Snowflake on The Enterprise AI Inflection Point: Why Data is the Moat

Sridhar Ramaswamy built a working app on his phone in about an hour using Snowflake's own coding agent, CoCo.

At the Six Five Summit 2026, Patrick Moorhead and Daniel Newman opened Day One with Snowflake CEO Sridhar Ramaswamy, one of four major CEOs kicking off the event.

Ramaswamy's argument centers on a shift in scarcity: AI has made software creation dramatically cheaper. Governance and provenance now sit underneath that shift, since even the most capable model will produce nothing useful on top of bad data. Snowflake's own internal system, which the company calls “Snow House,” feeds a sales agent used by 4,000 reps with data pulled from Salesforce, Snowflake's own consumption metrics, and Workday, giving the tool a 360-degree view no single source could provide.

Key Insights:
🔹 AI coding agents have made software creation dramatically cheaper, shifting competitive advantage toward whoever has the most trusted, governed data underneath the model.

🔹 Snowflake's internal “Snow House” system combines data from Salesforce, Snowflake's own usage metrics, and Workday into a single 360-degree view that powers its 4,000-person sales team's AI agent.

🔹 Ramaswamy recommends per-user spending caps, such as a $30-per-month ceiling, so employees get full use of AI tools without any single user running up an unbounded bill.

🔹 Snowflake's own sales AI agent costs less to run than the dashboarding software it replaced, which Ramaswamy uses as the internal benchmark for whether an AI tool is actually worth deploying.

🔹 46% of Snowflake's customers plan to increase spend, with roughly 30% penetration already inside their customer base and NPS climbing both sequentially and annually.

Snowflake's own data migrations have gone from taking three to four years to running in weeks, using the same AI tools now sold to customers, a concrete proof point Ramaswamy returns to when he talks about moving fast without waiting for a perfect data environment.

Watch the full video at sixfivemedia.com, and subscribe to our YouTube channel so you never miss an episode.

Explore more sessions from Six Five Summit: AI Unleashed 2026 at sixfivemedia.com/summit.

Disclaimer: Six Five Media is for information and entertainment purposes only. Over the course of this video, we may discuss companies that are publicly traded, and we may reference their equity share prices. Nothing discussed during this webcast should be considered investment advice or a recommendation to buy or sell any security. We are not investment advisors, and you should not rely on this content as financial advice. Six Five Media collaborates with technology companies and industry leaders to produce research-driven interviews and multimedia programming for enterprise technology audiences.

‍

Sridhar Ramaswamy:
Software has gotten significantly cheaper. But what has not gotten cheaper is what matters to your business. And data is the thing that matters to your business because it's not just how much revenue you made yesterday. It is how are your opportunities progressing through the pipeline of actually turning into customers.

Patrick Moorhead: 

Welcome to the Six Five Summit 2026. I'm Patrick Moorhead, and I am with my esteemed co-host and bestie, Daniel Newman. We are kicking off day one, talking about one of the most important factors in the enterprise today regarding enterprise AI, agentic AI. And something that's interesting, we've been talking about it for three years, but it's like, yeah, we have reached this inflection point in my trusted data. Governance and context may ultimately matter more than the models themselves. A big learning here. Also talk about what it takes to build AI-powered enterprises as these organizations move beyond POCs toward scaling. And I can't imagine a better person to have this conversation with than Sridhar Ramaswamy, CEO of Snowflake. Sridhar is alumni of the Six Five, but a first time at the Summit. Welcome.

Sridhar Ramaswamy: 

Thank you, Pat. Super excited to be here.

Daniel Newman: 

And I really appreciate you joining, Sridhar. And thanks so much for helping us kick off day one. I mean, what an esteemed group to have. And you're one of four. Mark Benioff also spoke here today. We've got enterprise apps. We've got data. We've got Matt Murphy kicking off CEO across the semiconductor space. And we even have a way for fabrication equipment with Gary Dickerson over at And we love having this broad perspective of CEOs opening up the days for us. And, you know, I want to get your kind of broader take. There's just so much going on and you're all in different parts of the market, but I want to hear yours. You know, you're in the middle of this data, this build out, you talk a lot about kind of provenance, governance, sovereignty, and so much going on. Where are you seeing, like, what's going on at the enterprise layer and sort of what do you think the market is still underestimating about all that's happening in AI right now?

Sridhar Ramaswamy: 

I think portions of the market actually get it, Dan, and that is that software is getting easier and easier to create. What used to be like a precious item is now more industrialized, more of a commodity. That means that if you have great data, And Snowflake has always been about getting our customers to have great trusted data governed the right way. The value that you can get from it is pretty immense. So it's not just, you know, me or any number of our customers finding out how much revenue did they make yesterday, but even a CEO can now slice and dice and look at the data. Why did this go down? Why did this, you know, go up? And that kind of access to data, combined with access to all of the information and the tools that we all use, thanks to technologies like MCP, is heralding a real revolution. And I would say the final thing is also that because we are making tools that make the act of bringing data into Snowflake go a whole lot faster, There's like this ever speeding up virtual cycle of you can get more done, which means bringing data into Snowflake, which means that you can get more value from the data, which means that you can begin to adapt what it is that you do. And we have lived through changes like this with both us, our sales teams, our finance teams, and so on, but increasingly with our customers. And I think that is what is really, really unique about the moment. It is that ability to both program and get stuff done with software, but also then get value from it in terms of how you get things done every day. All of the things that drive enterprises and business.

Patrick Moorhead: 

Yeah, it's interesting. In the run-up, I talked about even three years ago, data was discussed, but it seems like a lot of the benchmarks were, a lot of the conversation were the LLM benchmarks. And there's still that today, but it kind of missed the point. of where the value is, particularly for enterprise, right? And it's really thankfully starting to focus on what can you do with these models and the outcomes, but the big enabler is still data. And it's funny, this is a conversation where there's garbage in, garbage out. It's always been about data, but it seems like in this new age, it's as important as ever. And I'm curious, when you look at leveraging data, effectively getting security, the controls, the governance, big conversation there with the governance today. Why did this conversation change? Is it the reality set in, escape caged animals from the zoo, attacking data out there? Was it boards demanding ROI? Why and how did we hit this next phase?

Sridhar Ramaswamy: 

Ironically, it's the models themselves. And that was the first point that I made, which is that they have gotten so good at writing code. I don't think people internalize that enough. code was always like really, really hard. I mean, just look at your phone, look at all the apps that you have on your phone. It's probably three to five people at the very minimum. Obviously, with the big apps, it's thousands of people. They had to work incredibly hard to make something like that happen. Now, I've literally, in the space of an hour, taken an app that I have on my phone, like something like a simple metronome app, and I can get one of the coding agents, including our own Coco, to write me a new app that can do something like that. That's what I mean when I say software has gotten significantly cheaper. But what has not gotten cheaper is what matters to your business. And data is the thing that matters to your business because it's not just how much revenue you made yesterday. It is how are your opportunities progressing through the pipeline of actually turning into customers. Who is being most effective? Who is not? Is there a better pitch that somebody has? All of that analysis can now be done a whole lot faster and where governance and provenance and things like that matter is if you have bad data, no amount of software is going to help you. Like even in my own situation, once I realized that I'm not quite interpreting a data set correctly, I immediately stop all work because the smartest model in the world cannot make sense of truly bad data. I think it is that realization. that is driving this change, but I think we are also seeing the second order changes happen. Used to be, like a salesperson, when we gave him a sales agent, what they would do, it's like, oh great, I can spend less time looking up the open use cases so that I can figure out what to do. It went from there to, okay, let me use AI to figure out what the three best next things that I should be doing for this customer are. or if I have a new conversation, let me make sure I take the latest news about this customer into account as part of my pitch. It's that ability to do things like that, that is driving second order improvements on top of this data. And I think that's the excitement that you're hearing from everyone. But on the other hand, for every serious company, governance is a big deal. Like with the sales agent, for example, it better be the case that one of our account execs can see information only for their account, Not for all of the customers that Snowflake has. Getting that right, that's not an option. That is something that we absolutely have to do right. We also need to know where that data came from so that we can be sure about the calculations that go on with it. And I have an amazing data team that works on what are the certified data sets. that can be used collectively within the company. And so all of the things that Snowflake provides are the things that get massively leveraged by the data. So I would say, you know, going back to your question of why is data so important now, it's because AI is good enough at writing software that it can only realize value on top of great data and people are using it to better their businesses.

Daniel Newman: 

So, let me double click a little bit on this 1. you're basically if I heard you, you said the model is not the month, which, by the way, I agree their bottles are very good. They're very capable. They're also very democratized and becoming increasingly inexpensive to access. Right. So all of the above, you know, in many ways, you know, I often talk about access to compute being a moat, but you're actually kind of suggesting another moat here with the data. Enterprise data has become the real moat is sort of what you just said. So for the businesses, let's just let's break that down one more level. Like some companies are doing this well. Some companies are starting to try to do this. Some companies are probably struggling to do this. What does the continuum look like of companies that are being able to take advantage of their data and execute and get that kind of value and productivity out of it?

Sridhar Ramaswamy: 

I said all of the things that we have to solve if we want to get value from data. First of all, have to know what data. Does it have the meaning that you think it does? Most companies have multiple forms of revenue. At Snowflake, we have what's called metered consumption, but there's also gap revenue, which is subject to a different set of accounting rules, and you need to be able to tell between the two. And every company has its own complicated definitions of all of the things that matter to them. If they are sitting In a platform like Snowflake, where it is governed, where you can be sure that only the right people have access to the data, where there is meaning, semantics attached to the data, then AI comes super natural. It's easy to put AI on top of it because you have that shared understanding. But a lot of other companies, they have their data sitting literally in a box in a garage somewhere that a few people can access or it has a slow connection up to the internet. Those things are harder to deal with. And it's driving a whole renaissance in data modernization because people now understand how much more value that they can get from it. But the best customers, to be honest with you, are also taking an iterative approach. The world of data teams is famous for doing never-ending projects. I've been part of many migrations, three, four years. It's terrifying. But what is, again, I think amazing with the new tools, stuff like Cocoa that we ship, is that migrations can be done in weeks and months rather than quarters and years that it used to be before. And so even going through a prioritization process of figuring out what are the most important sources of data, who needs access to it? Is it a business user? Is it that you just need to do analysis on that? Taking that iterative approach is unlocking a lot of value for a lot of our customers, but for the industry in general, because things that used to take very, very long are now going a lot faster. I talked about how models are making software creation a whole lot easier. They're making things like system integration, or building pipelines, or migrations, all of which are just software with a different name. They're making those go much, much faster as well. And that's the buzz that you hear from the data industry. It's because people understand that there's now a fast path to getting value from data so that every company can be on this path to an agentic enterprise where more and more of the things that they do, which are wrote, can be automated. So the human's role shifts more into things like judgment. What's the right thing to do? As opposed to copying information from one tab to another. which unfortunately you, me and everyone else on the planet has spent many decades of their life doing.

Patrick Moorhead: 

Yeah, so Sridhar, when you were talking through that, the first thing that popped into my head is is something that my company has been pressure testing for a couple of years. And that was, I think we are fundamentally done with legacy environments. The automation to pull data and pull applications from legacy into modern, given the sophistication of coding agents, and data agents, quite frankly, is going to really change the world and unlock a lot of opportunities. And the other thing that is unique in this agentic AI era, which still includes models, is that it's a lot easier or the highest value comes from pulling data from different parts of the organization. I think you know, you might call it the agentic enterprise, right? But it's taking front end to the backend, you know, all your M's in the enterprise, HCM, HR, ERP, CRM, and pulling it all together. And as companies are starting to move in that direction, what are the foundations, the incremental, the newer foundations that they need in this era? before they can turn this into a system of action. Think of swarms of agents, a headless, right? There's still a human in the loop, but some of them, if you get them confident enough, they're just operating behind the scenes.

Sridhar Ramaswamy: 

I think first of all, an iterative approach to this, thinking about what's the most important thing that you need to modernize. is indeed very important. At Snowflake, for example, a lot of the energy that we put into AI went into the motions that drove scale for Snowflake. our software engineers, because creating great products is the lifeblood of Snowflake. So we spent a lot of time on that. Similarly, we spent a lot of time on sales, because getting our customers to know our product, to sell it to them, but deploy it to them is just as important. And so companies having a clear view of the functions that matter to them the most, the things that define them as companies. Figuring out how to bring that, those pieces of information together into a central place is super critical. At Snowflake, we've always been very lucky where for 10 years now, every piece of information about Snowflake flows into an instance of Snowflake that we call Snow House. And that 360 view of everything about Snowflake is what powers a lot of our AI at Snowflake. And so the sales agent, for example, not only has information from Salesforce, as it should, but it also has information from Snowflake itself, consumption information about our customers. It has information from Workday, which is where our HR information is. It has information from a system that we use for enablement, which has all of the material that we give to our sales folks. It's that ability to bring all of it together that gives us this 360 degree view, not only of what is going on with the customer, but what are the trends that we should be looking at. You talked about swarms of agents, Pat. I, it's kind of absurd to be able to think that I can do this. I will basically look at something, let's say use cases that we won and lost last quarter. I'll talk to Coco and ask, hey, what are the different dimensions that I should examine this use case data set with? And it'll tell me, oh, you can look at it by territory. You can look at it by G2K versus not. You can look at it by what product category was involved. Was AI involved or not involved? It'll give me 10 or 12 things, and I decide which are the slices that I want to look at. And it will go off and send agents that are now collecting that data in parallel, crunch for 10 minutes, and come back. I now even have agents forms that can run overnight, probably hundreds of invocations. to deliver this high quality report in the morning, but it's all built on that rock solid foundation of data with great meaning attached to it, governed the right way, because that really sets the stage for getting all of these more complicated things and optimizations done.

Daniel Newman: 

With all the opportunities you have to witness companies going through the transition, I'd love to get your read on kind of what advice are you giving? Because I mean, I'm listening to you talk about all the different things and it's a lot, sure. Like what practical advice right now are you giving leaders to be able to execute what you're able to do being inside a company that lives and breathes this every day?

Sridhar Ramaswamy: 

First one is, as I said, think iteratively. Don't get caught in old ways of, oh, we need to clean up all our data before we can get everything done. And it's not even things like consolidate on one single platform. I want Snowflake to succeed, but I also want the ecosystem of all the enterprises that rely on data to also succeed. So often we'll tell our customers, your data is your most precious asset. You should really keep it in an interoperable format that yes, Snowflake can help you manage, but you can use any tool to read that data. Having the clear understanding of what is the data foundation of the future is something that companies should get right. It's not as complicated as it sounds. It's the basics of our data is our data. We need to be owning it. And all of the applications that we use need to contribute to this basically company knowledge base of how it is operating. Once you go there, it's very quick. It comes down to what drives additional revenue for you. I mean, let's face it, in between cutting costs and making more money, every CEO, every company, Wall Street wants us to be making more money. And so what drives increased adoption of our products, what drives revenue growth is typically top of mind for people. Obviously, the other strategic things like, hey, don't spend too much on AI. That's a problem that we deal with, like everybody else. The fact that we can spread our traffic across models, both from open AI and from Anthropic, we also use open source models. These are the things that our customers want, and helping them deal with model costs, making sure that AI is inexpensive. is an important part of how we operate. I mean, I can assure you, for example, that the sales agent that my sales team, all 4,000 of them use, costs a lot less than the dashboarding that we used to pay for. And that's pretty stunning, because we give this agent to our sales team, like, hey, any question that you want that can be helpful for your business, it can drive a complicated plan. It's still not that expensive. Because we have gone in, we have squeezed costs, we have put in the best models, we have pre-computed things. So we have done a bunch of things that every customer can do. But all of those are downstream. I would start with, what are the things that matter the most to help our company go faster? And how do we get that done quickly with a partner like Snowflake that is all about delivering results? And that's where even things like our forward deployed engineering motion, where we work closely with customers to deliver the outcome. We are now sending our own engineers to be on site with our customers to finish the project because that's the speed at which people want to get things done. Going about it step-by-step, addressing the most important functions within an enterprise is how I would think about breaking it down. We're not just thinking and talking about it. We are involved in a ton of data modernization efforts because people want that step-by-step roadmap of how to be more effective. But the great thing about today is that not only can you get a project done, but your team can become way more effective using agentic tools to manage the data than what they used to be. That's also the joy of work today for me, because I can get things done with data and code that I could not even dream of like a year or two years ago.

Patrick Moorhead: 

Yeah, one thing I want to drill down on, and we did talk a lot about cost, but I want to make sure that we've simplified this, right? You've already talked about customer zero, how you balance cost and effectiveness, right? But I'm curious, how are you guiding them on balancing the two? I heard you talk about getting strategically aligned on the ones. I mean, but it's so funny, a lot of this seems common sense, but we didn't do it. We were using token maxing and basically giving the producers in the company kudos for spending the most money regardless of the benefit that provided downstream. So how are you guiding customers in the effectiveness versus cost trade-off?

Sridhar Ramaswamy: 

Oh, it's very simple. I mean, the tools that we provide for a particular function needs to cost less than what they're already using. I mean, I can assure you that the, as I said, the AI agents that my sales team is using cost less than what our dashboarding licenses used to cost. I'm very cutthroat about stuff like that, which is not, people cannot come in and say like, hey, ta-da, great functionality, spend a lot more. And so what we tell our customers is yes, we can help you roll out agents. Yes, we can talk to you about what drives transformation in how your sales team or how your finance team operates. But what you should also put into place are things like per user budgets that we support. And so it's perfectly fine if a customer says, you know, on average, I don't really want a user to be spending more than $30 per month on this critical tool. It's like, yes, it's a critical tool, but I want an upper limit on how much it is that they can spend. We support that out of the box. This way, they get the best of both worlds. They know that their employees can get answers to the questions that they want answered. any question far more than what a dashboard could answer. But they also have the peace of mind knowing that, you know, some over enthusiastic kid is not going to go blow through a million dollar budget. That's that's the kind of mindset that I think can be very, very successful. Today, it can never be, it was never about token maxing, I mean, like, I'll confess, I have shared a usage stats with my exec stuff. I used to even embarrass them by putting it up in front of every staff meeting. But what I would tell them was big numbers here. don't mean that you're very effective using AI. But trust me, zeros across the board sends a really loud signal.

Patrick Moorhead: 

It does, for sure. So it sounds like finding the right tools that allow you to manage, you know, almost like a rheostat button between effectiveness and cost, picking the right applications to do that, and of course, buying more Snowflake.

Sridhar Ramaswamy: 

Where appropriate, we have to deliver value. I never want our customers to be spending money if they're not getting that. And look, I am proud of the fact that with things like Cocoa, our customers can optimize their snowflake spend without involving my sales team.

Patrick Moorhead: 

I'll admit, I don't have too many conversations like this with people in your space that have this much discussion about cost. So thank you on behalf of all of your customers and users.

Sridhar Ramaswamy: 

It's what keeps us in the business in the long term. Like if we make money that we don't deserve to make, customers find out. I tell all my sales team, if your customer discovers an optimization that saves them a pile of money, they'll be happy minute one after they discover that optimization. Minute two is where the hell were you?

Daniel Newman: 

And I'll tell you what, it's, you know, to your credit is showing up in your numbers. You know, we track thousands of your customers and, you know, we run a regular intense survey for you. And, you know, I'm just looking here and now, you know, forty six percent of your customers plan to increase spend. And you already have 30% pervasion, like very high pervasion in the customer and high intent. Your net promoter score is sequentially growing, annually growing, which means people are getting value. And COCO, which I'm sure you care deeply about, is seeing significant growth in purchase, pilot, and planning. a little bit for everybody out there, people are digging it. So when Pat said, buy more snowflake, the market is actually looking like they're gonna buy more Snowflake. We'll see if it shows up in the next quarter. I will never ask you about that. But anyway, let's wrap this up. We started big picture and sort of with a crystal ball question, right? What's more fun than telling me what's going to happen. And I actually remember the first time I asked you a crystal ball question, you told me multiple years are too far off. So you might say something like this again, but you know, as we look, into the future and everything is changing so fast, but it can be three years, five years, maybe it's only one year because we're in like AI dog years, we call it. Everything's going so fast. But like, what do you see that's going to define the successful agentic enterprises? And what should CEOs be doing to make sure that they're one of them?

Sridhar Ramaswamy: 

The enterprises that succeed are the ones that make their organizations more effective by having AI take care of the drudgery of work, where people are more focused on the value and the judgment questions of what are the choices that we should be making. Life is infinitely complicated. As anyone that has negotiated a deal knows, there are like 20 different parameters that you can move. What is AI going to do? It's going to make sure that the legal terms that are part of some contract, they are acceptable to you. What it's going to do is ensure that there's no gotcha term somewhere. And it's going to ensure that there are no mistakes. And so the human is still left with, huh, is this the right trade-off to be making? I think the more we think progressively about how do we take the portions of work that are merely just transformation, moving, communication, overhead, things like that, and use the power of AI and data to handle it with the humans being left to deal with, okay, what's the framework? What's the judgment question? How can I do this more effectively? How can we collectively share the knowledge that we have so that we are a more effective enterprise? That's where I see things inexorably going. Models getting better at writing software, at thinking, at creating long-range plans. These are all means to that kind of an end. And going back, this is where things like data. Things like free access to all the information about your company are such critical building blocks for creating futures like this. Obviously, many companies are big, complicated, including Snowflake, so it's going to take some amount of time. But that's the part that is really exciting because we get to focus on the things that are intangible and hard and have AI take care of all the mechanical things that we all spend so much time doing.

Daniel Newman: 

Well, Sridhar, I want to thank you so much for helping us kick this day off here.

Sridhar Ramaswamy: 

Thank you, Dan. Thank you, Pat. Have an amazing summit.

Daniel Newman:

It's day one at the Six Five Summit 2026, AI Unleashed. And you just got a whole lot of insight there from the CEO of Snowflake and someone that's deep in what is going on here in AI and agentic enterprise. We hope you'll stick with us through the entire day here, and of course, the entire Six Five Summit. Check us out. We appreciate you. Stay here for more.

‍