Bringing AI to the Data Cloud, with Snowflake's CEO Frank Slootman

Frank Slootman, CEO of Snowflake Computing, joins Sarah Guo and Elad Gil this week on No Priors. Before scaling Snowflake to its blockbuster IPO and beyond, Frank was also the CEO from early to scale for landmark enterprise companies ServiceNow and Data Domain. Frank grew up in the Netherlands and is also the author of three books: Amp It Up, Rise of the Data Cloud, and Tape Sucks. In this episode, our hosts talk with Frank about the opportunity for generative AI in the enterprise, why Snowflake isn't really a data warehousing company, their acquisitions of Neeva and Streamlit, apps within Snowflake, and how AI relates to traditional analytics and BI. He also talks about his personal journey, why it's always a good time to do performance management, and why most leaders struggle to raise the bar for performance. ** No Priors is taking a summer break! The podcast will be back with new episodes in three weeks. Join us on July 20th for a conversation with Devi Parikh, Research Director in Generative AI at Meta. ** No Priors is now on YouTube! Subscribe to the channel on YouTube and like this episode. Show Links: Forbes: How CEO-For-Hire Frank Slootman Turned Snowflake Into Software’s Biggest-Ever IPO Amp It Up: Leading for Hypergrowth by Raising Expectations, Increasing Urgency, and Elevating Intensity Rise of the Data Cloud (Audible Audio Edition): Frank Slootman, Steve Hamm, Zach Hoffman, Snowflake: Books TAPE SUCKS: Inside Data Domain, A Silicon Valley Growth Story eBook : Slootman, Frank: Kindle Store Frank Slootman’s LinkedIn Sign up for new podcasts every week. Email feedback to [email protected] Follow us on Twitter: @NoPriorsPod | @Saranormous | @EladGil | @SnowflakeDB Show Notes: [00:06] - Frank’s Insights on Career Success as a three-time CEO [12:42] - The message of his book Amp It Up [25:01] - Future of Natural Language and Data [36:29] - Data Management and Industry Transformation Future [45:13] - Managing Resources in Changing Economic Environment [50:09] - Amping Up Energy and Intensity Amid Economic Headwinds

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:05Sarah Guo?

  Our guest today needs no introduction. Frank Slootman is the legendary three-time CEO of Data Domain, ServiceNow, and Snowflake, and one of the most looked-up-to leaders in technology for his relentless execution. We're excited to talk to him about what's on the horizon for Snowflake and how he looks at the AI opportunity. Frank, good to see you. Thanks for being here.

• 0:23Speaker 01

  Absolutely. Good to see you, sir.

• 0:25Sarah Guo?

  Let's start with just a little bit of personal background. You have had an amazing journey. You grew up in Holland, you're the first person in your family to go to college. What were you like as a kid and in college? And how did you end up in product management and computing in the US?

• 0:39Speaker 01

  Yeah, that's kind of a, you know, big, wide-ranging question. I sometimes have to, you know, go back and figure out what was the method to the madness because, you know, sometimes your life looks like a random walk, you know? In other words, it's just a series of events that kind of, you know, go from one to the other. But, you know, I was always a relatively focused, disciplined kid, if I were to describe myself in almost any realm, whether it was school or sports or any of those things. It's just the nature of the beast, you know, I would say. And, you know, definitely, you know, a bit of a chip on my shoulder, which I generally like in people, by the way. You need to have a reason to get up in the morning and have something to prove to the

• 1:24Speaker 01

  world or whoever. Those are all useful things. You know, obviously I ended up in the US because I think the US is obviously a much better, maybe not obvious, but it's obvious to me that it's a much better canvas for people like me. And obviously, we see that all around us, right? People that come from all over the world here because they have, you know, far greater opportunity and they would have where they came from. And it certainly is true for me. I mean, there's no doubt that I would have done in where I came from what I've done here. So I'm very grateful, you know, having had that opportunity. I always tell younger people, you know, it's very important where you decide to be. Don't just go where your friends are.

• 2:04Sarah Guo?

  To the point of choosing the right place, be it geography, yes, and thank you, America. My parents are also immigrants. You talk about being on the right elevator and some of the companies you worked at weren't the hottest companies at the time when you joined. Tell us about those choices.

• 2:18Speaker 01

  I just use the analogy of the elevator because there's this aspect of opportunity and circumstance that you can't change. It is what it is and you're going to be subject to it for better or for worse. And therefore you need to choose carefully. Some people think that I can will my way to anything. That's not true. So the choices you make, like we just said, where are you going to be? What industry are you going to be? What company are you going to be? What people are you going to be with are all very formative. And so you have to make very careful choices because if you combine good choices with a great execution, you get the perfect cocktail for opportunities, for future opportunities

• 3:02Speaker 01

  and for having a successful sequence of experiences. So it matters a whole lot.

• 3:08Sarah Guo?

  A lot, and I talk to a lot of people joining entrepreneurial ventures and they're always trying to figure out where to go. That is often where their friends go and sometimes it's where investor friends will direct them. What advice would you have for people choosing that company in terms of the things you can't change?

• 3:24Speaker 01

  It's a great question. I get asked a couple of times a year to speak to graduating classes at really prominent business schools and all that sort of thing. And they always ask me, is there one message that you have for the graduating class? I'm like, well, don't go working for some consulting firm out of school. Try to get a real job in the real economy, building real products, selling real products. You really need to feel what it's like to be in the dry stream of the economy as opposed to I'm just eating out of somebody else's trough. And I kind of sit on the vessel and glide along and I'm feeling good about myself, but you haven't really touched the real economy yet. And I really wish that for people early on in their careers to feel the heat of competition

• 4:08Speaker 01

  and also the cold winds of threat of markets that are disappearing because that's the real world. And a lot of people choose jobs that are very removed from the real world. I don't think that's helpful for people's development in their careers.

• 4:23Speaker 02

  How do you think about company versus industry versus role? Often when I talk to people as well, I kind of advocate for the... Choose the right industry and then choose the best company in the industry and the role is secondary. Do you think that holds true or how would you suggest that people actually find their way?

• 4:38Speaker 01

  Yeah, I totally agree with that. I think a role is not that important. You'll have many roles, okay? And roles come and go. And my first job, I took a role I really didn't want. But, you know, being an immigrant in this country, beggars couldn't be choosers. And I had to, I figured, look, I'll get in there and I'll make my way from there. You know, I was in a corporate planning group of like six people attached to the CEO of a large computer company. I was about as far removed from the real world as I could be. And I didn't want that. But that's all I could, you know, get into. These were the heydays of affirmative action. We didn't have a lot of picks. So, and in hindsight, I was right because, you know, once I got in there, you know, you spent

• 5:22Speaker 01

  two years doing typical MBA stuff, you know, M and A and all the presentations for boards and all this kind of stuff. But then after that, they pretty much gave me, you know, whatever I wanted to do was fine with them. And from there, you know, I made my way.

• 5:38Speaker 02

  You've had three just amazing CEO jobs, right? So, I believe you took Data Domain from less than three million in revenue through an IPO and a $2 billion acquisition by UMC. At ServiceNow, you took it from 75 million in revenue through an IPO and I think 1.4 or 1.5 billion dollars of revenue. And then Snowflake, of course, has just been an amazing run. And it's one of the really seminal companies in the data world. How'd you go from step one to step two with all these things? And in particular, you know, when you joined Data Domain, you had an academic co-founder, you didn't have a product that was commercially scalable yet. ServiceNow, you really turbocharged. Snowflake was growing, but, you know, was spending a lot of cash. So, A, what are

• 6:20Speaker 02

  the commonalities between those different experiences? And more generally, what kind of drives you? Like, what do you have to prove? You already had accomplished so much by the time you got to Snowflake. How do you keep going?

• 6:29Speaker 01

  So, let me first sort of correct the record on Data Domain. They had no revenue, no customers, nothing. There were 15 people there. And when we first started to, you know, assert the product, it was... It had one terabyte of usable space. Just imagine that. Okay, now it was a while ago. You know, under around thirty megabytes, you know, a second. So it was useless for for ninety nine point nine percent of of of applications. Um, so we're like, what are we gonna do now?

• 6:57Speaker 02

  Why'd you take the job?

• 6:59Speaker 01

  Well, I didn't know that. Um, you know, I I'll tell you why I took the job. First of all, you know, I I I got rejected numerous times for for C E O opportunities. And the ones that that they were interested in were like second and third string. And I know people really cautioned me at that time, don't hold out, you know, do not go for a second, third string, you know, deal, you need to have really good investors. You know, we were a startup, one out of hundreds at the time. You know, I'd be walking the halls of of N E A in Greylock and people looked at me, who are you? What company is that? Oh, okay. We were a no name and we were lectured on, you know, on other companies that in hindsight ended up being no name. So I mean, it's it's almost it's almost legendary.

• 7:43Speaker 01

  How J A Domain just manifested itself. And by the way, I I live for that kind of drama. You know, it was great. Um, but we we didn't have product market fit. We just didn't. And, um, you know, I I found a little bit of fit. I remember, you know, meeting with a C I O company that that that has been acquired since by by E M C. Uh, and I was, they were testing the products and the guy said to me, he said, you know, he said, that little product of yours was a real hero here on Friday. And I'm like, tell me more. Do tell.

• 8:16Speaker 01

  But he he he explained that, you know, they had their email database, you know, backed up on um on our device. And they had a massive corruption email databases as has happened back then, that's not common anymore. And it was four o'clock on the Friday afternoon and they're like, oh, my God, we're gonna be recovering from tape here all weekend long. We'll be sleeping on cots, blah, blah, blah. And then they remembered, oh, we have a we have a backup on disk. And by seven o'clock that evening, they were going home and and obviously you don't need to be a rocket scientist to figure out that there's a use case you can sell a few times more, right? So we stayed alive and we did do that three million dollars that first year. Um, but I still remember doing the very first contract with like a five thousand dollar uh service deal

• 8:59Speaker 01

  with Stanford University and they bitch and complain the whole way. I'm like, well, this is gonna be great business.

• 9:05Speaker 02

  You know, one of my, one of my favorite books, which I think is really a hidden gem in terms of go to market and sales and startups is Tape Socks. And I think you get into very great tactical advice that's lacking from a lot of other books. Like you get into different channel strategies and whether you should do them and partnerships and other things that I just don't think are addressed very well in a lot of business books. And you've now written three books and we can come back to the question in terms of, you know, what continues to drive you and all the rest. What drives you to actually share knowledge that way and write a book, it looks like with almost every formative experience that you've had?

• 9:35Speaker 01

  You know, I get an awful lot of inbound questions, you know, can we have coffee? Can you speak here? Can you do this? Can you do that? And I'm like, I really can't because it's just, it'll become a full-time job. So I'm like, look, I'll write a, and by the way, there's a day at a main book, the tape sucks, you know, self-published, it's home brew. And it's a very dense book, even though it doesn't have that many pages, you know, I don't spend a lot of time, you know, waxing poetic or having a lot of platitudes. That's sort of the difference between my writing and then everybody else's. There's no filler. Everything, it's super dense. Everything that I write is, I find meaningful and worthwhile

• 10:16Speaker 01

  sharing. But it's really, look, these books all have had different reasons. Okay, the last book that I wrote, I didn't want to write it. Okay, Denise Pearson, our CMO, really, you know, pushed me to write it. And she also made it easy for me to write it because I had a lot of help along the way. I wrote every word of it. Okay, no, in other words, it's not a, but I did have a ghostwriter who just went through it. It's not, look, you need examples here, or nobody will understand this outside of your business, you know, all that kind of commentary and, and explained this better. And so he helped me just make the book more consumable rather than this very narrow audience that we normally deal with. But the net of the, the reason why I wrote Amped Up was, you know, people said, hey, just like you just said, you've had three very successful experiences, different

• 11:00Speaker 01

  times, different markets, different technologies, different competitors, blah, blah, blah. You know, what's the secret sauce? I mean, Americans always think there's a formula that can be extracted. And if I just have my hands on that, I can just do it too, right? It's immediate gratification type of thing. And the book's really the answer to the question of What do you guys do? What do you think explains the success in these companies? It's my answer. It's not that I'm trying to sell that to people at all. I don't care whether you agree with me or not. I'm just telling you what my best guess, my best take is on the answer to that question, right? People sometimes go like, "Well, I don't agree with this." I don't care. I mean, I did kill customer success at every company I've been in. I think it's the biggest bullshit thing that goes on in Silicon Valley.

• 11:43Speaker 01

  It doesn't mean that I need you to agree with me. I'm just telling you what it is, right?

• 11:48Sarah Guo?

  So one of the core messages in Amp It Up is about the importance of urgency. And you talk a lot about how to create it. I guess maybe a more difficult question is, why do you think a bunch of CEOs and leaders don't push for more urgency or higher standards?

• 12:04Speaker 01

  Well, I know you guys have been to a California DMV before. You want to see a lack of urgency? You know, this is what naturally happens to human beings. It's innate. We slow down to a glacial pace unless there are people who are going to drive tempo and pace and intensity and urgency. That's what leaders need to do because people naturally slow down. They're like, "Well, I need to be here anyways." And their mind is wandering off on their next vacation or what they're going to do on the weekend. And it's like, you need to set high focus, high intensity, high preoccupation with what we're doing. I mean, people sometimes ask me, "What's the message of your book?" And I'm like, "Read the title, okay?"

• 12:49Speaker 01

  That is the message. Look, there is an X-factor. There's an enormous amount of room in the margin that is right under your notes, okay? And you have the opportunity to take it up in the next meeting, in the next podcast, in the next email, in the next Slack message. You can take it up. You can push the urgency. You can push the standards, right? You can push the alignment, right? You have all these opportunities. Are you taking them? It's an easy message, but it's really hard to have the mental energy to bring that to every single instance of the day, right? And that's the message of the book. There's a lot of room there. There's a ton of room there. And people don't realize it because, yeah, I've seen companies where you have young CEOs.

• 13:33Speaker 01

  They just think, "I hire a bunch of people and I sit back and wait for greatness."

• 13:39Speaker 01

  They have no idea that they have to relentlessly drive, you know, every second of the day, every interaction and seek the confrontation because, you know, CEO jobs are insanely confrontational, which is not human nature. We don't like it. We are naturally confrontational. We avoid it. I mean, I had a founder CEO once, every time somebody had to get fired, you know, he had a CFO do it and he stayed home that day because it's just so hard, right? And it's like, I don't have the disposition for it. We understand that. But there are people in the enterprise that have to do that stuff. Okay?

• 14:15Sarah Guo?

  That fully resonates. But another piece that strikes me is people are afraid, right? That they don't have the right people, that they'll lose in the talent marketplace. If they push hard enough, their people will leave, right? What would you, how would you respond to that?

• 14:30Speaker 01

  Well, if they leave, they should leave. Okay. I mean, this is the great thing. You know, culture sorts and sifts. You attract the right ones and you start losing the wrong ones. So it's actually quite perfect. If people are leaving, they're just not your DNA. They're not your blood type. And by the way, you need to create your blood type, you know, around you. Otherwise, you're correct. You have nothing but conflict. I mean, I remember having people after two weeks just said, you know what? I can't take the patient intensity of this place anymore. It wasn't me personally. It was like, everybody was like that. You know, they were all, you know, calling people out and driving these expectations they weren't used to. And they wanted to go home at four, four P M. And pick up the kids from school. I'm like, well, you need to go back to H P and sleep in your cubicle. This is not the place for you.

• 15:14Speaker 01

  So you need to, but like, culture can be incredibly helpful, you know, to a company. But culture is not a general thing. There's no such thing as general goodness. I mean, the culture needs to really enable your mission, right? And whatever, whatever enables your mission effectively is a good culture. There's no universal culture. That's good. You know, it depends on, you know, the type of leadership you have and type of business you have. And, you know, where you are in your journey and all this kind of stuff. But, you know, culture is a very powerful thing because if you don't, if you don't fill the void, somebody else is going to, you know.

• 15:46Sarah Guo?

  I want to switch over to talking about Snowflake and then what's going on in AI. Can you just give our listeners a sort of Snowflake 101? You know, what is the sort of scale and core innovation and use case of Snowflake today? And we can talk about how the, how the company has been evolving from warehousing to cloud, the data cloud and application platform in AI after that.

• 16:11Speaker 01

  Yeah, our founders probably would argue immediately with you that they were never a warehousing play. So they sort of want to.

• 16:18Sarah Guo?

  Forgive me.

• 16:19Speaker 01

  Yeah, you're forgiven. But there's a reason for it because, you know, they were dealing with semi-structured data right from the get-go. And sort of the workload types were more than just sort of batch analytical, you know, type of stuff, which is mostly associated with data warehouse. And that's also purely structured data. So there was, there was always a broader scope and focus. But our founders were two French guys, long time, you know, Oracle CTOs, technologists, architects. They were really responsible for breaking Oracle from the departmental level. You probably can't remember that far back, but Oracle at one point in time was a departmental platform to the enterprise platform that it became. So things like Parallel SQL, you know, were all things that came, you know, from them.

• 17:04Speaker 01

  So they, they left and, you know, they wanted to reimagine database management, you know, for lack of a better word, for cloud computing. And in other words, they didn't want to carry technology forward or as little as they could. They wanted to reimagine. So, you know, building a database or a data platform, whatever you want to call it, for cloud computing was very different than just sort of taking a Postgres SQL kernel forward and kind of hacking it up. For the cloud, I'm being very unflattering here, but there's plenty of people that have done that. So they did some really breakthrough things, you know, most notably that most people know is the separation of storage and compute.

• 17:47Speaker 01

  I mean, back in the day, people maybe not remember this, but you know, I mean, you bought storage and compute in combination. You couldn't buy one without the other, whereas in the world of cloud, you can come and there compute and, and, and storage independent of each other. And of course, it became a consumption model, not right away, by the way, that was sort of an, an evolution and you know, obviously today is by the machine second or compute second. But once upon a time, it was, you know, by the node and it was by the, by the machine hour and all that. Now it's so incredibly fine grained and granular that, that is completely different. But the other thing that they did is they took the control plane out of the cluster itself. So the clusters are now all stateless, you know, in other words, they're clueless, which

• 18:29Speaker 01

  is great because you can run tons of them, you know, concurrently, right? So there's not, there's not one master, the master lives outside of the clusters. So running jobs concurrently is another huge thing because in the world of data warehousing, just to use that word again, Sarah. I mean, the, the reality was, you know, you had to beg for a 2:30 AM time slot three months from now because you know, the cluster was consumed very quickly, very easily. Now it's like, there's no limit. So this is, this is what I often tell investors, like, I'm not creating the man, I'm just enabling it, okay? It's so pent up, it's, it's insane, right? And the architecture does that, right? And then I could also provision workloads either for economy, in other words, you run the cheapest possible, or I could run for performance blistering fast and you could make these optimizations and choices.

• 19:14Speaker 01

  So this is, this is beautiful stuff, right? Because we, we just, we just opened up the demand in that, you know, legacy marketplace. And then of course, we started migrating, you know, Teradata databases, I mean massive Teradata plants. And by the way, I mean, yes, we're still in the early innings of that because it's not easy to, to move those platforms at all. But you know, a ton of Hadoop, of course, which is sort of the, you know, what we used to call big data and now all data is big, so that, that descriptor doesn't make, you know, too much sense anymore. You know, and all the Cloudera and on and on and on, tons of Oracle's SQL Server, I mean, so that's, that's what we've been doing, but. You know, when I started, you know, the, the, the tagline, if you will, the positioning

• 19:58Speaker 01

  or core message was this is the day, this is the data warehouse built for the cloud. That was Snowflake's message. And I'm like, okay, well, I'm going to stick with that because, you know, you, you, you taint yourself with a brush pretty soon, you can't get it off you, which is pretty much what happened to us. I mean, you just started on, like, so here we go again. I have an allergic reaction every time I hear data warehousing because to me, it's just a type of workload now. It's no longer a market, it's no longer an industry where, and, and, you know, cloud data management platforms, you know, are, and certainly we are, you know, we're, we're seeking to become full spectrum workload capable, meaning from the most batch analytical to the most streaming, online, transactional, you know, massive, you know, uh, scale and,

• 20:42Speaker 01

  and, and, and extremely low latency from, from what you're used to and in all T P type of environments. And the reason is, we don't want the, the, the whole premise behind the data cloud is that the work comes to the data. The data does not go to the work. Now, why does that matter? You know, because historically, the data has always been pumped around to go to the work while you get massive siloing of the data. You, you, you don't even, you even have to work at it. You're going to get siloing, you know, whether you try or not, because you have a new app, you got a new silo, you know, because it comes with its own database, right? And the siloing prevents you from really fully exploiting the potential that, that lies within your data because there's now walls that, that exist between them.

• 21:22Speaker 01

  Um, so the notion of a data cloud is, is kind of a really new data strategy element in the mix and, and, and we advocate really hard. I mean, I've, I've said it to C E O of the large banks, I says, don't go re-siloing your world in the cloud, you end up with the same sort of problems, you know, you have right now and your data science, M L, A I, et cetera, teams are going to be, you know, very frustrated, uh, you know, trying to overlay and blend that data and fine tune and train and do all these fancy things we do now, um, you know, with data. So, you know, we're trying to create an unfettered, uh, data universe, data orbit, that's much bigger than your enterprise, by the way, because this is really an ecosystem, right?

• 22:05Speaker 01

  You, you have data providers, you know, in, in the world of, you know, financial services, you know, get FactSet and. Bloomberg and S and P and all these things. So in Hetron, they have hundreds and hundreds, you know, data flows, you know, coming in. So you really need to think of data management as a much broader orbit than than just your your your enterprise. And so in the world of of artificial intelligence or general intelligence around data, the ability to mobilize data, you really need to have a data call strategy. That's also why we are multi-cloud capable, because we don't think, you know, you can have a data cloud in a single public, on a single public cloud platform. By definition, you can't, right? So that's really the strategy.

• 22:48Speaker 01

  And obviously, things have taken off a lot, but there have been multiple iterations in the in the journey, you know, of Snowflake. I mean, started out off, started off just moving legacy, you know, systems to the cloud and taking advantage of the elasticity and the economics and the provisioning and all these things. But now it's much more broadly workload capable, and that's a journey that goes on and on. The other thing that has changed is no longer a database world, you know. Historically, a database was just, you know, a platform that was self-contained and it had standard interfaces like ODBC and JDBC that you that the application used to access the data. Now it's like, well, wait a second, you know, we don't want to operate that way anymore because you're breaching the governance perimeter.

• 23:30Speaker 01

  So the application needs to execute inside the perimeter of the platform, not outside.

• 23:37Speaker 01

  So we have a programmability platform called Snowpark, okay? And then that's where, you know, all the applications live. We have native application framework, all these kinds of things. So now you're looking at a very different platform environment, very different layer stack than historically what we've had in the on-premise stack that we've grown up with, certainly I grew up with. So that's kind of as short as a story as I can tell you.

• 24:02Speaker 02

  That's that's really great background. And obviously, Sniplic has accomplished amazing things and really become central now to the the enterprise data world and ecosystem. How do you think about what's shifting in AI? Because I think we went from a world where we had almost like this older version of AI models, CNNs and RNNs and things like that, where people were doing old school natural language processing or other things. And then more recently, we've had this big breakthrough wave of generative AI. And it felt like the starting gun for that to some extent was really when ChatGPT came out about six months ago. And then GPT four came out maybe three months ago and then suddenly everybody started building applications against this. How has that been showing up or has that been showing up yet in terms of the AI use cases that you see in the enterprise or your customer requests or has anything really shifted

• 24:46Speaker 02

  yet in terms of, you know, the broader enterprise ecosystem that you deal with, just given that often it takes six months for an enterprise to plan something if it's a very large business. And so I feel like the last few months have just, or last two quarters have just been a lot of big companies kind of planning against what to do.

• 25:00Speaker 01

  Yeah, you know, first of all, large language models are about language. Okay. No surprise. But, and it's a huge deal because, you know, I was taught the, you know, the basics of COBOL when I was in school and COBOL stood for Common Business Oriented Language. Well, there was nothing common or business oriented about it. It was extremely cryptic syntax and all that, but compared to assembler and machine code, it was amazingly, you know, the syntax was amazingly comprehensible. So this is all relative, you know, in the eighties, we had SQL, which was back then, you know, also positioned as something that mere mortals could use to query data. So this is all about what, how and what is your relationship with data, right?

• 25:45Speaker 01

  And over the years that has, you know, evolved, but it's been immensely frustrating, you know, for people to get, you know, access to data in the form that they want. And there's a lot of ad hoc and there's a lot of standardized reporting and dashboarding and all this kind of stuff, but it's been difficult. So, you know, going to natural language is like, it's like the last mile here. And that is, is an enormous thing. I mean, the effect on demand will be just enormous because every mortal, if you're semi literate, maybe you're not even literate, you can just talk, you know, you can, you can get value from data. Wow. You know, so it is an incredibly, you know, big deal. But, you know, the generative aspect in terms of content generation, that's very cool when you're trying to plan a trip to Yellowstone, but when you're in the enterprise, you're

• 26:29Speaker 01

  dealing with structured proprietary data. And, you know, they're not planning trips to Yellowstone. They're gonna, you know, they're gonna ask really hard questions, like in insurance, for example, they may say, you know, we had disproportionate, you know, bodily injury claims in Florida and the surrounding states didn't have it, you know, A, what explains that? B, are we gonna have it again next quarter? And C, what do we do about it? Do we stop underwriting? Do we change our pricing? Blah, blah, blah, blah. Believe me, you're not gonna get the answer to that question with the large language model. So you, you gotta sort of separate the issues of, you know, text to speak and all, all of that, you know, which I think are incredibly valuable from going to structured

• 27:12Speaker 01

  proprietary data, because that's a, that's a very different realm. So, you know, the way I'm trying to think about it right now is, yeah, we have language models, but we're gonna see all kinds of other models. We're gonna see business models, okay? Because the question I just asked, you need to understand business models. I mean, one of the big things that, just to stick with insurance for a second, one of the biggest things in insurance, in a specific type of insurance, like auto insurance, auto insurance is Skycall and Progressive and Liberty Mutual and all these people, you know, telemetry data is, is number one through 10 for them. Okay, telemetry data is the device you get in your car and it, it knows when you're speeding and all this kind of stuff. And by the way, that, that's how they now price risk and they're, they're capable of lowering their prices, yet increasing their

• 27:56Speaker 01

  profits because of their extremely sophisticated and refined use of that data. That data is extremely predictive, you know, in terms of, you know, what, what the claims are going to be. And it's the difference between winners and losers and people who make money and people who don't make money. So that's, that level of, of, and by the way, that's not even AI, that's, that's just, you know, machine learning, really data driven, and that's already in, in broad use in, in, in other insurance companies. That's, that is sort of, you know, where this is all going. And I need to be able to ask questions that analysts might take weeks and months, you know, to, or bring in McKinsey or Bain or whoever, you know, to go and study, you know, problems, right? The systems will be able to start giving you insight into those kinds of questions. That's really where we live, you know, proprietary,

• 28:41Speaker 01

  structured enterprise data, that's a totally different realm, you know, and, you know, and by the way, you couple that with language models. Having natural language to build is, yeah, that's pretty powerful. Sorry, you mentioned Marvel movies, you know, the way he directs systems. That's a nice model. But I imagine in medical, we have diagnostic models, you know, and we have all these different, you know, levels of intelligence that we can build, that as long as they have the data, I mean, they're going to be insanely lightning fast, providing insight. And we acquired this company called Niva, you know, very recently. I'm very excited about bringing the expertise

• 29:22Speaker 01

  into, into the company because, you know, they're search experts and I'm a search junkie. I mean, 25 years ago, I mean, I wish I'd had search, you know, earlier on in my life because it's such a huge thing, you know, I just can't help myself. I'm always. And search is so addicting because it lets you sort of explore everything that's known and ever been written or published or opinionated about and sort of process all that information. But the problem with search is it has no context, right? It just matches on strings. And, you know, if you search on Snowflake, you might get the company, you might get the weather, you might get the social phenomenon, because it doesn't know, it just knows the word. And it's incredibly, and so enrichment and context is really the name of the game

• 30:04Speaker 01

  in the world of data, right? We always like to say, you, one attribute can, can make a data attribute go from being mundane to being high octane because of the context that it creates, all of a sudden it becomes wildly insightful and impactful and predictive and all these kinds of things. So, you know, in order for, for, for search, you know, to, to get that context and become stateful, is, is, those are going to be enormous step forward. And, you know, chat and search, you know, it all becomes one, one natural language conversation after a while. So you combine that, you know, with having this, these new levels of intelligence specific to industries or just subject matters, you know, I think is, that's really where there's a world of opportunity waiting to unfold

• 30:48Speaker 01

  still. I'm certain that it will, you know.

• 30:51Sarah Guo?

  Yes, you know, Aniva is a dear former portfolio company. Do you imagine that the snowflake like interface for users changes a great deal over the next, you know, five, 10 years and in terms of like supporting more natural language or a broader user set?

• 31:10Speaker 01

  Yeah, both of those things. You know, I think that there still will be a future for for BI companies, business intelligence, sort of tabloids, lookers of world. And, you know, dashboarding is done for a number of reasons. Sometimes it's just, you know, basically providing data in a consumable format. But it's also done because it's a way to basically tell people, this is how I want you to look at the data. This is how I want you to understand. So there is sort of a guiding element to dashboarding. Not all analysis is ad hoc based. Now, a lot of it is. And, you know, for ad hoc, you know, nothing is going to be better than than natural language. At least I'm already using it.

• 31:54Speaker 01

  You know, we push Salesforce data into what we call Snow House. That's our internal snowflake data that we push everything into. And it's just incredibly easy to use already commonly available services and have, you know, conversational relationship with that data. You know, are my two top reps in this country or that market or this industry? Now it spits it out in a fraction of a second and but a beautiful graph attached to it and all of that. So it's very addicting because it just it's just like search, right? You just keep going and going and going and it becomes like a whole journey. So, yeah, I definitely democratize access. Anybody semi literate will be able to get, you know, way more value than they ever imagined from the data.

• 32:39Speaker 01

  And it will change, you know, how products get used. I mean, BI will not be the same. I think I see that as severely affected by. By this evolution.

• 32:48Sarah Guo?

  You made another acquisition of a company called Streamlit that I think we're also both familiar with. Can you talk about the rationale for that?

• 32:56Speaker 01

  Streamlit is a company that does visualization, animation, you know, for for Python applications, but specifically in the world of machine learning. The problem with machine learning is if you're not a programmer, it's pretty damn hard to consume, you know, what it is and how it works. But Streamlit is almost reflexively reached for by Python programmers to basically make a machine learning model consumable by a general business user, right? You can manipulate the variables and it just redraws everything. Visualization, animation, and that's the reason that we acquired Streamlit is, A, you know, that's certainly we have

• 33:37Speaker 01

  to have visualization and animation. And by the way, this also touches the world of BI because a lot of people use Streamlit, you know, for the same reason that they would use BI type of products, but this is just much more, you know, specific to all kinds of reporting and use cases and dashboarding. So what we wanted to do with Streamlit is to bring it inside Snowflake. We call it Streamlit in Snowflake. And the reason is you need to have that hardcore trusted sanctioned governance perimeter because otherwise people will not allow the business to use these kind of applications. Governance is a really big deal because the data needs to be sanctioned and trusted and the business should not be

• 34:20Speaker 01

  able to get in trouble with the data. And that's really what we try to do with Snowflake. We are a hardcore enterprise grade platform. It's really hard. I mean, you can bring Python to your data in two weeks' time, but the problem is, you know, people are downloading libraries every couple of weeks to their heart's content and people have no idea what kind of risks they are exposed to in terms of exfiltration and all that. We spent two years, you know, making Python non-ports and it was an enormous effort to do that.

• 34:50Speaker 01

  But you know, you go to large financial institutions, like we're not gonna let Python anywhere near our core data. It's just not even a conversation. And we're like, well, we're gonna do it in a way that, you know, the people that use Python, there are many, obviously, but they can do it in a way that they don't violate and create exposures to the enterprise. So that's really the role that we play. We talk about governance a lot. We talk about data quality a lot. And we get into this conversation, I don't know how many times a day, because in a world of AI, if you don't have highly organized, optimized, sanctioned, and trusted data, what do you want, you know, your models to do? Just kind of train on a data lake? I call it a landfill, you know? I mean, you have no idea what the hell is in there. You know, everybody dumps their stuff in there.

• 35:33Speaker 01

  You're gonna go train on that? It's just absurdity. So you're having highly organized, optimized, sanctioned data is really, it's a prerequisite for all, and people publish what they call data products. I'm sure you've heard that term before. A data product is essentially, you know, I've taken data, you know, out of a lake, and I've created it into a trusted, optimized, understood object that I can now give to the business and stand behind. That's really the role of the chief data officer, to make the data, you know, trusted, organized, and optimized. And then also that the business can't get in trouble, you know, whether either because data is no good, or because they're breaching all kinds of security and compliance, you know, aspects of using data. So that's, Stream is really important to us. The great thing about it, it's an open source project. So

• 36:18Speaker 01

  you know, so many people out there are reaching for when they want to publish something. And you know, we're like, okay, we're gonna bring that inside the enterprise perimeter and make it high trust.

• 36:28Sarah Guo?

  I go back to sort of the journey you described from not just a data warehouse, but only data warehouse as a first workload, to, you know, broadly, you know, more online analytics, other workloads, applications that sit inside Snowflake with, you know, unified data. What are the, what are the biggest challenges you guys face in making that vision come true? Is it convincing people to like move to, you know, customers to an entirely new architecture? Is it building the ecosystem? Is it just supporting the workloads? Because it's a, it's a very big rewrite of sort of enterprise architecture overall.

• 37:04Speaker 01

  Yeah, but it's, um, you know, we are rewriting anyways because of our, our, our migration to cloud. It's like the most disruptive thing ever. Um, and yeah, look, you know, when I was at ServiceNow, we basically had an on-premise architecture that we hosted in the cloud. And by the way, I'm, I'm, I'm not being, you know, unduly critical here. I mean, because it was very useful that we were, you know, a single tenant platform. It had all kinds of advantages and we were able to, uh, to manage it really well through massive standardization and, and, and, and things like that. I'll give you an example, um, you know, all, all, all the federal business that we had, um, at ServiceNow was all on-premise Oracle because, you know, you could not get in there with a cloud solution, you just couldn't. By the way, you still can't.

• 37:47Speaker 01

  I mean, the certifications in, on, on federal are so insanely demanding. Um, you know, federal is, is, is a very small part of our, our business because we spent, we're, we're, we're in the process for years and years and years, uh, to meet, to meet those standards. It's very, very hard, right? Um, but we are a pure cloud implementation. We can't run on premise. I get asked that by people, you know, like, I mean, I can't even conceive of it, you know, the way Snowflake works, right? Because it commandeers, you know, resources. It's not a, it's not a machine centric, uh, platform, you know? Um, so that it's, it's, um, it is a big change. There's, there's no doubt. And, uh, as I said earlier, you know, um, you know, we, we fight the siloing of data

• 38:32Speaker 01

  because we're that kind of a company. From a data strategy standpoint, we really tell people, you need a different data strategy for the cloud. Do not continue with what you've been doing because you've created a massively proliferated, bunkered siloed world and it will not serve you in the world of AI and machine learning and any, any level of data science. If you want to drive intelligence from data, you're going to be in a world of hurt if you keep siloing the data. And, and we tell that to application developers, to ISVs and say, look, don't have your own data container, okay? Because instinctively, application developer, oh, I want to have my own, my own data layer hanging underneath it. I'm like, you know what? Um, it's, you, you're gonna hate it because A, it adds no value to what you do because you're not a data management expert. It's just a, a utility function, uh, you know, for you.

• 39:18Speaker 01

  But then, you know, you're out of silo and the customer is now frustrated because they're going to start pushing their data into Snowflake and now we have pipelines and ETL process and all this kind of stuff and latency issues, governance issues, all this kind of stuff. So we just announced that this relationship with Blue Yonder, for example, has said, hey, we're going to fully re-platform, you know, on Snowflake. Because in the world of supply chain management, that's really important because we need to have visibility, you know, across all the entities that make up a supply chain. You only do that when you have a single data universe, when you have all these containers. It's impossible. That's why supply chain management has never been platformed. Because the data problem was unsolvable, literally, you know. So the other thing is the supply chain management.

• 39:59Speaker 01

  I mean, they run these extremely demanding analytical processes, right? And they run many, many, many times, you know, per minute, per hour. And they are very, very commanding of resources, right? So again, this is where, you know, our style of computing is very, very desirable, right? Because I can run the process, I can run them as fast as I need to, I can run as many as I want concurrently. So all these new architectural things are lending themselves really to use cases that have been there for generations. But, you know, supply chain management is an email spreadsheet business. I mean, it's still living in a world of Microsoft 30 years ago. That's insane, right? Because it's one of those use cases that should have been extremely optimized, but

• 40:43Speaker 01

  it isn't, right? So yeah, you're going to be doing re-platforming, re-architecting and re-imagining. That's what we did. Snowflake is a reimagination of data management for cloud computing. But, you know, as we get through our journey, it's looking more and more different than what it used to look.

• 41:00Speaker 02

  You mentioned some very large-scale evolutions in terms of just the data world there. What are some of the other future directions that you're most excited about or the big thrusts that you see coming in terms of data?

• 41:09Speaker 01

  Data is going to redefine whole industries, okay? And that's what I find the most interesting. And the reason I say that is... First of all, you know, nine out of ten conversations I have with customers are not technology and architecture and all that and migrations. It's about industry use cases. It's about call centers. It's about, you know, making medicine predictive, for example, because everybody knows, you know, healthcare is economically, you know, not viable at the scale that we need to deliver it in. So data can make us, you know, predictive and prescriptive, right? We can, if we have enough data, you know, we can tell who is at risk for what disease, when, and what they need to do. All data driven.

• 41:53Speaker 01

  This is not, well, we, this is not somebody's opinion. The data just, data doesn't have opinions, okay? Just, that's what it is. And it gives you the accuracy to go with it. The more depth and breadth of data that you have, the more that certain that stuff becomes. But this is, this is how healthcare will become much more effective, obviously, because you don't, you're no longer reacting to disease and symptoms, but you're getting ahead of it. And every, every healthcare institution, you know, that we talk to and are a customer of, this is where they want to go, this is where they need to go. They don't want to treat disease, they want to prevent it, and they want to anticipate it. So it will change, you know, healthcare as an industry, but, you know, I just mentioned, you know, auto insurance is a similar type of example.

• 42:37Speaker 01

  In the world of pharma, you know, it takes an average 12 years to, you know, to, to bring a drug to, to market. Well, then you've got five years left before your patent runs out. But what if I could compress that by one, two, or three years? Now you've changed the economics of the entire industry, right? So, you know, data is, is far more important to how the, the economics and how the industry functions than, than people still realize, you know.

• 43:02Speaker 02

  How, how do your investments in R and D reflect this? Or what are the big areas of thrust that you have right now from an R and D perspective?

• 43:08Speaker 01

  The, the hardest part, you know, for us is, you know, I have to massively enable We have to massively enable this platform to be incredibly broadly and capable, not just broadly, but also in-depth because if it doesn't do what people need it to do, or it doesn't do it well, they're going to say like, well, forget it, we'll just pump the data over here. And now we're back to, you know, fragmenting and siloing the data. So if we have the data, we have to enable the workloads. Okay, we have to. And that's really hard. That's really hard. I mean, you mentioned some of the workload types, but we do things like global search. Okay. Because in the world of cybersecurity, you know, that's incredibly important. Because a lot of cybersecurity companies that, you know, they're partners of ours, they are running on a data cloud.

• 43:52Speaker 01

  They don't, because they couldn't sell to their customers yet another database container. Customer didn't want it. They said, look, you know, bring the data here and then we can combine it with all these other data sources, you know, vulnerability and all. And then, you know, our analysts can search one day in a universe instead of 15 of them and try in their head to figure out what does it all mean and do something with it.

• 44:13Speaker 02

  Yeah, I'm definitely seeing a lot of people right now building in terms of Snowflake app so that they can just maintain the data locally within a Snowflake instance for a customer, but then provide enriched functionality on top of that or access to that data in ways that are really performant and combined with what the, you know, with what the company is trying to do more broadly. So I think that's been a really great innovation for the industry. I guess one last question is just around the macro shift. So obviously, we've gone from a zero interest rate environment where everybody was just buying software like crazy to a world where people are cutting SaaS budgets increasingly, they're rethinking spend. Does the macro environment change your point of view on consumption or credit-based pricing? Or, you know, how you think about the pricing and economic model in this new regime?

• 45:00Speaker 01

  Yeah, not really. You know, we have different stakeholders that have different opinions on this. Investors, of course, love it when you have customers over a barrel and you can keep a gun to their head and they're going to pay you no matter what. I don't particularly like that. You know, when I was at ServiceNow, you know, I always felt that it was not an equitable relationship that we had with our customers. Um, because oftentimes, you know, they they would sign up with us for many millions of dollars and it took them nine months to even get in production. They were paying for all their users all this time. I'm like, how how is that equitable? Um, so one of the things that I really liked about Snowflake and cloud computing and consumption models and the elasticity is that you pay for what you use. It's a utility model and and um, you know, i- is is that painful?

• 45:46Speaker 01

  Sometimes, yes. I mean, I I I talked to the CFO of a bank last week and he said, you know, this is my bank's growing three percent, Snowflake's growing twenty-two percent, you know, and it's that that can't go on forever. You know, the CFO gets in there and he goes, he starts calling bullshit on everybody and saying like, hey, people, um, you know, they basically say thi- this is the size of your bread box, live with it, you're not going to get a new contract. But it's and then people need to go back to the to the drawing board and go like, okay, i- it's a very fine-grained thing because you can go into your Snowflake workflow and say, okay, I'm going to downgrade the provision on this. I'm going to run this less frequently. I'm going to change the retention period on data. You can do all these things to to lower your your your consumption of storage and and compute. Does that hurt us sometimes? Yes, but it's it's a value to the customer because, you know,

• 46:31Speaker 01

  if if you're in a SaaS subscription model, they gotta wait for their next drill before they can start cutting off a limb here, whereas with us, you can do it in near real time. Investors don't like it, I understand, because they they love it on the way up, they just hate it on the way down.

• 46:44Speaker 02

  Yeah, absolutely. I guess related to that, a lot of the people who tune into no priors um are people who are running their own companies right now and they're at different stages. You know, we have everything from early-stage startup CEOs to executives at larger companies, um, researchers, engineers, et cetera. And one of the big questions on their mind right now is how to manage differently through this, you know, economic downturn or the shift in spend or the shift in the macro environment. You obviously are known as a CEO who is very good at making tough choices and, you know, prioritizing in both good times and bad times. How should people think through managing differently in this in this changing economic environment? What are the first things people should do?

• 47:19Speaker 01

  You know, I mean, I I see all these layoffs, you know, with Amazon and Meta and Google and all this kind of stuff and and um, we we don't do layoffs because we don't wait until there is a, you know, huge headwind. Um, we're always pruning the tree, so to speak, right? So we don't have to do it as some massive event that is super unsettling. Um, you know. Management of resources is something that should be happening on a, on a daily basis, not just performance, but also, you know, bringing supply and demand in sync with each other alignment. That should be happening constantly. But the, the culture is sort of evolved over the years where it's just, it's just unfathomable, if that's a word,

• 48:01Speaker 01

  where you just, they can't conceive of being so confrontational that we're gonna take somebody out of a job. So we just look the other way until we get a crisis and then we started ripping out, you know, tens of thousands of people. I just don't think that's, that's fair, as, as well as effective, right? I mean, so you're, this is the reason my world doesn't change all that much because I was already doing it. So, uh, these are just more, uh, sort of management practices and, and, and ways of thinking about, you know, how you run things, um, you know, rather than, oh gosh, we have economics that way now. We need to change everything we're doing. No, you don't. You just need to run things, you know, like you always, by the way, people are not used to living in downturns, you know, when you've been around longer, it's like, hey, they come around. Okay. It's part of

• 48:44Speaker 01

  life. And by the way, let's, let's, let's, you know, let's double down, triple down, put our game face on, put our boots on, you know, we're into fight now. This is actually gonna be a lot, I will say to you, this is gonna be a lot of fun. This is where it really happens, right? So in other words, you can get up for it, you know, we, you just amp things up. That's what you're doing. People are growing up like, oh, they, they only know, you know, that the trees grow into the heavens. Trees don't grow into the heavens. Okay. They don't. Um, so they need, everybody needs to grow up a little bit, you know, and, and just get a leash on reality and say, look, this is, this is part of life, you know? Do I have to start rethinking everything because economically things are now, you know, different? Yeah. To some degree, yes. I mean, we're, we're scrutinizing productivity much harder in sales organizations, you know, we might be a little bit quicker on the

• 49:28Speaker 01

  trigger. All that kind of stuff for startups, obviously, you know, raising money is a whole different ballgame and you guys are in that world. Uh, so they, they definitely need to think harder. I mean, when I was a data man, we, we, we would basically run the company from one fundraising milestone to another. That's how it was back then. That, that hasn't been the way it's been. I mean, Uh, you know, in recent years, people have not never had to raise money or run businesses that way to prepare themselves for a fundraising milestone. They've never done that before. Well, you should, you know, because that's how you stay alive. I, you know, I mean, fundraising is oxygen for a company, you know?

• 50:03Speaker 02

  Yeah, basically, I think gravity turned back on and everybody's like realizing it.

• 50:08Speaker 01

  Yeah.

• 50:09Sarah Guo?

  Frank, this is a great conversation. Is there anything that we missed that you think would be useful or interesting to talk about?

• 50:15Speaker 01

  Um, well, we, we already talked about amping things up and that's always the, you know, um, you know, when we, we have conversation like this and a lot of people are listening to it. I just, I just, I'm, I'm trying to get people to say, you know, my next meeting, my next message, my next encounter, my next situation, I'm going to amp it up, uh, because it's just a choice that you make. And, uh, you know, don't be afraid, you know, that people will react poorly to it. They won't. The good people will actually love it. And especially if you're in a leadership role and who isn't, you know, this is, this is really what people want. They want to inject energy and focus and intensity and quality so that the whole place starts to feel, you know, exciting, you know, and, and it's not like, oh, it's

• 50:59Speaker 01

  four o'clock or five o'clock or whatever. No, right? It's, it's much easier to live in an energized environment than one that's devoid of energy, you know?

• 51:09Sarah Guo?

  I love it. It's a very, um, it's a very courageous message. Uh, thanks for doing this, Frank.

• 51:12Speaker 01

  You bet.