Sridhar Ramaswamy，Snowflake CEO：DeepSeek 并非 OpenAI 的威胁，OpenAI 胜过 Anthropic｜E1258

I think it is terrifying to be a startup building on top of OpenAI. The line between an infrastructure provider and an application provider today is super blurry. There’s basically no guarantee that OpenAI, Anthropic, Microsoft, or Google aren’t going to go and create.

You’re not going to switch from ChatGPT to DeepSeek? It’s a product; it’s not a model. There’s a big difference, Harry. There’s a reason Anthropic hasn’t done as well, because it operates at the model level. It’s all of the additional things that go into ChatGPT that I think have staying power.

Sridhar, I’m so excited to have you on the show. It has been so long since we saw each other last, so thank you for doing this with me.

Thank you, Harry. Amazing to be here. 20VC is absolutely one of the iconic names in investing. I’m excited for this.

You’re very kind. Listen, I spoke to Margot before the show, and she said you’d be really missing a trick with Sridhar if you went straight to AI and straight to CEO-ship. She said, “You’ve got to get a little bit more personal.” So I said, “Okay, Margot, tell me more.”

Start with this: as your younger self, in college or in your first job, did you ever imagine yourself becoming a CEO, and why?

No, I did not. Certainly not when I was getting my bachelor’s degree, and certainly not when I was getting a PhD. In fact, when I was getting a PhD, becoming a professor sounded more like the ideal. At some point, I wasn’t that excited about doing research, and that’s when I switched over to software engineering.

Even in software engineering, it’s very much about aspiring big but taking little steps. When you’re at the right place at the right time, great things happen.

I totally agree with you in terms of being in the right place at the right time. Says the man with a fund and a podcast.

My question to you is this: when you look at graduates coming out today and entering the workforce—I know you also have two boys, who Margot told me about—what would you advise 20- and 21-year-olds coming into today’s workforce who are scared, in some ways, looking at AI and asking, “Where are there going to be long-term jobs, and where are there not going to be?” What would you advise a 20- or 21-year-old entering the workforce?

My kids and I used to talk a lot about what they should work on. They’re 25 and 23. Roughly, my advice to them was: find something that you’re passionate about and that society values.

I was like, “Don’t pick these noble professions that we all pretend we value, but in fact we don’t.” Being a teacher is really hard—God bless them. Hopefully, by the time you have graduated, you have skills and passion in something that society values.

Obviously, everything is undergoing a lot of change. The best advice I have for navigating the environment is to be open and nimble about where change is coming. Embrace change. Don’t be one of those people who says, “The internet is never going to be a thing.” AI is a thing.

See what is possible, and put in the time and dedication to excel at the job while thinking about where things can go. That core formula of drive and malleability is the advice that I give to a 21-year-old. Honestly, it’s also what I look for when I’m trying to hire a senior executive into Snowflake.

I see so much debate around software engineering and engineering. People say it’s the first place where AI is really going to dominate, and you’re starting to see that with Copilot and a lot of the tools around it.

Do you disagree and say there’s immense value in software engineering moving forward—that students should keep going—or do you think it will actually be the first profession to be impacted?

I think all kinds of knowledge professions are going to be impacted in a very big way. My simplistic but fairly accurate description of what AI can do is that it’s an incredible translation layer between all kinds of knowledge—unstructured and structured. It can make sense of it all in a way that only humans were able to before.

I would say any knowledge worker, absolutely including software engineers, should embrace this technology and see where it’s going to go. I think it’s too early to jump to conclusions about whether software engineering becomes an apex profession or a thinned-out profession.

What I mean by that is, after the internet, there wasn’t a particularly good reason for every city in every country to have foreign bureaus sitting in Washington. Why? The big outlets, like The New York Times, could reach every part of the world. You could get pretty good coverage. Maybe you didn’t like it, or maybe you wanted another viewpoint, but journalism went from being a broad-based profession to one that was narrower.

The same thing happened to music, and the same thing happened in other fields. It’s a little hard to predict how much impact AI is going to have. But to the extent that software will continue to be a huge part of everything we do, I still see tons of opportunity in applying software and AI to more and more areas.

I’m not down on software engineering as a profession any more than I’m down on analysts or CEOs as professions. I joke to people that I’m an email machine. That’s all I do. I talk, write, and read.

Embracing the future, seeing where it can go, and being nimble is what it’s going to take. I don’t think software engineering is going to disappear anytime soon.

I love the simplicity: “I’m an email machine.”

People always think you just get better at CEO-ship and become more phenomenal as the years go by, and you present this wonderful façade. What have you gotten worse at over time?

That’s a good question. Physical skills are harder as you grow older. That’s just how it goes. Once upon a time, I wouldn’t think twice about a 20-mile run with no preparation. It was like, “Yeah, okay, let’s go run for a bit.” That’s not going to happen today.

The mental side is very much under your control, and that’s the part I’m happy about. I’m relentless about pushing myself to learn, measure, and adapt. When it comes to physical things, it is harder.

The other thing we have to be realistic about as we grow older is that our ability to sink a huge amount of time into a brand-new area is limited. You have to have humility about things like that.

I joke that one of my sons is 23 years old, and I’m like, “You’re not really going to be a concert pianist in your life. That part is done.” We have to accept what is changeable and what is not changeable.

But especially when it comes to mental things, for the agile, driven person, the world is your oyster. You now have all of these language models that can create custom programs for you and do amazing things that you would never have been able to do before. In that sense, I think there is even more opportunity than before when it comes to the mental side.

You mentioned 20-mile runs and relentlessness. By the way, I think you’re still an Adonis, so I don’t know what you’re talking about.

I spoke to many of your friends, and they said you operate with an incredible intensity. I thought, “Gosh, it’s another praise of Sridhar.” But they said, “No, because people around him struggle to keep up.”

I’m by no means running 20 miles and as athletic as you, but I am intense, and people struggle to keep up with me, too. I struggle with that. What advice would you have for me and other high-intensity CEOs who need to bring people with them, but find it difficult?

I think it’s important to make sure that people understand the big picture, the fleeting nature of opportunity, and how quickly it can disappear.

I run a company that is making, call it, $3.5 billion roughly a year. If you think about things like world GDP, adjusted for inflation, it grows by 1%, 2%, or 3% a year. It takes 100 years for something growing at 1% to double. We’re talking about doing that in 2 years, give or take.

The first thing I tell people is that we are extraordinarily lucky to be in environments that can generate this much value. Sustaining and accelerating that takes extraordinary people. It’s not for everyone, and I’m okay with that.

But let me tell you what it takes to succeed at Snowflake: the expectations are high because the returns are high, the opportunity is big, and the opportunity cost is even bigger.

It’s a question of what people want to be part of. At Snowflake, we absolutely want to be the data engine for every single enterprise on the planet. I think that’s a pretty amazing mission, plus we get to build an iconic company along the way. It takes special people. I don’t have to make excuses for that.

When you hire someone thinking that they’re able to scale through different stages of company development, and you’re wrong, why have you been wrong?

These things are unpredictable. Life is very long, and people also change over time. There’s a certain something that, take your pick, £10 million or £50 million is going to do to people.

Part of the reason I joined Google in 2003, a year before the IPO, and honestly part of the reason I succeeded, is that a bunch of people who joined in 2001 and 2002 suddenly became worth $500 million. They said, “Life’s great. I’m done.” Honestly, I don’t blame them, but that’s part of what it takes to have a driven team.

Not everybody can scale. I’ve had lots of conversations with people about what it takes to double your team. Roughly, my advice has always been: every time your team doubles in size, all of the things that made you wonderful for your previous team are typically massive inhibitors to succeeding in your new job.

I’m not sure people really internalize that. It requires a lot of reinvention in terms of what people have to be and how they have to operate. There are quite a few people who can make these transitions, adapt, and thrive.

I almost see it as my duty, job, and responsibility to give people those kinds of opportunities. But I’m also relentless and cutthroat in that I give people time and opportunity. One of the things I’ve learned as an adult is not to shy away from unpleasant conversations.

I talk to people about what is working and what is not working. I give them time, and if it doesn’t work, it doesn’t work. I have demoted people, given them half their job, and said, “This is better for you.”

That’s one of the hardest things you can do in your life. Do you have any tips on how to force the hard conversations?

Whenever I have an email that I don’t want to send or don’t know how to send, I force myself to go into the lift in my apartment building. Whatever I’ve written by the time I get to the other end gets sent. That’s the deal I make with myself.

Do you have any tips on how to force those conversations?

The first thing to remember is that conflicts typically do not resolve themselves. You have to internalize that postponing things is going to make them worse. The more you postpone, the worse it gets. Just telling yourself that is important.

The other thing is that avoiding the hard conversation is typically doing the other person a disservice, because you’re not helping them. Not all hard conversations need to be, “You’re out of your job.” They can be, “Here are the areas where I think things need to go better. Let me tell you why, and let me tell you what my expectations are.”

It’s also important to be respectful. I find it important to be painfully humble in these conversations. I’ll typically go into the conversation and say, “This is going to be a difficult and unpleasant conversation for us. I’m just coming out and telling you that. I’m not pretending that I know all the answers. This is super awkward for me to have this conversation, but I feel it’s important that we have it.”

To me, that combination of respect, straightforwardness, and humility goes a long way.

One final question before we move on. You mentioned the demotion element, which is a really hard thing to do. If you’re willing to demote someone, does that not just mean you should fire them?

Someone once said to me, “If you’re willing to take less, you should never do the deal.” I apply that here: if you’re willing to have them but give them less, should they not just go?

There are situations where that makes sense, but remember that all of these are fast-moving environments.

An area that you thought could be handled fine with 20 engineers can suddenly blow up, and all of a sudden it needs 100 people. The person who did the job well with 20 people is now holding a portfolio of 40 people that needs to grow to 100, and they’re struggling.

The business can’t wait. It’s not that they’re bad people or incapable of doing a good job. To me, finding the right context and framework to help somebody succeed at a given point in time is okay. It’s okay to have those sorts of conversations.

I had one director who accidentally became responsible for two different areas. Part of what I told him was, “I’m taking one of these areas away. By the way, the thing that you have in your hands is going to grow to be a $40 billion business.” I think I promised him that in 5 or 6 years.

This was Google Shopping. I said, “This is what we’re looking at. It’s a big opportunity, but you get to focus on a single job.” That person took the job, eventually became a VP, and was enormously successful.

It’s a lot of selling, because whenever you have a conversation like that, the first reaction people have is, “I’m a failure.” No, you’re not. You’ve been given an impossible job. Let’s shape it to make you succeed.

Again, that is real leadership: going and selling something that people don’t want, painting a vision for the future, and getting them to stay. That’s hard.

Do you think richer leaders make better leaders? Are you more convicted and less worried about the downside? Are you less fearful about what happens if you lose your job or something goes wrong? Do richer leaders make better leaders?

No. I think it can make them callous, and it can make them too tolerant of massive amounts of risk.

It’s a balance. You always have different constituents, and it’s important to remember that as a leader. It’s not just you. It’s the employees in your company, your shareholders, and your customers.

We went through a pretty difficult time as a company last year, and that’s when it became clearly focused for me that this affects a lot of different people. It isn’t always about swinging for the fences. Sometimes you have to make that dramatic 90-degree left turn to get somewhere.

I don’t think being in a position where you’re personally unaffected by the outcome, from a dollar perspective, is necessarily a positive thing.

I want to move into the world of AI, as I said we would. I want to first get your advice. We walked around Hyde Park together, and I’m looking at this market today as an investor asking where sustainable value is going to be generated.

It seems like models get commoditized faster than I thought. DeepSeek has shown that in the last few weeks. A lot of application-layer material seems pretty light in terms of just building on top of commoditized models.

How do you think about where sustainable value is being created in today’s market?

I had a conversation with some people from OpenAI and an application company, one of these coding platforms. What I was telling them is that the line between an infrastructure provider and an application provider today is super blurry.

In a weird way, if there is a brand-new application to be created, there’s basically no guarantee that OpenAI, Anthropic, Microsoft, or Google aren’t going to go and create it. All of these companies are motivated. They see a coding assistant taking off and say, “That’s fine, I’ll do a coding assistant.” They see people generating much better legal documents and say, “Maybe we should launch a legal ChatGPT.”

To me, that is part of what makes value creation in today’s environment really tricky to figure out.

I would say that the place where there absolutely is value creation is with companies that have a set of customer relationships, deliver clear value, and are able to embrace AI fast enough that a disruptor is not going to unseat them.

This is why I feel good about Snowflake. We are a data platform. We help people collect data, make sense of data, run analytics on the data, and run predictive workloads such as machine learning on top of the data. AI is a massive accelerant for the data life cycle and for data access.

Will somebody who starts with these foundational new capabilities as their starting point be better than Snowflake? Starting from zero, mind you, the bet is that they won’t be. I see AI clearly creating value. You see Salesforce doing this with some pretty cool work with Agentforce. They’re saying, “We have relationships. We are going to disrupt ourselves. We are going to be out there.”

But in terms of all-new value creation, I’ve got to tell you, it looks very murky in terms of long-term value.

There is value in product. It’s really important to understand that I think OpenAI will be successful not because it will always be the best and cheapest at creating foundation models, but because it has a product experience that, by some accounts, has half a billion loyal users. That is really hard to create. I think value endures on the product side.

Do you really think they’re loyal? DeepSeek got to number 1 in the charts in a day. You’re not going to switch from ChatGPT to DeepSeek? OpenAI isn’t stupid. If it can host DeepSeek to power some part of ChatGPT, it will shamelessly do it, as it should.

Why would you not switch to DeepSeek? It’s free, it’s as good as ChatGPT, and it comes with a bunch of additional features.

It’s a product. It’s not a model. There’s a big difference, Harry.

There’s a reason why Anthropic hasn’t done as well. It operates at the model level. It’s all of the additional things that go into ChatGPT: the ability to create images, the ability to upload something, and the ability to run a little bit of code. It’s a put-together product that I think has staying power.

I’ve had Sam Altman on the show, and he said, “We will steamroll startups.” He put on a 20VC jumper, which made me very happy. I was thinking about branding, and then the first thing he said was, “We’re going to steamroll startups.”

That’s not great for my venture fund. But my question to you is: if you were Sam Altman today, what would you do? We had the founder of Groq on the show last week, and he said that OpenAI has to open-source.

Sam is succeeding in having a consumer product that is approaching Meta and Google scale. Some of that is because of the incredible publicity machine that Sam and OpenAI have, but which was the last company you heard about that got to 500 million users without really paying for advertising in recent years?

That’s not a social network. What is happening is truly remarkable.

There has always been a little bit of mystique around OpenAI and its closed-source models. It’s also becoming clear that they are good at misdirection in terms of where they tell people their hard problems are. It’s good that people like DeepSeek went and busted some of those myths.

But I would separate the success of OpenAI the company from what’s going to happen tomorrow. Obviously, there are other things that they want to do, such as betting big on AGI. There’s a new company in the mix.

One point that Sam made to you that I completely agree with is that I think it is terrifying to be a startup building on top of OpenAI.

There are two points where I’d be concerned. Number 1 is when you look at NVIDIA beneath you in the stack. To what extent do you worry about NVIDIA moving into your segment and becoming a competitor?

I have to have the same worry about the CSPs. I always have to be watchful about things like that.

AWS has Redshift. Microsoft has Fabric. Google has BigQuery. Oracle has its data-warehousing platform. You should always be worried about being a little mouse in the land of giants. If they sneeze, you might get blown off. I get it.

On the other hand, I tell people that it’s OpenAI and Anthropic that have created the best models on the planet for the last 3 years. It’s not Microsoft, Amazon, or Google. Gemini is good, but let’s face it: at best, it’s a fast follower.

The thing that I tell people—and you know this better than anyone else, Harry—is that a startup with product-market fit is a living, breathing thing. It’s magic. It’s not as if you and I know how to create it with a few words and ideas.

Snowflake is a manifestation and realization of that magic. It is hard to create a data product by copying. You have to keep innovating.

Databricks has emerged over the last 5 years as a very credible player in the data space. More kudos to them. But there are things that we at Snowflake are very proud of and that are very different from how Databricks operates. We need to continue to innovate and be up there.

Money doesn’t buy you amazing foundation models. It doesn’t buy you Snowflake.

My question on Nvidia was one side of that. The other side you mentioned was Databricks. I spoke to friends, board members, and investors, and one of them said that Databricks might potentially be ahead in AI workloads, with MosaicML and MLflow, and with direct model hosting.

How do you think about that? Do you think Snowflake is behind in that respect?

We should separate machine learning from AI. I joke to people that all of the machine-learning people rebranded themselves as classic AI. They were first to market, that was their sweet spot, and they’ve been good at it for a while.

We were very much in catch-up mode. But AI is a much younger field. It didn’t really exist, let’s call it, pre-ChatGPT.

I feel incredibly confident that we are not only cutting-edge but ahead of them on a number of fronts. We have thought methodically but moved very quickly to make AI a reality at Snowflake, in everything from how you do data transformation and data engineering better to how you make sense of unstructured data and reliably get to structured data.

We are putting all of this into an agentic framework called Snowflake Intelligence that is coming out.

When it comes to AI, I feel very good about where Snowflake is. That isn’t an issue. I have work to do in terms of getting the word out and getting more and more customers. We have amazing marquee names—Amazon, Elevance, Bayer, and a ton of others—that are using Snowflake AI in production.

I feel good about where we are, and it’s a fast-moving space. The thing I pride myself on, and the AI team as well—many of whom I’ve worked with for a long time—is that in fast-moving spaces, my attitude is: I can run as fast as anyone else. Bring it.

You kind of remind me of Nik from Revolut. When you’re talking, I’m thinking, “You know what? I don’t want to fight you. You do you. I’m going to support you. I’m a believer. Keep going.”

Which structure do you think aids innovation and aids winning better: being a super-late-stage private company, like Databricks, or being a public company, like Snowflake?

Innovation is not an option. It’s just something we have to do.

Do I have more constraints than Databricks, being private? Absolutely. It’s very clear that they’re buying a bunch of business that they don’t have to worry about things like free cash flow. They have doubled the number of salespeople that we have.

But as I also tell people, it’s easy to lose a lot of money. It’s very easy to be uncalibrated about spending money. You should be careful what you wish for.

Do we operate under constraints? Absolutely. But that’s what innovation is about: how can you drive in the face of constraints?

DeepSeek is yet another illustration that having rich uncles is not always a good thing.

Where have those constraints helped you, and why have they hurt you?

With AI, for example, we said we weren’t going to be unreasonable about how large the team was. I didn’t ask for a blank check. We said, “We know what we need to do. We need to focus, have clarity around the things we need to build, and be very good at doing it.”

With a modest investment, we were able to catch up very quickly in AI. I think having those constraints drives clarity. You’re not trying to do everything, and you’re not trying to chase every possible unscalable business.

Constraints can be problematic when there are dramatic market reactions, because people can overinterpret changes that happen quarter to quarter in the numbers. In some ways, the Snowflake stock journey—the ups, the downs, and then the ups—is a reflection of the scrutiny that a public company faces.

In an ideal world, I would be able to tell my team and investors, “We’ve got this covered. Don’t worry about it. Everything will be great.” But you absolutely have to worry about it, because people are making mortgage payments based on what their stock is worth.

You get into externalities that are difficult to manage. It raises the bar on operating responsibly. That’s something we had to learn quickly last year.

Being a public company means you have to worry about dramatic reactions, because they can end up having a bunch of second-order consequences.

If you could, would you rather take the company private and remove those constraints?

No. I think the visibility is a good thing. You don’t get that anymore when you’re private. It provides liquidity, and being able to grade yourself is an important part of the process.

It’s very easy to fool yourself into thinking things are going to be great when you don’t get any feedback. We’re in the world of almighty dollars. They are well calibrated. Either you make money or you don’t. Either there’s free cash flow or there isn’t.

I think that dose of reality is helpful. Even if you look at the competitor we talked about, for a few months it was all about, “Amazing AI is going through the roof.” I haven’t heard much from them about AI recently. All of a sudden, it’s, “Amazing SQL is going through the roof.” It feels like a bit of three-card monte to me.

Being a public company forces you to be open, show your work, and do that all along the way.

I want to touch on enterprise adoption of AI. I was sitting with a bunch of CEOs at a summit the other day, as painful as those can be, and they said, “We’re still waiting for you people from technology land to deliver much ROI.”

Do you think we’ve seen excitement followed by a plateau, as we did with self-driving, or will we see exponential acceleration of adoption within the enterprise? Will it be a very smooth hockey stick up and to the right?

I think it’s going to be gentler in terms of growth. On the other hand, I can tell you very confidently that AI is creating value today, and it will continue to create enduring value.

Many things that used to be incredibly hard for you and me to deal with are easier. If a CEO is telling you they’re seeing no value from AI, I would question how much they know about it.

Let’s get into examples. What are the obvious, easy ones?

I was at Davos and did 30 meetings. I absolutely used things like dictation and transcription tools to write notes for myself. I also handwrote a whole lot of notes.

Someone on my team said, “25 pages? Really? I don’t want to read this. How about you give me a summary?” I just took all of my notes, put them into Claude, and said, “I want a 1-line summary per meeting.” Out came a beautiful, concise summary. It was magical.

Similarly, we have internal chatbots that have access to structured data. They can handle questions that would otherwise require you to click around in dashboards and provide a lot of context. Again, it’s incredible value.

Davos is an interesting collection of the world’s biggest CEOs. Were there any enduring takeaways for you from the sentiment you saw there?

I took a very utilitarian view of Davos rather than focusing on the bigger picture. It was my first time, and I was terrified by the idea of spending 5 days away from work, especially with the board and earnings coming up. But it was well worth it—an amazing number of meetings with an amazing number of key people from companies.

The message I heard from CEOs was much more, “Help us create utility. Tell us what is possible.”

When I talked to them about what we can do with unstructured data—being able to create a chatbot on a document corpus quickly—or with structured data, they got it. Then I told them, “Now you get to mix and match them and create agentic platforms where you can access a whole lot of related information in one place.” People were excited.

When I talked to them about automating parts of underwriting by bringing together all of the structured and unstructured information needed to underwrite insurance for a building, people said, “Oh my God, that is amazing.”

You’re correct that we need to continue focusing on utility, but I didn’t face a lot of rampant skepticism about AI. I just whipped out my phone and showed them the 5 stupid and fun things I had done with AI the day before.

In terms of the speed of innovation, it feels like we’ve never seen incumbents innovate at the speed they are now. I’m friends with Scott Belsky, the former chief product officer at Adobe, who recently moved on, and everyone has moved so much quicker.

The whole point of TaxJar was that incumbents were super slow. They’re not slow anymore. What’s happened? Have you ever seen incumbents move this fast?

I think it’s the result of 3 cycles of disruption. All of us remember our history. We all know that no one wants to be the person known for the IBM deal with Microsoft.

People understand how dramatic a shift that was. Companies like DEC disappeared. We took over SGI’s buildings. They were the darlings. The first buildings Google took over in Mountain View were the SGI campus, with that funky purple SGI Plex.

For about 5 years, I used to tell everyone who joined my team, “SGI built these buildings,” and then I would look at them to see if they understood the point. We learned from that.

I would actually say that mobile was a big platform shift in which incumbents did pretty well. Mark Zuckerberg went after the mobile web for Facebook and then said, “Forget it, I’m doing an app.” Google did the same thing: it took search and moved it over.

I led my team for 5 terrifying years as we took mobile monetization from being 10% of desktop to being 100%. I would say that this generation of companies had already become smart about platform disruptions and understood that they were going to be a big deal.

That’s why you see these companies making crazy investments into the future. They see something, and they have the money to do it. This is why Facebook became Meta. Mark doesn’t care that augmented reality fell flat on its face. He says, “Bring it. I’m on to AI now.”

I think that is companies truly learning from the past and innovating hard.

Where does this go? I hope you’re my friend, so please advise me.

I’m seeing Mark invest $65 billion in data centers, and we’re seeing the $500 billion announcement of Stargate. I know there’s equity and debt involved, and it isn’t a straightforward deal, but we’ve never seen money thrown at anything like this.

This is an arms race in AI. Where does it end?

With a bubble bursting, as all bubbles do.

There was a funny and insightful conversation between Ben Thompson from Stratechery and Nat Friedman where they discussed whether this is a good bubble—the bubble in the 1990s that eventually laid the fiber for the whole world, which Google, Facebook, and you and I took advantage of—or whether it’s the dumber Webvan stuff that just burned money in the early 2000s trying to deliver groceries, and we had to wait 15 years for Instacart to show up.

Let’s face it: you and I don’t know. If the bulk of those investments are going into things like power and buildings, you can say that they’re creating infrastructure for the world and something good will come out of it.

On the other hand, if the investments go into rapidly depreciating hardware, that is value disappearing in a puff. I think it’s early.

But back to your point about what we do: there’s still scope for innovation. An OpenAI product cannot possibly handle every workflow that exists. What you and I need to do is identify the things that are still ripe for disruption and where value is being created.

I think Harvey is a great company. I don’t think it’s going to get disrupted by OpenAI. Finding niches like that and investing in them is our formula for success.

Money doesn’t always buy you everything.

Speaking of money not always buying you everything, the street is saying that the issue with Snowflake is growth. To what extent can you buy growth, and how do you think about an M&A strategy going forward to add the extra 10 points of growth the street wants?

First of all, $3.5 billion going on $30 billion sounds pretty cool to me. Obviously, we have to sustain it, but it’s not a bad place to start.

What we’ve done, and I said this in our last earnings call, is significantly expand the aperture of what Snowflake is taking on. We played at the gold layer and said, “You do analytics with Snowflake, and, oh, by the way, we do a little bit of machine learning.”

We’ve gone from that to saying we will help with data ingestion, data engineering, analytics, machine learning, and end-user access with AI. The aperture just got a whole lot wider.

We have a team that is not only driving things like getting more of the analytics market, which is our sweet spot, but also disrupting other very large segments of the data space.

Absolutely, we will continue to look at companies. But an inorganic, unrelated acquisition strategy is not something I’m excited about. I think it would be a distraction.

Snowflake is a product-led innovation company. We are not a private-equity shop. There’s nothing wrong with being a private-equity shop; it’s just a different skill set. That’s not us.

A lot of our growth has to be driven by product innovation. Will we do smart acquisitions here and there? Snowflake spent roughly $150 million on Neeva, and I think it has more than paid for itself. Those are the kinds of things we should be doing.

What line of revenue is small or insignificant for Snowflake today that, in 7 to 10 years, will be a dominant line of revenue?

AI, 100%. I think it’s going to be a big deal. It offers the opportunity to disrupt business intelligence as we know it today, and it could be more direct-to-consumer.

I would actually say that the thing that can turbocharge growth for Snowflake is people building on Snowflake. We have companies like FactSet, JPMorgan, BlackRock, Fiserv, and Zscaler building data applications on top of Snowflake.

We let them do some pretty nifty things when it comes to bringing together their own data and data from their customers to create these data applications. I would say that is small today, but I fully expect it to become a massive overall revenue opportunity for Snowflake.

The really cool thing is that you go from being an expense item to actually being part of the top line of these companies. That creates much better partnership dynamics. I want to be able to say, “We make money when you make money.”

If we expand that to the model landscape, I look at it and say, “I don’t know where this is in 5 to 7 years.”

Do we have a world of many specialized and verticalized models, or a world of fewer, generalized, very large models? I’m intrigued, especially given your unique perspective from your Google experience.

I remember reading that people historically thought search would eventually be verticalized, and they didn’t anticipate these huge, multi-purpose behemoths really owning everything.

This is a really good question. Unfortunately, my answer is going to be: predict early and predict often.

If we deconstruct Google Search and why it came to dominate, first and foremost, Google made a bunch of deals to become the default search engine with a whole pile of players—Yahoo, AOL, and then soon Firefox and all the PC manufacturers. It was a very deliberate strategy to become the center of search.

Microsoft was asleep at the wheel for a lot of those early years. That’s how Google got started. Organic marketing was part of it, but honestly, it was a small part. It was that gravitational pull that enabled Google to knock out every other vertical.

It will amuse you to know that Live.com, a predecessor to Bing, had better image search. It had the best image search; Google did not.

With something called Universal Search, Google basically said, “Oh, you’re looking for images? We’ll put them right on the main search page. You don’t need to go anywhere.”

That became the vector by which Google conquered pretty much every other vertical, whether shopping, videos, or Maps. Through that central, sticky property, Google conquered the consumer world.

If you look at AI, it isn’t obvious to me, especially in the enterprise, that there is a single entry point. On the consumer side, I know you disagree, but I think ChatGPT is becoming that entry point.

If I were a betting person, on the consumer side, to the extent that there are variations in models and specializations, I think that will accrue to the incumbent. The incumbent is 100% ChatGPT and OpenAI.

I think entry points for other models will be fragmented. At some point, I used to worry about whether there would be a single enterprise entry point that would conquer them all. I still worry about it. I think there is still that opportunity.

But on the enterprise side, I absolutely see there being all kinds of specializations. In the last 50 years, there hasn’t been an equivalent of Google Search in the enterprise.

I also love the historical analysis of Google’s distribution strategy through partnerships. I didn’t know it was to that extent.

All of us think Google was an immaculate conception: one fine day, we all decided we should use Google. Which partnership had the largest impact on Google’s distribution ability?

Yahoo and AOL early on. Those were the entry points. AOL didn’t know how to do search, and Yahoo didn’t think search was important.

Does being first really matter?

One hundred percent. If I remember correctly, we paid AOL more money than we made from it. It was like $100-plus million per year.

Why were those the most valuable users?

You have to talk about both. The Google founders are amazing. They made a bunch of incredibly smart business decisions.

When you go back to those days and reflect on them, is there anything you take from that experience now, as the CEO of an incredible $3.5 billion-revenue public company, and think, “I can take that. I can learn my lesson here”?

Relentlessness. I still remember that every discussion with Larry and Sergey used to be exhausting. They never ended. They just argued with you about every single topic.

But good things came out of that. You examined every possible angle, and you pushed yourself really hard.

I also joke to people that by the time I became head of ads, working with the legal team was a whole lot easier because they had been battered around by Larry and Sergey pushing the envelope with things like Google Books and YouTube, which were legal nightmares early on.

That is my lifelong takeaway: pressing hard, being first-principles-driven, and remembering the Dutch auction. What other company has done a Dutch auction for an IPO? Google did it.

The amount of energy they brought into every discussion, and the relentlessness of driving toward the truth and the right business outcome, is absolutely something I carry forward to this day.

Listen, I could talk to you all day. I’d love to move into a quick fire. I’ll say a short statement, and you give me your immediate thoughts. Does that sound okay?

Sounds good.

How do you feel about founder mode? It’s eulogized in the Valley. You’re a CEO, and you’ve also been the founder of Neeva. How do you feel about founder mode?

I think it’s just a shortcut for describing effective people. Anyone can be that.

You have a 23-year-old and a 25-year-old, and you have a great relationship with them. My brother just had his first baby, so I’m an uncle for the first time. What advice would you give my brother, who has just had his first baby?

In my humble opinion, parenting is 90% presence and 10% luck.

You’re done. Sounds like you worry about the books that you want to buy, the cool software. I’m like, the best gift that you can give to your child is you yourself just being there.

What do you believe that most people around you disbelieve?

The possibility of change. People assume that too many things cannot be changed, including things within themselves. We put ourselves in straitjackets in terms of what we can do and what the people around us can do.

I had a great statement: “The heaviest things in life are not iron or gold, but unmade decisions.” What unmade decision do you think about most?

I followed my instincts generally. It was crazy for me to come to the Valley because I was a successful researcher in New Jersey. It was crazy for me to quit my first company when I was running a team of 120 people and go from that to being a software engineer.

I don’t have many regrets. I have a wonderful family and a spouse who let me make these crazy decisions. We moved house, and she was very against it. It was much closer to our friends.

I just do things, and I fall flat on my face very often. I pick myself up and move on.

You can be CEO of any other company for a day. What would you be CEO of?

I love my job, but Snowflake.

I’m knocking that question back to you. What’s your favorite brand, and why?

I have to admire the aura that Benioff brings with Salesforce. It’s not just the brand; it’s the brand in the marketing. I think he’s a remarkable human being.

Do you think brand and marketing are important for Snowflake today?

Absolutely. Too many of our own customers who love Snowflake think of us as, “It’s a very good warehouse. I like that.”

We absolutely have our work cut out for us.

What’s the hardest thing about being CEO of Snowflake at the end of the day?

You’re dealing with people. You want to be a good human being, but you also want your team to succeed.

Making sure that you’re driving while still being a good human being is hard. It should be hard. It’s not something you should ever take lightly.

Final one. You’re interviewed by a lot of people. You ask questions all day—to team members, employees, kids, family, and everyone else. What question does no one ask you that you think they should?

One interesting question could have been, “What are you most proud of?” I don’t think anyone has asked me that.

What are you most proud of?

Probably being a good dad.

Can I ask how you define being a good dad?

As I said, my wife and I focused very much on presence and love. We were there when they needed us most, and we set good examples for them.

Both of us worked. My wife has never missed a day of work in her 30-year career—not one day. To me, it’s about setting a good example, being there, and helping them become thoughtful, independent people.

She’s never missed a day of work? Does she want a job? I’ll hire her. That sounds fantastic.

Please, hire away.

The other side of that is, “You have a cold? Really? You can go to school. It’s all fine. You look fine. That fever? Not a fever.”

Sridhar, I so appreciate you doing this. I so appreciate our friendship. Thank you for being so open. I’ve loved doing this.

Harry, this was a fun, fun, fun conversation. Thank you.