like Vercel. They have a great new product called v0, which is their coding assistant. You can develop just beautiful React apps with very, very little programming.
We announced a private preview with them just a few weeks ago, and we are racing to get it out, where you can build an app in a Vercel environment but deploy it securely into Snowflake. Just a push button that says, deploy to Snowflake, and then you now have a modern React customizable app that is running within the Snowflake security perimeter, obeys rules, obeys permissions, all of that stuff, and your teams then don't have to worry about, well, do I have to manage a separate hosting environment? Do I have to worry about permissions? It's increasingly that kind of stuff that we want to do. There are many others in the space, whether it's a Replit or a Lovable. We see a slew of these kinds of partnerships coming that marry the best of app technology with the incredible staying power and gravity of data and secure governance.
Okay. Why don't we leave it there? I think we're out of time. Sridhar and Brian, thanks so much for coming to Arizona for our event.