All right, fantastic. We will go ahead and kick off the Snowflake session at the Goldman Sachs Communacopia Conference.

I'm Gabriela Borges, I lead our software franchise. Delighted to have with me on stage Sridhar Ramaswamy, CEO, Brian Robins, CFO.

Thank you both for being here.

Thank you.

Thank you, excited to be here.

Sridhar, I want to start with some of your conversations in the field. Tell us a little bit about what you're hearing.

If we look at the modern data tech stack today versus 2024, 2025, the pace of innovation has changed. Tell us a little bit about what you're seeing in those conversations and hearing in those conversations about the customer journey to go from old to actually new, meaning AI-enabled.

Yeah. There's always been a lot of demand for data modernization.

The word migration strikes terror in the hearts of pretty much every CIO or CDO, honestly everyone in engineering. Just takes forever, highly uncertain outcome, and so on.

I think AI is having a pretty profound impact on how quickly you can get those done, and expectations not just from me, for my team—I've talked before about how I want migrations to be mostly automated—but even customers are expecting it. We have a very large manufacturing client, for example, do a Teradata migration planning aggressively to finish it in less than three quarters.

This is not something you would have heard of in 2024. That's part one of a lot more is possible, let's go get it done.

What is equally interesting is now the ability to have conversations at a data platform level, at the level of Snowflake, which has stayed out generally of a lot of these kinds of conversations, is having them about how we can deliver business value. It is everything from how can we automate invoice processing at a really large energy manufacturer, because processes like this were always super manual, super spotty.

They would do spot checks here and there. They estimate that on $10 billion that they pay out every year, they'll be a percentage point more efficient, except that that's an astronomical amount of money.

It's talking about that or talking about supply chain optimization or talking about how do you implement a custom CDP a whole lot faster. It's having conversations like that that I think are very, very distinctly 2026 compared to the previous one where, honestly, most CEOs wouldn't even bother to talk to me.

It's like, "Oh, data vendor, who cares?" I think that change is what is remarkable about this moment.

I want to come back to custom CDP, but let's stay on the migrations thread for a moment. When we came to your conference in June, a lot of the system integrators we were talking to spoke about how migrations can now be fixed cost instead of variable cost because of coding tools.

Tell us a little bit more about the shift from variable cost migrations to fixed cost migrations, and how you think about the impact that coding tools, and then that segues nicely into CoCo specifically, can have on that pace of migrations.

I would say this is a larger trend. AI fundamentally is making software industrialized.

I wouldn't underestimate, even now, the threat that it poses to every tech company, every software company. It's a profound, profound shift.

What it has also done is it has obliterated the distance between what a data platform like Snowflake is and what applications running on top of data can be. Now, stuff that people build on top of Snowflake will not look like your standard packaged SaaS app.

It will have its own look and feel. We can talk about that.

The other thing that it's done is it has also squished the distance between a data platform and actually what used to be called services, because you can begin to automate a lot of things. Inasmuch as software represents what we, humanity, call intelligence, that's what they do.

They put workflows, they put data structures into place, they organize our thinking. AI accelerates that massively, which is why a number of folks, because they can now bring the power of coding agents, are basically saying, "Wait, I can compress the time of something like a migration massively and also feel very confident that the weird problems that will come up during any real migration, the little odds and ends, can also be fixed equally quickly."

They're sensing an opportunity because the majority of the industry is still operating on time and material, a lot of time, and a lot of materials, and a lot of money. The progressive system integrators are going, "I can guarantee outcomes."

This is what we do as well, where we're saying, "We can deliver outcomes for our customers simply because the ability to get things done fast is much better now than before, but also the ability to deal with unknown things is just also a whole lot better." It's the combination of these that I think will drive a massive change through the services industry.

Not that I think services will go away. It's just going to look dramatically different, a lot smaller than what it did before, but one that is much more tied to what are the outcomes that customers want.

Let me ask about CoCo specifically, because, Brian, then we can bring you into the conversation as it pertains to how you think about guidance. Sridhar, we started talking about, look, it's not just the number of customers in the install base that are using CoCo, but it's actually the depth of usage and these net new use cases that you're also solving for.

I guess part one would be, tell us a little bit about how you as an executive team push to deepen and strengthen the usage of CoCo within any given customer.

Yeah. A lot of it is what we have learned ourselves.

I think I've talked about this previously. Part of a huge unlock for Snowflake, the company, was having a coding agent available to every single person within the company.

It was not a specialized tool, and so we saw bursts of creativity and innovation in every department. That's been very helpful for us just to understand what is possible with AI.

The nice thing about harnesses in general, and I think the reason they're going to have a profound impact on everyone, including all of you, is the work that you do now is visible, observable within one system, which also means that it is optimizable, it is automatable. We have a lot of insight into what is a customer doing?

What are we doing with CoCo? Are we doing repeat things?

For example, we now make recommendations for here's a skill you should be building because you seem to be doing this very often. It's a quick hop from there to here are a set of skills that your colleagues are using, this is something you can use to make yourself more effective at work.

We can understand things like the depth of usage and then tie it back to things that we can do. At the end of the day, life is about what's an action that you can take that can produce an outcome that you want.

We have things like hands-on labs that have proven to be highly, highly effective. This is basically a 2 to 3-hour tutorial run by one of our more technical people with the customer.

Because of that, the customer gets more effective in what they do, their data teams are happier, they get more work done, simpler to debug annoying problems that are a part and parcel of their life. It also gives us a clear road map for this is what it takes to drive truly deep adoption with each of the customers that matters to us.

I'll just add on to that. One of the things that CoCo has done is it's opened up the aperture to who we sell to from a persona perspective.

A year ago, when I joined Snowflake, I hardly had any customer conversations. Today, every week I meet with 3 to 5 CFOs and talking to them about what we're doing internally on CoCo and what the art of the possible is.

I think there's no better way to actually expand CoCo or Cowork adoption by showing people how you're using it internally. As Sridhar said, these skills that we're making can be applied to vast sets of data to our customers to actually get them started to use that.

I think once you show them what you do internally, the art of the possible, and how quickly you can speed up things, they're extremely interested.

Brian, Sridhar used the word burst there. Look, it's a new product cycle, and you're going to have customers that are experimenting, you're going to have questions around gross retention durability.

I think you've already said gross retention has held stable even as CoCo has scaled. My question for you is, as an analyst, we try to model CoCo, and we also try to model the impact that migrations and the speed of migrations is having on your business.

What advice would you give us as we try to think about some of the blue sky scenarios over the next 18 months, and how do you derisk the forecast from customers getting really excited about CoCo, but ultimately it's still very competitive and you've got experimentation and could that usage pattern actually fade over time?

Absolutely. It's something that we struggle with internally as well when you launch a new product.

If you ask anybody how to model that, when you don't have that much data, it's difficult. Fortunately, we have an amazing team internally that has been doing this for a very long period of time and have built very sophisticated models to understand what new product adoption would look like, and then compares that to historically what new products have done.

We'll then take that and then basically model up what a scenario is. Then we have a very wide and deep group that discuss what we'll put in guidance from that perspective.

When we do come up with guidance for the core platform, like the migrations, it's based on observed behavior. We have years and years of data with that and can get pretty close.

For the new products, we try to be a bit conservative, we don't take a month worth of data and extract that out and say this is what it's going to be like all for this year and next year. Now that we've had two quarters of data, we feel more confident in what we can extract from that observed behavior.

When you and I first met, we started talking about the customer cohorts and how, look, it sometimes takes year one as the initial ramp and then year two is really when a customer gets into full swing Snowflake adoption. Has anything changed as you look at the speed at which these customers are ramping?

I think from a cohort perspective, if you look at all the verticals across the company, we still have the same sort of vertical penetration, if you will, into financial services, manufacturing, government, and so forth. As far as ramping, with the use of CoCo and AI, we're seeing customers ramp much quicker than what they've ramped historically.

Whether it's our partner network or what we're doing internally, what Sridhar talked about as outcome-based pricing, has really built the credibility with our customers. If someone comes to me and says, "I can guarantee you X for this set price, and I know historically that took a lot of time and a lot of materials and so forth," like I'm all in, and we're seeing that from a customer perspective.

We do track internally how long it takes for them to get up to their consumption run rate, and we're seeing those curves get steeper and steeper and steeper. Customers are deploying quicker, they're consuming quicker, they're using partners, us, and themselves are using our agents to actually do that.

It's really exciting to see.

That's a lot of where our go-to-market teams have to go. Our CRO often talks about shifting right towards outcomes, which just means that Snowflake as a company, and their team in particular, has to focus a lot more on how do we get use cases live with customers?

How do we get it to scale within customers? What do we have to do?

It's increasingly a result of many things that used to occupy lots of time: getting ready for a meeting, doing research about what does the customer have, what's their data estate, how do you maintain it. All of that getting easier, faster.

Similarly for solutions engineers, they would spend a lot of time building a little demo that would take forever. Now they can be conjured up kind of on demand.

There's this big shift. We've even created basically new job functions.

One of them is called an activation solutions engineer, that are expressly charged with what Brian said, which is how do you get a new logo to go live on Snowflake much faster than what they would otherwise? That is a trend that we want to keep leaning into and pushing.

I would also say selling into or actually describing what we do to all these personas, if you go to a CFO and show them what the art of the possible is, they're immediately saying, "How do I get that up and running, like yesterday?" When Sridhar goes and talks to a number of CEOs, they want it yesterday.

The sense of urgency around getting these results are also super fun.

Let's talk a little bit about custom apps. Sridhar, you started talking about custom CDP, and certainly there's been a debate on the application level on what is the future of applications.

How do we think about packaged apps versus headless architectures versus some of the interesting things that customers are building on top of Snowflake? Some of the things that you announced at Summit: workflow orchestration, agents, and on the application layer as well.

Tell us about how you see the app layer evolving.

I think this is a time of just a lot of innovation with what is possible, because, let's face it, building any kind of meaningful application before, again, was just a hard thing to do. Now pretty much any of us can pick up a coding agent and say, "Hey, I want not just a web app, like an Android or an iOS app."

Dealing with the app stores is the most time-consuming part of doing something like that now. The apps themselves are easy.

We are playing around with lots of different formulations, including things like should the notion of an application be rethought as a handful of self-evolving skills running on top of a data substrate? Let me give it to you with a specific example.

Let's say, "Oh, you want an internal survey application." Easy enough to imagine.

What do you want to do in a survey application, especially if it's just internal? You want to know who your employees are.

You want to know how you target, let's say, a particular group. You want to manage visibility into the results that are coming.

It's not that complicated. Now, if you have a Snowflake deployment like we do, absolutely our Workday hierarchy is mirrored in Snowflake, so that's where you get that from.

You set up a couple of tables to store surveys, to store results, and figure out notifications. An admin does some things, and then an ops person says these people are allowed to send out surveys, and ops people go to send out actual surveys.

Now, I didn't really talk about a UI. You can conjure that up on demand by saying, "Oh, if somebody clicks on this link, bring up this React app for them to respond to the survey."

What would be an actual SaaS procurement is a handful of skills that can be installed on top of data that is already sitting in Snowflake. It's governed.

You don't have to worry about, "Hey, does everyone have access to this data?" You can manage the visibility.

You can also do follow-up analysis on it. If it's free-form text and you want to run AI on it, that's not an issue.

You see where this is going. I think it just makes many, many more things possible.

This is not to say that this is the end-all, be-all solution for everything. To the extent that our vision, consistently for multiple years, has been we want to be the place where you can bring together all of your data and get a 360 view, analytic view, of that data, it sets us up very nicely to rethink what is an application of the future.

For what it's worth, we sync our Salesforce data also into Snowflake. If I want to create an annotation application that's not part of Salesforce, but acts partly on Salesforce data and can push it back, that too is possible.

It's a very different way of thinking about what's an application. The first time I tell people an application can be reduced to a handful of skills, you get a blank stare, like, "Really, what does that mean?"

Similarly, these skills don't have to be static. As you look at how people are using them, they can get additional functionality over time.

They can self-evolve. That's how a lot of support systems are, like my team's SRE systems are evolving.

They built the first version, they built some skills, put some data. They looked at how they were analyzing it, then they created newer skills.

They said, "Oh, half of these things can be automated by agents looking at stuff first as opposed to having a human look at it." Now that's a very different notion of what an application is.

It is something that is evolving as it goes along. I think the world is rich with possibility, and there's just going to be a lot of innovation everywhere.

As a platform, we focus on what are these little nuggets that we can lay out there that is going to convince some bright person in one of our customers to say, "Here's something else that I can build with it." We learn from them, and then figure out how to make it more broadly available to other customers.

You see this feedback loop and where it's going.

Let me ask you on the skill side. This is a little bit of an orthogonal question, but it's so top of mind in the last three months.

Do you have a view on how the ecosystem will evolve between open weights and frontier? Does that impact how you think about your business strategy and dynamics like skills?

Let's say it's the same for you. If Anthropic is the only one making great models, you're in trouble, I'm in trouble, we're all in trouble.

The fact that OpenAI is creating amazing models is good for the world. It's good for them, it's good for us as well.

I look at open source the same way. I think innovation here presses the foundation labs to innovate even more.

I think as a phenomenon, this is great for us. It also feeds really well into the Snowflake narrative of we are truly a cross-platform solution.

We're one of the few people that can tell you you can run on AWS, and you can effectively run exactly the same deployment on Azure with not a whole lot of work. If you want to do disaster recovery between these instances because some regulator is on top of you saying, "You can't go down if AWS goes down," that too is possible with Snowflake.

We look at models the same way. It offers up lots of options for choice, for optimization.

The thing that I think is unique about this moment, and it's not a good or bad, it's just a strange happenstance of the moment, is that none of the model makers so far, other than something like a ChatGPT which does have true consumer lock-in, have been able to create that kind of lock-in well into a pretty massive investment cycle. Every smart engineer knows that they can move instantly from a Claude Code to a Codex, not a problem, or vice versa.

My team moved over from being heavy Cursor users to CoCo users on the solution engineering side without me having to harangue them, which is normally how things work with situations like this. I think that's also pretty unique, which just means that something that can interoperate between these models is a pretty cool thing to have.

I think skills themselves are the great equalizer because they're English. It means that every model is immensely capable of taking a skill that perhaps was written for a different harness, for a different model, and figuring out how to tweak it to work in a new situation.

I think this all makes up for robust choice for all of us.

I think this next question is for both of you all. Strategic value of selling inference pass-through into your install base versus potentially a lower gross margin, how do you think about that trade-off?

Hate to start like this, but it depends on what inference is. If it is merely reselling undifferentiated capacity from a large supplier, you're not creating any value.

That's fake news on the part of people that are doing this trying to pretend that they have a business. On the other hand, if you say, "I have a gateway that actually can do a meaningful job of helping my customers optimize spend," in other words, I am creating value on top of my suppliers and it has some amount of stickiness and redeeming value, that's a meaningful new category.

We look at inference, for example, in the context of, can I offer my customers choice, because we buy capacity in bulk both from OpenAI and from Anthropic, and we have the capacity to do things like run open weight models ourselves. In that context, inference becomes a little more interesting.

We also generally take the lens of it's important for us to understand what our strengths are. Our strength is as a data platform, and inference as a component of a modern data platform absolutely makes sense to us.

We also like to sell at the highest value-creating point. In other words, my order of preference is always, if I can convince a customer to use CoCo or Cowork directly, that's what I want them to be using.

If they say, "No, all I want is model capacity from you, I'm going to run my own harness," yes, we will do it. If they say, "I want neither of those, I just want the data platform, and you can be a back end to Claude," we will do that as well, somewhat more reluctantly.

It's important that you have your Maslow's hierarchy of where you're creating value and acting according to that. Brian and I are super aligned on what are the business outcomes that we want to drive.

Let's say if CoCo adoption were to go up massively and it has an impact on our gross margins, I'm happy to come and explain that to you all day long. That's not an issue, because it will drive a meaningful acceleration in our overall business.

I can also tell you, as we grow in scale, as we do these things, we get better at optimizing. We get better at running open source models, which will have much better margins.

We obviously buy at a bigger quantity from the suppliers, like we do with AWS and the $6 billion contract, which gives us better economics. There are good answers to things like gross margin, but it needs to make strategic sense.

What I have little appetite for is being a blind reseller of someone else's intelligence.

100%. I think the analogy is what we do with the hyperscalers, where we put our software on top of them and basically deliver a value-add service that has good ROI for our customers, and that's really key.

From a gross margin perspective, we have a lot of control over that. We announced, I believe it was last quarter, a big deal with AWS.

We constantly work with our hyperscalers to do stuff better, faster, cheaper, and we'll continue to do that as it relates to inference. When we launch new products, the number one thing we want to do is make incredible products that people will adopt, get value out of, and will drive revenue.

Then we have demonstrated that we can actually show leverage in that once we get economies of scale and a number of customers on that. That's on the gross margin side.

Regardless of that, we're very committed to getting operating leverage in the model overall, and that's what we guided to for the year. The framework that we do go do our gross margin, we have very accurate models internally is based on the uplift that we see in our AI products, which is phenomenal, and we like that.

That's what we guide to for the rest of the year. We're confident that we can actually continue to get leverage in the overall model and do things around gross margin.

I think this is an interesting thread to pull on just for two more minutes here. Let's fast-forward and imagine that, this time this year, CoCo we think is a home run.

Brian is coming on the earnings call, and gross margin is not where the Street modeled because CoCo was fantastic. Talk to us a little bit about the guardrail on CoCo gross margin.

Sridhar, you sort of alluded to it there. You can explain this holistically as part of a much larger value proposition.

What I would say first is we are a consumption business, and so it does take time to ramp. We also know, although we're bringing that down, and we also build models to understand what this consumption's going to be.

It's really Sridhar and I don't want to surprise folks, and so if we were seeing that massive CoCo adoption beyond what we're already seeing, we would have the ability to communicate that to you within a quarter or two to manage that.

I want to ask a technical question on agents. Sridhar, there is a school of thought that the systems and architectures of today are not going to scale for real-time agents, because agents have so much more volume, for lack of a better word.

Talk to us a little bit about how you would address that concern if investors say, "Well, Snowflake was founded X number of years ago and is designed to scale for humans, not for agent queries."

There's a little bit of a meme-ing and attempted category creation by the folks that say things like this. But 100%, there is richer data that comes from agents and things like trajectory analysis for all kinds of purposes.

The good part is, how do you make your team more efficient? The bad part is, is there someone in the company that's actually doing research on bioweapons?

As a CEO, you really want to stop that very quickly. There's the good and bad aspects of just needing to make sure that you do better with that.

In all of this, the overall criticism that we have not addressed super low-latency data well is very legit. I'm a big fan of laying it bluntly to my team and accepting things when we need to do better.

I said this to you folks, I think it was two years ago, at what we had done in machine learning and notebooks. It was not up to par.

Fast-forward to now, you're not going to hear that from any of our customers. Not only is the offering really good, we have also, thanks to technology like CoCo, massively accelerated the process of migrating, let's say, from whatever set of notebooks that you have onto Snowflake, or being able to create not one, but dozens of machine learning models as part of an experiment that you are running.

Making sure that we deal with data that say has 500 milliseconds or less of freshness requirement is not something that we do really well right now. There's a team that's actively at work on this.

It basically comes down to things like what are the trade-offs that you want to make in terms of cost and efficiency, and also just the querying speed that you want. We've been putting into place things like interactive tables for much lower latency serving, and our streaming solution has brought things like the freshness down to the 2 to 3-second range.

There's active work underway to bring that further down to the 500-odd milliseconds, at which point it stops being an issue. It's an opportunity.

It's a threat. We are well aware.

We absolutely are working on it.

I want to ask the switching cost question both ways, which is,

we've

talked already about speed of migrations accelerating to Snowflake. We've also talked in the last year about things like standardization of data tables and data gravity becoming less grave, for lack of a better word.

How do you think about the longer-term implications from switching costs potentially going down in the data infrastructure world?

Just bluntly, and I tell my teams this, I'll admit this to you: if migration into Snowflake can be made a whole lot faster, migration out of Snowflake can be made a whole lot faster. That's the world we live in.

It applies to data platforms, it applies potentially to consumer software, it applies everywhere. That is something that we all have to understand.

Then there are also broader industry trends, like a lot of CDOs and CIOs simply saying, "I don't want my data to be held hostage by anyone." You don't grudge them.

You can't grudge them for saying that. We support open formats.

We want to increasingly make it painless for our customers to deal with open formats. We have now something called Snowflake-managed Iceberg tables, which is a fancy way of saying you can have your cake and eat it too, which is you can store data with Snowflake, but have it be queryable in Iceberg format by any other engine.

The place where we create value has to be further upstream. It has to be in, do we provide better governance?

Do we provide better disaster recovery? Is it easier to create and run agents on top of Snowflake?

Do we provide a better observability solution? There's this whole stack of things on top of the data platform that is open that we have to be providing.

To a large extent, for a lot of these, my attitude is bring it on. It's not that easy to create the platform that Snowflake is.

If it were, the hyperscalers would have eaten our lunch 10 years ago. It is hard to do, and I think AI actually accelerates what's possible with Snowflake.

I think the Snowflake of old, which used to hang on to a set of what it thought were inviolables that could never change, I think that company has also changed. This is a company that's much more attuned to, where is the world of data going?

Where do we create value? What is strategic value that we could be creating in a way that's still faithful to what our customers want?

We feel good about how we are positioned. Absolutely, what can migrate in can migrate out.

You need to both be paranoid about that, but also seize the opportunity while you can.

We really, really drive internally this customer-first obsession. People can come in and leave easily, but if you have a customer-first obsession and really focus on the business cases, outcomes, and ROI, one of the top 10 skills that we have is around cost optimization.

We want people to be fully optimized. We want people to use the product.

We want them to get value out of it. We're constantly going back.

If we see an anomaly with a customer, we'll actually go alert them and say, "Hey, your bill is running higher than it's been before. Were these jobs that you meant to kick off or not?"

Really being obsessed with the customer, I think really is a key priority for us.

Let's stay on the pricing implication here. I think about architectural enhancements in Snowflake like the Gen 2 instances, for instance.

You've got this natural pricing deflation in your business, like any good technology business, and yet your role is to also abstract value and price one level above the core components of that, you can capture gross margin. Maybe, Brian, just tell us a little bit about how that philosophy is coming into play in 2026 with things like Gen 2.

Yeah. Part of the business is you got to be better from a price-performance perspective than the last generation was, and you always have to give your customers the ability to get more out of your product.

We price that in, but we're seeing the offset of that with volume and new jobs coming into the company. As we go and price stuff, we carefully take that into consideration so there's not big step-downs in revenue.

We also want to get our customers to get the benefit and the value of these performance enhancements that we're doing on the platform.

I'll perhaps add on a quote from one of my previous bosses and mentors: revenue solves all known problems.

Fantastic. Let's leave it there.

Please join me in thanking Sridhar and Brian for their time. Thank you, everyone.