Title

Presentation

Snowflake Inc. beats earnings expectations. Reported EPS is $0.25, expectations were $0.15.

Operator: Good afternoon. Thank you for attending today’s Q3 FY 2024 Snowflake Earnings Conference Call. My name is Hannah and I will be your moderator for today’s call. All lines will be muted during the presentation portion of the call with an opportunity for questions-and-answers at the end. [Operator Instructions] I would now like to pass the conference over to our host, Jimmy Sexton, Head of Investor Relations at Snowflake. You may go ahead.

Jimmy Sexton: Good afternoon and thanks for joining us on Snowflake’s Q3 fiscal 2024 earnings call. With me in Bozeman, Montana are Frank Slootman, our Chairman and Chief Executive Officer; v, our Chief Financial Officer, and Christian Kleinerman, our Senior Vice President of Product, who will join us for the Q&A session. During today’s call, we will review our financial results for the third quarter of fiscal 2024 and discuss our guidance for the fourth quarter and full year fiscal 2024. During today’s call, we will make forward-looking statements, including statements related to the expected performance of our business, future financial results, strategy, products and features, long-term growth, our stock repurchase program and overall future prospects.

These statements are subject to risks and uncertainties, which could cause them to differ materially from actual results. Information concerning those risks is available in our earnings press release distributed after market close today and in our SEC filings, including our most recently filed Form 10-Q for the fiscal quarter ended July 31, 2023, and the Form 10-Q for the quarter ended of 2023 that we will file with the SEC. We caution you to not place undue reliance on forward-looking statements and undertake no duty or obligation to update any forward-looking statements as a result of new information, future events or changes in our expectations. We’d also like to point out that on today’s call, we will report both GAAP and non-GAAP results.

We use these non-GAAP financial measures internally for financial and operational decision-making purposes and as a means to evaluate period-to-period comparisons. Non-GAAP financial measures are presented in addition to and not as a substitute for financial measures calculated in accordance with GAAP. To see the reconciliations of these non-GAAP financial measures, please refer to our earnings press release distributed earlier today and our investor presentation, which are posted at investors.snowflake.com. A replay of today’s call will also be posted on the website. With that, I would now like to turn the call over to Frank.

Frank Slootman: Thanks, Jimmy. Welcome and good afternoon. Q3 product revenue grew 34% year-over-year to reach $698 million. Non-GAAP adjusted free cash flow was $111 million, representing 7% year-over-year growth. Results reflect strong execution in a broadly stabilizing macro environment. while Snowflake’s global revenue mix is highly diverse in terms of industries and geographies, the company derives an ever larger revenue share from mainstream enterprises and institutions. This, as compared to a newer crowd of digital natives, have made up many of Snowflake’s early adopters. We added 35 $1 million plus customers during the quarter, 9 of our top 10 customers grew sequentially. Generative AI is at the forefront of customer conversations, which in turn drives renewed emphasis on data strategy in preparation of these new technologies.

We said it many times, there’s no AI strategy without a data strategy. The intelligence we’re all aiming for results in the data, hence the quality of that underpinning is critical. Meanwhile, Snowflake has announced and showcased the plethora of new technologies that let customers mobilize AI. We’ve introduced Snowflake Cortex to leverage AI and machine learning on Snowflake. Cortex is a managed service for inferencing large language models. This opens up direct access to models and specialized operations by translation, sentiment and vector functions. Business analysts and data engineers can now use AI functionality without the fractured highly technical challenges of the AI landscape. Last summer, we introduced Snowpark Container Services, which also serves as the second pillar of our AI enablement strategy.

Developers can access any language, any library and flexible hardware inside the governance boundary of Snowflake. More than 70 customers are already using container services in preview with many more waiting in line. Snowflake makes the common AI use case is easy and the advanced use case is possible. We are well positioned for AI based on the scale and scope of our data cloud programmability and governance framework. There are hurdles challenging enterprise adoption of AI and ML. The first is broad access to quality data. Snowflake addresses this challenge through its data sharing architecture. 28% of all our customer share data, up from 22% a year ago and 73% of our $1 million-plus customers are data sharing up from 67% a year ago. AI models can only be as smart as data they are trained on.

Security and governance present another challenge for enterprise adoption of AI and the now Snowflake Horizon offers a unified security and governance solution built for AI. Horizon strictly and consistently enforces user privileges on data across use cases, including large language model applications, traditional ML models and ad hoc queries. As part of Horizon, we introduced universal search, which enables customers to search the data cloud. Customers can now discover data and metadata that exists across their accounts and in the Snowflake marketplace. Snowflake continues to win new workloads outside of its traditional. Snowpark’s consumption grew 47% quarter-over-quarter. Consumption in October was up over 500% since last year. Over 30% of customers use Snowflake to process unstructured data in October.

Consumption of unstructured data was up 17 times year-over-year. Our newest streaming capability, Dynamic Tables entered public preview earlier this year. Approximately 1,500 customers are using the feature and initial adoption is outpacing expectations. We have a number of major new capabilities becoming broadly available in Q4. Our native apps framework will go GA, UniStore for transaction processing, Snowpark Container Services and Apache Iceberg Tables will all enter public preview. These products unlock substantial new workload expansion opportunities. We are campaigning globally to expand our audience. This fall, our Data Cloud World Tour traveled to 26 cities worldwide. In-person attendance at these events reached 23,000 nearly double from last year.

Next up is our Build Developer Conference in early December, where we anticipate 35,000 registrations. Build is focused on building apps, data pipelines and AI/ML workflows. We hope to see you there. With that, I will turn the call over to Mike.

Mike Scarpelli: Thank you, Frank. In Q3, we saw strong consumption from a broad base of customers. Our performance was evenly split between large and small accounts largest customer stabilized as expected. Migrations drove growth in Q3. Our two fastest-growing customers who are both migrating workloads from a legacy vendor. One of these accounts is in their second year on the platform, the other in their eighth year on the platform. We added four customers with more than $5 million and two customers with more than $10 million in trailing 12-month product revenue in the quarter. Growth in September exceeded expectations. For three weeks, consumption grew faster than any other period in the past two years. Consumption continued to grow in the month of October.

Q3 represented a strong quarter for bookings execution. Remaining performance obligations grew 23% year-over-year to $3.7 billion. Of the $3.7 billion in RPO, we expect 57% to be recognized as revenue in the next 12 months. APJ and SMB drove growth in net new bookings. We are making significant progress in delivering margin expansion. Non-GAAP product gross margin of 78% was up approximately 300 basis points year-over-year. Improved terms from the cloud service providers have contributed to margin expansion. We also benefit from increasing consumption of higher-priced additions of Snowflake. In Q3, price per credit increased 4% year-over-year. Non-GAAP operating margin of 10% was ahead of expectations. Operating margin benefited from revenue outperformance and increased hiring scrutiny.

Non-GAAP adjusted free cash flow margin was 15%, benefiting from favorable timing of collections. We ended the quarter with $4.5 billion in cash, cash equivalents and short-term and long-term investments. Our strong cash position allows us to opportunistically repurchase shares. In Q3, we used $400 million to repurchase 2.6 million shares. Year-to-date, we have used $592 million to repurchase 4 million shares at an average price of $147.5. Now let’s turn to guidance. For the fourth quarter, we expect product revenue between $716 million and $721 million, representing year-over-year growth between 29% and 30%. We’re increasing our full-year guidance to approximately $2.65 billion, representing 37% year-over-year growth. Consumption trends have improved.

We are seeing stability in customer expansion patterns. Our guidance is based on observed patterns and assumes continued stability of consumption. For the fourth quarter, we expect operating margin of 4% and 360 million diluted weighted average shares outstanding. For the full-year, we are increasing our non-GAAP product gross margin guidance. We now expect non-GAAP gross margin of 77%. We still expect a product gross margin headwind in the fourth quarter associated with new products. We are increasing our fiscal 2024 non-GAAP operating margin guidance. We now expect non-GAAP operating margin of 7%. We are increasing our fiscal 2024 non-GAAP adjusted free cash flow margin. We now expect non-GAAP adjusted free cash flow margin of 27%. For the full-year, we expect diluted weighted average shares outstanding of 361 million.

We are on track to add more than 1,000 employees this year, inclusive of M&A. With that, operator, you can now open up the line for questions.

Operator: Absolutely. [Operator Instructions] Our first question is from the line of Mark Murphy with JPMorgan. Please proceed.

Mark Murphy: Thank you so much. I love the 11-minute earnings call, and congrats on a fantastic result. So Frank, we are hearing broadly that conversations are starting with generative AI and they’re stopping at data, because they find their data estates aren’t in good enough shape. Are you sensing more tangible uplift there around that concept that Snowflake might be on the front edge of AI projects and perhaps seeing that spill over into customer conversations or drive more pipeline for some of your other products like Snowpipes and Snowpark and data sharing?

Frank Slootman: Generally speaking, yes, one of the interesting things is that customers are now getting preoccupied with their data estates because they have to get them into shape where they can productively take advantage of the newer technologies, which we are now also showcasing and delivering where they can just turn it on and have well-governed frameworks to run them with all the things that they’re used to from Snowflake. So it’s definitely true that the frenzy and the high degree of interest in AI has a knock-on effect on the interest in data strategy, data platforms. And people are also not just looking at the quality of their data and the optimization of the organization curation of data but also what kind of data they need to be able to have access to. So people are taking a much broader view of their data estates as well in terms of what’s in it and what should be in it.

Mark Murphy: Thank you very much.

Operator: Thank you, Mr. Murphy. Our next question is from Keith Weiss with Morgan Stanley. Please proceed.

Unidentified Analyst: Great, thank you guys. [Indiscernible] on for Keith. Maybe just start off with a quick question on sort of you mentioned several times on the call, the stabilization that you are seeing in your — with the new growth. And we’ve obviously heard that from sort of your hyperscaler peers as well. When you’re looking at an account level, could you give us sort of an idea like how much of the stabilization is coming from cost optimization, scrutiny alleviating versus sort of existing customers who slow what migrations leaning in more versus customers leaning more into new products? Is there any way that you can kind of give us more color, what parts are playing out already and which are yet to come maybe into the next year?

Frank Slootman: Keith, one of the things, it is Frank. What I mentioned is our customer base has evolved in recent years to include much larger enterprises and institutions who are typically not prone to over consumption and unbridled expansion that they then later have to reset and rationalize. Because of that, the exposure to these drastic resets and optimization that we saw earlier in the year is getting less and less with each incremental quarter. Secondly, people have really driven themselves through these processes and rationalized themselves and are now in a good place to move forward. You can only optimize and rationalize so much. At some point, it’s diminishing returns. People get tired of us and they’re moving on to things that are now new and interesting, namely preparing for enabling AI and ML technologies.

Mike Scarpelli: I’ll add to that, too, that why we see that stabilization is nine out of our top 10 customers all grew quarterly sequentially. And the other point I’ll make is we are seeing a shift, as Frank mentioned. Our biggest customers are mature enterprises we’re seeing now and mature enterprises have always scrutinized cost. They always will. And so there’s nothing new there, and that will continue. And that’s just the way anyone should run a business.

Unidentified Analyst: Got it. That’s helpful. And then maybe one quick question on your sales and marketing headcount. Obviously, we noticed it’s basically flat quarter-over-quarter this quarter. But then again, it’s only sort of one number. Can you shed some light in terms of like what regions are you may be leaning more into versus getting more efficiencies or any sort of areas that you’re investing in still? Because obviously, your growth seems pretty healthy. So just some color on where you’re investing, would be helpful.

Mike Scarpelli: Yes. So first of all, in general, with sales and marketing heads, most of those typically are added right at the end of Q4 or even more so at the beginning of Q1, so people can get involved in our sales kickoff. And what I would say is we are continuing to invest very heavily in our sales and marketing function, in particular, in Europe. We talked about six months ago or so, we added a new leader there. We have been changing out some people and investing, and we’re continuing to, and we will continue to prune underperformers globally and invest more in the right people as we go forward. And APJ is another one that we continue to invest in.

Unidentified Analyst: Excellent. Thank you.

Operator: Thank you, Mr. Weiss. Our next question is from the line of Raimo Lenschow with Barclays. You may proceed.

Raimo Lenschow: Thank you. Congrats from me as well. Guys, you have like a crazy amount of new exciting products coming out. How do you think about the sales setup here now going forward? Because you’re going to be able to address lots of different areas from like classic data to kind of more AI. Does that mean your sales approach needs to change, Frank, here going forward? And it doesn’t sound like you are going to have a crazy sales force expansion here. Like how do you want to do that going forward and ensure that all these new products are actually finding the market? Thank you.

Frank Slootman: Yes that’s actually an excellent observation. We have historically had sort of one dominant selling motion that we sort of deployed everywhere and anywhere. It has served us well. But as you correctly observed, the market has really changed. When you go back to 2015, Snowflake really swam in swim lanes that were very narrowly defined and very well understood. Now we’re in the mega market, right? These are very, very broad-based platforms that are incredibly capable in many directions. And we’ve been working very hard, as you’ve seen in recent years, in delivering just an absolute ton of capabilities to enable these platforms in all these different directions. I mean, our drive towards applications and the whole programmability framework, the onslaught of AI and AM and our capabilities, all of that is coming to fruition.

Now we have a ton of irons in the fire, and they’re now all getting hot. So from a sales standpoint, we have much more specialization happening and that is going to literally going all over the world because it is impossible for a single person to be to be expert in all these technologies and all these disciplines. So we’re going to have people that have general purpose capabilities, sort of core skills, if you will, and then we will have teams of specialists that will augment these groups wherever they’re needed. So our basic selling motions and how they are supported will evolve rapidly in the coming year.

Raimo Lenschow: Okay, perfect. Makes sense. Thank you.

Operator: Thank you, Mr. Lenschow. Our next question is from the line of Kirk Materne with Evercore ISI. Please proceed.

Kirk Materne: Yes, thanks very much. Congrats on the quarter. I guess, Mike, could you just talk about the impact that some of these newer unstructured data workloads are having on consumption at all, meaning I assume it’s still a very, very small part of overall consumption when you look at it, but is that helping with the stabilization? Can they turn into sort of catalyst for acceleration as we get into ’24? Anything you to sort of dimensionalize the impact that has as we think out to next year? Thanks.

Mike Scarpelli: Well, they’re definitely helping with stabilization. I can’t quantify exactly what unstructured is doing, but it’s not just that. It’s also we’re starting to see the effects more of Snowpark that’s doing very well for us right now. Some of the new things we’re already seeing very early uptick in streaming and dynamic tables. We really, I would say, Streamlit is built into our forecast for next year. The other products because they’re new, we really don’t build much in for that because we need history before we can do that. And clearly, we think a lot of these new products that are just going into public preview now and GA next year will be a catalyst for growth for us in 2025.

Frank Slootman: One follow-on, Kirk. Document AI is a technology that we’ve been working on for quite a while, and it really was on the basis of an acquisition we did over a year ago. That is now going into preview. And it is incredibly popular out there and what Document AI does returns to an unstructured document into a semi-structure document. So it can become a full participant in analytical processing. There is a ton of interest in that. And so that really brings the entire estate of unstructured data, which is 80% of the world’s data into the analytical sphere. So we do expect that to become very important, and especially because AI and LLMs are so heavily focused on unstructured data, textual data, at least today, this will be something that will be a driver in our business.

Kirk Materne: Thanks. And Mike, can you just give us an update on where the federal sort of opportunity stands? I know you guys were waiting to hear back on FedRAMP. Any update on that? Thanks.

Mike Scarpelli: Federal is a huge opportunity, and I would say it’s upside for us because it’s such a small piece, but we should have our FedRAMP high authorization literally any day. I actually thought I might have had it today. So stay tuned. You’ll see an announcement on that very soon.

Kirk Materne: Thanks. Appreciate it.

Operator: Thank you, Mr. Materne. Our next question is from the line of Brad Reback with Stifel. You may proceed.

Brad Reback: Great, thanks very much. Just a quick technical question with Graviton 4 now being announced by Amazon, should we think of any potential headwinds there?

Mike Scarpelli: Well, as we told you before, every year, we count on a 5% headwind associated with software hardware improvements. Graviton 4 was just announced recently. We really have not tested that and not all hardware improvements benefit our software. We expect there will be some, but we just don’t know. It’s too early, stay tuned and we’ll update you when we have more information.

Brad Reback: Great, thanks very much.

Operator: Thank you, Mr. Reback. Our next question is from the line of Alex Zukin with Wolfe Research. Please proceed.

Ethan Bruck: Hey, guys. This is Ethan Bruck for Alex Zukin. Congrats on the nice quarter. My question is, if you’re assuming the consumption trends persist through 4Q, and that’s how you’re guiding, I guess, what would be the biggest puts and takes to accelerate growth next year? And in the same vein, when we think about NRR, just given the stabilizing consumption trends, I guess, directionally, when should we think about when would we expect NRR to stabilize trough and around like what level would you expect?

Mike Scarpelli: So I’ll start with — I’m not going to guide to NRR. I do see it stabilizing. It could dip a little bit more. I do expect over time, NRR, as we’ve said, we’ll converge with our revenue growth. at the size we’re at. And clearly, the biggest puts and takes for next year is going to be the — to continue to see the stabilization we have we’re seeing right now and what the impact of a lot of our new initiatives are going to be next year, and it’s just too early to tell right now and to guide to that. So stay tuned for our February call when we give guidance for next year.

Ethan Bruck: Great. Thank you very much, Scarpelli.

Operator: Thank you, Mr. Zukin. Our next question is from the line of Brent Thill with Jefferies. You may proceed.

Brent Thill: Thanks, Mike. I think you mentioned there is a higher utilization on the higher tiers that you’re offering. Can you just extrapolate what you’re seeing there?

Mike Scarpelli: Well, typically, large enterprises are the ones using our business critical in BPS, and those guys are becoming bigger and bigger customers. And as a result, we do see more of our revenue being derived from these very large companies who are using our higher SKU, which has higher margin for us, and that’s what we’re seeing there.

Brent Thill: Okay. And quickly for Frank, when you think about just the overall AI impact, do you feel that this is a bigger tailwind in the back half of ’24? Do you think you’ll see it coming in early ’24? How are your thought process in terms of adoption in this way, helping you? When does that impact hit from a monetization perspective?

Frank Slootman: There’s going to be a lot of nuances to how AI is going to unfold and translate itself into the effects into our business. It’s not just like turning on the switch and all of a sudden, you see incremental consumption happening. As we said during the prepared remarks, we’re already seeing that the interest in AI is also driving interest in the data strategy, which then has a knock-on effect on consumption. It’s also the expansion of the data universe that people are bringing in quality of the data initiatives. All of that is going to bring incremental workload utilization to Snowflake, even though you would normally not necessarily characterize that as AI, but these are things that are going to enable AI, and it might well be that a ton of the workload of AI is actually the proverbial iceberg where only the tip of it that’s ticking above the water is really AI, but everything that has to happen to support us is happening below the service then we’re going to be a huge beneficiary of that.

We think that Snowflake is super well prepared because our data estates are in a very, very advanced state because our customers have spent years, years and years. And in some cases, literally decades from prior legacy platforms to building these estates and curing the data and organizing and optimizing for the data to be completely trusted and sanctioned in their environment, and that is a huge value when you start tackling AI and ML models.

Brent Thill: Thank you.

Operator: Thank you, Mr. Thill. Our next question is from the line of Patrick Colville with Scotiabank. You may proceed.

Patrick Colville: All right, thank you so much for taking my question. In your prepared remarks, you talked of very strong consumption in September. Consumption continued to grow in October. We’re now 29 days through November. Can you just talk about trends thus far this month?

Mike Scarpelli: What I would say is trends are good, but you have to remember, it also has a big holiday in the U.S. and the week of Thanksgiving is typically a slow week. With that said, I’m happy with the consumption we’re seeing, and that’s reflected in our guidance.

Patrick Colville: Perfect, thank you so much.

Operator: Thank you, Mr. Colville Our next question is from the line of Tyler Radke with Citi. You may proceed.

Tyler Radke: Hi, good afternoon. Thanks for taking the question. I wanted to ask you about container services. So it sounds like you’re seeing some good early momentum there, more than 70 customers in private preview. I’m curious how many of those are actually deploying large language models directly on the Snowflake data? And for those customers that are doing that, what type of uplift or how you kind of characterize that consumption when they do that?

Christian Kleinerman: Yes. This is Christian. Yes, we have seen a variety of use cases, but a good number of those prior preview customers are leveraging GPU instances in silver and payer services for use of large language models. So for sure, it’s a meaningful part of the early adoption use cases and we’ve got lots of encouragement and excitement to continue the rollout of the preview.

Tyler Radke: Great. And just a question on go-to-market. So I think earlier this year, you talked about some execution issues and it seems like that’s been partially resolved to get good results in APAC. I guess, do you feel like you have all the right key sales leaders in place at this point? Or are there still some roles you’re looking for? And I guess, ultimately, do you feel like you’ve turned the corner on those execution issues as well?

Frank Slootman: Yes. It’s Frank. Look, in a global sales organization that we’re running and we’re running all the way from an on-demand selling motion that’s unattended, if you will, by salespeople to SMB, to an ISB to mid-range large and then the extremely large customers. The — it’s always a work in process. There’s always opportunity for improvement. That’s been true for as long as I’ve been here. That said, I feel we’re getting incrementally better, stronger, deeper in terms of our ability to sell. We’re becoming much more consistent and more and more productive every day. So a lot of stability and a lot of progress. But this is something that’s never over. Anybody who’s been in this business knows how this works. It’s a very dynamic mix of things.

But on the whole, we have to, as a company, a $3 billion or something of that order of magnitude run rate, we lean hard on our organization to show up and deliver those results every quarter. So it’s becoming a formidable enterprise in terms of our go-to-market capability.

Tyler Radke: Great, thank you.

Operator: Thank you Mr. Radke. Our next question is from the line of Brent Bracelin with Piper Sandler. You may proceed.

Brent Bracelin: Thank you and good afternoon. Frank, we’ll start with you here. I think last quarter, unstructured data was mentioned twice. You’ve called it out here more than a dozen times. You talked about October trends being up, I think, 17x increase year-over-year. How much of a game changer is this, particularly as we think about Document AI and Snowpark Containers coming out next year? Clearly, a focus here. It certainly hasn’t been an area you supported in the past, but it feels like there’s a sea change movement here. How much of a game changer is unstructured data support relative to the growth opportunity going forward?

Frank Slootman: Look, unstructured data, as I said earlier, is the majority of the world’s data. And until relatively recently, it’s been borderline unusable for analytical purposes, because it is unstructured and you can’t reference it in analytical workloads. It’s ironic that both through the on slide of large language models that is also extraordinarily of dealing with textual data as well as things just Document AI that Snowflake developed, and it’s bringing to the market that this data is going to become a full participant in these types of workloads. It’s super exciting. It’s going to really enriched and really unlock insights and outcomes and all of that we haven’t had before. So this is going to be a driver of our world as we know it in terms of this type of computing.

Brent Bracelin: That’s helpful. And Mike, if I look at average consumption revenue per customer, those metrics improved on a growth perspective, slightly year-over-year for the first time in over a year. What are the driving factors there? How much of this is just the optimization headwinds now starting to ease versus net new workloads coming on the platform?

Mike Scarpelli: Well, what I would say, and I’ll repeat what I’ve said at numerous times to investors, optimizations are part of our life. They’ve been happening at Snowflake from day 1, they will continue to happen. Nothing is new with optimization. I don’t see any big ones happening now, but that’s not to say they won’t happen in the future, because history has shown they happen all the time. Most of the growth that we’re seeing within our customers is we talked about two of our biggest growth customers was on-prem legacy migrations into Snowflake. So there’s initial migrations, but we’re also seeing new workload expansion within existing customers as well, too. So it’s — there’s no one thing that’s driving it. It’s just general consumption we’re seeing. And I will say Snowpark is starting to kick in for us. Still not 10% of our revenue. That’s a long ways to get there, but it’s still meaningful for us.

Brent Bracelin: Great to hear, thank you.

Operator: Thank you, Mr. Bracelin. Our next question is from the line of Derrick Wood with TD Cowen. You may proceed.

Derrick Wood: Great. Thanks. I guess I wanted to dovetail off of that on-premise legacy migrations. And I guess for Frank, I mean, when we saw the macro hit, I think it did cause a slowdown in customers looking to re-platform from on-prem to cloud. Just curious, I mean, I know you guys are kind of highlighting the consumption trends improving here. But wondering what you’re seeing in the pipeline for new Global 2000 accounts? And if you’re seeing legacy replatforming initiatives start to kick back up now that the macro environment has gotten a little bit more stable?

Frank Slootman: Well, I mean, there’s no doubt that the legacy replatforming is sort of the hardcore of our business. And almost surprisingly there is just an enormous amount of workloads still sitting on-premise that is still waiting to get migrated to the cloud. So I expect this to continue for a very long period of time. But as what Mike’s said is very important. Once you get those data states into the cloud, our new architectures and our new technologies are now enabling opportunities that people haven’t had before. And so that drives workload expansion. So it’s not just like, okay, we’re going to be doing in the cloud where we used to be doing on-premise, and that’s the foundation of the business. It is definitely foundational but the opportunity is really what grows from that.

That’s been the Snowflake story from day one because of what’s now possible. We don’t have the capacity constraints and people can run unlimited numbers of workloads. And now with all the new technologies in terms of programmability, AI and sky is the limit. I mean, in our conversation with customers, I was telling us is that your problem is no longer the technology. Your problem is your imagination and your budget, right? Not that those are easy things, but technology is not holding us back anymore. It’s our ability to harness the technology. And then, of course, you have to pay for it as well.

Mike Scarpelli: Yes. I’ll just add to that, Derrick, too, on your question about Global 2000. Yes, we added two Global 2000 last quarter. And as I’ve said many times before, selling into a Global 2000 as a campaign, it’s a one to two, sometimes three year sales cycle. With that said, we have a number of Global 2000 in our pipeline for Q4. And what I also want to see, too, is not every Global 2000 starts with an on-prem migration. Many of them start with a first-gen cloud solution they had purchased to migrate to Snowflake. Yes, almost all of them have an on-prem estate, but that doesn’t always happen at first. It really varies by customer. But with that said, we still see a lot of on-prem migration to be done over the coming years, and it’s going to be for many, many years.

Derrick Wood: Got it. And Mike, a follow-on for you, really around RPO. And we’ve heard of more companies looking to take a pay-as-you-go approach. So could you speak to what extent customers are shifting to that approach versus annual or multiyear commits and how that may play into your RPO growth trends looking forward?

Mike Scarpelli: No. I don’t see that, that much. The only time you see pay-as-you-go are ones who had signed a 3-year contract and then they run out of capacity. And then they just pay as they go. Actually, one of our top 10 customers is like that. They have until April to continue before they have to do another contract, if they want to get the same pricing that they have. I actually expect Q4 is going to be a pretty significant bookings quarter with a number of renewals that are up or customers are going to do something. And we continue to push for three-year contracts with our customers. Payment terms, I do expect though that is one of the things I’d rather give up on payment terms and discount to price per credit. And that’s really, I’ve said it all along, I anticipate longer term, customers will want to do more monthly in arrears payment terms, and that is available to customers.

It all comes down to what price you want to pay per credit. And to date, most people want the lower price per credit and are willing to pay annually in advance.

Derrick Wood: Helpful. Thank you.

Operator: Thank you, Mr. Wood. Our next question is from the line of Joel Fishbein with Truist. You may proceed.

Joel Fishbein: Thanks for my question and congrats on the great quarter. I guess, Frank, this one is for you. Snowflake Summit, you announced an expanded partnership with Microsoft. And I’m just curious how that partnership is going, and you also announced some joint product integration. So I’d love to get an update how that’s going and how you feel about that going forward?

Frank Slootman: This is Frank. Yes, we actually saw quite a bit of energy coming from the Azure platform this quarter. The things that we worked on in the renewed relationship with Microsoft is really much better alignment in the field from a compensation standpoint that is just super, super important in our world. And we’re seeing the effects of that. And the Microsoft platform really upticked during the quarter. I don’t know the exact numbers at hand, but it measurably ticked up. So we’re encouraged by the behavior we’re seeing in the field, and we’re encouraged in the overall sentiment that’s developing in the field. So it’s definitely healthier than the relationship with Microsoft has been historically. So we’re pleased and optimistic about it.

Joel Fishbein: Any milestones we should be looking for there from the partnership?

Frank Slootman: Not really. We’ll let you know.

Joel Fishbein: Thanks.

Operator: Thank you, Mr. Fishbein. Our next question is from the line of Mike Cikos — with Mike Cikos. Please proceed.

Mike Cikos: Thank you for getting me on here. And good to hear some of the earlier comments around Snowpark, which is consistent with some of the growing momentum we’ve heard in some of our checks. I think the question is probably more for maybe Mike or Christian. But is there enough empirical data or a sizable enough customer base yet for Snowpark to start talking about how these Snowpark customers, I guess, what adoption of Snowpark does for consumption versus non-Snowpark customers? Or maybe how we should think about usage building over time as Snowpark becomes a larger part of those customers’ workflows?

Mike Scarpelli: I think it’s still too early to tell. I will say — we have one customer doing a significant migration, which will increase their consumption on Snowflake to roughly $1.5 million a year. And there are a number of those that we’ve identified in POCs, but we haven’t done the migrations. I see that 1 customer going into production right now, which, by the way, has saved them significantly on their legacy vendor. So clearly, the — with some customers, it’s quite meaningful, the consumption we’re starting to see, and we expect that trend will continue into next year.

Mike Cikos: Understood. Thanks for calling that out, Mike.

Operator: Thank you, Mr. Sikos. Our next question is from Patrick Walravens with JMP. You may proceed.

Owen Hobbs: Hi, this is Owen Hobbs on for Pat. Thanks for taking the question. Congrats on the strong quarter. So starting off, I guess, how much consumption comes from the different hyperscalers? Is there 1 that kind of is it split evenly between the three or does kind of have the majority of compute share there?

Mike Scarpelli: No. AWS by far is our biggest, followed by Azure and then GCP. GCP is up to 3% right now. Microsoft Azure is the fastest-growing one, but AWS is still 76% of our business with Microsoft being 21%. As I said, GCP is 3%. And I will tell you, 1 of the reasons why GCP is not as big as just so much more expensive for our customers to operate in GCP than it is in AWS and Azure. And as a result, our salespeople are really not inclined to do much in GCP.

Owen Hobbs: Thank you. And then going bigger picture, maybe this is a question for Christian as well. How is the role of the data scientist you sort of kind of changed as we move through the era of big data, machine learning and now into AI? And I guess where do you see that going in the future?

Christian Kleinerman: Yes. We see actual interest on both the traditional data science and ML platform. And we’ve had a number of announcements at Snowflake for those such use cases, but we obviously see lots of interest on generative AI and large language models where we have also expanded obviously capability to do both hosting via Snowpark Container Services, but more seamless inferencing via the new Snowflake Cortex. So we see strong demand from our broad customer base for both types of use cases.

Owen Hobbs: Good. Thanks so much.

Operator: Thank you, Mr. Walravens. Our next question is from the line of Gregg Moskowitz with Mizuho. You may proceed.

Gregg Moskowitz: All right. Thank you very much. Frank, since you speak with a lot of large company execs, I’m curious to hear any thoughts you have on IT budgets next year as well as whether there will be incremental budget dollars approved for AI in calendar ’24 or if the AI spend will primarily or maybe even entirely come from other areas of IT?

Frank Slootman: I don’t even hear the words AI and budget in the same sentence. In other words, they’re going to make resources available to enable it. But if anything is holding them back is really understanding how to do it. It takes time in tech to mature and evolve deployment architectures where everybody involved is fully confident and comfortable that these are the right ways to do it. So you see a lot of benchmark and compare and contrast experimentation, testing all these kinds of things. And they’re going to go through many, many iterations of that. We will as well. And I think that the field will become very proliferated with large foundational models and many, many, many subsector specialized models that are very, very, very deep, but also are very, very narrow in purpose.

So it is going to become an enormous field. I will tell you that when you talk to the C-suite in large enterprises, people are looking for a reset of economics, like, for example, in contact centers, pricing optimization, supply chain management, I mean, really very, very big impact opportunities. These are not sort of marginal incremental and that has the attention. Data has always had a promise, but it’s really on steroids now under the influence of the newer technologies that we’re all excited about.

Gregg Moskowitz: Very helpful. And then for Mike, really encouraging, of course, to hear that for three weeks, consumption grew faster than any other period over the past two years. Over which three weeks though, did you see that particularly strong consumption? Was that post Labor Day? Or did it span a different time period?

Mike Scarpelli: Post Labor Day.

Gregg Moskowitz: Right, Perfect. Thank you.

Operator: Thank you, Mr. Moskowitz. Our last question is from the line of Will Power with Baird. You may proceed.

Will Power: Okay, great. Thanks. Mike, you referenced that nine of your top 10 customers, it sounds like we’re nicely quarter-over-quarter. I guess I wonder, were there any common use cases or products or common threads kind of driving that? Any other color behind the strength there?

Mike Scarpelli: No, these are just all very large customers with massive data estates that continue to — one was doing a big migration, but the others are just continuing to move workloads to Snowflake.

Frank Slootman: That’s very industry-specific as well.

Will Power: Okay. And maybe just a quick question on product gross margins. Last couple of quarters, you’re kind of already at your longer-term target, I think. It sounds like there may be a couple of headwinds there in Q4, but is this kind of the general level to expect going forward? And maybe any color on kind of the upside surprise, I guess, if you will, given it seems like you’ve gotten there faster than you might have expected?

Mike Scarpelli: Yes. I don’t really see a lot of upside in our long-term guidance and for a number of reasons. The big headwind that we’re seeing this quarter to our margin is with all these new products, that are going into public preview. We have to start to amortize the costs associated with the software development costs that we’re required to capitalize under GAAP that are now going to start to be amortized on top of that. In particular, something like Unistar, a lot of times when we introduce new products, many times, we actually have negative contribution margins until it takes us usually up to half a year to nine months to actually fine-tune the software to get the — to take costs out of operating those new features. And so we just have so many new products that are coming out this quarter that is going to have a headwind on the margin.

Will Power: That’s helpful. Thank you.

Operator: Thank you, Mr. Power. That concludes the question-and-answer session as well as today’s call. Thank you for your participation. You may now disconnect your lines.