Excellent. Thank you, everyone, for joining us this afternoon. My name is Keith Weiss. I run the U.S. Software Research Practice here at Morgan Stanley. I'm very pleased to have with us from Snowflake both newly appointed CEO Sridhar Ramaswamy and CFO Mike Scarpelli, longtime CFO Mike Scarpelli. So thank you, gentlemen, both for joining us. Before we get started, for important research disclosures, please see the Morgan Stanley Research Disclosure website at www.morganstanley.com/researchdisclosures. Again, so thank you for joining us at the conference. It's really great to be able to talk to you right as you're sort of embarking on the CEO position at Snowflake. I guess the first question that I was really interested in hearing from you, when Frank Slootman was talking about the transition on the call, he's basically saying, like, listen, this is the right guy for the next stages of evolution.
The challenge that Snowflake is going to see over the next couple of years, what makes you the right guy? Which is a weird question to ask you, but, like, what makes you the right guy for the next 3-5 years for Snowflake?
First, excited to be here. Was here last year. How much the world has changed in a year. You know, I've been at Snowflake for several quarters. It's been a fun experience because I got to know the exec team, the board, the legend that Frank is himself. We spent a whole week together just doing customer calls, meetings from 8:00 A.M. - 10:00 P.M. every day. So I really got he got to know me. I got to know him. Just a little bit of background. I have an academic background. Joined Google very early as an engineer, in fact. Got to run the search ads team. Then gradually more and more of ads at Google. I was a lead of the ads and commerce team for over five years.
You know, helped grow that business from the $1.5 billion that it was when I joined to over $120 billion when I left. So I've seen massive amounts of scale. And Ads at Google is very unique because the way performance advertising works, you know, the business is actually run jointly by the sales team and the product team. The sales team gets people to commit. But similar to how consumption works, the product team then decides how that money is actually spent. So deep background in both the ads business but also other areas like payments, where we helped launch Google Pay. So I've seen the whole gamut of how businesses operate at massive revenue but also massive scale, a team of over 10,000 people when I left Google.
But now coming to Snowflake, I'm incredibly excited about Snowflake because I think a cloud that puts data at the center, it's just a very, very different way of looking at the problem of, you know, how do we provide an amazing data platform for enterprise. And that already shows through in all of the customer conversations that I have. And I think there is incredible potential, starting with the fundamental observation of what does it mean to be a cloud that has data at its center. And we're also at a pretty unique moment in technology with the rise of AI and the disruptive power that it is going to bring.
And again, I have long and deep experience in this from, like, the biggest machine learning systems ever built on the planet but circa 2006, where we built these prediction systems for ads click-through that were instrumental in making that into a truly efficient sort of platform. Fast forward all the way to building the first AI search engine at Neeva, my startup, which I spoke about here last year. So sort of lived through all of these sorts of evolutions, helped companies make major transitions in business. One of the key accomplishments that Philipp Schindler, the current head of sales at Google, and I had is we really helped transform Google from a desktop ads company to one that was both mobile and desktop focused.
There was a lot so it's all of these things that I think gave the board confidence that I would take all of the great qualities of Snowflake, the drive that Frank exemplified, the customer obsession that our founder Benoît exemplified, the collaborative culture that all of the exec staff exemplified, and add on that element of an increased sense of urgency, much more competitive paranoia, and just a desire to get things done fast. Because back at Google, we were way ahead of everybody else. But still, my team went all in because opportunities can be fleeting. It's really these kinds of qualities: ability to operate at scale, ability to handle a big business, a vision not just for what Snowflake is today.
I signed up because I think Snowflake can be a $100 billion revenue company, not because I believe, like, you know, in what we announced last year or what we are going to do this year. I think it is a similar kind of potential for Snowflake. It's that combination of proven ability to execute at scale plus the technology chops that, you know, truy an awkward answer, but that's my long-winded answer to your question.
Got it. So from the investor perspective, when Snowflake IPOed or when Frank and Mike joined Snowflake, the thought process was, this is an innovative company with a good cloud solution addressing a really big market. And what we need is great operators to put the foundation in place for scaling this company, for going after those customers, for getting that great go-to-market motion in place. And Mike and Frank have scaled that company. And what investors are thinking now is, with you coming on board, it's more about innovation, right? More about picking up the pace of innovation, being able to expand out the solution portfolio, being able to get into potentially adjacent markets that are going to further enable you to and here's the question.
Is it further enable you to penetrate the existing TAM or go into adjacent TAMs that are getting more and more interesting with what we're seeing in the innovation cycle?
I mean, like, look, I don't think of myself as just a product leader. And that part is important. I worked extensively with business teams of all kinds, helped products go from scratch to massive scale. So for example, I took over the YouTube Ads business in 2015, nearly quadrupled it to over $20 billion of revenue, started shopping ads on Google from scratch to, I think it's like, it's $25+ billion today. And so it is the ability to take products and drive them to scale. And we had our share of pain at Google trying to convince a team that sold search, which, like, honestly, sells itself to selling YouTube, which was just a vastly more difficult sort of endeavor. So I bring a holistic approach to how we solve problems.
And having said that, I think there is enormous opportunity in product innovation, in being able to, you know, provide additional depth with our existing customers but also significantly wrap up new kinds of customers, new kinds of use cases. Obviously, at any given point in time, there need to be priorities. We can't be everything to everybody at the same time. And, you know, we're sort of we have our work cut out in my mind for, say, the next quarter or two, where really it's about taking the slew of things that we have in public preview, whether it is transactional tables or interoperable storage or an application platform, all the things that we have announced in AI, getting them into GA, getting them into the hands of our customers. That is going to drive momentum for this year.
I think it is that, you know, it is that feedback loop of great products, getting more customers, getting adjacent use cases, being able to expand, and then a nascent but vibrant partner strategy of people building on top of Snowflake. Those are all the force multipliers that I feel very excited about being able to exploit at Snowflake.
Got it. So maybe taking a step back, you Snowflake chose you, and you chose Snowflake at the same time, right? And from what I hear, you had a lot of opportunities ahead of you. You saw an interesting opportunity in Snowflake over the next five years to kind of make this investment. What was so interesting about the Snowflake opportunity? And is it beyond kind of just what we were thinking as investors of, like, hey, listen, there's a lot of money spent in data management. These guys are the new thing in data management. Is it a broader opportunity than what we're thinking?
I do genuinely think it is a broader opportunity. You are, you know, absolutely right. I had other public company CEO opportunities, other, like, big tech opportunities. I had a certain, you know, a fair number of choice. Snowflake stood out in terms of a technology company that had the ability to create enormous value today with the products that we have with customers. It's the existing opportunity plus the ability to add on so much more. We already have startups that have pretty much bet their future building on top of Snowflake. We are expanding that with things like container services and native applications. We have thousands of startups that are going to participate in a contest a few weeks from now for, you know, like, the best idea to be built on top of Snowflake.
It is that sort of platform mentality plus the ecosystem that is possible plus the fact that it is an incredibly rich vein to be mined for many years to come. It's really the culmination of all of these things that made me think that this was the place that I wanted to be, you know, spending many years. And as I was saying, yes, there are the immediate goals. And, you know, we need to be hitting those numbers. But I think of this company as having the potential to make tens of billions, $100+ billion, in a world where we are still at a very early stage of things like cloud adoption. And that core concept, so easy to state, so hard to realize, which is, how do you provide a cloud that has data at its center?
If you're using Snowflake, you don't have to worry about how your data flows from AWS to GCP. We take care of that. We take care of replication. We truly provide a, you know, like, this transparent framework almost that is able to bridge a lot of these things. I think it is that core platform. It is also the mentality of people like Mike, like Frank, like Benoît to think forward, four steps ahead, and invest in things like a platform ecosystem that make this opportunity really, really exciting for me.
If I could add too, Keith, you know, there's this perception of many investors that Sridhar is a technologist. I can tell you, yeah, he's an amazing technologist. But he is a driver. And he is an operator just in the time that I have spent with Sridhar. And, you know, Frank, he was brought onto the company to really put discipline in place, help build the go-to-market. We're now 7,000+ people. This is actually the biggest company Frank has ever been at now. When he left ServiceNow, it was, I think, $2.4 billion. This was $2.7 billion last year. This was the right time for Frank to hand over the reins to someone to take it to the next level. And he is 65, too.
Got it.
I don't plan on doing this when I'm 65.
I don't plan on doing this when I'm 55. So, like,
we're all on the same page here.
Missed out on scintillating conversations? Like, yeah.
Yeah. It's more the earnings season side of the equation that wears on us, which is a segue into talking about Q4 earnings.
Love that.
Yeah. It made me bring Mike into the conversation. Let's just hit it head-on. Like, the big investor debate right now is the level of conservatism in FY 2025. In the FY 2025 guidance, it came in below kind of what consensus expectations we're looking for. Can you talk us through kind of it does seem like, from the conversation, that there is an increased level of conservatism in the guidance into FY 2025. Why was this the right time to apply that extra level of conservatism?
Well, I'll start with Q4. You know, Q4, we had a good bookings quarter. But to be honest, we still fell short of expectations on the revenue side from what we thought internally. And going into this year with the CEO transition, last year was a rough year. We were hoping, from a consumption standpoint, we'd see customers return to consumption patterns pre-2024. As a reminder, we're in fiscal 2025 right now. And we didn't see that. Yeah, consumption trends are good right now. But with the new CEO coming on board, we have so many new products coming to GA that we just we're not going to forecast those until we start seeing a history of consumption. So I do think there's call it conservatism. I call it prudent guidance. And we'll take it one quarter at a time.
We would really like to see Sridhar succeed as our new CEO.
Got it. In terms of the consumption falling short of your expectations in Q4, any postmortem that you could give us in terms of why that was? Was it just, like.
Yeah, we talked about it. It was really the holiday, the prolonged effect of the holiday this year. Remember, in Q4, you have two big holidays. You have well, there's more than two big holidays. But you have Thanksgiving in the U.S., which does have a dramatic impact on consumption. But this year, with New Year's and Christmas, it was pretty much a two-week vacation for people, it seemed. That was a little bit more prolonged than we were expecting. It wasn't until the second half of January that we saw things really pick back up again to where we expected.
Got it. Got it. Then in terms of the overall spending environment, one of the senses that I've been getting through both kind of Q4 prints and also here at the TMT conferences, I feel like investors got a little bit over our skis. Like, we saw a second derivative change. We saw things stopped getting worse and maybe getting a little bit better. But getting a little bit better still is in a good spending environment. So that's my perception of kind of the spending environment. What's your perception of kind of what's going on in enterprise spending?
Well, actually, the spending environment is pretty good. You see that based upon the commitment customers are making with their bookings. I would say the one area that still is a little bit challenged is more the venture-backed companies. They're still buying, but they're buying in very small increments. There's a lot of companies that I know a lot of our customers are working on funding rounds right now. They're not happy with valuations right now. I think you're going to see more of that. It's a small piece of our business. I would say about 10% of our customer base revenue is associated with venture-backed startups, pre-IPO companies. The spending environment is actually pretty good with our customers.
I just think a lot of the customers we have now that are ramping on Snowflake are a lot more disciplined, more mature companies than a lot of the digitally native companies, where they had these massive valuations and money being thrown at them that they just spent euphorically. The customers we're bringing on today and ramping up, big telcos and banks and stuff, have always been very cost-conscious. They're going to do things at their pace. Those are the ones that are becoming our top customers rather than the digitally native companies.
We are also a lot more disciplined about how we work with customers, how they're spending their money. We know that at the end of the day, the CIO has to go justify the Snowflake line item to their CFO. And so there's a lot more attention that we are paying also to things like, you know, are things being implemented well or, you know, making sure that the customer is getting enough value out of the spend that they are getting. And that work is better done upfront than to go through, like, up and down sort of thing. It's also our mentality that I think is working more closely with customers.
Got it. So in the near term, there's a more prudent approach to ensuring that you're going to be in a position to beat and raise on a go-forward basis. You also took away the longer-term targets. A lot of investors saw a very similar construct to kind of how you thought about long-term targets at ServiceNow and how you're thinking about at Snowflake. It was a very P times Q equation of how many large customers you can get and what they're going to be spending. If we think about the perhaps less confidence in that long-term target, was it more on the P side of the equation or the Q side of the equation that's come under a little bit more caution, if you will?
What I would say is, everyone we talked about this at Investor Day, and I said, to hit that $10 billion target, we need to see consumption trends improve in the second half of the year last year. And they never did to the extent to get back to that pre-2024 level. And what I said as well is, we are still charging internally at that $10 billion. But you can do the math to get there. We need to see a massive re-acceleration next year in our business. I'm not forecasting that to the street. But there's a lot of new products we have coming out. And we're doing everything we can to get there. And I've said, let's take it quarter- by- quarter. And we'll revisit it later on.
Got it.
Another thing to keep in mind is that unlike, say, SaaS subscription businesses, which can bank the money and then just dole it out by quarter because the subscription is already paid for, we recognize revenue when there is consumption. And so there's very much, like, an earn-it mentality that is built into the model, which is why the predictions that Mike's the forecast that Mike's team do is based on just historical consumption. We literally have to show the money to you and to ourselves in order to earn it. So I would say the comparison to something like ServiceNow is not quite apt in the case of Snowflake. There is power in this model in that our customers only pay for what they're using for.
As we think about things like applications built on top, think of all the SaaS subscriptions your companies are using and have them, you know, go to a model in which it's like consumption. If only 1/10 of a team is using something, that is all that you pay for. So there's power in this model. But it is also subject to sort of earning it as you go along.
Right. And you guys made a change in the sales compensation model to align more towards the sales incentive model to align more towards consumption versus bookings.
Yeah. I'll tell you. We made, you know, when I've been at the company now almost 5 years, it will be the summer, one of the first things that when I sat down with Frank and we were thinking about what do we need to do on the go-to-market, the change was, is the company previously used to just pay reps on a booking. And they just sold a 1-year deal. And customers could have unlimited rollover. So you had early on a lot of customers that I would say were oversold because customers didn't care. They could use it whenever. And reps got a big commission all upfront. And one of the first things we said is, we need to change that. We really only want customers to buy what they think they're going to consume. And so we put in place that there's no rollover.
Our sales team said, well, that won't work. And I'm like, we'll get a customer buy a 3-year contract. And we'll give them rollover. And if they renew for an amount equal to or greater, we'll let them rollover any unused stuff. And then we also need to change that incentive because we need our reps. It's one, you can sell the customer. But that doesn't drive revenue. You need to get the customer to consume. So we started four years ago this transition. And I told the team, eventually, we're going to get to 100% commission on consumption. And one of the things last year, we had our reps were anywhere from 10/90-90/10, whether it was growth and revenue. They were being paid on. But what happened, and this is the one thing we did not do. We had a small spiff for landing new customers.
But reps, they had a growth quota that could come from a new customer. It could come from an upsell. But the reality is, upsells happen on their own. When customers consume, they have to buy more capacity. And reps gravitated toward just upsells because it was easier than spending the time going through this security review and getting a new customer. So our comp plan this year is 35% of our reps, their only job is to land new customers, get that initial contract. You sign anything within 12 months after that, you'll get paid on that, too. And then 55% of the reps are just on revenue only. Their job is to be in the customer, walking the halls, finding new workloads to drive more revenue. That's how they're going to make money.
And then 10% are what we call a hybrid that have a mix of new customers they have to go after, which we've identified in their territory, and some small number of existing customers that are driving revenue.
I mean, that's mostly small theaters that can't afford the specialization. But the bulk of the revenue plan, you know, has the specialization of you're either getting new customers or you're focused on increasing consumption with existing customers.
Got it.
I want to shift gears a little bit and talk about the.
Let me just say one. This was not a surprise to the Salesforce. They knew this was coming last May, June. We told them.
Yeah. And I mean, I think the interesting thing from the investor perspective is, we all think of salespeople as being coin-operated. If you now incent more heavily the salespeople to go after consumption, shouldn't that have a positive impact on consumption? Like, aren't they going to be pushing that hard?
If SaaS hadn't entered, we wouldn't have done it.
Excellent. Shifting gears a little bit more to the product side, in particular, GenAI services, definitely top of mind for investors, top of mind for CIOs. Actually, hit the top of our CIO priority list was AI/ML this year. Can you talk to us about what we have in the pipeline coming up? The Neeva acquisition did sort of broaden out the perspective. But there's also Container Services and Cortex. Can you just talk to us about what the full portfolio is and what our expectations should be on when we're going to start to see more of this stuff coming into production?
Yeah. Yeah. So when we thought about AI, we wanted it to be integrated deeply into Snowflake so that it works out of the box. This is one of, you know, like, the most amazing quality that the Snowflake team has in product design. And so when we announced Snowflake Cortex, it was essentially a platform AI layer, meaning if you are accessing Snowflake, you had access to all of these things. It was nothing special that you did. We run a model garden, which is a collection of GPUs and models as part of every deployment. Or you, like, you know, it's just built natively into Snowflake. And the benefit and the way we integrated all of this AI functionality is, we make it available at sort of every level of the stack.
What this means in practice, for example, is that an analyst that wants to write SQL, most of us can write SQL, can access language models. This very morning, a product manager that I was talking to said, "Sridhar, I wanted to just take a look at the new use case data that we have sitting in Snowflake. This is our reps entering use cases for all of their customers. What drives consumption?" He said, "Oh, I was looking to see if by running sentiment detection on the text that they enter for their comments, whether I could predict the win rate or not." Sure enough, it turns out, when people have positive comments about use cases, our win rate is, like, close to 99%. On the other hand, if it comes out negative, then the win rate is far, far slower.
Now, this person knows nothing about language models, nothing about deployment, how they should use it. It was just like a query that they ran. 15 seconds later, they had that insight. So that is power. In addition, Cortex also has a Semantic Index. You can think of this as a Vector Index. So startups are all the rage these days, built natively into it. And our aspiration here is that you should be able to build a chatbot on any corpus, a bunch of PDF documents, maybe your help text, maybe some customer support cases. You should be able to do that in 5 minutes and stick it into a Streamlit app. So now you have basically a RAG-powered solution that can do citations for you, that can produce reliable answers based on your data. And then this is what is enabled sort of at the platform layer.
It just hit Public Preview a few hours before. We also announced a big partnership with Mistral, one of, you know, these folks came out of nowhere to become one of the top model makers of the world. We are going to be hosting their models natively within Snowflake. And, you know, that gives you an idea of what is being enabled at a platform level. We're also working on key applications. One is called Document AI, which is basically for extracting structured information from documents like PDF. And there's also a Copilot offering, which is the holy grail right now of people that have data in SQL databases, which is, I want to be able to ask questions in English. And you figure out the SQL query, run it, and produce me a tabular answer or a visualization without my needing to know SQL.
This is one of the most difficult problems that there is. Just to give you an idea of how difficult, you know, Claude 3 from Anthropic came out as the best model yesterday. They provide benchmarks, for example, where they score 85% on a pretty hard benchmark. Another way to look at 85% is, like, 1/6 of the answers are just wrong. Is that a business application that you and I want to put? No. This is where, on things like the Copilot API, we are placing an extra emphasis on things like, how do you make sure that the answers you generate are actually the right ones? How do you play this trade-off against complicated questions you should not answer versus the questions you should be confident of answering? As I was saying, Cortex hit public preview today.
We hope to have it be in GA by summit time, which is early June. We think this will be a, like, meaningful addition to our AI portfolio simply because it enables any person that can write SQL on top of Snowflake to also become an AI person out of the box.
It's interesting because I think that's one of the unappreciated aspects of Snowflake that customers appreciate and investors perhaps don't as much. We all think about Snowflake as highly performant. It is a much more performant data warehouse than what you're doing on-premise. We think about it as highly secure. This is where you go when you want highly governed, secure data. But it's also ease of use. It seems like that's an aspect you're really leaning into, particularly when it comes to the GenAI portfolio.
Well, it is. It goes into every product that we develop. We do the hard work so that our customers don't have to. We provide, for example, this is in the land of machine learning. But we provide functions for things like anomaly detection, for time series prediction, which, by the way, like, language models cannot do things like that. You still need traditional machine learning. But we ship that as part of the core SQL libraries that we ship so that one of our big customers, one that spends, like, multiple tens of millions dollars a year with us, they're super excited because they go, wait, I have thousands of analysts. And all of a sudden, they begin to do data science. What I told you earlier about doing sentiment analysis, that's, like, it's new age analysis.
But again, somebody that can write SQL is able to do that. I think that tight integration with the core product is a huge strength of Snowflake. One of the VPs of data science that I met, you know, I've met over 100 customers, been to, like, five continents in the past few months meeting with customers. One of the things he told me was, like, "Sridhar, I've used Snowflake for three years. I've not filed a single support ticket.
Impressive. So Streamlit went GA a couple of months ago, I want to say.
Yeah. I think that's right.
We're talking about Cortex going GA in June. We're going to have Iceberg Tables coming GA. But the only new solution or relatively new solution that's part of the guidance is Snowpark, right? And Snowpark had a really nice year of acceleration, expected to be 3 percentage points of overall revenues in the forward year. Can you talk to us about the main drivers of adoption of Snowpark, where you're seeing success with that, and what is bringing it into the fold for customers?
I can take a first crack at it. The core functionality in Snowpark has been around for a while. Really, what drove adoption was us understanding that this was a new technology and a new skill to be learned by our sales engineers and account execs, just in being able to go through a process of working with customers to figure out where we can be doing these things. It is that pattern matching, you know, where they are talking to their contacts in different departments, prioritizing which use cases should come to Snowflake, and then deciding on these. I would almost say the success of Snowpark so far, which we are going to take as a lesson going forward, is that it is, like, that efficient sales enablement. A lot of our sales engineers, sales folks, have been with Snowflake for a while.
They know how to use the core, you know, SQL product. They know how to use collaboration. But Snowpark is a different beast. You're writing this in Python. This is, like, data engineering. It's different from what people have done. And the lesson that we are taking away from it is, yes, there's a slew of stuff that's coming to GA. But we need to almost bring a new lens to how we think about, you know, sales enablement, whether it's in the field CTO office or the sales engineers or even the account exec in being able to confidently pitch and say, "Aha, that use case, you can do, like, very, very quickly on top of Snowflake." You know, while we like what Snowpark has accomplished, I would actually say the value of Snowpark is much more in it telling us what we need to do in the future.
Got it. So there's new products that are coming into general availability that are potential upside to the guidance. There's also price performance improvements in the overall platform that we have to think about in terms of compressing consumption a little bit. Automated data warehousing sizing is relatively straightforward. It's making sure your customers are in the right size data warehouse to avoid them being just kind of overspending on a go-forward basis. Tiered storage is one that I struggle with a little bit more to understand the impacts. But I don't know where customers are today in terms of what tier and where they could potentially go to.
Yeah, yeah. So historically, we have discounted a few of our very large customers in storage. We've renegotiated over the last few years our pricing with AWS, Google, and Azure such that we had a fair bit of margin in storage. That was never our intent, to have a big margin in storage. So we rolled out we started doing it, played with it in Q3. But in Q4, we rolled out what we called tiered storage pricing. It's only impacting our biggest customers. You have to spend a minimum of $1.2 million, I think, a year to get there. So all of our big customers have this now. And the more you commit, the bigger the discount on storage you get. And that is you're getting the full impact of that in Q1.
It started a little bit in Q3 and more towards the end of Q4 when customers sign contracts. You're going to get the full benefit or impact of that this year. That's factored into the guidance.
It's kind of like a volume discounting methodology for the largest customers?
Yes.
OK. That makes sense. And then the one that.
I didn't see that much downside in doing it because most of those big customers are ones over time that want to go to Iceberg Tables anyways.
That's a great segue because Iceberg Tables is kind of in the middle. It's something that can weigh on consumption in the near term. But you guys see it as, like, a large opportunity longer term. Can you talk us through why that is? Like, what's interesting about Iceberg Tables? And why do you think that can spark higher consumption on a go-forward day?
I'll start. Then I'll let Sridhar talk. But first of all, most of the impact from customers moving to Iceberg Tables will be next year. It doesn't go GA until mid this year. And then it's going to take customers some time. What we do think is new workloads will start to be accessible to Snowflake later, as soon as that goes GA, so that that can actually drive compute in our engine. Hard to forecast what that's going to be. We haven't forecast that. I do think some of our larger customers, because they've told us they want to, will move to open file formats in Iceberg. That will take storage out of Snowflake. And when you remember, when you ingest data into Snowflake, that takes compute up. When you take it out, it takes compute, too.
But we will lose that compute associated with moving the data into Snowflake. But we'll be able to run queries about that. And it will open up immensely the amount of data available to run queries on within Snowflake. So net-net, we think it's going to be a positive. But I'm not forecasting that to be positive this year.
Got it, got it. I'm going to open it up for questions so that the mic runners could get into position. There's one question, actually. Actually, we can take the question back there.
Thank you. From a competitive standpoint, Databricks is presenting at this conference tomorrow. So perhaps we can get kind of an early rebuttal on the competitive concerns. I assume they're going to talk about accelerating growth and share gains in the market. So I just kind of want to hear the response to that ahead of time. Thank you.
It's predictive research.
You know, all I can tell you is when it's been very consistent since day one, when we are going into getting a customer to migrate from on-prem Teradata or with their Hadoop or whatever, we're generally competing with Google, Azure, and then AWS. We do see Databricks in many of our large accounts. And by the way, we helped bring them into many of those large accounts pre-me, when we were kind of partnering with them more. And we do compete with them for workloads. And they do very well in the data science persona because they have a very good notebook that people like. But what I will tell you, a lot of that growth of our Snowpark has been at the cost of growth in Databricks within our accounts for Spark workloads.
We are talking to a number of customers and showing them that we are dramatically cheaper than them, too. So this whole notion that we're expensive, all I'll say is, get real data. Don't look at benchmark data and real customer workloads. And that will speak for itself. I'm not going to say they're going to do well, just like we're doing well still, too. And all I know is my numbers are actually reported numbers. And next year is actually guidance as if I'm a public company. So I can't comment on what they're going to say they're going to do next year. And I almost guarantee you, if they were public, they wouldn't say that's what they were going to do next year.
True.
I would also point out, too, we've been very efficient in terms of operating margin expansion and free cash flow.
Yeah. Just on the competitive side of the equation, you've always spoke to Google, BigQuery, AWS, Azure as more of the competitors. The fact that there's such an affinity now, like Anthropic is with AWS and the GPT models with Azure, has that changed the competitive dynamic at all?
So I would say this is where models like Mistral or even what Facebook did with the Llama model are important. In many ways, Llama set the stage for the explosion of other companies, including us, that look at this area as something where we can make a difference. Replicating the performance of Llama 2 is now, like, you know, feels more and more like an achievable benchmark. As we all know, like, replicating GPT-4, which nobody has done, just felt that much harder. So I think, especially with the rapid progress that, like, companies like Mistral, with whom we have a partnership, are doing and the ability to expand on that, I think we're in a pretty good place when it comes to that.
And over time, I think you will see less and less of, like, pure model affinity as things go from, you know, it's like there's infinite dollars in everything in the world of AI to, oh, maybe demand is actually somewhat limited. I think that's when you're also going to see more players get their hands on models as well.
Excellent. We can take one more question in the front.
Yeah. I just to complete the competition landscape. So Databricks claim they have better products for the unstructured data. So we talked to some of the early AI startups. They also say the same thing and wonder, like, how do Snowflake respond in terms of product offering? And how should we look at this? That's the first question. And second question is that, so yeah, we understand. So next couple of quarters is going to see. But if we're taking a longer term, let's say 2025, 2026, should we expect Snowflake back to the higher growth trajectory to, like, $10 billion target, like, I mean, like, 30%+? That's the growth rate that investors expect on longer term. Thank you.
That's right. I'll let Mike take the second question. I'll start with the first one. Snowflake actually does just fine with unstructured data. We support storing data in XML, being able to access it, being able to write SQL queries on it, being able to access it from Python. So I don't feel that as a competitive gap. It's more one of companies trying to highlight their strengths. For example, with the Databricks notebook, which I have used a lot of, as I have with Snowflake, your typical modus operandi is that your data is sitting in S3. And you fire up a notebook. And you try. And you try and access it. From within that notebook, you might write Python or SQL code to be able to do it. But what was remarkable about Snowflake is our ability to handle data at scale.
So for example, we loaded the entire Neeva index, which was multiple petabytes. It was all of our search pages, into Snowflake. And we were able to run interactive queries on not a very large warehouse. And so it's, I think, like, we offer excellent performance for both structured as well as unstructured data. We also support these things called external tables, where you can keep the data in S3. And you can create an external table that references that. And then finally, for example, we have published benchmarks on things like Icebergs sitting in external cloud storage, in which we show that the performance on those tables is pretty similar to that of performance on top of regular Snowflake. So we feel pretty good about just, like, how competitive the product offering is in a number of cases. We are behind on the notebook.
We acknowledge that there's a product in private preview. Just like with AI, where we went from not quite there to being world-class in six months, we will make sure that we ramp up with that much veracity and velocity.
On your second question, I think I answered it. I'll say what I said. Internally, we're still charging towards the $10 billion. Clearly, we'll need a re-acceleration in our revenue growth next year. We'll take it one quarter at a time.
Excellent. Well, thank you, gentlemen, for joining us. A great conversation. We'll see you again here next year.
Thank you.