Episode 453: Snowflake CEO Sridhar Ramaswamy on Transforming AI & Powering the Data Cloud

Chrisan, excited to be here. Looking forward to our conversation.

Our founders, and this is amazing that it was over a decade ago, had a singular vision that was based on a really cool insight that out in the world of cloud computing, you could scale storage orthogonally from compute. In other words, if you made the compute engine flexible enough that you could replicate it, then you got some amazing characteristics. What does it mean? Like when you buy software in a box, if you want to do lots of things in that box, you're no longer able to do it. You have to buy a new box, a bigger box. Snowflake was infinitely scalable because of its unique architecture. But over the years, we've built on it. We are the premier collaboration platform for enterprises wishing to exchange data with one another. And over the past few years, we've all been adding on additional machine learning and AI capabilities. I describe Snowflake as an AI data cloud. We are the platform that enterprises, small and large, use to store their data, get insights about how their users are using their product, and then make predictions about which parts of their product their users will want to use, leading to an end-to-end platform that tightly integrates with the products that companies ship. And AI, of course, is going to massively accelerate the power and value that companies can get from data. And that's the reason why Snowflake has been growing and thriving so much, and why we and our customers and partners are so excited about what's up ahead.

Well, you know, the New York Stock Exchange is the most studied stock exchange ever on the planet, and you're thrilled and honored to be on it. But there's also a very deep collaboration between the firms. It's at the level of us being the data platform for ICE and for the New York Stock Exchange, and being an important distribution mechanism for all of the amazing data that NYSE wants to share with its customers and partners. We get a lot of great feedback about where Snowflake should evolve from Lynn and her team. And things like the tree lighting and the judging are the icing on the cake that demonstrate how closely the companies work with each other to partner, but just as importantly, create success for all of our giant customers. It's a deep partnership that I'm very, very grateful for.

Yeah, you know, Neeva was born of an idea that the world of search needed to be changed in a dramatic way. Back in 2019, when we started, we did not know that GPT-3 was in the offing or that language models would become the force that they are widely acknowledged to be today. But we knew that there was a better search engine to be created. But four years into the journey, honestly, we had built an amazing product, but we were struggling to achieve commercial success at scale. Which is when we started talking to Frank Slootman, Chris Lynch, Christian Kleinerman, Benoit and Thierry, the founders of Snowflake. And we realized that this was going to be a great opportunity to take the technology, the search, the AI that we had built at Neeva, but apply it in an enterprise context for it to have massive impact in the enterprise. I tell people that AI is already magical because these language models can do things like understand fluid language, like the language you and I are speaking, and make sense of it, extract context from it, and do that effortlessly. We knew that how people were going to consume information was going to be changed forever. Just like back when the first smartphones came, most people in the world kind of simultaneously realized that there was a before and an after, that everybody was going to have these amazing devices with them. I think we're going to think about AI the same way. There is a before and there's an after. The before is clunky little interfaces that you and I had to type information into very, very carefully. And the after is fluid language that we write or speak, and every application, every phone is able to understand exactly what we mean. We thought this was a match made in heaven. And both the folks at Snowflake and the senior engineers and my co-founder at Neeva felt that these were people that we could work with, that we'd have a lot of fun doing it and have big impact. And I'm pretty happy with how the past 18 months have turned out, not just for me, but for the entire company. My co-founder is the head of engineering at Snowflake now, is having a broad impact. But most of the key engineers at Neeva are influencing the roadmap and product of Snowflake in a very big way. And in as much as I said there's a before and after with AI and data, we feel very proud to be a part of charting where enterprise data, where this AI data cloud is going to go.

Google was an amazing place to work at, and I was lucky enough to be part of possibly the best ever business on the planet, ever, which was the search ads business at Google. But there are many characteristics that made my time at Google and my learning very, very special. I'll just touch on a few of them. First and foremost is we always knew that success could be fleeting. And one thing that I, you know, because you learn this is the land where the most iconic of founders went on saying things like 'only the paranoid survive,' or you could see what happened to companies like Silicon Graphics. And so the land of companies that succeeded and didn't succeed was deeply imbued in us. And so we had this relentless drive to excel, to exceed, and that was part of the mentality that created what was eventually a 100-plus billion dollar business. There are not many of those on the planet. But I think that is an important quality that I bring to the table at Snowflake, which is yes, achieving success is important, but achieving excellence, which is a way of living on an ongoing basis, is just as important to create iconic companies, to drive lasting success. I would say that was an important characteristic at Google. I went from an individual contributor, an engineer that wrote code sitting in a cubicle talking to no one, to running a team of over 10,000 people. I got to work with many iconic people, people like Eric Schmidt, who was the early CEO, the founders of Google, whose drive and intellectual curiosity was just amazing. But also people like Bill Campbell, famously the coach of Silicon Valley, who taught me what it was to be a human leader. Bill always knew what was happening with my children. He cared about what I did at work, but he cared just as much about whether I was doing well in my life. These were very humbling people to learn from. I try to bring that humility, that care, that want to the relationships that I have with the people that I work with. And balancing those two, how to be a human, how to be a good human being, how to make sure that we take care of the people around us while simultaneously creating a culture of excellence and creating lasting success, I say those are big, big lessons that one learns quite often just by watching and imbibing. And I'm very grateful for the time and opportunity that I got at Google.

Yeah, it's a great question. And initially, I set myself three goals. We knew that we were in a rapidly changing world, and so making sure that we had very good product velocity was something really important, was a huge priority for me. I worked very closely with the product and engineering teams to make sure that we were iterating, that we were rapidly shipping, that we were getting feedback from our customers, that we evolved different go-to-market motions than the ones that we had used to make the cloud warehouse successful. So that was a huge, getting new products right and getting them out to the market quickly was a huge priority. I also spent a lot of time on the road. I was out on the road probably every other week this year. This is to learn from customers and getting direct feedback from them about what we were doing well and what we needed to do better at. And by the way, like being on the road is also a great way to meet your field team, whether it's sales executives or sales engineers or sales leadership. So you learn traveling the world, traveling through the cities in the country, meeting customers. And I would say that has stood us in good stead. So we laid the foundation for faster product delivery, we laid the foundation for being able to take new products to market efficiently, and then also learning from our customers. You know, things take a little while to manifest themselves, but I'm exceptionally happy with how things came together for our Q3 earnings, for example. And also, honestly, the confidence that's given the entire company that they can step up and deliver at even greater scales than what they're doing right now.

Yeah, I would say yes, AI is changing very quickly, but there are also a set of fundamental observations that you can make. I talked earlier about how AI is going to take the language that we speak and the language that we write and be able to make sense of it, be able to extract structure from it. It spans the gamut, by the way. If there is a table in an analyst report that somebody puts out, the language model can extract that quantitative data from the table and put it for analysis into a Snowflake table. It can also fish out interesting tidbits about feedback that an analyst has about a company, which is just text. In addition, one of the most interesting things about AI is that it can do these data transformations with much more ease than before. I'll give you a simple example. Like five years ago, if you had access to a bunch of clinician notes and wanted to know what are the symptoms that patients had, that would be a custom project. You would need to go get software engineers, they would need to know machine learning and AI, and they would work for many months before they built you a system. Right now, somebody that uses Snowflake can write a single SQL statement and be able to do exactly this. And so we went about it very much from the viewpoint of how do we put a foundation of AI that is accessible from every part of Snowflake, whether people are ingesting data or transforming data or building reports with data. And then on top of that, we've started building out the key primitives, the chatbots as it were, that can then be stitched together. This way, it becomes almost like assembling a Lego set where the important pieces are there. Yes, you might get new pieces, but you're able to flexibly combine them into systems that create value for you as a customer. I'm very pleased with how far we have come. And we announced this agentic platform called Snowflake Intelligence that's going to push the boundaries of what is possible even more. It's going to let people compose chatbots on unstructured text, like on unstructured documents like text documents or PDF documents, but also on structured data that they store within Snowflake. And then also be able to write updates to other systems like a Workday, like a Salesforce. It's this gluing ability of AI models that we are currently exploring. But hopefully, this gives you an idea of how you start from the bottom, the world of data, then let customers use the power of AI to transform this data and then get value from it. And part of what is really exciting is that business users, like my CFO, can get answers to questions without needing to know SQL, without needing to know complex technology, in very, very flexible ways. And these are some of the things that we are creating. But you're right, we have to keep pushing the boundary of what is possible with AI models and data.

It's a great question. I mean, and we went through a lot of these questions ourselves when it came to Snowflake using AI products. And so we've set a, we have a set of guidelines, we have a set of principles for how we think about AI and putting it into Snowflake's products. The first, an ironclad guarantee that we offer our customers is their data is theirs. Your data as a Snowflake customer is yours. We are never going to use it to train anything that is shared across customers. That sanctity of data is the thing that gives our customers comfort that their data is safe. And obviously, customers both care about the data from a business perspective but also from a regulatory and doing the right thing perspective. So that's the bedrock of how we think about data and AI. On top of that, we provide a set of tools to make sure that AI products that our customers build cannot be misused to generate harmful output. For example, you have something called Cortex Guard that can be used to filter away malicious input that bad actors put into these applications to try and trick the models into doing bad things. So we build protection like that right into the core product. And then how AI is going to be combined with other technologies like machine learning is a further topic that we are actively exploring. As you know, in the world of machine learning, there are fairness laws. We need to make sure that these models are not biased. And so we are providing a set of tools to let our customers be able to easily judge these themselves so that they can both stay compliant with laws and regulations but also just as importantly do the right thing from a larger perspective. There's a very active dialogue, as you pointed out, in a pretty fast-moving space.

Look, simplicity is a topic that's near and dear to our heart. It goes back to the founding of Snowflake. Our founders had previously worked in a company where product complexity was rampant, where you needed armies of administrators to make sure that things were well tuned. At Snowflake, they took the opposite approach. They said we want to make this product easy to use and easy to get value from. And this is where we really shine through compared to our competitors. Snowflake is easy to use. You can load petabytes of data incredibly quickly into Snowflake, but magically, you can start running queries on them interactively right after you load them. And it is that simplicity that drives broad adoption. AI is great, but if you need to be deeply experienced in AI to get any value from it, that's going to limit its application. Our dream is more that business users, for example, that know what makes revenue for them, that know what creates profit for them, are able to get at the data they need without complex business intelligence tools, without the need for an army of analysts that need to interpret every single thing that they said. And I said that's what we love doing. We love taking complicated technologies but making it easy and safe for people to use. One worry that a lot of people have about using AI is whether things like access control will be preserved. I mean, think about it. All of us have documents that are stored in either Google Drive or in SharePoint. And putting these into an AI system needs to guarantee that only the right people can still see it. It cannot be the case that you stop obeying access controls. Things like that we build right into Snowflake so that our users don't have to worry about whether access controls will be followed. They absolutely will be. I think these are some of the things that people need to keep in mind as they're thinking about how they get value from data. And at Snowflake, we try very hard to make sure that things work as expected, things work quickly, things are easy to try out. I tell people that you can build a chatbot on Snowflake in 10 minutes flat. And then you can see is that something that you want to deploy at scale to your whole company or not. Simplicity and ease of use drives value creation.