Hi, everyone. Welcome to Summit, and welcome to Investor Day. Thank you for joining us here, whether you're joining in person or virtually. We appreciate you making the effort. I'm going to kick things off with a quick overview of our agenda today. We will have presentations from Sridhar, from Christian, and from Brian. Sridhar will give an overview of really his vision for Snowflake, and what that means for both our Core Data Platform opportunity as well as our AI opportunity. Christian will then take the stage and go over a lot of the product announcements that you heard from us this morning, and really detail how those are fulfilling the vision that Sridhar will lay out. Finally, Brian will come up here and share our financial outlook and the implications of that vision. We will wrap with a Q&A, so we'll take questions from the audience.
Sridhar, Brian, and Christian will all be on hand to answer your questions. As a reminder, we will be making certain forward-looking statements today. This is a statement on our non-GAAP financial measures as well as our safe harbor. Both are available on our investor relations website. With that, I would like to pass it over to Sridhar.
Thank you. Thank you, Katherine. It's great to see all of you. Summit continues to be an event whose scale I have trouble absorbing. One of my friends who came last year, Sarah, thanked me for inviting her to a rock concert. It's kind of funny to be in the world of data and be able to have that kind of excitement and impact. I'll start with a big-picture view of both the disruption and opportunity that AI provides for many companies. Snowflake is definitely in that list. For most of us, work is an endless sea of tabs, and we are responsible for figuring out how to organize our time, how to organize the information that we consume, and then to figure out what to do with it. I joke to people that Command-Shift-A on Chrome is life-changing. That's really hard.
What we are beginning to see is AI changing the very nature of information work. What this means is that the data that your AI agents have access to, we'll get into what an agent is and so on, is critical. Integrations with the different pieces of software that all of you use, that I use, that is also critical. Overall, for an organization, governance over all of this data, security, is also a big, big deal because these coding agents are immensely powerful, but also sometimes don't have good judgment about what's okay to do with what data. This means that for an analyst, somebody that wrote SQL for a living, things are just very different. They go from effectively creating dashboards or writing one-off SQLs to creating what looks closer to software.
Within Snowflake, for example, we deployed skill packs that were specialized to different departments within Snowflake in a matter of four weeks. We've been continuously iterating on them. Using these kind of agent products, it's super intuitive for all of the non-technical folks at Snowflake. Brian is going to talk to you a little bit about the CFO experience on doing that. For our data scientists and our data engineers, they now think in terms of how do you automate creating an entire pipeline. For them, even adding a single column in a table used to be this endlessly tedious work of making stuff propagate across hundreds of files manually with people looking it over. That stuff is getting automated. For a lot of these end users, myself included, when I need to look at sales data, I don't want to be writing SQL.
Data integrations are seamless, which means that all of the information that I want is just available, kind of thought analysis at our fingertips. Deliverables for the smartest people that know how to take advantage of these products is no longer limited by how many hours they work. It comes down to how effective are they at using agents to get their work done. In fact, one of the idioms that we are trying to teach our software engineers at Snowflake is that they really need to be thinking of their work as being a tech lead of agents rather than an individual contributor that writes code one line at a time. It's a huge mentality shift. I'm not claiming that we saw all of this. I don't think anyone saw all of this.
I've talked previously about just the transformative power of even being able to access data faster, even in a pre-coding agent era. We started investing into this in earnest starting early 2023. One of the things that Christian, I, and many others have consistently believed is that AI is going to make the value of an easy-to-use, connected, and trusted data platform like Snowflake even more than before. These are our growth rate numbers for the past many quarters. As I said, during much of this time, our focus very much was on how do we create the definitive data platform that people would want to use if they wanted to get value from AI.
All along this journey, we also worked really hard as a company, I mean it, in discovering the basics of what does it take to create great products and launch them. Two years ago, I talked to you folks about how we were basically rethinking how we took new products to market, about forming v-teams that brought every specialized function back into a small collapsed team that could sit in one room and take a new AI product to market. I talked last year about how it was really critical that not just software engineers, but all of the solution engineers within our teams. These are the pre-sales folks that show the art of the possible with our customers, that help them get projects done.
I talked last year about how it was really important that they become AI native, because they could just get more things done faster. A lot of our success as a company, this was a trend three years ago, it was going down. Clearly, it's not. If anything, it's going up and going up well, has come from this back-to-basics approach of we need to create great products, and we need to figure out, as a company, as a team, how we take them to market. Even in this pre-agentic world, we are seeing the results of that effort, which a lot of people had to painfully reinvent how they worked, because people are happy sort of doing their own specialization.
It's awkward to suddenly say, "You're responsible for the whole, and you need to iterate a lot faster." All of that work has been paying off in things like productivity numbers. We measure the productivity of our expansion account executives in terms of how many quality use cases do they win per unit time, per quarter, per month. Similarly, we measure the effectiveness of our sales engineers, solution engineers, by how many use cases did they help their customer take to production. The number of use cases won per AE, this is not the size of the AE team increasing, but it's per AE, is increased by 86% year-on-year as we look at the quarter that just transpired. The number of use case go-lives per SE has increased by 58% year-on-year.
These are hard numbers to move because the average AE wins a handful of use cases per quarter. Even now, the way I think about scale processes within the team, and it's often an awkward conversation, is I routinely boil it down to what's the top decile doing, even within a population that on average clearly is doing better. We press very hard on what is the top decile doing that the rest of the team needs to learn. It's the process of continuous self-improvement that we think is really important for us. In many ways, that's the structural transformation of Snowflake as a company. Now fast-forward again to now. I have talked about, this is what my keynote was about, this is what Christian covered a lot of.
I think the future of work is very much all of us, all of you, me included, living in a new kind of environment. Just like all of us got used to living in a browser for most of our work life, or using our phones 24/7. What we see happening very clearly is that there is a new category, the agentic control plane, and that is going to be at the center of how work gets done. A lot of companies are going to be competing for it. For it to be effective, it needs to have amazing enterprise data and context. It needs to have all of the applications that that particular user is using and has context for. These are the Salesforces and the Workdays and the ServiceNows and the SAPs of the world.
Obviously the awesome models that seem to have no bound in their capabilities for what they can do. What we are very proud of is we have created products that can capture what this work is going to be, but in a way that is true to what Snowflake is. I'm under no illusions that competing with Anthropic on the quality of large language models that my team can create is a winning strategy. It's not. It's a failing strategy. On the other hand, we can go head-to-head with Claude Code when it comes to CoCo and say, "Here are the reasons why we are actually an important part of every customer's, and increasingly every partner's, data ecosystem." Most of our partners are here.
I've spoken to several executives already about how do we effectively have CoCo as a de facto implementation platform for all of the data work that their teams do, and this is hundreds of thousands of people in some of these organizations. On the CoCo side, we have measurable proof on a product that is very young. Our services team delivered a Spark migration 60-odd% faster working on behalf of a Global 2000 hospitality customer. A financial services firm saved over 500 hours on a job that they were doing. We often hear about migrations. At this point, the number of things that are possible with CoCo honestly exceed our imagination. We hear of people doing migrations that honestly we would not have thought about.
What we did do with it was sweat the details for creating a product that would truly be great when you worked with Snowflake. To me, this is the other reality of the current moment, which is a little bit of what the judge says about what they saw. You know a quality product when you see one, when you use it. Having that bar for creating amazing products matters more than ever. CoWork is even more ambitious. Obviously, it has its origins in Snowflake Intelligence. Back when we first launched Snowflake Intelligence, which was November of 2024, we saw it as a place where all analytic data came together. Part of what we are realizing, again, driven by large-scale use within Snowflake, is that it can be so much more.
Once you are able to access all of the common applications that you have, whether it's a Gmail, a Drive, and now even things like a Salesforce, and you have a platform in which work can be abstracted, work becomes very different. Again, is available right in your pocket or your laptop. We are earlier with these kinds of very large deployments of CoWork, but customers like WHOOP, tech-forward companies are figuring out how to use a combination of CoCo and CoWork to transform how their teams operate. In the analytic world, CoWork has already proved its mettle with any number of large customers. Folks like United Rentals or Domino's in Australia, or one of the largest banks, which is delivering a personalized solution for all of their exec staff using CoWork.
As I said, an important element of all of this product work is leaning into what is possible, starting with Snowflake as customer zero. We have talked about Snowflake being customer zero before, but I think we are practicing it at a very different scale and speed right now. Things are being co-developed. I want to show you one glimpse of how we are using these agentic platforms to transform how work gets done internally by our teams. This is an example of our support team using CoCo to transform itself. Let's watch the video.
AI has fundamentally transformed Snowflake's internal operations with Cortex Code, also known as Snowflake CoCo. CoCo understands your data better than any third-party solution. Since integrating CoCo across global support, solution engineering, and services delivery, we've unlocked opportunities that were previously out of reach. Before CoCo, customer support teams had to manually review cases, search across several systems, identify missing information, and route issues to the right expert. Here's one scenario of how it's transformed our customer support function. Each handoff meant reestablishing context for the new case owners. It was time-consuming and inefficient. With CoCo, that work moves faster. Let's look at how CoCo has changed key elements of the flow. Let's ask CoCo to investigate a specific case. CoCo securely connects to the systems behind the scenes, accessing case records, telemetry, Jira, source code, docs, various systems of records, and more.
CoCo understands what's missing, finds related cases, and recommends next best actions. It successfully retrieves detailed information from previously conducted analysis, all before the case was even picked up by the support engineer. Confirming the initial summary with the support engineer, CoCo proceeds to verify the issues in the case. Upon completing the verification, CoCo provides a recommended customer response and posts a case comment in with a suggested reply to the customer. The results, engineers are more efficient, handling almost 25% more cases. We are resolving cases 25% faster. Customers are self-serving even before they file a ticket. This is what happens when AI is built natively where enterprise data already lives. This is just one example of how CoCo is reimagining work. Every enterprise sitting on Snowflake data has untapped potential waiting to be unlocked.
Benefiting from data gravity, we think we occupy a key position in the world of AI, and we are very cognizant of continuing to be world-class in this. A lot of what Christian announced today was around continuing to be that trusted data platform, that governed data platform. Things like the Natoma acquisition are going to make that even more true in this world of agentic AI and agentic control planes. We think there's a significant amount of opportunity. Areas like observability are data-intensive problems that are ripe for disruption from people that are willing to think from first principles about what software should be. Honestly, we also get inspired by customers like Emmanuel that you saw yesterday, who came to us and said, "Hey, this is our data.
We want to rethink how my salespeople should interact with that data. I have the guts to say, I am willing to do that from first principles. I think that's the disruption. That's the opportunity that's there in front of us. I've stressed this in the previous two investor days that I've done with you folks. Strategy is fine. We think supercharged, great agentic AI products, AI control planes built on top of this incredible data foundation can be a great company. I stress execution a lot. That execution manifests itself in how are we able to move quickly in creating value. It's not lost on me or on Snowflake that we need to rethink speed when it comes to software. Living it is really important. I'll give you folks another example that's literal.
Right here, there's one engineer, just one, that's been working on a CoCo mobile app. This is one of these remarkably productive people that knows how to juggle eight balls at the same time. Yesterday, before we met a set of reporters, I had this momentary pang of doubt that I didn't know all of the launches that were going to happen at Summit. I had a list. It's a long list. I asked Christian, "Hey, where can I look?" Christian's helpful, and he gives me four documents. Not one, not two, four. Is that what he did? Four. I pasted into CoCo. MCP support's not yet been added, and so CoCo's, "Well, I got no MCP support." Thankfully, I put it into CoWork and that part worked. I had the list of launches, which was cool.
The more cool part was I go back to the Slack channel after we did the interview with the press folks, and I tell this person, "Hey, when's MCP support coming?" They go into slight panic. I usually like prefacing all my Slacks with low priority because people act faster than you really want them to, and you're the CEO. He's like, "No, no, we'll get it to you." Two hours later, he's like, "MCP support's added. Just update the app. You got it." Sure enough, I paste the same prompt back into the CoCo app, it gives the same summary. Execution really, really matters right now. Things like durable advantages need to be thought through. Christian speaks to some of these things, we pay attention to that.
If software is truly easier to create, what does that mean for the future of Snowflake? Where is the ongoing, enduring value? What are the products that we can create, for example, that can make CoCo much better out of the box than a Claude Code? How can it make it even more better for everyone else in the company if a set of folks use it? These are the feedback loops. I tell people, it's clear to me that Airbnb doesn't care about the cost of creating software going down because they create a network in the real world. Companies need to be thinking about what's the additional value, what makes this product better with usage. We spend a lot of time thinking through how do we execute to that kind of a vision, in addition to being a great data platform.
We want to be efficient on the go-to-market side. It's an enormous team. We get enormous leverage. We have talked to you many times about things like new logos. This was a remarkable quarter for us, because both the number of new logos that we won and the ACV, the total contract value that we got out of these new logos, both went up significantly year-on-year. That's because there are a set of people who obsess about this motion, who obsess about getting these customers onboarded, getting these customers live. That's the efficiency that I push for, that we push for. Part of the happy accident that happened with CoCo, CoWork and AI in general with Snowflake is the act of making these products broadly available to the entirety of employees at Snowflake, basically led to this explosion of creativity and ideas.
We didn't tell people, "You can't use CoCo because, well, you're not an engineer." Anyone can use CoCo. You can only access the data that you are supposed to see. We have governance controls on the data, sure, you can build anything. We saw amazing things like JB, our head of sales, he built a Streamlit app to look at his travel and entertainment expenses because he was sick of emails from Brian complaining about it. He's like, "Let me just look at it." That clearly changed his mind about what is possible. If you folks talk to JB, he will talk excitedly to you about how we can shift right in a massive way and have more of Snowflake sales team focused on delivering projects for customers. It's like we need to create outcomes faster.
Software engineering, as I said, is undergoing a complete revolution. Anyone that thinks that software engineering is about vibe coding is firmly stuck in early 2025. We are producing a set of, not we, like the world is producing a set of rocket scientists that are way smarter, can get way more done than the ordinary software engineer or even the excellent software engineer could do last year. By focusing on the basics of what Snowflake is about, what do we do? We make software, we sell software, we run software. That's the SREs. They've gone through a complete change similar to what you saw with the support team. They've completely redone how they look at operational problems, again, built on top of CoCo. We did this without buying new software. That's the magic also of the moment.
We are focused heavily on how do we make deployments go faster, because I see that as a final remaining hurdle. Obviously, we work with partners, we are also investing into a small team. I don't think of them as thousands of people. These are folks that know the best of what Snowflake has to offer as a platform. Go deep to understand what it means to solve a customer's problem and solve it as quickly as possible. You saw the results of some of that with Sanofi yesterday on stage. These are among the healthiest collaborations that we've had with a very motivated customer. We're doing similar things for large banks.
We anticipate that we'll be leaning into something like this as the impact of products like CoWork becomes obvious, and people realize that their data teams, like our own, have to modernize themselves for them to be relevant in this age of AI. The final comment that I want to make is that because we have invested so much in transforming ourselves, in being more effective as a company, and because of our ability to increase non-GAAP operating margin but also bring SBC firmly under control, we feel confident enough to say that we'll be reaching profitability at the end of next year. Brian will walk us through more of the details of that.
I see this as the culmination of the work that we have done over the past three years, to reinvent ourselves to be a more driven, more product-focused, more quality-obsessed, continuously self-improving company. With that, I'm going to hand it off to Christian.
Hello, everyone. How's it going? So good to see you. Whoa. Oh, wrong show. That was this morning. No, awesome to see so many familiar faces. I assume most of you attended the keynote this morning? Okay. Oh, some clapping. Good. CoCo is the answer. You got that. I will recap some of those innovations that we're launching at the conference, but I will also contextualize it for what is probably most interesting for all of you to think about it and how do you think about us as a company. To get started, I use the exact same diagram visual that Sridhar started with because it is truly a set of innovations that reinforce what we're trying to do here.
The more we've thought about this picture of the agentic enterprise, the clearer we are that the elements are data, AI models, connectivity to enterprise systems, and something that drives it. Sridhar, in the keynote last night, said something that is resonating 1,000% with customers that we talk to, which is the differentiation is not the access to the AI models. The differentiation is the access to the right data. That has created a sense of urgency in many of the customers that we talk to on, "Oh, I really need to go get my data estate in order." We have been saying for a number of years, you have all heard us say it consistently, no AI strategy without data strategy, and we're living it more and more on a regular basis.
The question that I think many of you are usually trying to infer or to get us to provide color is, okay, how do we differentiate? How do we stand out from the alternatives that customers have? Sridhar just mentioned it, but I cannot emphasize enough the easy, connected, and trusted. The keynote this morning had some fancier words, but it was the exact same, easy, connected, and trusted. I've arranged the set of launches and announcements that we have into these three buckets. With that, starting with easy. You know the answer, right? CoCo. Someone is whispering CoCo. It is true. It is not only on brand how we've thought about differentiating for a long time. You've heard us talk a lot.
We may be willing to give up some use cases where someone wants to turn knobs all day long because we just want people focused on productivity, business outcomes, business value. What has happened with CoCo is truly just we materially change that. I would like to say 10x that, 10x doesn't quite capture it. I shared this morning at Openflow, we added all these APIs, now we went from, "Oh, Openflow is cool, it's hard to configure," to, "I just ask CoCo, you configure it for us." Something at the encouragement of Sridhar, give him credit, every single launch that we're doing has to come with how is the experience simpler with CoCo.
In some instances, in many instances, we're starting to use CoCo first, then you go build the UIs and the APIs and all of that, because in reality, if you can just ask, "Hey, give me governance, give me interactive analytics," and CoCo figures it out, it's easier to build for CoCo private interfaces or internal interfaces than go and make it easier from a user experience or a UI. The emphasis on CoCo is not unwarranted. There are parts of a product that today, at Summit, they're only accessible via CoCo, and in reality, in instance, probably you'll never need any other way to access it. I cannot emphasize enough the role that it's playing for us. As we established in our earnings call last week, it is that nature that is helping the entire usage and use cases for Snowflake.
We announced a number of capabilities this morning. The way I would think about it, I don't want to go too deep into the technology. We're trying to eliminate the differences between the form factor. At the beginning, CoCo has a command line version, which is incredibly powerful, but it's accessible to a smaller set of users because not everyone is comfortable with a terminal window and a bunch of shell commands. On the other side of the spectrum, we have CoCo in Snowsight, which that one, the usage is quite broad because it's in the face of all of our users, but it's not as powerful because it didn't have the right sandboxing and security guarantees. A lot of what is in here, and I'm happy to answer questions at the end, but I don't think we need to go into those details.
A lot of it is eliminate that friction, bring the power of a command line interface, bring it into a desktop experience, bring it also into the hosted version of Snowflake. Now what we say is you get full power, but you still can sleep well at night in terms of the scope of actions taken by CoCo are constrained, whether it's on-prem or on your machine or whether it's hosted. The other thing that we're very excited is the CoCo Desktop. For those of you, I know that you actually are tracking very much or very closely a lot of what we do. Initially, when we had announced this research preview called Snow Work, it was all about we released the desktop internally to Snowflake. It took off like wildfire.
Everyone transformed how they work, that's when we said, "Okay, maybe this is a different way of working." At the end of the day, we have clarity. That desktop experience is CoCo. I think we're going to also follow with something like that for CoWork, which is the more governed experience. The desktop form factor has product-market fit inside of Snowflake. We made it available in public preview a couple of weeks, at the conference, it's generally available. We expect this to drive some additional usage of CoCo. Maybe I'll highlight here the Excel form factor, the VS Code form factor, as additional ways and surfaces for our customers to be able to get value of CoCo. It's not on this slide, I mentioned it this morning on the stage, which is we also put a CoCo plugin into the Cloud Code marketplace.
Of course, we would like to say that for data management operations, you don't have to use the indirection of Cloud Code to CoCo. If someone is already very committed to Cloud Code, there's a very easy way to plug in, and we've already heard from some customers in that situation, "Hey, this is ideal. I can use Cloud Code for application development completely unrelated to data or Snowflake, but I can delegate to CoCo all of the data management activities." Sridhar just mentioned migrations, and the way to think about what's going on in migrations is it's truly a reboot based on what AI has enabled. You all have quizzed us for years now on how quickly customers consume the contracts, how quickly they start with consumption, and we are seeing a massive acceleration of time to migrate.
I'll caveat, not everything in a migration is just what the technology needs to do. That piece, material acceleration. Sometimes there's things like, I will not be able to run this test because it's end of quarter, end of year, or I do a production freeze in the Q4 of my fiscal year. There's a number of constraints outside of the pure technology. Those were working, and some of the efforts that Sridhar mentioned help, but at least the pure coding, testing, all of that is materially changed. The middle column in here, actually, I talked about the middle column there. The one on the right, Spark. We've been working on more and more compatibility. I've been the one sharing with all of you that our engine is amazing. People that want to move from Spark, they want more compatibility. Guess what?
In a world where migrating from one type of API to another type of API is borderline free, we're starting to see different reactions. I mentioned this morning there's a customer that wanted some legacy Spark API, and it's a pain to support that legacy Spark API. We've been working on it, and we know that at the end of it's going to be compatible, but not super fast. We're busy doing that. In the meantime, the customer said, "Oh, we tried CoCo. We converted to Snowpark, and we're done. We're good," and it's five times faster and cheaper by implication. We have a renewed push on Snowpark.
We mentioned it to some of you that we are seeing increased momentum of Spark migrations use of Snowpark just because, hey, the pain that represented converting code is no longer as painful as it is. It's just easier to do. The last piece that I'll mention is the first column here, is the productization of an acquisition we made, a company called Datometry. What that company does, and what is now part of our migration suite, is lets us virtualize a Teradata experience. We hear from many customers, "I want out of Teradata, but I have all this stuff around it. I have scripts and applications and reporting systems," and changing all of that takes time. That's why some migrations of Teradata we've done are two years, three years.
What this virtualization lets us do is say, if this is Snowflake and these are the apps, we put a layer in between. That layer lets everything else in the enterprise environment think that it is Teradata. It looks like Teradata. It has Teradata SQL, it has Teradata scripts, and Teradata is a really rich functional product. To the rest of the ecosystem, it looks like Teradata. What it's doing behind the scenes is translating into Snowflake. When we did the acquisition, the converting it to support Snowflake took a few months. I don't know how long ago that was, six months or so. Now we're ready to start accelerating migrations through this, so we're excited about the opportunity to help customers move off of Teradata quicker. Sridhar mentioned it, CoWork is super important.
We do believe that it is the enabler to change how people go about their jobs. I think both Sridhar and I talk a lot about these tabs because at least personally, we use a lot of different apps, and you need to know where to go for what, as opposed to let's be more user-centric, and let's ask the questions for the request and something that you need to take action on. Just say it and then let the systems figure out the details. That is what's being enabled. That's also a big part of the shift of what we're doing here with the personal work agent. It's not about even more systems. Oh, they move from tabs to different agents. No. I have one entry point. It knows me, learns about me. It learns about what I like, what I don't like.
I was mentioning to a few folks that now there are so many things that I go do some interesting analysis, and then I say, "Hey, I would love to get a refresh of this once a week." I have CoWork doing all these things all the time, and I'm just getting regular reports. Now we introduce automation. Now I can say, "Oh, by the way, if you ever see this type of condition, just go take an action, email someone else, or do something like that." You can see how workflows are getting reinvented with the power of CoWork. I have one slide on artifacts and dashboards. This is, in my mind, what BI should look like if you were to start from a pure AI-native perspective.
It's not the goal of, "I have a dashboard, and then I'll see if it went up and down, and then I have to click 100 times." No, you ask a question, and then you get the visualization that most helps you understand what happened, and that's what then you go and share with others. It's not the other way around. By the way, BI was amazing for when it was introduced, right? It changed the accessibility to data, but it's still, here's a set of static views, and you need to figure out the answer. It should be the other way. You ask a question, we give you the answer, and here is a visual that helps you understand that.
That's what we're doing with Artifact Repository, live data, governed data, authored in CoCo, published with CoWork or made available to business unit CoWork, and we have the way for you to pin them down and say, "Oh, arrange these tiles effectively." It looks like the modern version of a dashboard, but it's curated for each user and what they want. Cortex Sense, we introduced it this morning. I'll be the first one to say it is early on, but the insight behind it is we have the ability to gather a lot of information that can help both CoCo and CoWork produce better results out of the box. I say out of the box as a contrast to today, if you curate enough semantic views and enough information, you can get all the results that you want with CoWork and with Cortex Agents.
What we're increasingly seeing is customers wanting, "I need answers now. I need to be able to roll something live as soon as possible." That's what Cortex Sense enables. In a few of the conversations we had last week, there was this question on how does it compare what you're doing relative to what a coding agent does? I'll caveat, this is one evaluation set. It's not fictitious. It's a real customer valid use case, but all of this is to say mileage may varies. I still pride on not generalizing where you don't have the power to generalize. What you see in the front row is leading coding assistant trying to interface with MCP for SQL and asking questions about the data.
Second one is CoCo and CoWork as we know it right now, the last one is CoCo and CoWork with this runtime context that Snowflake gathers to say, "I know enough about the user, the data," all of this to say, "Here's additional information," you see both material improvement in quality, but a lower cost. If you're thinking, "Well, how can it be lower cost?" There is a huge amount of cost and tokens that go into, "Oh, yeah, my bad. I didn't get it right. Sorry. Let's do it again." That burns a lot of cycles, as opposed to if you know what the question is, how do you answer the question, it translates to better economics.
We're extremely excited about this, and again, I'll caveat it to we're in the early process of testing the different scenarios, different customers, et cetera, but this is a big differentiator for how we think about out of the box, our agents, CoCo and CoWork, just produce better results for customers. Sridhar mentioned the acquisition of Natoma. This is one of those three key elements on the agentic enterprise. Important differentiators both for data administrators as well as for users. Data administrators, one, it connects to 100-plus business systems out of the box. Number 2, it enables those administrators to have policies on what agents using these MCP connectors can do. We were sharing the example. At Snowflake, the way we configure the email connector of Natoma was you can ask your agent, CoCo or CoWork, to send an email.
If the email is going to an internal recipient, it sends it. If the email is going to an external recipient, it puts it in your drafts. That's a way to force humans to do it. That's a policy we chose. The cool thing is Natoma lets our customers decide what policy they want. If they just want to spam people, that's fine. If they want to be more conservative, even for internal, they can do it. Third benefit is see and audit everything that's happening with these agents. Because if this is the new way for data to get pushed out, if, "Oh, we just sent some of your sensitive data to a connector that was going to do, I don't know, Slack or email," you want to know all of those things.
From a user perspective, instead of me authenticating with 100 different systems, I authenticate one to the gateway, Natoma, and that gives me authentication to the other 100 systems. We introduced Datastream this morning. The goal of this is to move Snowflake upstream such that it can have a streaming solution, capture data when it's created, whether it's a sensor, a device, a website, and be able to land it into Snowflake with almost no administration, very little management, and very low latency. We're very excited about this. It's in the category of it's early but quite promising. Pillar number 2 is connected. The integration, the interoperability that we're showing with Iceberg, I would say it's second to nobody's. That's a hard statement to make, I do it based on facts. We are truly committed.
We are steering the Iceberg standard, but we're also being among the first at implementing it. Right now, we're the broadest in terms of the implementation of the V3 spec, and we're steering the V4 spec. I mentioned this morning, we integrated all the REST catalog APIs into Horizon to make sure that we can interoperate with data regardless of where it sits. Even if it's on Databricks, we can read and write data. Other engines with Glue and others, they can read and write data that sits in Snowflake. This whole notion of I'm locked in and I put my data, that's not an excuse. Customers, please use whatever gives you the best experience, the best performance, the best economics. Okay, this one is super important for all of you in this room.
When we introduced Iceberg, and I think some of us regret how much noise it caused, but one of the things that was factual was with Snowflake, you store it in our format, and you pay for storage. With Iceberg, we always said it's customer-managed storage, but there's no reason for that trade-off. We introduced, and it's generally available here at Summit, Snowflake-managed storage for Iceberg. You can still be interoperable, but we'll do the management of the storage. We'll give the economics. That I think is going to be an even better tailwind relative to at least how we thought and modeled the adoption of Iceberg. Sharing, I think all of you know I'm personally passionate about the network effects, personally passionate about the unsiloing of data and helping organizations connect. I am very excited that we're finally breaking out of the, it's two parties, unidirectional.
How do we do multi-party collaboration and symmetric? It starts all with our Data Clean Room. This is productization, an evolution of an acquisition we did a couple of years ago. We're starting to see very interesting media use cases, advertiser, buyer type of collaboration, the use cases that helps collaboration, and this obviously gives us some structural advantages, the more parties you have exchanging data via Snowflake. We also talked about zero-copy partnerships. The marquee one that went GA actually last month was with SAP. We have a few initial deals of people buying into this integration. Today at the conference, we announced the expansion of some of the Workday integration we've done, new integration with IBM, a new integration with AVEVA. Then I could spend as many hours as you want on the importance of trust.
We gave it a decent amount of air cover this morning because I think this is what changes how Snowflake fits into the adoption of AI for enterprises. Rolling out AI, easy. Rolling out AI in a way that people can truly sleep well at night is not as easy. Horizon, the catalog, is where all this comes together. Horizon Context is where we brought all the explicit semantics and information for agents to be able to work well together. Again, there was a slew of announcements. I left a number of things out from this morning. There's a lot more during the conference on how do we help govern agents, which is security policy, identity for agents, data movement, exfiltration protection, all of those. We also talked about Adaptive Compute. I put it in here just because there's something I want all of you to be clear on.
Massive performance improvement. We're doing the exact same thing that we did with Gen2. We priced it in a way that we're aiming for revenue neutrality. Our customers get the benefit. That is materially faster. None of you need to go change your models. We're still good, even though we're trying to push very hard for the adoption of all of this. Interactive analytics for responsive experiences. You can say this gets into the ClickHouse type of workloads. I am always very careful to make absolute statements. It's incredibly competitive to the analytics customers have out there, and we're going to be making a big push to get adoption of this. With this, there's the Data Cloud as a whole.
This is the data for enterprise data managed govern part of the solution, it is part of the bigger picture that we share. Happy to chat more at Q&A. Hopefully, this was useful. Now we're going to turn it over to Brian. Thank you.
Thank you, sir. Thank you, CK. Super, super fascinating, all the product releases and product velocity that we're seeing. Thank you to each of you for coming out today. There's some of you that have been around the story for a long time. There's some of you that are relatively new. I'll walk in and talk about the market, some of the revenue drivers, get into GAAP profitability, capital allocation, so forth. The market today is roughly $225 billion. We expect that market over the next five years to more than 2x to over $460 billion. AI is expanding our market opportunity. Last year, we went through this. Over the five-year period, our market has grown roughly 30%. We really have conviction in this when we look at our large customers.
The top 25 large customers, they spend on average $34 million a year with us. That's grown over the last two years from $22 million a year. When we look at our Fortune 2000 customers, G2000 customers, in FY 2026, they on average have only spent $2.4 million with us. We feel that we have the right to actually increase those customers up to our large customers' spending. We'll jump into the core growth drivers of the business, primarily in the Core Data Platform and our AI workload. Let's jump into the Core Data Platform. Landing new customers is absolutely essential for us. When we land new customers, they don't add that much in year one or two from a revenue perspective, but it's fundamental to long-term durability of the business. When we land those customers, expanding them are really important.
We have one of the best-in-class net revenue retention rates. This really drives stability and expansion into our customer base. We're able to expand our customers on a lot of the AI workloads that Christian and Sridhar talked about. When we go to our customers, we sell business outcomes, which is really helpful. I've got a lot of favorite charts in the deck. This is one of my favorite charts. With the high gross retention and all cohorts expanding, you can see from FY 2019 to FY 2020, FY 2021, those customers are delivering the majority of their revenue this year. With the land motion, they expand over time with a high gross retention rate. This is really a powerful revenue engine for the company. We do that through a number of different ways. I want to touch on migrations and use cases.
Migrations from FY 2025 to FY 2026 grew 1.9 times. Use cases grew 1.7 times. Very meaningful increase. We actually are able to get this wallet share from a number of different sources. Where the real benefit comes in for us and our customers is when they're consolidating all this into a single platform at Snowflake. All right, let's talk a little bit about sales compensation. We use sales compensation to incentivize growth. You can look in FY 2024, we did not compensate on new customers. We made that change in FY 2025, it really paid off in FY 2026. We actually will go through and continue to make tweaks to the sales compensation model to get the most out of the sales organization to deliver the most for our customers. This is the fundamental pillars of our sales incentive compensation.
We announced a new CRO in first quarter, JB, as we call him. There's really two core focus that JB has. One is stability, and two is AI. From a stability perspective, JB's been with the company for over 10 years. We joke about it internally that JB actually bleeds blue blood, because he's been here so long. He actually pioneered the use cases to customers, which is used throughout the entire sales force today. He's taken that and other things that he's learned and actually using that to leverage AI. Every rep today within the company uses CoCo and CoWork. When we go to a customer, we're actually using synthetic data and creating applications to deliver outcomes for our customers. We've actually changed the way that we're selling to our customers and doing outcome-based pricing.
Our reps have first-hand knowledge of the capabilities of our products and how to deliver that. We're extremely pleased with the first quarter that JB delivered and look forward to many more quarters. AI accelerates growth. I got three charts up here. If you look at the chart to the left, the sales cycles are accelerating. When I went back and looked at average days sales cycle for this last quarter, it was the lowest in the last four quarters. You would expect with more choices and more evaluation out there that the sales cycles actually expand and actually take longer. They're actually doing the opposite of that. The sales cycles are accelerating. We constantly talk about how we are using our AI tools to actually get new customers to consume faster.
We've taken that from 10 months down to seven months, and we're continuing to see how we can decrease that. For all customers, migrations are accelerating. We've shown a 40% improvement. AI across the business is having a dramatic improvement on our time to ramp. Let's talk about a couple customer examples. Before I dive into specific examples, I'll talk about some broad trends that we're seeing. One of the things that we talked about on the last earnings call was there are secular tailwinds that we're actually benefiting from in the overall industry. Second is CoCo is actually expanding our personas that we're selling into and allowing more people to consume.
Thirdly, in our base business, we saw an acceleration in the base in the last quarter as it relates to CoCo and the secular tailwinds. This particular customer was a large customer that was an equipment company, and they adopted Snowflake CoWork. They adopted CoWork because they had over 1,600 locations, and the reps were having a tough time getting the information out to the customers and answering them in a unifying way. They use AI agents to actually build the responses with CoWork, and they were able to answer with greater consistency across all the 1,600 locations, faster answers, and increase the customer satisfaction. If you look at the next example, this is a semiconductor company. They originally bought deployed CoCo to optimize their queries and save money. Their supply chain department then actually picked up on CoCo and started using it.
Within the supply chain department, they were having a problem because it was really complex manual calculations on ordering inventory. Through the use of CoCo, they're able to decrease the time, increase the accuracy, and reduce the cost. This is another example of how we've seen CoCo play out in our customer base. In both of these, as you can see, CoCo is actually increasing the consumption of those customers. All right, let's talk about how this is working at Snowflake. I joined Snowflake a little less than a year ago, and when I first got here, I started playing around with AI tools. I can tell you that CoCo has dramatically changed the way that I work. I work so differently today than I worked a year ago.
I think each of us with AI are trying to retrain the way that we actually work. I'll give a couple examples on this. I've talked to some of you about my Good Morning CFO skill. Every morning when you come into the office, there's probably a number of websites you go through, reports you look at, structured data, unstructured data. When I get in in the morning, I go into CoCo and I type Good Morning CFO. It basically takes all this data from structured and unstructured sources and actually puts it into an easy-to-read format within minutes. I'm able to go through changes in the sales forecast, new hires that join the company, customer releases, and what's going on from the news perspective, major account wins, a number of different things.
Not only can I do that, I can turn it into visualization, automation, within CoCo. Our Natoma acquisition then allows me to connect it through our MCP servers to Gmail, Slack, and so forth at the application layer. Now I'm using CoCo really as a destination spot to actually work out of in the morning where it's bringing all this stuff together, saving an immense amount of time, and allowing me to actually send emails out around the world, understanding all this data in one simple place. It's also changed the way that I actually work with my FP&A team.
Historically, you'd have a list of reports that you periodically get, and on those reports, there'd always be one number that you would look at, and you would say, "What's going on with this number?" You'd go back to your power user and say, "Could you go extract that data and actually give me another set of reports so I can look at this data?" That process would go back and forth for probably two or three different times until you could draw a hypothesis about what the conclusion was. Now, with CoCo, if I see something, I actually, through natural language, not through a SQL query, it does it for me, go and query the data and inquire what's going on. I can drill all the way down to an account level, to a product feature level, and understand anything about the forecast.
It takes something that would take weeks to do down to minutes. We're also using CoCo in a number of different ways in the CFO organization. We have over 139 use cases deployed today. As Sridhar said, there's heroes popping up all over the company. People love to use CoCo and see what they can do with it, so they're automating and transforming the way they work across every aspect. Let me talk about how this is resonating with customers. I have the good fortune of speaking to a lot of customers. Two weeks ago, I was in London. I met with over 20 CFOs. There's two trends that are actually emerging. One is there's a different persona that we're selling to, and two is there's a larger sense of urgency.
When you take some of those use cases into a CFO and show them what you can do through natural language, not through a list of static charts through a BI company, but what you can do through natural language and automatically drive that into visualization automation, like the light bulb clicks like that. There's a couple accounts today where it's a CFO of a $10 billion company where they're going to do a big contract renewal, and I'm actually the lead with him on the purchase. We're selling to a new persona. We're constantly going in and selling to CFOs. The second thing is the sense of urgency.
When I go to our customer executive center and meet with customers, not only are you seeing just the chief data officer come in, you're seeing the entire management team, in some cases the board, and in some cases, they're actually bringing in a whole list of partners. It really is changing. I think when the newer models was released and now people are deploying AI more broadly, I really do believe there's this greater sense of urgency around deploying something. All right. Let's get into margins and capital allocation. Sridhar and I are 100% aligned. You can grow while getting operating leverage in the model. You can see in FY 2025, we delivered 6.4% non-GAAP operating margin.
On the last call, we just guided to over 13.5%, over 2x within the 2 years. We're doing that by keeping non-GAAP product gross margin flat at roughly 75%. We talked about the AI products have a lower gross margin than our core, how are we doing that? We're doing it really 2 ways. One is we're being extremely disciplined on headcount. Last quarter, we reported, absent of the Observe acquisition, net headcount increased only by 17. The quarter before that, only 37. We're changing the way we work by using AI tools and necessarily not adding heads to get work done, but transforming how we're working. On the other hand, cloud spend has gone up a little. The offset of these 2 is giving us operating leverage in the model.
The billing payment terms for the company have been really consistent. Over 80% of the people pay in advance. You can see, last year it was 86%, two years before that was 82%, but really remains consistent. There is some noise between billings growth and revenue, though. You can see on the far left-hand side, in FY 2024, revenue grew 36% and we had 29% billings growth, but in FY 2026 it is actually reversed. When our customers actually come to the end of their capacity or use the capacity that they purchased, they really have two choices, and we allow them to do either of the following. They can actually do an addendum and actually extend the current contract when the contract renewal comes up, or they can actually do an early renewal. This causes some noise between billings and revenue.
Over a long period of time, this actually normalizes out. If you look at the 3-year, they actually are normalized. All right, let's talk about GAAP profitability. We're super excited to announce this today, and the leverage that we're getting in the model while seeing the revenue growth that we have. We announced that we'll be GAAP profitable in 4Q FY 2028. There's 3 levers to do this: revenue, operating expense, and SBC. We actually went and just played with the bottom two, operating expense and SBC. This is not a discussion about FY 2028 revenue, we're seeing greater operating efficiency and operating expense. For modeling purposes, to help you out, assume the same trend in SBC that you've seen for the following few years. We're at 41% of revenue.
Last year, we were at 34% of revenue, and this year we said we'd be 27% of revenue. That should help you from a model perspective on how we're going to reach GAAP profitability in 4Q28. With that as well is, we don't expect to do any large M&A. Okay, we'll jump into capital allocation. Three primary areas. One is organic growth, R&D, and sales and marketing. You heard Christian talk a lot about, from a product perspective, what we're releasing into the marketplace and the velocity that Sridhar talked about in his presentation. We'll continue to do that. From a sales and marketing perspective, it's really adding capacity to the capacity model, where needed. We have about $800 million left of authorization, our $4.5 billion that was announced earlier.
From an M&A perspective, we have typically done small tuck-ins on a buy versus build, more of an acquihire perspective. In conclusion, we have a very large and growing market. We have durable growth drivers with the land and expansion motion that we have. Our customers are fanatical about the products and services that we deliver. AI is accelerating all aspects of the business, and we've given you the framework today for us to reach GAAP profitability in 4Q28. With that, I'll invite Sridhar and Christian back up on stage, and there will be some mic runners running around, and we have roughly, call it 25, 30, 35 minutes, clock's still going up, for some Q&A. If you have some Q&A, please fire away. Got one up here, and Keith is going back. We got Carl.
Okay, great. Yeah. Happy to kick it off, and thank you for today. Maybe this is for Sridhar and Christian. OpenAI did an event this morning. I'm sure you were too busy to have listened to the livestream, but they announced a new data analytics product. The spirit of the question is, how ambitious do you think the frontier model companies will be over time in vertically integrating down into the data layer? Do you feel like the way this is going to play out on the next three to five years is that they'll partner with firms like Snowflake and your peers, rather than go after it with first-party products?
Yeah, I can take a first cut at this. I think the market in front of them in the enterprise, which is to actually get every company to rethink how work should get done, starting with things like software engineering, is very large. I suspect that that is where the bulk of their attention will go. Running products like Snowflake is a whole new set of both practical and operational skills. Having said that, as I emphasized, that software is changing so rapidly that people should not be in the business of making long-term predictions about what is possible and what is not. That's my current best answer.
I don't think I have much to add other than a lot of what we talk about that, yeah, querying data, not so hard. Doing so with correctness, with trust, with all of that takes some more time. I share the alertness that Sridhar has instilled in all of us, which is just simply to pay attention to what's out there, what's working. In many ways, I am seeing the dynamics with the cloud-
The AI model provider is similar to what has happened with the cloud providers, where, yeah, there may be some overlap, but at the end of the day, we're more complementary than not at many customer sites, and so far it seems to be very similar dynamics.
Yeah. If anything, just building on what Christian is saying, absolutely. The cloud providers, as you know, have data platforms. They also quickly get into this mode of, yes, we both need to partner and compete. In certain sets of customers, we will be competing, and we'll sort of stay separate in that and be in our lanes. While in others, we collaborate. We have an excellent working relationship with both the model providers. I actually think that the world is headed to a place where most companies want a certain amount of model independence. You don't have to squint that hard to understand that being reliant entirely on one model provider introduces the same kind of dynamics that sitting on exactly one CSP does for your business, especially if it's large and varied.
We didn't get as much into it, but we spent a fair amount of time making sure that both CoCo and CoWork work effectively across all models. As others, you saw the partnership with SpaceX, but we also watch the open source models carefully, where if their performance rises up, that's actually very positive for us because we rent GPUs from the hyperscalers, and we have excellent infrastructure teams that can help us run that at scale. It obviously produces just different margin profiles than working with the large model makers. It's pretty early for all of us.
Let me add one more. Sorry, but we're riffing on each other. Your comment reminded me. We're starting to hear customers tell us, "Oh, I made a big commitment to this AI model company, but now I want to use the other one." That dynamic we saw with the cloud providers, and it's starting to benefit us, which is, hey, you may have made a commitment to Snowflake. We'll give you model choice. That, I think, would make me pause on, do I want to go all in with one company in a world where nobody knows what the world looks like three months from now?
Thank you both.
Right here. Yeah. Thank you for taking the question. Sanjit Singh with Morgan Stanley. I think as a management team, you guys have been very front-footed in terms of acknowledging that the world is changing, and it's changing fast. Even with the presentation today, I think you gave us a clear sense of where you're making your bets and where you're going to invest behind products. I'd love to get a sense of, having been at multiple of these analyst days, you guys have reached a tremendous amount of innovation in terms of products. Can you give us sort of a real-time view? I got a good sense of where you're focusing going forward.
Are there parts of the product portfolio that maybe we've discussed before on just container services, Unistore, the data engineering portfolio that you're pulling back because the world is changing and this is where you want the team to focus? It's more of a kind of like a portfolio allocation question, Sridhar and Christian, give us a sense of where we're headed.
Yeah. I think one of the principles I live by is all of us can have theories for what's a great product and what's going to achieve product-market fit. None of us are, in fact, capable of willing that into existence. That's just how it goes. By the way, all of the usual instincts that people have for how to will PMF into existence is usually some variation of, I'll give them more attention and I'll give them more people. Some unhappy combination of both of these typically produces the opposite outcome of actually trying to get product-market fit. Even in the world of AI, the thing that I told Christian flatly in July, I was the sponsor, the first sponsor of the CoCo project. I told him if by September, October, we didn't have traction, we should walk.
Because, as I said, everybody talks a big game about their ability to do things, like super app announcements galore, but PMF is something very special. In areas where we perhaps had a thesis for what could be, I would put something like Native Apps into that, where we made a substantial investment. You might mind substantial investments in early products are a mistake, but you can't change the past. If we basically deconstructed that into what are the core capabilities that come as a result of that way of thinking. A Native App is just another way of saying, I want to share both data and code from a provider to a customer and have some rules for who can see what data and who can see what code.
We deconstructed that into a set of capabilities, and we're not pushing Native Apps as an end-all, be-all concept for applications quite so aggressively. It's a very slimmed-down team. We are being thoughtful about where do we pull away from. Vivek is actually really good at extracting leverage from the teams. I mean, the SRE project, that's like the ta-da part of the project, which is, hey, you don't need to spend so much time dealing with annoying pages at 2:00 A.M. in the morning. This concept called KTLO, Keep the Lights On, a bunch of our teams have this.
That number has dropped by a lot. What am I getting for it? Do we need to move some people from this team over to this other team where we think there is promise? That kind of reallocation is very active. This is not product, and where there are tougher decisions to be made about disciplines changing. Christian and I came to the unfortunate and joint conclusion that tech writing didn't need to be this independent job thing anymore. We effectively disbanded the team. We got a bunch of grief because of it. Our take is like, LLMs are better at writing documentation today using a coding agent than a specialist whose job it is to write documentation. It looks obvious in retrospect, but when you make it's still painful.
I think we're being pretty flexible about where we are allocating, where we need to pull away from, and we'll continue to do that.
Actually, 100%, I have more examples. For example, SPCS as a third-party customer bringing your workload, yeah, has not gone the way we envisioned it. I just shared the cloud agent sandbox that is enabling a lot of power to CoCo. That is the same team and borrowing technology from what was built there. The momentum with Notebooks, the momentum with Streamlit came from, hey, instead of pushing SPCS so hard as something external, make sure that you have a great runtime environment. Some of that reallocation is happening for sure. I'll give you one other example. The interactive workloads, that borrowed so many changes that had been done for Unistore, which is why we were able to go turn something around in way less than a year, and the performance of that technology is amazing.
Yes, there's for sure reallocation, and there is reuse of technology we built.
We'll go right here, and then-
I just looked up the data analytics announcement from OpenAI. These look like lightweight skills that run on top of Snowflake. We knew about this. I didn't quite make the connection. I think of this more as a set of skills that let you answer analytic questions, and can generate SQL for Snowflake, but also a bunch of other platforms that were mentioned. First of all, it's not under the data layer. It's more about how do you use these from Cortex Code. This is something that they've actually talked to us about. I have to give both the model makers credit for being really good partners. You saw that with Daniela yesterday. We work effectively together. There will be some cases where they would prefer, obviously, to have Cloud Cowork rather than CoCo or Snowflake CoWork. That's fine.
I think there's an element of maturity that we have about how we approach these relationships.
Yeah. That's 100% calling into our MCP SQL connector.
That's correct.
We gave them a quote on Friday.
Right here.
Hey, guys. Alex Zukin with Wolfe Research. You guys are riffing on each other. I'll riff on Carl and Sanjit. The question I think a lot of investors, and even customers have. We watched the presentation. It's full of innovations. It's full of new products. I think we're having a little bit of a hard time seeing or understanding the collapse of functionality and the consolidation of functionality across the models, the hyperscalers, the apps. They all seem to be in this world of delivering you the answer that you want from any question that you ask. I guess, I think I know the answer. The answer is CoCo. The question is CoCo your way of driving a lot more consumption of the core, or is it more about expanding beyond?
Like when Brian is talking about new personas that you're selling to, how those look, how those feel, those sales cycles, those lands. When is CoCo the right answer versus Cloud Code versus Codex versus whichever model Microsoft launched today?
First of all, I think, CoCo and Snowflake CoWork are really two sides of the same coin. They share an enormous amount of infrastructure underneath. It is tailored for different personas. All companies, definitely Snowflake, need to have a clear-eyed view of what they're good at. Everyone can aspire to more, you need to be very clear about what you're good at. We are amazing at being a data platform. CoCo is all about how do you get value faster from the data platform. It's pretty much how do you bring, it's equivalent of your AUM. How do you bring more assets to be managed by Snowflake, either directly into Snowflake or on Iceberg, we're increasingly indifferent to those things. How do you take data through its value life cycle?
We feel very confident, and we have published benchmarks comparing Cloud Code to CoCo on things to do with Snowflake, of CoCo being a really good, perhaps the best solution in the world for working with Snowflake. You can say, "That's not a big deal. It's your platform." Yes, it's not a big deal, but it's not like everyone has the equivalent of CoCo for the products that they are creating. Still is a lot of sweat, still is a lot of work. While we have a lot of work to do in terms of driving CoCo adoption by each and every one of our customers, it's still pretty early. When we talk about the 7,000 customers, it'll be more the case that there's one user that's gone and done it.
It's not that they have switched over to this agentic way of operating, and it's not like we have made migrations easy enough that anyone can migrate from anything into Snowflake. There's work to be done. I would say, in the near term, there's just, that's where there's enormous potential for us, but it's sort of playing our game of be a good data platform, CoCo activates everything that you can do with that data platform, including creating products like CoWork that people can get even more value from. In that sense, CoWork is an expansion play from where we are. It's early. Our aspiration there is that we get some mega deployments. When we first created AI products, and I'm positive I said this last year here. I usually like having clarity about priorities.
When we talk about AI, well, I've always said create world-class products first. That matters more than anything else. Create products that customers love. Get marquee folks to adopt the products that you create. Any such breakthrough is inevitably exceptionally difficult because you have to prove, and you have to get the customer to trust you. Then drive, scale that option, then drive revenue. Margin will come if you have done one, two, three, four right. CoWork has to go through this kind of a motion. Snowflake Intelligence, its predecessor, which is mostly an analytic product, has done well. It placed us firmly on the AI map. It generated a lot of momentum and relevance for Snowflake as a product, but CoWork is like a giant step ahead in terms of the things that it can do.
We very much have to prove ourselves, in terms of getting the product deployed by large departments like we have with Snowflake. The fact that I can get Brian to talk to every CFO and look them in the eye and say, "This is how I transform how we operate," is a huge asset. It's the same for JBNT. We have to translate that into the logos, into the 10,000-user deployments. We have to get customers happy with the cost of spending money on AI. We have to convince them that the per-unit cost of using CoWork is a lot less than paying for some amount of subscription software. That's the potential, but I'm the first person to say that is early, and we have to prove the scaled use cases to you. Between the two, CoCo sells itself. Why?
We have, whatever, 14,000 customers that love Snowflake. I just go to them and say, "Everything you do with Snowflake is going to be 10 times faster." They'd be foolish to not go try it out. CoWork, we have work to do to sell it and prove ourselves. No riffing.
We got one third row right there. Can you just try to get the microphone a little bit back?
Thanks. Ittai Kidron from Oppenheimer. Thanks for the presentation. Super interesting. Maybe I'm going to ask a little bit of Alex's question and the opposite of Carl's question a little bit earlier. Going back to CoCo, Sridhar, the first thing you said in your presentation right here, right now, is that data gravity is a major advantage that you have compared to everybody else. Add on top of that context, which again, you have compared to everybody else. If we skip forward in time two, three years, how do you think about what CoCo can really evolve and develop into becoming? Carl asked the question of what happened if the model companies go down into the data. I would argue that based on what you said, you have a far greater improvement and advantage right here, right now.
Why not go aggressive upwards, not into the model itself, but more general coding agents and many other things that you can attach to the data and the context that you bring to the table that others don't have?
It's a great question, but execution eats strategy for breakfast. I have the strategy, just have to get the other part right.
What are you planning for us in two years from now? Open up the kimono a little bit.
You should not be in the prediction game when things are improving by 20% every month. I'll just honestly tell you.
That's tough to sell for us. We're analysts. That's kind of what we have to do.
That is, I think, part of the conundrum of the world right now. I read this in this amazing book, where he basically says all history writing is teleological. People usually write history by assuming that the path that you took to get there was preordained, and then they write the history. I think it's just really hard to tell right now. I think we have clear ambition. We have a team that is willing to execute and live up to what we think other companies can be. We have value to show. The rest of it, whether we do it with our own people that can help with deployment and set the stage for what is possible or, whether we come up with a set of effective partners that can drive change through a lot of customers. I think that is part of difficult execution.
We'll go over here to Rob.
Hi. Thank you. Rob Owens from Piper. Great day from a new product perspective. One thing I'd love for you to double-click on is just the data streaming opportunity. Is this something that's customer-driven? Is this part of your bigger vision as to where Snowflake is going to fit in the future? The answer can't just be CoCo. It's got to be something else.
Actually, Dave, this is a very unique and differentiated-
Go for it.
Please go for it. Go for it.
It's both. It's part of our core direction of travel, which is we want to help customers through the entire life cycle of data. Those are not empty words. It's truly the entire life cycle, and there was a big gap upfront. Today, customers deal with technology that is frankly complex to manage and expensive to go from web logs and sensors and devices and apps and mobile phones all the way to Snowflake. One piece is direction, but the other piece is we do meet with our closest customers. We have a forum, we call it the Black Diamond Council, and the signal was they're very clear. If you guys were to solve this Snowflake style, we'd love to do it. As soon as we started sharing details, the interest is very high.
Maybe the third piece that I'll say that is very interesting in that space is there's a technological disruption happening there, which is the original streaming systems all kept the data in memory, which made it insanely fast but also incredibly expensive. We've seen a number of entrants and players delivering something that has a separation of storage and compute. The data is in cloud storage. It's materially cheaper, a little bit slower, and many customers have said, many of you companies that are represented here in the room saying, "We're totally fine with that trade-off." All of those things come together. That's what led to the opportunity, and what Sridhar not only he reminds me on Lit Pro 101, but just say, the thesis is there. Now it's on us to make sure that it's a great product and delivers on the thesis.
I'll stress that point again. I think the act of recreating something is people tend to look at it much more as there's something over here, it's a product, it's some system that works, it's a company, and they think they can essentially deconstruct that and somehow get to that point. As many of you know, I spent a lot of time at Google. I used to have endless arguments with Larry about what innovation meant. Part of the thing that he drilled into my head that stays with me to this day is, you'll never win by aspiring to be someone else. You have to find your path. It's the journey that matters. That journey usually starts with a brilliant new insight.
Much of streaming was designed for all of you, was designed to minimize the amount of time that it took to get the data from the stock exchanges of New York to the data centers in New Jersey. It's just like all of your teams wanted to squeeze that millisecond out. Everything was in memory, everything was RPC, remote procedure calls, machine to machine. People optimized the heck out of it. Obviously, it's sort of expensive. That insight that's been had a few times before is that most people don't really care that data shows up in 20 milliseconds. If it's 400 milliseconds, they're like, "Eh, it's fine." Is it 10 times cheaper? That's the core underpinning of data streaming, which is still a bright, new insight.
By the way, if the model companies were to want to disrupt Snowflake, it has to be some new thesis like, "You know, I can think about this just very differently." It's not, I'm going to compete with Snowflake and operate at 22.5% margin, whatever the margin they want to operate at instead of something else. Maybe it works for Bezos, once upon a time with the retail industry that was unwilling to see the internet. It's just not something that works in a general way. It starts with this bright insight for this is how you rethink. This is the origin story of Snowflake. All of you folks know it. It started with that one core kind of thesis. Still, it's the journey that matters, whether we can create a successful product, get people to adopt it. There's a lot of hard work ahead.
Got the next question up here.
Great. Hey, guys. Brad Zelnick, Deutsche Bank. Great summit. It feels like an innovation blizzard this year, hats off to the entire team. I guess my question is, in a world where enterprises are over-consuming tokens, beginning to question ROI, and even putting in curbs and usage limits, it was great to see, Christian, your slide comparing CoCo plus CoWork and the promise of Sense ahead versus general-purpose code gen. How does Snowflake position itself to be insulated from an inevitable wave of token optimization to come, and even be part of a solution and to be able to benefit from it? Thanks.
Yeah. I think absolutely how much tokens are used, what models are used, what cost is a big issue. I also think there are lots of really good technological solutions that we feel confident. This is where having control over the harness, the thing that's actually executing the plan is so very important. One of the things that one of the engineers and I co-developed together a few weeks ago was a Skill that would generate a plan for how do you solve a complex problem. Part of what you can do when you do things like that is you can have sub-agents work with smaller models. Similarly, I'll point back to our advantage, if there's an open source model that's perfectly great at some job that also happens to be hosted by Snowflake, we can use one of those models.
We don't always have to use the marquee name models for every job. In fact, we make these models available within every Model Garden, and people run a lot of jobs using much smaller models as well. Similarly, I think techniques like Skill compilation. A Skill is an English language recipe, but 90% of the time, the Skill is actually doing something fairly deterministic. You don't need a fancy LLM to do the deterministic part. There's an experimental project for the most common use cases of value Skills. How do you compile that thing down into code, where basically the code is invoked instead of the big giant LLM to interpret the English? I see a slew of techniques like this show up as there are concerns about cost.
Honestly, we also want to put them into products like CoCo and CoWork as we continue to innovate with them. One of the things, ironically, that Christian and I are quite happy about is that things like Snowflake optimization is a lot easier with CoCo. Even though people optimize more with CoCo, we actually get more consumption anyway because they just do a whole lot more. That's the benefit of a general-purpose, Swiss Army knife-like tool that CoCo is. We have already worked on things like per-user limits, per-account limits for how much some tool should be used. In fact, I'm having a conversation with a company about deploying CoWork for 3,000-odd sales folks within that company, and part of the guarantee that they want is a per-user limit.
Our model, which is pure consumption, we don't charge a per-user fee for CoWork, is actually very beneficial here because customers end up getting the best of both worlds. They can both place a limit on how much one user can consume, but if users don't consume anything at all, they spend 0. You see little innovations like that actually be helpful, especially in a consumption model that starts at 0, that does not have a seat-based license. Many of the coding agent providers do a blended seat plus token pricing, and I'm actually pretty happy that we've stayed away from those.
I'm 100% happy that we stayed away from per seat. The other thing is, it doesn't fully inoculate us to your question. It's actually very insightful and valid, we did learn a lot on when someone is consuming with Snowflake, let's go and have a conversation with the customer and make sure that it's valuable consumption. I have lots of medium-sized regrets, if there's one big one from my time at Snowflake is when things were going amazing and all of you were revving your models, we never went and asked the customer, "Are you getting value out of this?" We learned that, we're not going to let it happen. Is there a risk of some technical disruption that changes? Sure. We'll cross that bridge, at least correlating value with spend matters a lot.
That's why the controls that Sridhar talked about matter so much for us.
We'll go over here. There's a microphone right behind you.
Brian, I like your socks.
I got them at the swag store.
Sridhar, if all these products take off, low 30% growth doesn't really feel right. It feels like this is a 40% market. Your primary competitor is growing twice as fast as you. When you think about ultimately what you think these new solutions can do to help accelerate growth, I know you're not giving guidance here, it doesn't feel like you're at cruise altitude from where the rest of the industry is at right now.
I don't know what to say. Absolutely. We aspire for more, but showing is the new doing.
Right.
Our guidance is based on rooted observed behavior, and we did considerably guide up this last quarter. As we see things happen, that's when we'll update our guidance.
Having said that, I can't resist the ask for a GAAP profitability guidance from said person you mentioned.
We'll go right over here. Microphone.
Hey, thanks for doing this. Mike Cikos with Needham & Company. Given the larger number of personas you guys are addressing, the applicable use cases, can you talk to the population growth within your existing customers? You're obviously not a seat-based model, but if I was trying to make an analogy to your NRR, how is that seat growth trending within the existing customers? Are we actually seeing an acceleration in the number of seats for those organizations?
I do think that we're seeing a broadening of the reach of Snowflake. As Sridhar said when he was talking about CoWork, it is harder to go beyond our core audience. Those examples that we're quoting on, you have customers saying, "Hey, I'm going to put this in front of 500 users, 600 users, 3,000 users." It is happening, and it reaches different functions and disciplines. I still think that it's early on in our journey with CoWork. Convincing customers to deploy something to every employee in an organization takes time, takes effort, and we're very early on. From a number of people in an organization that leverage-
Snowflake. We're very under-penetrated. To be clear, we just started on that journey.
This is something we track pretty extensively internally in the context of Snowflake CoWork, which is the number of unique users for each of the customers that we have. How do we drive that up? How do we get entire departments to adopt it? That journey is still early. Hopefully, we'll have more updates for you in the coming quarters.
We've got time for one last question. Keith, if you could give the microphone to someone for the last question, then I'll wrap it up.
Thanks, guys. Appreciate it. Adam Tindle, Raymond James. I recognize the announcement on GAAP profitability is going to play well to this audience, but I want to ask a challenging question on that, Sridhar. You outlined a generational growth opportunity. Brian talked about the TAM here. Your business is accelerating. Why is GAAP profitability important at this juncture, and why not pour more investment in now versus being governed by this promise?
It isn't clear that simply throwing more humans at problems gets more things done. That's my honest assessment. I think more of my energy should go into having each and every one of my engineers think and act like the person that reliably got the feature out in two hours. That is going to drive more leverage for us than simply hiring more people. The other thing that you should also take into consideration is, similar to the point about tech writing, there is a transformation of the workforce itself that is going on, where pretty much every team, this is not just engineering or sales, is going to be quite different. We are going through a process internally for what does it mean to have a particular function operate in a true AI-forward way? What does that team structure look like?
What do the job definitions look like, and how do you go from here to there? Beneath what looks pretty conservative, there's a massive amount of churn and reinvention that are going on. I think, as I said, I will end with scale no longer needs to be driven by the number of humans that you have working on some problem. Getting super linear scale from highly effective people is what business is going to be about.
With that, I want to thank our IR department and all the other Snowflakes who make this event possible. Thank you for your support. Have a wonderful day. Enjoy the rest of the Summit.
Thank you all.