Snowflake Inc. (SNOW) Earnings Call Transcript & Summary

Hey, guys. Thanks for joining us. I'm really looking forward to having Mike here.

Mike, to get everyone back on the same page. You guys had like pretty good, solid, healthy numbers. Just from your perspective, what were the highlights? How did it play out for you?

We saw good performance and it was very broad-based from large customers to small customers. We talked about on the call that we really saw a stabilization in consumption patterns going back to a little bit more historical consumption patterns. We called out that September after Labor Day had some of our largest week-over-week growth within our consumption trends -- within our customers continue to see strong daily consumption in October and November, and I feel good about where we're sitting right now. I'm not saying this is a recovery and you're going to expect a lot of upside. But really, we've seen more stabilization, no big optimizations. I'm not hearing of any big optimizations. I will say, and I've said this so many times, though, optimizations will always happen in our business as a normal cycle. Like I know 2 of our big customers that really started to ramp last quarter, although they've been ramping over the last few, and they were doing big on-prem migrations. It is very typical that when a customer starts migrating, a lot of their consumption spikes and then they go into production and it kind of moderates and comes back, then they move more data and it goes up -- you'll see those trends. That is a normal trend within our customer base. So no, I feel good and it's broad-based across industries, financial services continues to do well. Health care continues to do, technology is doing well, telco. I will tell you the one area that has a lot of upside. There's a little downside because it's such a small piece is FedRAMP high. And literally, the PMO from the federal government has said, we have our FedRAMP high and it will be -- you'll see it on our website. Every day, we're checking. And that will open up a lot of new opportunities for us, and we have a healthy pipeline within the federal business that we've been working on for the last 2 years are just waiting for this.

Yes, yes, yes. Okay. That's interesting. And then we've had in London last month. Well, it's actually in October. Yes, it's like time flying. The -- you just did the big customer, Global Tour Events. I think you were in New York and London. What was the customer conversations like there?

So it was New York, London and Stockholm. I would say New York and London was very much financial services, and I actually did a dinner with a bunch of the chief data officers at the big banks in New York. And their excitement around Snowflake was streaming Dynamic Tables and Iceberg Tables. They all want Iceberg. They all want to be able to have their data. And you can understand why, so that they don't have to pay for the cost of storage twice. And it actually will make Snowflake cheaper for them too and they can do more work on Snowflake by having all of their data in open format Iceberg. Nobody wants to get locked in on their data. Everyone wants to have their data in open file formats, that it's easy to move the data. It's easy to move the data in. It's easy to move the data out and it makes it cheaper to run your queries on that.

Yes, yes. And then I mean did you notice what are differences in terms of how far customers are down the adoption curve, if you think about your versus...

Absolutely. Yes. I would say the U.S. customers in general are further along in their adoption of Snowflake across the enterprise. I would say in Europe, it's more early. Certain -- there are customers that are very well established there, but in general, I don't think I have a single European customer in my top 10 yet. I don't have a single Asian customer in my top 10 yet. I have some in both that will be close to getting in there, but they're not there yet. You'd expect to see at least a couple Asian and European ones over the long term in there.

But do you think it's just timing?

Timing.

Timing. And is there -- do you know when you were having the client conversations, is there differences on end demand there at the moment? Or is it like is it globally, it's tougher times. Now everyone is kind of living in tougher times.

Yes, I would say the European customers in some Asian by region are a lot more concerned about data residency and sovereign clouds because of their -- it's reviving its head again. I heard this when I was at ServiceNow. The whole company turning over all of their customer data to the U.S. government. That's been revived a bit lately. So -- and we're working on strategies with different sovereign clouds.

Yes, yes, yes. You talked a little bit about the last quarter as well about the shifting mix in the top customers that you have more established guys coming in there. Like where are we on that journey?

It's continuing. I will have at least one new top 10 customer this quarter because of a bank that is growing really fast, potentially another one. It doesn't mean those other customers aren't doing good. They're just -- I'm just getting such big customers and it's a lot more mature. If I look at our top customers, they're -- early on, we used to have the likes of Laceworks and Instacart and Coinbase, who are early adopters of Snowflake in our top 10. Now I just have Instacart is still in my top 10, they continue to grow. There was that big stuff 6 months ago that everyone those guys continue to grow with us. It's just I have more mature -- I got a telco that's moved up into there. So we'll have 2 telcos in there, more banks in there, large banks.

And if you think about it, what does it mean from a predictability perspective? Like obviously, earlier this year, it's like those digital natives having to kind of do almost like emergency kind of thinking about their stuff. Does it...

What I like about large enterprises that are more mature is they've always been cost conscious. So I don't see them -- their costs getting out of control. There are always a lot more disciplined, and it is more predictable, their business. They are also a lot more methodical in how they do their migrations, and they share that with us. So I know like I talked about next quarter, I'll have a bank move into my top 10. We have a road map over the next 3 years. We've been working with them on migrating data, and they have hundreds of on-prem data warehouses.

And so is it -- okay. And -- from your perspective, visibility, like I'm now kind of referring more like to the CFO is like, is it -- is there a contract with it? Or is there kind of more like you talk about a road map and...

Well, the road map I'm talking about is on the technical level with our PS people and our sales engineers with the customer on the migration. There are contracts with these customers, which is their commitment they're making with Snowflake.

Yes. Okay. And then the -- if you think about it, like where like -- I would agree with you, there's tons of like data master warehouses still inside. Like the one question I get a lot is like, oh, like but Teradata was never that big. So why Snowflake is still kind of growing. Like where does -- where do you use these workloads coming from? And what -- is there any new workload in there as well as well as a...

Yes, yes. So a lot of people think we just replaced Teradata. When we're doing a Teradata migration, we're generally replacing a lot of Hadoop, a lot of SQL server, a lot of different things. And there's a lot more new use cases that customers determined and I talked about this one telco that's been ramped. They've been a customer for almost 8 years now, I think, or 6 years. It's before my time. They were never -- they never started with an on-prem migration. They started doing some new use cases in the cloud and in the last 6 months, they've decided they're doing their Teradata migration. So it's all over the board.

Yes, yes, yes. Okay. Okay. Perfect. Yes, that makes sense. And it looks like that's the thing when I used to cover them like I think Teradata was only like a #4 in the market, like -- because, yes, they were a high end that everyone knew them, but like there's so much Oracle, Microsoft...

It's kind of like when I was at ServiceNow whenever it really is very similar to the analogy that everyone always just thought ServiceNow was a help desk market to help desk market is $1.6 billion market. Look how big it is today? It's clearly a -- there's a lot more -- we're much broader -- data warehouse is just one workload. Data sharing is one of the key differentiators of Snowflake and I think people do not appreciate that. That's why financial services and the financial service providers think of the BNY Mellon, the State Street, the Fidelities, the BlackRock, DTCC are all standardizing on Snowflake and doing data sharing with everyone. And it's a huge network effect to get new customers because of that. And it was funny, 4 years ago, we were trying to get Bloomberg data into Snowflake and get Bloomberg and now took so long, but now Bloomberg data, not all of it yet, we're working on that, is available for customers to use data sharing and get it in Snowflake natively.

That's the thing like data sharing is like the one thing that everyone underappreciates there, yes. The -- Okay. Yes, makes sense. And then the -- shifting gear a little bit, like so there's the snowflake now we have like Snowpark. Like how do you -- like maybe where are you -- what are you trying to achieve at Snowpark, and where are you on that maturity curve?

I think we're still very much in the early innings of Snowpark now. It's available in multiple languages to work with now that we have Python. There's still more though, there's still different versions of Python. Snowpark is about a $70 million run rate right now for us today. So it's starting to get meaningful. Really, what we're -- our sales people have been focused on is Spark migrations. And more, I would say it's almost a low-hanging fruit because we can see the workloads customers are running in snowflake whether you're using open source Spark, Databricks, EMR, and those are the ones that we've gone after and doing migrations. As Snowpark container services comes out next year, it will be used more heavily in app development for people. What you got to understand with Snowpark and there's a migration involved and that takes time. So we can quickly go in and win a POC and then you need a customer to get lined up to do the migration. And some of those migrations take months and months to do depending on how complex the system was. So it takes time. And we're now starting to see some of those things go into production today.

And then the -- I'm glad you were here because like Frank would kind of kick me out of the room like container services, like why is that so interesting? Like I should know that but...

Yes. The big thing about container services, it will enable people and your container to bring your own application into Snowflake to run natively in Snowflake, and you don't have to worry about all the security and other things because we provide all that security. I'm giving it a very simple form but it also will enable technologists either I'm an accountant. You can also bring your own large language model of choice into Snowflake to be able to fine-tune that model on your data. That's what's really exciting. And we talked about Cortex coming out, which makes it easier for the average analyst to be able to do that. If you're a heavy core data scientists, you'll use Snowpark or container services to do that directly. But it will also enable new applications to be developed very easily in Snowflake. So as Frank always talked about, it's about bringing the applications to the data versus the data to the applications. And that way, you know the data stays in a secured governed environment.

And then you have Streamlit as well, which kind of help you kind of build the...

Streamlit will enable people to develop applications. It also gives you a visualization layer. Within my finance department, there's a lot of dashboards that I look at on a daily basis, but it's more than just a dashboard, it's dynamic. But I used to do everything in Tableau. And now I've switched everything over to Streamlit. And that's the first thing I do every day when I wake up as I go into my Streamlit dashboard and I look at my revenue for the prior day. And then I look at all of our customers to reforecast, I can drill down into anything. I can see how many Snowpark credits were consumed yesterday, how many streaming dynamic table credits were consumed by who.

And the -- on that note, I wanted to ask that later, but I might as well just squeeze the question in now, like if you think about the forecasting and the maturity of your forecasting, I appreciate it's tough because consumption trends, so you can't predict them. Like how does that -- like the forecasting cadence kind of changed for you over the last couple of years with all this?

Yes. We've developed an in-house model for doing our forecasting and your model is based upon historical data, and if your model has never seen a downturn, it's hard for it to predict that in that. And so over the last, actually 9 to 12 months, we knew our model had outgrown -- our business had outgrown our model, and we developed a new model that we just went live with it at the beginning of -- I guess it was the end of Q2, we were running it in parallel. We went live with it at the beginning of Q3, which was a series, we call it our ensemble model. It's a bunch of different models, looking at the different cohorts to predict our business, and it seems to be a much more accurate model at the specific customer level than what our old model was.

Yes. And then the -- but...

And now we have all the data around the downturn so you can better predict when you see customers slowing down what's happening.

Will it ever be like...

No model is ever going to be...

Yes, it's like you never know what's coming next. Yes. Okay. Yes.

And by the way, you still have to have human intervention with models. Even when we had our model before, there were certain things I never believed and we would make top side adjustments, I would say, with our ensemble model and make a lot of adjustments.

Okay. Like -- so in the downturn, you kind of put another layer buffer on in a way. Is that kind of the way to think about it?

I would just say, I think it's a more detailed model that better splits out our business, and I will say, as we become bigger, it becomes easier to predict.

Because like one customer coming down it's like...

One customer is not going to have as big of an impact today as it did 2 years ago.

Yes, yes, yes. Fair enough. The -- I wanted to switch gear a little bit. So if you think about new products, there's a lot of stuff coming out from your perspective. You mentioned some of that already. Like if you kind of -- I don't want to -- it's like picking your favorite child, but like if you kind of rank what's coming out there? What's kind of the one that we should pay most attention to?

In the next 12 months and over the last 3 months is probably the most new product feature releases that we've ever had in Snowflake's history. We now have Streamlit just went GA a few weeks ago. Streaming Dynamic Tables is in public preview. You have Iceberg Tables is in private preview. You have container services is in private preview. Unistore is going into private -- or public preview. We actually have 10 customers that are actually using Unistore in production environment. Iceberg Tables will be -- a lot of those things will be GA by Summit next year in June is what our goal is. And I think the one that I'm the most -- I continue to be very bullish on Snowpark and what it can do for us longer term. Container services is a big one because as we talked about, that's going to enable both native app developers on Snowflake to easier develop. And we -- by the way, we have a lot of customers that are re-platforming their business on Snowflake. I think of like [indiscernible] there and there's a bunch of others. They all want container services. We also have a number of customers that want to be able to fine-tune their large language models on their Snowflake data in Snowflake and container services will enable it to happen easily in a secure governed way where you know your data is never leaving Snowflake.

How does that -- if you think about the new products, you're kind of broadening your addressable market by quite a bit. Like do you think about -- do you still think like in the old IPO base of my TAM? Or like is it -- like do you try to put numbers on? Or how do you think about it?

Yes. Well, we laid it out at our Investor Day, it's over a $200 billion market opportunity we're going after and growing. And there's going to be many, many winners in this space. It's not one person take all, but data continues to grow. And if you want to get the benefits of generative AI, you're not going to do that on-prem, you're going to do it in the cloud. And so you're going to have to have your data in the cloud. And in order to get good results out of your large language models, you have to make sure you have clean data, you're running it on, we think Snowflake is the ideal place for that to be done. Will it be the only place, absolutely not. But we think it will be a prime choice for customers to do that.

Yes. I mean how do you think about like if you -- we had Workday, ServiceNow and Snowflake on -- not Snowflake, Salesforce on downstairs on the stage as well. They always -- they pride themselves because they have these systems of record with a lot of data in there and then kind of see that as like a good starting point for doing AI. Like in a way you could be the consolidator of data, like how do you see that playing out for you guys?

The reality is most customers do use CRM, Salesforce. They use ServiceNow, they use ERPs, whether it's Workday or [indiscernible] SAP, whatever. All of those people are putting their data in the Snowflake because when you're going to run any type of model, you generally are going to want to look at a lot of different data from different sources and Snowflake's -- customers will choose what the right place to do, and we see them choosing Snowflake. We already have a lot of customers taking ServiceNow data and putting it into a snowflake. CRM data, the most common type of data going Snowflake, CRM data. We have a lot of large customers that want to have all of their SAP data, some big German companies are really pushing to have all of their data, their SAP data in Snowflake.

Yes. It wouldn't be in HANA. Okay. Yes.

It's funny actually, ServiceNow took their -- and I remember, I was the one that made the decision to go with HANA there. All that HANA data is now in Snowflake.

Okay. That's funny. Yes. Okay. And then how does it change the competitive -- like -- or how do you see that playing out over time in the competitive landscape then? Because it does feel like everyone a little bit comes together because data is the gravity.

It's funny. All of those companies you mentioned they are customers of Snowflake too. As I said, I think there's going to be many winners in this. It's not one person, and there's going to be a lot of coopetition happening. Just like today, how we compete with AWS, Azure and Google, yet we partner really well with AWS and Azure. Actually, I think we just went -- AWS re:Invent, we just won their partner of the year for their data. And so we clearly partner with those, and we'll work with -- we partner well. At the end of the day, I think all of those people are going to want to do what's in the best interest of their customers. And if their customers want that, we'll work together. And we work very well with all of those companies.

Yes. Yes. Okay. And then you mentioned AI already a little bit. Like are there -- how do you see that playing out for you? Or just -- are you kind of seeing this like as the -- you are the vehicle to consolidate the data and then AI gets driven off that? Or are there other initiatives for you guys?

There are other initiatives that we're working on. Some we're not talking about, and I won't. But there's a lot of -- at the end of the day, the value of what you get out of your models is going to be based upon how good a data you have. And we think Snowflake is the right place for that. We are working on offering an on-demand GPU model where even though I have to do it through a reservation, it's for 30 days, I'm working on structuring that. I think that will be an attractive thing. And because people do struggle, especially small companies getting access to GPUs. With AI, most customers, when you talk to them, still are trying to figure out their AI strategy, but they realize that they have to have their data in the cloud to really get the benefit of it. Everyone talks about a copilot. We announced a copilot 6 months ago, too. Everyone has one. Those are simple things with AI. But wouldn't it be nice to -- and we have a vision -- where a business analyst can just ask any question to Snowflake in plain English and be able to get an answer? That's the type of stuff we're working on.

Yes. I mean -- and follow-on question. You've always been like a -- we know you have for years, and you've always been a very sober CFO. How do you think about the monetization, especially around these co-pilots because there is like there's some firms that, go, well, like, oh, I can jack up the price by 60%, then off you go for running more queries or creating more queries. How do you think about that monetization?

The beautiful thing about stuff like, we really have one product with 4 different flavors: Enterprise, standard enterprise, business critical and VPS. Most customers are on business critical. And every new feature we have, you don't get charged for it. It's the consumption, the compute that it's using those different features is how we monetize. And so the more queries you run, the more times you refresh your queries, the more often you want your data updated using streaming and dynamic tables, it's going to use more compute, you have to pay for that. And so it truly is a consumption model. And the more features we have, the more the customers use those is going to drive revenue. It's not about -- we're not going to increase prices. And by the way, if you buy more, you commit more, you'll get better pricing too on your price per credit. We now have tiered storage pricing that if you make a bigger commitment, we'll give you a better price on storage. I'm going to protect the price per credit. But the storage pricing, we get a better price on.

Yes. Actually -- and I only have a few minutes left, so I need to kind of get a couple of hedge funds question in. Like you mentioned consumption.

I don't answer hedge fund questions.

So that's why I do. If you think about consumption, like how sensitive do you think you guys will be if the economy gets better? It's like what we've seen so far is like people were thinking about optimization or maybe I have less data in Snowflake and then save me a little bit of money. If it gets better, again, like how sensitive do you think you will be?

In a consumption model and what Snowflake has, customers can really ramp very quickly, but they can also slow down if they choose. And so I think a lot of the new products we have coming out next year could be a reacceleration in our business. I am not forecasting that right now, and I need more time to see how these things play out before I look at changing any guidance or anything. I will say a lot of these new things that we have coming on next year will be a headwind to product margin expansion. And there's a number of things from unit store -- the economics on that are not as good as our core product just because of the way the system works to get the performance. There's having to store the data twice and there's costs we need to take out, we'll get there, but it's going to take about 6 months. GPU's trying to do an on-demand model is more challenging and will be an economic headwind, but I think we have to have the GPUs to offer to people because we want to be early to the game on that to get people on Snowflake. There's a lot of things happening with certain customers around the world with sovereign clouds that are a headwind to margin expansion as we're opening up new deployments. It takes time before you can get those to scale to get to your ideal contribution margin.

And then since we are on margins, like so that's more on a gross margin level. Like how do you think about the OpEx spending and then operating margin as part of that?

Raimo, how many years have they known you? 12, 13 years? We're seniors, and every year, I've always showed operating margin expansion.

Yes, you did actually.

We will continue to show operating margin expansion. It doesn't mean we're not going to invest in the business, but we're going to invest efficiently. And every year, we're going to show operating margin expansion. I've said that at Investor Day, and we're very -- I think we can grow this business very fast and without sacrificing growth, continue to show leverage. The one piece where we're not going to see any -- on an individual -- we have a lot of R&D investments we're making next year and continuing to make with these products, a lot of the stuff around AI, what we're doing. I got another $31 million hit to R&D next year just with what we're doing on GPUs internally.

Yes. Okay. And then last question on cash, like you're nicely cash flow positive. You had a very strong cash position. Like how do you think about usage of cash?

We talked about last end of February, beginning of March, we -- the board authorized a $2 billion buyback. We bought back just under $600 million -- was at $598 million to date at $147. We will continue to opportunistically repurchase shares. As I said, we're going to try to match it to our cash flow. Does that mean I'm going to use all of the cash? Or are there may be quarters where I go over, but over the period of time -- we -- a lot of companies say they're going to do buybacks, but then never do them. I always like to do what we say we're going to do. So we're going to continue to do that. And we continue to look at M&A and -- but all of the M&A we're looking at continues to be this smaller M&A and is more of the teams of people that we can bring on. And I would say Neeva was an amazing acquisition for us with the talent we got out of that. Really, really strong talent there.

I mean you're in a good position. You're growing organically really well, and there's nothing to -- like what do you want to buy?

Yes. We don't need to buy a revenue stream, but any company that has revenue, I really don't want it.

Yes, yes. Perfect. I think our time is up. Mike thanks. I really enjoyed our conversation again.

Thank you.