 Welcome back, everyone. The Cube's three days of wall-to-wall coverage of Snowflake Summit 22 is coming to an end, but Dave Vellante and I, Lisa Martin, are so pleased to have our final guest as none other than the co-founder and president of products at Snowflake, Benoit Dajaville. Benoit, thank you so much for joining us on the program, welcome. Thank you, thank you, thank you. So this is day four, as you guys started on Monday, this is Thursday. The amount of people that are still here speaks volumes. We've heard close to 10,000 people here. Could you ever have imagined, back in the day, 10 years ago, that it would come to something like this in such a short period of time? Absolutely not. And I always say, if I had imagined that, I might not have started Snowflake, right? This is somewhere scary, I mean, and, you know, it's a, yeah, it's huge. And you can feel, you know, the excitement of everyone. I mean, it's like mind-boggling. And the fact that so many people are still there, you know, after four days is great. Your keynote on Tuesday was fantastic. Your energy was off the charts. It was standing room only. There were overflow rooms. Like we just mentioned, a lot of people are still here. Talk about the evolution of Snowflake this week's announcements and what it means for the future of the data cloud. Yeah, yeah, it's a, so evolution, I mean, I will start with the evolution. It's true that's what we have announced, you know, this week is not where we started, necessarily. So we started really with, very quickly, with, you know, big data combined with data warehouse as one thing. We saw that the world was moving into fragmented, siloing data. And we fought with Thierry. We are going to combine big data and data warehouse in one system for the cloud with this LACCT and the service, simplicity. So simplicity, amazing LACCT, which is this multi workload architecture that I was explaining during the keynote and really, you know, extreme simplicity with a service. Then we realized that there is one other attribute in the cloud, which is unique, which doesn't exist on premise, which is collaboration. How you can connect, you know, different tenants of the platform together and, you know, Google showed that with Google Docs. I always say, you know, to me, it was amazing that you could, you know, share, you know, document and have direct access to document that you didn't produce and you can collaborate on this document. So we wanted to do the same thing for data and this is where we created the data cloud and the marketplace where you can have all these data sets available. And really the next evolution, I would say, is really about, you know, applications that, you know, are powered by that data, but, you know, are, you know, way simpler to use for, you know, all the tenants of the data cloud and this is the way you can share expertise also, including, you know, ML module. Everyone talks about ML and the democratization of ML. How are you going to democratize ML? It's not by making, you know, necessarily training super easy such that everyone can, you know, train their ML for themselves is by having, you know, very specialized application where data and ML is at their core which are shared, you know, through the marketplace and which are leveraged by many, you know, tenants of this marketplace that have no, you know, necessarily knowledge about building this ML module. So that's where, yeah. When you and Terry started the company, you know, I go back to the improbable rise of Kubernetes and there were other, you know, more sophisticated container management systems back then but they chose to focus on simplicity and you've told me before that was our main tenant. We're going to worry about all the complex database stuff. You knew how to do that, but you chose not to. So my question is, did you envision solving those complex problems over time yourselves or through an ecosystem? Was this by design or did you, as you started to get into it, say, let's not even try to go there, let's partner to go there? Yeah, I mean, it's both, it's a combination of both. You know, it's like the simplicity of the platform is really important because if our partners are struggling to put their solution and build solution on top of something, they will not build it. So it's very important that number one, our platform is really easy to use from day one and that is really has to be built inside the platform. You cannot build simplicity on top. You cannot have a complex solution and all of a sudden realize that, oh, this is complex. I need to build another layer on top of it to make it simpler, that will not work. So it had to be built from day one. But you're right, what is going to be snowflake? I always say, in 10 years from now, we just turn 10 years old or we are going to turn 10 years old in few months, actually few months, yes. So for the next 10 years, I really believe that most of snowflake will not be built by snowflake and that's the power of the partners and these applications. When you are going to say I'm using snowflake, actually probably you are not going to use directly code developed by snowflake. That code will leverage our platform but you will use a solution that has been built on top of snowflake and this is the way we are going to decouple the effort of snowflake and multiply it. It's an interesting balance, isn't it? When I think of what you did with Apache Iceberg, if I use Iceberg and I'm not going to get as much functionality, but I may want that openness but I'm going to get more functionality inside of the data cloud and I don't know but if you know the answer to what's going to happen. That's a super good question. So to explain what we did with Apache Iceberg and the fact that now it's a native format for us. So everything that you can do with our internal formats, you can do it with Apache Iceberg, including security, defining masking, data masking, all the governance that we have, fine grain security aspects, the replications, you can define, you can use search optimization on top of this. But there's a but, right? So, but if I do that inside of native snowflake tools, I'm going to get an even greater advantage, am I not? Yes, yes, so that's what I'm saying. So that's why we embraced Iceberg because I think we can bring all the benefit of snowflake to people who have decided to use Iceberg. I mean, open formats, Iceberg is a table format. So, and why it was important because people had massive investments in open source, in Adoop, and we had a lot of companies saying, we love snowflake, we want to be a snowflake customer, but we cannot really migrate all our data. I mean, it will be really costly and we have a lot of tools that need direct access. So this is why we created Iceberg because we really think that we can bring the benefit of snowflake to this data. You have customers optionality. Okay, you know I use this term super cloud, you don't use the term, but that's okay. And I get a lot of heat for it, but to me, what you're doing is quite a bit different than multi-cloud because you're creating that abstraction layer, you're bringing value above it. My question to you is, the most of the heat I get is, oh, that's just SaaS, are you just SaaS? No, I mean no, absolutely not. I mean, you're right, we are super cloud, I mean it's a much better word than saying we are multi-cloud, multi-cloud is often viewed as, oh, I have my system and now I can run this system in the different cloud providers. Snowflake is different. We have one single platform for the world which happens to have some regions are AWS regions, some regions are Azure, some regions are GCP, Google, and we mesh them together. We have this no-grade technology that connects all our regions together such that we have really one platform for the world and that's very important because when you talk about connections of data and expertise, applications, you want to have global reach, right? It doesn't exist, we are not siloed by region of the world, right? You have a lot of companies which are multinational that needs, have presence everywhere and you want to have this global reach. The world is not an independent set of regions and countries and that's the realization. So we had to create this global platform for our customers. And now you have people building clouds on top of your data cloud, but that to me is the next thing. In your keynote, you talked about seven pillars, all data, all workloads, global architecture, self-managed, programmable, marketplace, governance. Which ones are the most important? All of them, you know, it's like when you have kids you don't want to pick and say, this one is my preferred one. So they are really important. All of them, as I said, without data, there is no snowflake, right? So all data is so important that we can reach every data wherever it is and that's part, iceberg is a part of that. But all workload is really important because you don't want to put your data in one platform if you cannot run all your workloads and workloads are much broader than just data warehousing. There is data engineering, data science, ML engineering, all these workloads, applications. So that's critical. Programmable is where we are moving, right? We want to be the place where data applications are built and we think we have a lot of advantages because data applications need to use many workloads at once, right? It's not that the data application will do only data warehousing, they need to store their states. They need to use this new workload that we define, which is Unistore. They need to do data engineering because they need to get data, right? They have to serve this data. So they need to combine many workloads and if they have to stitch this workload because the platform was not designed as one single product where everything is consistent and works together that you have to stitch, it's complicated for this application to make it work. So Snowflake is, we believe an ideal platform to run these data applications. So all workloads, programmable, obviously, such that you can program. And programmable has two aspects, which is a big part of our announcements. It's both data programmability, which is running Python against petabytes, terabytes of data at scale and doing it scale out. So that's what we call data programmability. So both Java, Python and Scala. But also running applications like UI and we had this acquisition of Streamly. Streamly now has been fully integrated in Snowflake. We announced that such that not only you can have these data programmability but you can expose your data through this nice UI, interactive UI to business users, potentially. So it goes all the way there. Global is super important as we say we want to be one platform for the world. And of course, as I said, the last pillar which is somewhere critical for us because we are cloud. We need to have governance, we need to have security of our data. And why it took us so long to do Python? It's not because it's hard to run Python, everyone can run Python, it's because we had to secure it. And I talk about it, creating this amazing sandboxing technology such that when you include third-party libraries and third-party codes, you are guaranteed that this third-party code will not reach to its filterator data. We control the environment that Snowflake provides. Can you share some of the feedback from the customer? You probably had many customer conversations over the last four days. That's mild. You know. Actually not, because I was so busy everywhere that unfortunately I didn't speak to many customers saying that I had everyone stopping me and talking about what they heard and yeah, there is a huge excitement about all this. What's been the feedback around the theme of the event, the world of data collaboration? Data collaboration is so critical as every company these days must be a data company to compete, to win. What's been, from just some of the feedback that you've had customers really embracing data collaboration, what Snowflake is enabling? Yeah, I mean every company, I mean almost every company which is using Snowflake, is collaborating with data. You have heard the number of stable edges that we have and there is a real need for that because your data alone, you cannot make sense of your data if it is just a loan. It needs to be connected with other data. You haven't not generated so all data when you say the first pillar of Snowflake is all data is not only about your data but it's about all the data that are created out around you that's put perspective on your own data and that's critical and it's so painful to get, I mean even your data is difficult to have access to your data but imagine data that you didn't produce and so yes, the data collaboration is critical and then now we expanded it to application and expertise, sharing models for example. That's going to have a huge impact. All data includes now transaction data, right? That's a big part of the announcements that you guys made. So and that's the motivation for that was really, if we want to run application, full application, we announce native applications which are fully executed and run inside the data cloud, right? They need all the services that application need and in particular managing their states and so we created Unistore which is a new workload which allows you to combine transactional data which are generated by this application and at the same time being able to do analytics directly on this data. So we call it hybrid table because it has this hybrid aspect. You can do both transactional access to this data and at the same time analytical without having data pipeline and moving data and transforming it from the transactional system to the analytical system, right? Snowflake is one system again in the spirit of simplifying everything, you know. This is the Snowflake. I'm going to ask the same question I asked it first all over again. When was the aha moment that you and Thierry had that said this is not just a better data warehouse, it's actually more than that. You probably didn't call it a data cloud until later on but did you know that from the beginning or was that something you kind of stumbled into? No, we, so as I said, we founded Snowflake in 2012 and Thierry and I, we locked in my apartment and we were doing the blueprint of Snowflake and trying to find what is the revolution with the cloud for this data warehouse system, analytical system, both big data and data warehouse. And the aha moment was, you know, but of course cloud, okay, what is cloud? It's elasticity, service, and later collaboration. So in the elasticity aspect, we, you know, when you ask, you know, database people, what is elasticity? They will tell you, you have a cluster of nodes, you know, like if it is a rock or it will be a rock cluster and elasticity is that you can add, you know, one node, two nodes to this cluster without having too much impact on the existing workload because you need to shuffle data, right? It's hard and doing it online, right? That's elasticity. If you can do that, you're elastic. We thought that that was not very interesting to do that. What is interesting with elasticity is to plug new workloads. You can plug a workload like that and that workload is running without having any impact on other workloads which are running on the platform. So elasticity for us was having, you know, dedicated compute resources to workloads and these compute resources could start and be powered as soon as the workload, you know, starts and, you know, will shut down when the workload finishes and it will be sized, you know, exactly for the demand of that workload. And we thought the hard moment was, okay, if we can do that, now we can run a workload with let's say 10x more compute resources than what you would have used or 100x more. Okay, let's say 100x more because we paralyze things. Now this workload can run 100x faster, right? That's, you know, assuming we do a good job and we scale which is, you know, our IP. And if we can do that, now the compute resources that you have used, you have used them for 100 times less. So you have used 100x more resources because you have more, you know, nodes but because you go fast, you use them for less time, right? So if you multiply the two, it's constant. So you can run an accelerated workload dramatically, 10x, 100x for the same price. Even if we are not better in efficiency than competition, just having that is what the magic, right? Did you, you know how Google founders originally had trouble raising money because who needs another search engine? Did you get from original, like, when you started going to raise money, Amazon's got a database. So who needs another cloud database? Did you get that early on or was it just obvious, Spicer and companies as well? Spicer is a little bit on the crazy side and ambitious and so Spicer is Spicer and of course he had no doubt. But even him was saying, Benoit, Thierry, you know, Adoop, right? Adoop is, everyone is saying Adoop is going to be the revolution and you guys are betting actually against Adoop because we told Spicer, Adoop is a bad system. He's going to fail. But, you know, at the time, you know, everyone was so bullish about Adoop, everyone was implementing Adoop that it didn't look like it was going to fail and we were probably wrong. So there was a lot of skepticism about, you know, not leveraging Adoop and not being an Adoop, okay? Something built on top of Adoop. That was number one. You know, there was no cloud warehouse. At the time we started, Redshift was not started, was the pioneer somewhere when Sloan Fleck was founded. So creating a data warehouse in the cloud, you know, sounded crazy to people. You know, how am I going to move my data over there and security and what about security? You know, the cloud is not secure. So that was another, you know. So you guys predated that Par Excel move by, okay. So that's interesting. We, and you know, I thought when Redshift, I mean, Amazon announced Redshift, I was sure that Mike Spicer will come and say, guys, you know, it's too sad, but you know, they beat you guys and they build, you know, something and actually it was the reverse. Mike Spicer was super excited and so it was interesting. Wow, that's amazing. You know, cause John Furry and I were early with theCUBE. When theCUBE started, it was like the beginning of Adoop. And so we would, we brought theCUBE to, I think it was the second Adoop world and we was, you know, rubbing nickels together at the time and I had, and I was so excited bringing compute to storage and it made so much sense. But I remember, I won't tell you who it was, but an early Adoop Committer told me, this is going to fail. I'm like, what? And he started going eight spaces, crap and all this stuff. And then, and I was sad because I was so excited, but it turned out that you had the same permission. Because of complexity, okay. Adoop failed for two reasons. One is because they decided that, oh, a lot of these database thing, you don't need transaction, you don't need SQL, you don't necessarily, you don't need to go fast. It will be batch, you know, no more, you know, real time interaction with data. No one needs that. Cheap storage. So a lot of compromise on the very important technology and at the same time, extreme complexity. And complexity for me was where I was, I knew that it was going to fail, big time. And we bet, you know, snowflake on the failure of Adoop indeed. And there was no cloud, you know, early on. There was no cloud too. It was not built. You know, that was like, it doesn't. You're right. And the model that Adoop had for data didn't work on blob storage. Blob storage is not as efficient as HDFS. So that was also another failure. Do you ever sit back and think about, so you think about how much money has poured in to separating compute from storage and cloud databases. And you started it all. Yeah, yeah. No, this is pretty amazing. Yeah. All right, so that's good. That means that you're onto a good idea. A lot of people get confused that, again, they think that you're a cloud data warehouse and you're not. You're much more than that. Yeah, I hate that. So I have to say, because from day one, we were not a cloud data warehouse. As I said, you know, it was all about, you know, combining the big data, you know, massive amount of unstructured data at beta by stored as files. Okay, that's very important, you know, stored as files, where it's very easy to, you know, drop data in the system without, you know, very low cost to combine with data warehouse, you know, full multi-statement transaction. When people will tell you today, oh, I invent, you know, now we are data warehouse, they don't have multi-statement transaction, right? They have, you know, so we had, you know, from day one, multi-statement transaction, you know, really efficient SQL. You could run your dashboard. So, you know, combining these two worlds was, I think, the crazy thing that's the crazy innovation that Snowflake did initially. And I know it's really easy to build data warehouse somewhere because if you don't think about big data, petabytes, semi-structured data, you know, you remove a lot of complexity. This is why, Lisa, when you get excited about technology, but you always have to have a, somebody really deeply understands technology to stink test it, all right, so. Awesome, thank you for sharing that story. Fantastic, so over 5,900 customers now, I saw over 500 in the Forbes G2K. Over, almost 10,000 people here this year, if we think back to 2019, there was about what, less than 2,000 people. What do you think's going to happen next year? I don't know. I don't want, I don't like to think about next year. I mean, where I always say, you know, Snowflake is so exciting to me because it's like a TV show, right? Where you wait the next season, and you know, we have one season every year. So I'm really excited to know what's going to happen next year. And I don't want to project what I think will happen, but you know, all these movements to the, you know, Snowflake being the platform for that application, I want to see what people are going to build on our platform. I mean, that's the excitement. Season 11, coming up. Yes, yes, season 11, yes. Oh, binge watching here, Benoit. It's been a pleasure to have you on the program. Thank you, thank you. Congratulations on the incredible success of Momentum. The energy is contagious, we love it. Thank you so much. Thank you, bye-bye. For Benoit Dajaville and Dave Vellante, I'm Lisa Martin. You're watching theCUBE's coverage of Snowflake Summit 22. Dave and I will be right back with a wrap.