Snowflake Inc. (SNOW) Earnings Call Transcript & Summary

Thank you, everyone, for joining us. My name is Keith Weiss. I run the U.S. software research effort here at Morgan Stanley, and very pleased to have with us CFO of Snowflake, Mike Scarpelli. Thanks for joining us, Mike.

Good morning.

Before we get started, a brief disclosure. For important research disclosure and personal holdings information, please see the Morgan Stanley website at www.morganstanley.com\researchdisclosures.

So Mike, again, thank you so much for joining us. First time at the TMT Conference, I think, as CFO of the public Snowflake. Coming off of a tremendous year, 106% revenue growth, net dollar expansion rates that we thought were incredibly high to begin with, actually improving throughout the year, ending the year at 178%, I believe.

Correct.

On the cohort expansion. Maybe to start out, just a big picture. Can you talk to us about like the core opportunity that Snowflake is going after, what's really hitting the mark? Because we hear it from customers. We've seen the results. What is it on the product side of the equation that you guys have gotten so right that you're able to on tap that opportunity at such a rapid pace?

Yes. Well, a couple of things. We've been very successful with doing these big on-prem migrations. And once the data gets into the cloud, customers are realizing so much more that they can do with the data. And it's really because of the performance of Snowflake and the ease of use. I would say the #1 thing that we got right is ease of use. You talk to our customers, you do not need to be a very sophisticated person to use Snowflake, unlike some of the other products out there you need to be. And quite frankly, a lot of other vendors have failed at these on-prem migrations. And we got that right. Data sharing. Data sharing is one of the #1 things that gets customers excited about Snowflake. They may not use it Day 1, but they all have aspirations for data sharing and what they can do. And what I mean by data sharing, it can be data sharing within the organization or it can be data sharing between a vendor and a partner relationship or it may be data sharing to consume data that you're buying data through a customer that they do data sharing. So there's all kinds of data sharing relationships, and that's what we got right that excites our customers.

Got it. When I think about the Snowflake opportunity, I see kind of 3 acts emerging, right? Sort of Act 1 is there's a lot of on-premise traditional data warehouse workloads that are moving into Snowflake because the performance and the price/performance is just so much better. Act 2, once they get there, they have the ability to share that data, but also commingle that with other data types. Act 3 is you're going to build the analytical applications, both sort of your end customer but also OEMs on top of that data. Is that a good framework to think about sort of the product strategy over the next 3, 5 years?

I think that's good, especially the getting developers to build applications on top of Snowflake is that key strategy that we have. We have all the data in Snowflake. People want to build applications with real data. We offer that. And we offer it in a secure, governed way. That's the other thing that we've really got right is the security and governance around Snowflake. You know exactly who is accessing data and who is doing what with data. I think we have some of the best security and governance there.

Got it. Got it. And just in terms of, put some parameters around sort of the size of the market opportunity. You can look at it from sort of the, like the top-down perspective, the IDC categories, and analytical database is close to a $20 billion market today. So that in itself is big. I think equally as interesting, if we look at it from a bottoms-up perspective of what Snowflake customers themselves have been doing with the platform, how big individual customers could get, you have 184 customers over $1 million. Can you talk to us about some of those biggest customers, how big can they get? Is there still room in those customers? And if we think about the broader customer base of approaching 6,000, how many of them have the ability to get to that $1 million scale?

Yes. Well, we talked a little bit about that at our Analyst Day last June on what it was going to take to get to $10 billion and the 1,400 customers paying us north of $1 million a year. But going back to the 184 customers we have today, not one of those is fully saturated. A lot of people ask us, what does a mature customer look like? One of our largest customers 1.5 years ago, if you told me were they a mature customer fully saturated when they were doing $32 million a year, I would have said yes. Well, that same customer is doing $42 million a year run rate right now, and there's still opportunity within that customer. So we really don't know what a mature, fully saturated customer looks like because we haven't gotten there. And yet when I compare that customer with other customers out there and I realize how some of the other banks that are much, much bigger could be consuming way more than that, it's just going to take years to get there. It is a very slow process. A lot of people think that we land an on-prem customer to migrate to the cloud and it just happens in 6 months, and they shut down the legacy system. The reality is, is we have signed hundreds and hundreds of customers to migrate from on-prem to the cloud. The reality is, only about 50 of our customers have fully shut down their legacy system and about 30-plus of those have been Teradata. What most customers do is they move their key workloads where they're having challenges first into the cloud. They start using it and then they gradually move more and more. And as we improve our price/performance, they move more workloads to us. This is going to be a many, many year journey. And by the way, this is no different, and I've said this before, than what I saw at ServiceNow. There was customers that took 10 years to shut down their legacy system before they were fully deployed on ServiceNow, and we see that happening here.

Right. And I think another interesting part of the equation is the ability to be a $1 million-plus customer isn't just limited to the Fortune 500.

Correct. If you look at our Fortune 500, the average revenue of our Fortune 500 is only about $1.3 million today revenue. And if you look at the average revenue out of the customers paying us north of $1 million, it's $3.5 million. So it tells you there's a lot of growth opportunity within the Fortune 500. And by the way, we've landed Fortune 500s over the last 6 months that have zero revenue today because it takes time to roll it out. They may have a tiny bit because they're doing some testing and stuff on it, but they're nowhere near their contracted rates.

Okay. It's a good segue into sort of the consumption model and the consumption model aspect of Snowflake. I'm a big fan of consumption models broadly, right? To me, this solves one of the original sins of software, which is shelfware, selling customers product that they're not going to use. Snowflake doesn't get paid until people start hitting the data, until certain people start querying. So you're matching up the value to the customer to value to Snowflake. On the other side of the equation, there's a learning curve for investors to understand sort of how these consumption models work. We got to understand the seasonality. The other item that came out in the most recent quarter is the price/performance curve that Snowflake is incented to show their customers better price/performance over time. Can you talk to us a little bit about the price/performance or sort of platform enhancements that you guys have put into place, what that does in terms of price/performance? And we'll start with kind of the financial impact that Snowflake gets upfront on that.

So first of all, on the consumption model, it is a different model. It's harder to forecast the business. We literally forecast revenue on a daily basis by customer. And you see seasonality in the weekends. There's less consumption. Holiday is another thing. So it is more challenging to forecast the business. But the beautiful thing about Snowflake is I'm able to look on real-time every day what was actual consumption yesterday. What does that mean? We literally reforecast on a daily basis. We have always factored in every year a 5% revenue headwind associated with performance improvements, whether they're coming from software, whether they're coming from hardware. For instance, this last year, we had a revenue headwind associated with storage compression that reduced the amount of storage customers were consuming. We saw some customers decrease as much as 20% their storage. It all depends upon what you were storing. And that's a revenue headwind, but that's good. What we did see is when our storage is cheaper, people put more data in it, and that drives more consumption on the compute side because you only put data in Snowflake if you're actually going to do something with it in Snowflake. We have also seen too that when you make the, and that doesn't really improve the price/performance, it's just making storage cheaper. Price/performance is what we look at all the time. We rolled out our warehouse scheduling service. And by the way, these happen all the time throughout the year. This was the most profound. Six months ago, when we were talking to our engineering team, they didn't think this was going to have as big of an impact. And by the way, it doesn't impact all customers. It only impacts some customers. And it's counterintuitive to the CPU enhancements on Graviton2, and I'll explain that. With warehouse scheduling service, what this makes is it makes your processing of your queries more efficient when you're scheduling. So if you in your warehouse, you just have one massive warehouse of data, and you're running just really, really large queries in a few of them, warehouse scheduling service has no impact on your performance. But if you're a customer where you have a lot of data, but hundreds and hundreds of small queries you're running, that's going to have a dramatic impact on your performance and what you're using. Now switching it to a CPU improvement when with the CPU improvement we're seeing with Graviton2, AWS' proprietary chip. We see, it really varies by customer. As an example, Snowflake, we saw 5% performance improvement. We see other companies with zero and yet we've seen a few smaller ones 30% improvement, depends upon what you're doing. If you're running really, really large queries that are really compute-intensive, you're going to see a huge impact. But if you're running a lot of small queries, where there's still a lot of read and write that has to go back to memory, you don't see the same, you don't get the full benefit of the CPU because you're I/O bound. So it's not as easy to predict until you really put it into production, but we feel that every year we'll have these 5% revenue headwinds associated with these things.

Right. But on the flip side of the equation, you improve the price/performance. You bring down sort of the...

It opens the market opportunity.

Exactly. So there's from a P and Q side of the equation, it opens you up to more quantitative workloads coming on over time. Do you have any visibility into that in terms of sort of within your customers that there are certain workloads that right now Snowflake wouldn't be price/performance, but they start to come in at different levels?

Yes. It's kind of a build it and they will come, but not quite. We know there's a number of customers that we run POCs in the last 12 months that we know that those customers have said, we'll move these workloads over to you. But until you get your performance at the speed at which you can do things up to this level because the latency is a bigger issue, we can't move those to Snowflake. So we know with all the performance improvements we're doing within our existing customers, our salespeople just need to go back in and demo and show what we've done and run POCs again that there will be workloads moving to us, but it takes time because once again, customers don't just, our performance improves and they move more workloads on that. They still want to test it and do a number of other things. And that's why we think there's about a 6-month lag for every. And historically, we've seen that, too, is every performance improvement we do, we may have a revenue hit, but then within 6 months, those customers are consuming more.

Got it. And just longer term, I would think there's a more strategic imperative to ensure that your customers see continual price/performance gains that you open up for more and more types of workloads over time. Because in the Ps and Qs equation, like you want to maximize for the Qs near term. You want to maximize for bringing as many workloads on as possible because we know those workloads are extremely sticky and then over time that they're going to grow. So that you're incented to bring on as much quantity of workload as possible.

Correct. And history has shown that if you make things cheaper and better performance, people do more in you from a technology standpoint.

Got it. So if we take a big step back just to close on the consumption topic, the better-than-expected price/performance gains, the growth trajectory of Snowflake, has this impacted the longer-term growth trajectory in any significant way or is this just sort of part of the normal course?

This is really the normal course. I was telling you last night, when we went public, we built a IPO model. Our IPO model for 2023 had us doing, I think it was $1.680 billion. You can go back and look at it because you had it. In May of 2021, when we had our earnings call and then we did our analyst call in June, we called out this new chipset with Arm when we were talking about the hardware compression technology. Our revenue forecast for 2023 then compared to the $1.9 billion we guided on the high end was $120 million less. So our business is strong and continuing to grow. We had a phenomenal Q4, and we've managed through all these price/performance improvements.

So the net of that, basically a year ago at the IPO, we were expecting $1.68 billion, now we're going to do over $1.9 billion, even with the price/performance?

Correct.

Okay. So in the...

I got to be honest. I almost wish I never said anything about this price/performance. I'm very transparent.

But I think it's important that you understand the consumption models over time. I want to switch gears and talk about the competitive environment. I think when we think about the landscape in terms of 3 kind of core competitors. One is the legacy vendors that were dominant in the on-premise environment. Two, the other hyperscalers out there like a GCP or an AWS and Azure. And then three would be kind of emerging new more private companies and a lot of stuff coming out of the open source world. What do you actually see in reality? What do you see in the field?

Yes. So when we're competing to do an on-prem migration, it's always Google, Microsoft and AWS. We're not competing with Teradata. We're not competing with Oracle. We're not competing with SAP HANA. Those customers already made the decision. They want to get off prem and go into the cloud. And I would say Google is the most competitive and their product is probably the closest, BigQuery. They're extremely competitive on deals. They do a lot of things with giving very free, really cheap, throwing in a lot of things for free to get them. And we've been very successful there. Doesn't mean we win every deal. Technically, we almost always win, but they will go in up at the top and balance of trade. I would say AWS, we partner with the most. Doesn't mean we don't compete with them, too. They still compete with Redshift and stuff. But we co-sold $1.2 billion with the cloud vendors last year, and the vast majority was AWS, the vast majority. Zero was Google. A little bit was Microsoft. And then within our existing customers, we do see some of these emerging, call them, like Databricks. They coexist in many of our accounts. I would say 80% we don't overlap. Like for instance, Morgan Stanley, you use both Snowflake and Databricks. You talk to the guy at AT&T who's very, you talk to customers and they don't see the competitive nature between the 2. There's about 20% where we kind of bump heads a little bit, but the vast majority of customers we coexist, and they see a very defined use case where they use Snowflake and a very defined use case where they use Databricks. Databricks tends to do very well within the data scientists, a more technical user. You need to be a more sophisticated user. The one thing that I mentioned at the beginning that Snowflake has, Snowflake is just easy to use. You don't need to be a heavy, tech sophisticated user to use Snowflake. And that's why they tend to do well in more tech companies, where we still do well in tech companies, but we do well in, think of big companies in the Midwest that don't necessarily have as much technical talent as companies on the West Coast do. We tend to do better. And as I said, we coexist. In the vast majority of what they do and we do, we don't overlap. It's a good company and it's a massive market. There's going to be many winners in this market.

Right. Exactly. Just touching on the co-selling arrangements that you have with AWS and Azure and don't really do much on GCP. Can you just, you started citing those numbers, I think, in Q3. You talked about $500 million in Q3. And then that number come up to $1.2 billion. Were those in existence in prior years and it just started scaling now or is that more of a newer phenomenon?

It's been in existence for the last 2 years, but it really is scaling now we're seeing. And one of the reasons why we see that, and actually, there was an article in CRN with AWS and Snowflake about the relationship. AWS is realizing the amount of other software services they can sell around Snowflake. And they would rather a customer be in AWS than a customer be in GCP or Azure and vice versa for those others. But there's a lot of SageMaker that gets sold in AWS around Snowflake. And if AWS didn't partner with us to be in there, there's other vendors of those types of services that are partners of ours that we would bring in. So when AWS partners with us, we're obviously going to push their software services versus a third-party software services that could still work in AWS.

Right. Got it. I want to shift gears a little bit and talk about sort of the evolution of the product strategy over time, kind of shifting from Act 1 to Act 2, if you will. In terms of sort of the data sharing, we started talking a lot more about the Data Marketplace, how many different data sets you have in there. Where are we in the usage of that, both in terms of your end customer understanding what they would do kind of with the Data Marketplace and to kind of the monetization of that from a Snowflake perspective?

Sure. So the Data Marketplace customers have had, for some time, the ability to consume data through the marketplace that you weren't able to transact directly in the marketplace. To buy that data, you have to transact outside. We have in private preview right now. We've built the infrastructure for that to happen with Stripe as the processing on the back end. That will be rolled out fully GA this year. But I don't want you to think of that as being a big revenue generator for us incrementally, the selling of that data. We're really just looking to cover our costs. It's no different than marketplace. Our list price is 10%. But the reality is, is on larger customers, as long as we get 2%, it more than covers our costs. It's like AWS Marketplace or Azure Marketplace. They try 10%, but the reality is you pay 1% to 3% depending on the size of deal going through it.

Got it. And then in terms of customers seeing sort of the opportunities and seeing the potential sort of new businesses that they can stand up. Where are we in that of your customers in that and what they can do?

I still think we're early, but we have a number of companies that have built their businesses on Snowflake. We have, I think, 285 powered by Snowflake customers. A lot of those will have, you will be able to procure through marketplace. But marketplace is more the data versus the software today. We do see a day where people can actually transact to buy software applications through marketplace, but that's still some time away. And that's one of the things that we think it's important that we have Snowpark for Python in GA and Streamlit will make it much easier for people to write applications on Snowflake.

Got it. Great segue into sort of Act 3, which is building out sort of the ability and building on Snowflake as a application development platform. Can you talk to us about where Snowpark fits into that? What's so important about Snowpark in that initiative? And then Streamlit was a recent acquisition that you just made. How does that further kind of catalyze the opportunity, if you will?

Sure. First of all, I'm not a technologist and I couldn't write an application. But I will tell you, Snowpark enables people and developers much easier to write applications, directly accessing the data within Snowflake. And the first Snowpark version we rolled out was Java, but the more common language that people are using today is Python. And you may say, well, why didn't you do Python? Well, it took us over 2 years. We were working on Java. That time, Python wasn't as popular. Python has really taken off. Python is in private preview now. It will be GA at the end of the year. That will be important. Streamlit, what's really important about Streamlit? Streamlit really, really, it's open source. It's a really good following of developers. There's over 40,000 developers who are using Streamlit today in the world. Many of those are on Fortune 500 companies. There's over 1.5 million applications that have been developed using Streamlit open source today. And one of the things that Streamlit was working on doing was to actually build out a cloud within Google that they could have people build applications directly in that cloud. The problem was they didn't have the data. The other problem is they struggle with security and governance. And that's where Snowflake is really good at security and governance, and we have the data. And so by building this, we're still going to support the open source community and do this, so anyone can still do it. But if you want to be able to use Streamlit in a secure and governed way, you'll use Snowflake. And all of our customers will be able to use this. And why? This is going to drive more consumption. This is not going to be something we're going to charge our customers for. But the more you're using it, it's driving consumption in Snowflake.

Right. And across the solution portfolio, I think the common theme from like the pricing model perspective of trying to lower the barriers to entry, lower friction to using Snowflake, catalyzing kind of more activity, whether vis-a-vis data sharing or applications, but all the monetization is going to take place vis-a-vis compute and driving more compute.

Correct. There is some storage, but the storage runs about 10%, low margin. It's really compute. That's the beautiful thing, too. That when we come up with a new feature or product, whatever you want to call it, we don't have to go back into our customers and try to sell it through procurement to license them for something. They just automatically get it, depending on what version you're on. But something like Streamlit, we would make available to every version of Snowflake. And so you don't have to engage with procurement and procurement is something that slows down people all the time.

Great. Got it. So there's 2 prices, prices for compute, prices for storage and a simplified price list and limited friction. So I want to shift gears a little bit to go-to-market. Today, international is about 18% of the overall business. Where are we in terms of expanding out the distribution capability, enabling sort of international to get to a larger scale, if you will?

Yes. Well, what I would say is, our international business is doing extremely well. It is growing very quickly, but it's being overshadowed by the strength in North America. North America just had a phenomenal yearly figure. I would say we're very mature now in the U.K., France and Germany, and we continue to build out more in other countries. For instance, we just opened up this year Israel in the Middle East. We're pretty much in every country in Europe with zero exposure to Russia and Ukraine. Actually, we have no employees physically based there either. We have a lot in Poland, and that is something we're concerned about with what's going on over there. We have a lot of engineers in Poland and Berlin. On the go-to-market side, what I would say is, 2.5, 3 years ago, the customer base in Europe was really a commercial customer base. It's really shifted to be more of an enterprise-focused and large. It doesn't mean we're ignoring. We're still building out the commercial. We're hiring reps that know how to sell into, and we have been for the last 2, 2.5 years, that know how to sell into larger enterprise. That's where Frank spends a lot of time in Europe in front of these larger customers. And we're seeing that traction, and I feel really good about the EMEA business. We've made a lot of investments in Asia. Historically, Australia had been our strongest market there. We're seeing very good traction in Japan starting to pick up where we're seeing big deals. South Korea, we landed some big marquee accounts in South Korea. I know it's going to take a while to ramp them from a revenue perspective, but these are customers that can be $10 million, $20 million a year customers. We're building up quite a bit in India right now on the sales side, too, because a lot of nice, interesting customers there to go after. And so internationally, I feel very good that, that business is going to grow. It's just North America where I think we're going to slow down our growth. We haven't seen that. Every geo last year, I think, for the full year exceeded their plans.

Excellent.

North America exceeded by more.

Okay. Good problem to have. Also on the go-to-market, can you talk to us a little bit about sort of the cloud verticalization in that it's not really a product initiative. It's more of a go-to-market initiative in terms of better understanding your customers' problems. Can you talk to us about how that's been executed to in terms of the sales force?

Sure. We've been working on this for quite some time. And you're correct in saying, we talk about we have a media and ad cloud. We have a financial services cloud. The reality is, it's the exact same product. It's really, we have groups within the sales organization that really understand the business problems of those industries and how Snowflake can help with those business problems to help drive their business. And that's why we have the vertical focus. The type of people we're hiring. We're hiring people that actually come out of the ad tech industry. We're hiring people that come out of, we have SEs that we're hiring that come with biotech backgrounds. We just hired one that's a genomics expert. Why? Because this person is going to focus on health care customers. So how can Snowflake help health care customers to in essence solve problems to invent cures for certain diseases and stuff. We're dealing with one of the largest health care providers in cardiovascular. They have all this data and it's in all different silos. And they want to figure out a way, how can Snowflake help them solve cardiovascular disease. And you need people who understand those industries. It's not just about technology. It's understanding the type of data for how Snowflake can help you.

So selling a solution rather than just selling a technology?

Yes.

So running short on time, I want to squeeze in one margin question. Earlier, we talked about sort of the outperformance versus the initial IPO targets on the top line in terms of what you're targeting for FY '23. The bottom line is exceeding kind of the IPO targets as well. The ramp to free cash flow positive was faster than initially expected. Is that durable on a go-forward basis? Is that sort of faster pace of profitability improvement something that's structural or was it kind of more onetime in nature and we're going to have to give it back?

Yes. Well, I think on the operating margin, I do think we saw the benefit of COVID on the operating margin that we weren't expanding our offices as quickly during COVID even though we were hiring people. We didn't have people traveling. So we did see margin, operating margin improvement there. In terms of next year, obviously, we're returning to the office. We're returning to in-person events. That does impact our margin, but we're still showing expansion on an annual basis. And free cash flow, we've been very focused on free cash flow since the day Frank joined the company. And I'm very focused on that. I feel very good about the 15% free cash flow next year. And I do, as I said on the call, we are very focused on growth, but it's going to be efficient growth. And we will continue to show leverage every year because I just don't think we could spend the money faster and get the productivity we'd want to get out of it.

Right.

If you ask any of the leaders in the company, no one is starved for headcount in the company.

Right. Got it. So a lot of inherent leverage?

That's correct.

Outstanding. Unfortunately, that takes us to the end of our allotted time slot, but thank you very much for joining us, Mike.

Thank you, Keith, and thank you, everyone.