Fast in the cloud is free: How $40B gets built

"Fast in the cloud is free." This counterintuitive insight became the foundation of a $40 billion revolution in data warehousing. While most entrepreneurs chase trends, Benoit Dageville and his co-founder spent months in a cramped apartment dreaming about what seemed impossible: making data queries 100 times faster while costing exactly the same.

Dageville, co-founder of Snowflake, stumbled upon this breakthrough not as ambitious entrepreneurs, but as Oracle engineers who "never wanted to do a startup." They simply wanted to build something amazing. Their reluctant journey from database developers to cloud computing pioneers offers rare insights into how transformative ideas actually emerge—not from market research or competitive analysis, but from deep technical understanding combined with fearless dreaming.

In this interview, Dageville reveals the exact thought process behind discovering a killer idea worth billions, the art of making revolutionary technology feel familiar to users, and why starting with 100% trust (not 0%) is the only way to scale a company to 9,000 employees.

Benoit Dageville: As you can see from my thick French accent, I am French by origin, but I lived in the Bay Area for many, many years. I came here in '96. I did my PhD in database systems, but I felt that in France there was no real opportunity to create a new system. I'm a creator. I'm some way an artist. It's stupid to say, but software engineers sometimes feel of themselves a little bit as artists because they build new things and France was not a good environment to do that.

I did research in France but then I left to join Oracle, the core development. Myself and Thierry Cruanes, who is the other co-founder of Snowflake, we were both working in the core database division at Oracle, and the funny thing is that we never wanted to do a company. We never wanted to do a startup, really. We wanted to build an amazing product.

We thought the cloud could be a revolution for data access. Traditionally people would put data in their data centers and they would use systems to access this data. We thought with the cloud we could create a solution so simple where people would not have to manage infrastructure. They could just ask questions to their data and Snowflake would give them answers and to do that we had to do a startup, we had to do a company.

Benoit Dageville: So Thierry and I, we left Oracle and when we started with Thierry, we had no idea what we were going to do and actually we had no idea if we were going to find a great idea. In terms of finding a great idea, it takes time. We started in August 2012 and we worked until the end of 2012 to start to build a team, so I had a small apartment on El Camino. Thierry was coming every day to my apartment. We bought a whiteboard. We had two laptops, and we fought completely from the ground up of what cloud means for these data systems, how cloud can be impactful.

How can we build Snowflake in this completely different environment than traditional data centers, and we were dreaming. You have to dream to find a big idea. You have not to be scared to dream about what does it mean if only that was possible. What would be the impacts? And then you work back from that if statement and you see maybe we can make it happen. It was a big if. It was a little bit of a dream, but maybe it's not.

And let's look at the new technology and that's the other important thing is technology unlocks new possibilities, and I think it's important to understand what new technology can do to an industry and think back a little bit and say wow, cloud, that's what cloud can do and that's really what we were looking for is how cloud can have a dramatic impact on this industry which was super old and super boring somewhere the world of data databases, it's like boring.

Benoit Dageville: And what was this great idea was about, you know, in the cloud you can provision a lot of resources, servers, machines. So if you have say 10 machines or if you have 100 machines with 10 times more machines, basically you can go 10 times faster. So we say wow, and we know how to do that. So we said wow, in the cloud we can allocate maybe 100 servers, 1000 servers, and we can go 100 times faster, 1000 times faster.

Actually, because we rent the compute resources to the cloud provider, when we are done, we can return that. And so in the cloud, if you go 10 times, 100 times faster, you pay exactly the same price because you use these resources 10 times, 100 times, 1000 times less. Even though you have more resources because you use them less, you pay the same price. So you say fast in the cloud is free and if we can do that, and this was of course technically very challenging, but if we can do that, we can basically build a system that can go 10 times 100 times faster than any system that exists today at the time and yet it will not cost our customers more.

So that's the killer idea. And when we had this idea, we said if we can make it happen, if we can build it, then we have an amazing company. So we knew that this was going to be a revolution.

Benoit Dageville: And actually it became a revolution. We started to have one customer, 2 customers, and they were blown away by what you could do with Snowflake and without complexity it was simple to use because we removed all the friction. We were able to get more and more customers and people would talk to each other and then it gets to this viral state where it exploded.

We were very worried about that because we said if we show that Snowflake is so different, people are going to be worried too. At the same time, if you say it's exactly the same as what you are using, people are not going to adopt you. So you have to find a good balance. Sometimes you have to explain yourself in comparison to existing technology. We, for example, use the same terminology that was used in the old days, so we had a system called warehouse. Everyone knew warehouse. That's a familiar concept.

So you have to use existing concepts such that you can help people go your way and your revolution. So somewhere if you say it's a revolution, it's possible that it will not work. So you have to bring them slowly, not too fast. So it's tricky. It's a very tricky. It's a lot of explanation how you explain it, how you make it look familiar somewhere, even though it's very different. So it's an art.

Benoit Dageville: The number one quality of an amazing product is simplicity. If you can use a product without knowing a lot about this product, this product is going to be amazing. Think of cars, Think of iPhones, think of simple products. These are the ones where you say, ha ha, wow, and a product which is simple but at the same time so powerful.

People who have complex products but expose this complexity to the end users are not making great products. So for me it's first of all, simplicity, but underneath again, if you take the analogy of the iPhone, the iPhone is nothing but a simple piece of technology is probably the most advanced piece of technology. But when you use this product, they are simple. So that to me is probably the number one aspect of a great product.

Benoit Dageville: Take 3 fingers and list all the things that you could do potentially, but try to limit it to a few, very few important things, and you rank. You just say our product obviously needs this and that. It's always tempting to say, oh, I'm going to appeal to everyone, and then you do a bad job because you try to make everyone happy. When you try to make everyone happy, no one will be super happy, I can tell you.

So the way to streamline is both to really focus on a few things and also to make these few things focus on a few persona. Don't try to appeal to everyone.

Benoit Dageville: We have a saying at Snowflake, go direct. We are human beings, so we have conflicts, we have disagreements, and that's very important to. Amazing technology. I always say is built from arguments and discussion, and I disagree with you. I think we should do this. No, we should do that, and these discussions are the things that really makes the best product at the end, because you get to this consensus. You understand different opinions, so.

It's important to be able to do that in a very secure environment where people feel that they can speak up. They can say, no, I think this is not going to work, and it doesn't matter who you are to say these type of things. So I think you have to encourage, to encourage creativity, you have to encourage everyone to speak.

Benoit Dageville: The most important to grow is indeed to hire someone that you trust, otherwise you become the bottleneck and you don't want to have this full control of everything. So how you gain this trust is interesting. I remember actually the first hire we had at Snowflake, our first one. He was asking a lot of questions before he wanted to sign. He was asking tons of questions like, OK, how is it going to work, suspicious.

And he said, Benoit. He told me one day, you know, Benoit, I'm asking all these questions. He was from Europe. He said, You know, I'm from Europe, and Europe, you don't trust people. You start almost with zero. And then as you know these people, as you start to trust them. You know, you give them points, so to speak, and then you fully trust them, but you start at zero and you build that trust over time.

He said in the US it's the reverse. You start with 100%. You've never met this person. You give full trust and then you remove if you see that it doesn't work, and I would say that the biggest thing for me is that from day one you give full trust. If you hire someone, you have to absolutely trust them from day one. You don't build that trust. It could not work.

We made some hiring mistakes and we discovered that, oh, this is not working for one reason or another, then you can remove this trust, but you start 100%. You have to trust and give someone else the opportunity to do their thing, and this is why you hire them. So this trust is a super critical aspect, and my take is give it 100% to them from day one. Don't do the European thing.