Snowflake Inc. (SNOW) Earnings Call Transcript & Summary

If we have such a large gathering on the last day of a 4-day conference, it has to be the company.

I don't know about -- I think it's you, Kash, that hairdo, they all want to see. I'm just jealous so...

And if it's not just the company, it's Scarpelli. Your last name is like -- people don't refer to you as the CFO of Snowflake. They don't even call you Mike, it's Scarpelli. Your last name is like legendary in software CFO circles. Mike, welcome back.

Thank you.

It's great to have you here with us.

Thank you.

Yes. So can you share with us -- the news in the last 12 months has been a new CEO. You and Frank have been paired up on quite a few occasions. What is it like to have Sridhar? What is he doing at the company? What are the changes...

What I would say is, first and foremost, Sridhar is hyper-focused on getting engineering to deliver product faster, but product that customers want. And so he's really driving more alignment between engineering, product management and the go-to-market functions and making sure that engineering really owns a revenue number for what they're building. And he's really driving that a lot more. I don't think our go-to-market and product and engineering was as tightly aligned as it is right now. Sridhar is also super focused across the company. Just to scale the business, you need to have very good processes in place. You need much better cross-functional working going on in the company. And the guy is -- I've never met an individual who can retain as much data as he does. He is like in the weeds into everything, but he can go up to a very high level, too. He's a really impressive individual and a workaholic. I can attest to that. It's funny, I was joking with Chris Degnan today, too, and we're both like, I haven't worked this hard in 10 years. Yes.

I'm pretty confident that I would not survive. I work hard, but this conference the past week, it's just -- working hard, but also it's been just very gratifying, fulfilling to see all of our companies come together on stage and having these conversations. Maybe a bit more. Is it fair to say the culture of the company is changing from sales-dominated to equally balanced between sales and engineering? I mean, how would you say...

Sridhar is super focused on go-to-market and it's customer first. It's really listening to our customer. And so I would say, no, but he wants engineering to take more ownership in the go-to-market, too, to make sure that the go-to-market is -- has the right product. And I think that's going really well. He's also trying to, as we're getting bigger as a company, to make sure decisions can be made faster and empowering more teams to make decisions so there's not bottlenecks and waiting for things.

He's in Tokyo, as we speak, I believe...

He is in Tokyo right now, doing our keynote for Data World Cloud Tour there right now.

Nice, excellent. So Mike, your consumption business model, I mean, any broader trends you'd like to share with us?

I don't really comment on inter-quarter trends, but I would just say as we gave guidance, we like the trends we were seeing, and our guidance reflects what we're seeing at that point in time.

Okay. And so guidance for second half, you guys -- some of you smiled or I heard some -- Mike usually talks more about consumption trends. He's not saying anything, but I take that smile as a...

I'm pleased.

Okay. Right. So guidance for the second half went up, which is good. I mean, rare to see in software, a solid beat and a raise that was more than the magnitude of the beat. But some investors concluded that maybe the magnitude of the beat is slowing down and maybe there was a conclusion given the adverse reaction to the stock price that there is a step down in the growth here of the company. We've all been kind of spoiled by closer to 29%, 30% type product growth, which is good, right? At the same time, you've got several new initiatives, Copilot, Cortex, Document AI. They're not in the guidance. Where could we be surprised if we are surprised?

So I'm going to...

There are so many things in there. You can throw that into an LLM.

Just as we weren't super excited about a 5.5% beat in Q1, I wasn't that excited about a 2.4% beat in Q2. This is a consumption model. And consumption changes by customers on a daily basis. Some go up, some go down, that is a normal thing that we see in our business. And it's much harder to predict. We try to give guidance that is meaningful guidance. And we've said since day 1 going public, a 3% to 5% beat is a good beat. I'm not saying a 2% beat is bad or a 6% is great. I try to give it so that it's in line with that 3% to 5%, but there's so many variables in that. What I would say is, and I know a lot of people are asking is, what is Cortex and Iceberg and all these things going to do for you next year? And I would just say, I need to see a few quarters of history before I can put that into our guidance. And I would like to be in a position when we give guidance for next year that we're going to have better visibility because clearly, it needs to be part of our guidance for next year. And it will. I just don't know what that is going to be yet until I have a couple more quarters of consumption. What I will say is we're very -- we feel very good about the number of customers that are starting to -- are already using Cortex. We have 2,500 customers as of the last quarter, and that continues to grow. I'm excited what I'm seeing in notebooks. We had 1,600 customers in notebooks and that's going to go into GA soon, knowing we have more development work to do on all of these new features. There's always more development work that happens. But we think that will be meaningful to help drive Snowpark even more beyond the 3% we told people it was going to be this year. And Snowpark is right on track to do that through the first half of the year.

Yes. And it's not like it's going to stay at 3%. If it could go from 0% to 3%, it's going to go from 3%...

We would like it to be higher, but I'll wait and see what we -- exiting the year, where we're at, we'll guide based upon that. But it's clearly growing faster than the core of the business. But the core of our business is very strong. Data warehousing and data engineering are doing very well for us.

That's good to know, good to know. NER, finally, after going through a rough period because of the macro rate increases, it seems to be stabilizing. It was down a little bit, but it's down a lot less than it has been before. What's driving the stabilization, if you -- if I could indulge in using the term, stabilization?

What do you mean? Stabilization in what?

NER, the net expansion rate.

As we've said before, we do expect over time that net revenue retention will converge with our revenue growth. And remember, net revenue retention, you have to be a customer for 2 years before you get included in that cohort. And the reason being is if we don't do it that way, that number is much higher because of all those new customers that are growing so much faster in the early part of their journey. And so I do expect that, that will converge over time, and I'm pleased with that stabilization there.

Got it. We do have a cRPO question similar in here because it was very solid. But I know that you're involved in the products of the company. And we don't need to get super technical, but the company has a goal for massively increasing the number of new products this year to 11 to 12. It was 4 last year, 3 the year prior and 2 per year, the year prior to that. How is engineering keeping pace with this list?

So what I would say is, you have to remember, many of these things, we've been working on for years. They're not something that we've just started working on last year. And some of these projects take much longer -- like Iceberg, we've been working on Iceberg for a number of years. I would say the one that was fastest to market was probably Cortex. Notebooks was pretty fast. But the other ones we've been working on for many, many years, and there's more to come as well, too. We tend not to talk about future road map until we have those into public preview or private preview.

So I think you touched upon this a little bit, but we have Snowpark, we have Cortex and unstructured data contributing to growth this year. Where are we -- and tell us more about your conviction since you're targeting 3%. What are the things that you're seeing from a customer activity consumption standpoint that give you the conviction that the noncore can continue to be a good driver of the business in the years to come?

Well, the biggest thing that we're seeing is just the uptake in the number of customers that are starting to use the new features. Clearly, Snowpark, we're seeing that, we're seeing a lot of migrations of customers. We can see customers -- existing customers who are running Spark workloads. They're taking the data out of Snowflake, running those smart workloads and putting them back. We've been doing a number of those migrations, and there's a lot more in the pipeline to do. We've learned a lot in that, that, that is a migration. It takes time. But I'm really excited about Iceberg. It's still very much in the early innings, but we already had, as of last quarter, 400 customers that are now net new workloads, data that was never in Snowflake, they're starting to run things in Iceberg with Snowflake. So I'm pretty excited about that, that I think that will open up a lot of data that otherwise would not have been in Snowflake.

That's great. Can we talk about Cortex? I think there was the case study of Penske Logistics, and there was also a financial services call center case study that you talked about. How do you package and price Cortex? Or is it just a way to get customers to consume the platform...

It is priced separately. But at the end of the day, we sell a customer a credit and capacity. And the normal data warehousing or data engineering that you get charged literally by the second of compute time use. In the case of Cortex, Cortex is really by -- it's a token. And this is an industry thing that others are doing, in that you get so many words or those interactions are running LLMs, that's all done by token, and that really is -- we look at the pricing of that and we're comparable with what others are doing. I will say the margin on that is much lower and you see that in our -- and this is really the pricing keeps dropping on these things, yet the underlying GPU costs have not come down.

Could there be relief when Blackwell comes up? Is that an avenue to potentially lower costs of compute?

Potentially, but we're not on that yet. So the biggest thing that's going to impact the GPUs is really -- we still are in a little bit of supply constraint out there. We have more than enough in where the majority of our customers are. There are some regions around the world where you still can't get GPUs through the hyperscalers. They just don't have availability yet. But in the U.S., I have no issue with GPUs.

You should have been here yesterday...

And I'm not going to buy any more GPUs until I see the revenue to support it.

Okay. Okay. Well, since you said you're supply constrained, you should have been here yesterday morning at 7:20. Jensen Huang was up on stage and it was crazy. We had like overflow, overflows, overflows. So -- and he did acknowledge the supply situation, and he said relief's on the way in Q4...

I do think that is going to come, and then I think you're going to see pricing come down.

Yes. Yes. It's good for software companies.

Yes.

And what I've been talking about is, it's about time that the whole GenAI activity percolated up from infrastructure to platforms and applications. I had not planned on asking it, but do you care for -- any thoughts there? I mean, when are we going to see just GenAI? I think it's been so much of an infrastructure build-out, LLMs, et cetera, et cetera. And I've been of the view that this is going to amount to nothing, unless platform companies like you and applications companies can start to do some useful things for end customers. How are you thinking about this, if you have thoughts?

I'm not a technologist, so I'm not going to say. All I will say is I think it's still in the very early innings. I think everyone is talking about trying to figure out GenAI use cases within companies, but the reality is that very few are using it en masse today, other than there's a lot of technology-focused companies. But I'm talking brick-and-mortar large enterprises across the world. They're still not -- have not figured that -- I was in an EBC yesterday with a customer, a big German power grid company, one of the biggest in Europe, and they're still trying to figure that out.

Are you guys internally doing anything with GenAI at all, just exploring use cases?

Oh, internally, we are doing a number of things. I will say, one of the things that we are -- and there's a lot of use cases you can do. Like for instance, in development, in your QA, writing your automated test scripts and running the test scripts to do things, I'm seeing a lot of GenAI use cases and cleansing data much faster. We're looking at how GenAI can help migrations go faster with tooling and stuff. We're working on it as well as some of our partners here. So there's a lot of use cases there. We have not really deployed GenAI yet in the support function and how we can get better leverage out of support by use of that. And that's something we are exploring now. So what I would say is I think GenAI is really in the early innings of most enterprises.

Migration of applications? Or what...

Migration of applications.

Okay. From...

From workloads, from -- yes.

Got it. And so you're seeing some automation benefits?

We're working on that, as I said.

Okay. So I'll share with you that I've asked this question of other company executives in the last 3 days or so. Almost everybody has some GenAI project going on internally. And I think the example of Twilio was very particularly resonant. They've talked about how a project that originally was scheduled for 3 months with 100 engineers, it seemed too good to be true. But with GenAI, they got it done in 10 seconds. And I asked the CFO immediately, so are you going to dial that into the guidance? And she said that ultimately, yes. And I looked at the CEO, and I said, so this has to have some impact on your business. He said, it will be material very soon. So it kind of stood out in every company not quite to that magnitude of internal operational cost savings, but it is going to be material, is my...

I was talking to a data provider on Friday in my New York office, and he was telling me how they are using large language models, where they used to send the data offshore to be cleansed by humans. And now they can run it on large language models and have that done in a matter of hours that was taking weeks before.

That's amazing. Thank you for sharing that data point. Those are things that -- it's just -- because the question we get from investors, when is this thing going to be returning on the massive investment?

I would just say, I think it's in the early innings. But as you said, every customers talking about it, but they're trying to figure it out, how they're going to use it. And there are a lot of customers that are concerned about what data can be subject to a large language model in their company. And it's really putting in place the policies around the use of GenAI in your organization.

Yes. Mike, another thing that I've been leaning towards the view in the last 3 days or so, we talk a lot about LLMs being trained on, on unstructured data. But what you have is precious structured data that cannot be the reason why the LLM would hallucinate, right? So have you given thought to how valuable is -- this data that's sitting in Snowflake is incredible. And if you want to train domain-specific elements, this data is going to be accurate. There's no hallucination. It's going to give you good business insights. What are we missing? The market doesn't see Snowflake that way. Snowflake is not unstructured data. The other guys have unstructured data. There's not been a discussion of how your data is going to be valuable.

Yes. So first of all, Snowflake really got its start in semistructured data, not structured data, and then we went to structured data. We do support unstructured data now, where we take that unstructured data and put it into a structured format so you can get those insights. And we do think the key to AI is your data strategy, and many of our customers do realize that. It just takes time.

Exactly. Are we at the point -- coming back to growth, are we at the point where the core growth is stabilizing? You have all these new products. Could they be additive to growth? Because Iceberg, Native Apps, Snowpark Container, Streamlit are all going to be -- either they're out or out very soon this year and expected to drive growth next year. So where do we get that inflection point where the core is stabilizing and then you have all these new things that could be additive to growth?

Well, first of all, I'm not going to guide to next year. But clearly, we are working, and Sridhar is very focused on this, and we're working very hard. And we would like to reaccelerate the growth in the business, and we are looking at different plans as to what we can do. And I would just say stay tuned for that. Chris, our head of sales, he is working his a** off right now, and that we're really trying to understand where can we invest more in the go-to-market that will drive the yield we need. And I think he's doing a very good job there, and I'm exploring right now an acceleration of sales hiring in the second half of the year. We have asked people, though, to do effective performance management as well, too, of people. You're not just going to add people if you haven't gone through the process of performance management. And that's just a normal thing that you should be doing all the time in an enterprise.

Yes, you talked about at the earnings conference call, your decision to accelerate sales and marketing, hiring in the second half. Tell us more what drove that thought process? Is it the new products that are coming downstream?

No. Part of it is we're seeing certain teams and regions do very well on the acquisition of new customers. And we're realizing that we have a lot of opportunities, opportunities that we're just not following up on as quickly in certain regions and certain geos even within the U.S. And we're realizing we can throw more headcount at those on the sales and SEs in particular, what we're interested in to close those opportunities sooner. Part of our realignment in our comp plan this year was we realized we need people that are just focused on new logos and get them to focus on that. And we're pleased with the changes we've seen coming out of that in the activity. We're actually really good. Chris and the sales team now are much better at actually tracking activity from the number of meetings a rep is taking on a weekly basis. And you can drill down now and see all that data and really understand who are the reps that are working very hard and engaged and who are not. It's not just about...

And Sridhar is focusing on this, too?

Sridhar is focused on data. I'm telling you, the guy works 16-plus hours a day, 7 days a week. And that's not an exaggeration.

So we -- the thing that I hear from investors, he's an engineer. So Frank was the operational sales-oriented leader. And there has not been any conversation with Sridhar that I've had that he did not exhibit an incredible understanding of its customers, the business results. And there's this healthy level of impatience for, let's go get this thing done, right?

I would say, Sridhar is a very data-driven individual, which I think is very good. He doesn't shoot from the hip, and he's just different. They're both -- Frank is great, and Sridhar's great. They're just different individuals.

Great. Let's talk about -- sorry to jump back and forth, this is such a good conversation with you always. It's always good to have you here. Unistore. Let's talk about Unistore. The thing that was exciting, I think, Summit 2023 if I'm not mistaken. I think it's been a theme for the past couple of summits. You said less than 10% of customers are using Unistore, and you've got revenue contribution target for fiscal '27. Why has it taken that long? And could it be even better than what we're...

Well, I would say one of the reasons why it's taken so long is no one has ever done this before. That's why it was not easy from a technical standpoint...

Does everybody know what Unistore is? I don't want to go into a deep...

It's really combining transaction and analytical workloads in one database. And it tends to be more for -- I don't see migrations going to Unistore, it's more for net new workloads that are transactional in nature, but have heavy, heavy data volumes and querying that's needed in that data. We're -- we've been in public preview with it. We are about to go into GA. I'm just -- I'm one of the gating factors for when we say it's GA. And the reason being is the -- they've had targets they've been working on, on the cost structure to get it so that it was going to be margin positive. And I will just say that spending an hour on that yesterday, that will be very soon that, that will be GA.

That's great. That's great. And to just round out the discussion on that, you talked about the sales incentive structure. We talked about how gathering new logos is going to be a focus. We also launched a new program at the start of the fiscal year, where you're going to be shifting a certain percentage of the compensation being driven by consumption. Talk to us more about how that is playing out. Are you satisfied with where we are? How long of a room for improvement do we have for that trajectory?

What I would say, first of all, is comp plans are always being tweaked every year to drive the behavior you want. We started 5 years ago when I joined, moving more of the sales rep compensation to revenue tied to revenue. And the reason being is reps would just get a deal, and they didn't focus on getting the customer to consume that. So we've moved more where 55% of our reps are predominantly, and I want to say predominantly, compensated their quotas based on revenue. They have a lot of other spiffs and stuff where they can make money as well, too. And what I would say is in the second half of the year, we rolled out a spiff that compensates them more on the bookings as well too in the total contract value that's being signed to incent them to also get customers to renew at bigger amounts. And we just rolled that out for the second half of this year.

Okay. So as I promised there was an RPO question. RPO has been outpacing cRPO the last 3 quarters by a healthy magnitude, right? That gap only continues to widen. What's driving all of this outperformance?

What I would say on the RPO side, we're seeing more and more -- as our top customers are changing and we're getting more traditional large enterprises, large banks, they are making longer-term commitments on Snowflake. And most are doing 3-year contracts now. When I joined the company 5 years ago, they had never sold a multiyear contract. It was a 1-year contract only. And I would say most new customers, there are exceptions, but most new customers always start with a 1-year deal, and then they'll do a 1- to 3-year deal. We're seeing more and more of our large established companies do big 3-year deals. And like last quarter, we had two $100 million-plus renewals with customers. And I expect that trend will continue, but they're going to be lumpy. It all depends upon where that customer is in its renewal cycle. We talked about at the end of Q4, we had a customer that did a $250 million 5-year deal. In Q1, we had a customer that did a $120 million or $110 million deal. Last quarter, we had $130 million and $100-something million...

You guys are throwing around $100 million deals like it's like a common...

And I have, between now and the end of the year, I know I'll do 2 to 4 of those. But those are -- I don't expect those customers to come back again for another 2 to 3 years.

Yes. Yes. What is driving these big commitments? I mean, I looked at the backlog, $5.2 billion. I don't think many companies of your size, if any, have that kind of backlog relative to your revenue. What's going on?

I actually don't necessarily disagree with that. We're going to do about, what $3 -- whatever, $3.3 billion, $3.4 billion in revenue. This year, $5.2 billion relative to that and backlog is not that -- remember, it's multiyear, and I don't think that's out of whack at all. And I just think customers, large customers, and people are saying, well, why are customers signing multiyear deals? Well, a, when they're choosing Snowflake, this is not a 1-year decision. These are -- remember, we're replacing systems that are 10, 15, 20 years old. It's very disruptive to replace these systems. And these are enterprises that have made their bet on Snowflake being their platform for a long time. And it is painful for large organizations to go to procurement every year to get new contracts in place. And so a lot of our big customers want to deal with procurement as little as possible.

Have you been easing on your procurement?

I have a lot of people internally complain about my procurement. When they're complaining, it tells me they're doing a good job of holding our vendors on pricing.

Yes, that's great. I wanted to ask you since you brought up the vendors and pricing and procurement, where are you in understanding how much capital you need to put in? I know you've guided to lower margins a while ago. Where are you in understanding how much investment the company needs? At what point could we start to see an abatement and recouping of returns, investments in GPUs that you put in?

Yes. What I would say is, there will be no more GPUs in R&D. We have more than enough. And in terms of the -- what's running through COGS, until we see revenue that supports the investment we've made today, there is going to be no more material increase in GPUs in the COGS line. And why do I say material? There are some regions around the world where we may want a few, depending on customers where we don't have them today, but that's not going to be material.

Okay, good to know. Let's talk about Iceberg tables. How is it playing out? I know you had some thoughts on -- besides a whole bunch of other factors, Jimmy Sexton was right to correct me, he said, it's not just Iceberg, but several other factors that go into the self-induced compression to the revenue growth rate. How -- if you were to just isolate Iceberg tables out, what is the actual demand for Iceberg? And is it playing out per your plan or could it be a little bit slower or faster?

No, I would say all along, we knew Iceberg, the impact in Iceberg was going to be the second half of the year and more skewed towards Q4. What I would say is, and I mentioned earlier, we have 400 customers that are already using Iceberg with net new workloads. They're trying it out, and the feedback has been pretty positive. We know we have more development work we need to do around Iceberg for supporting that. I do know that a number of our large customers want to adopt Iceberg. Not all of our customers, some of our top customers have said, "No, we like it all running in Snowflake. We don't want to deal with managing that." I do think a lot of our customers, it will be a hybrid environment where they'll have some data in Snowflake and other things they move there. It's still too early to tell. I have not seen any customers move data out of Snowflake that was being stored in Snowflake as a result of Iceberg. And even some of our big customers who have said they want to adopt Iceberg, we have it out there, yet I'm still seeing them do big migrations to Snowflake. I think it's going to take a number of years. Remember, a lot of the customers who want -- that have been the ones we've been working with for a number of years that want Iceberg open table format are really a lot of the highly regulated banks, but these are also the ones who move very slow on doing things. So I know they're going to do it. It's just a matter of when, and they're the ones that control when that happens. And they tend to want to do a lot of testing and stuff before they do anything.

Got it. We just have a minute. And if there's one question, we can take one question. And if not, Mike, I just wanted to give you the opportunity to close the presentation by opining on what do you think is missing from an investor perspective? What are we getting wrong with our assessment of Snowflake stock? Not me. I think I have the right assessment. I'd say so...

You know what I would say is, first of all, I never comment on share price or anything.

No, no, not the price. Not the...

And I'm not one to tell you what you have wrong or anything, but I just want to remind people, we are playing in a very large market. There are going to be many winners in this market. And I just want to leave with you, you have to remember why do customers go to Snowflake. They go to Snowflake because of ease of use, that you don't need to be a highly technical user. We have very strong governance in place in Snowflake. And what I mean by governance is, it's very easy to control who has access to what data within the company, and it's very -- a really good audit trail around that data. The way we do data sharing, no one else has data sharing. Financial services have really, really adopted Snowflake because of data sharing. And we're the only one that has the ability very easily to do cross cloud for customers, and that's becoming a much bigger issue with especially regulated industries around the world, knowing that they can quickly fail over from one cloud to another cloud. And I'm not talking region. I'm talking like from AWS to Azure or Google or whatever, very easily. Not saying it's cheap to do, but they have that and it's available to do. Because most people are realizing that for real resiliency, it can't just be from one Azure region to another Azure region because what happens if Azure goes down totally, you're exposed. And we're the only ones that can do that. And that's a real concern with regulated industries, especially banks.

Got it. Well, on that note, can we give a round of applause for Michael Scarpelli?

Thank you.