- 6 hours ago
Category
🗞
NewsTranscript
00:01So just in the last couple of months,
00:03you had a big milestone.
00:04You raised $5 billion in equity at a $134 billion valuation.
00:10We have the, you know, we're kind of lucky
00:12because we have an opportunity to talk to you
00:15as you're making decisions, and maybe you've already
00:17made them, about how to allocate this big windfall,
00:21if you will.
00:22Can you walk us through, you know,
00:24and you named several things that you want to do,
00:26that you would plan to do with that,
00:27that was giving employees exits.
00:30There's development of your products.
00:32You mentioned Lakebase in particular,
00:34which is your database purpose-built for AI.
00:37Accelerate Genie, your conversational AI
00:39that lets people have conversations with their data.
00:42You mentioned AI research, a handful of other things.
00:45Tell us about how, give us, kind of draw us a mental picture
00:48of how you're thinking about prioritizing
00:51that important milestone, $5 billion.
00:54Yeah, that's a great question, and thanks for having me.
00:55I'm excited to be here.
00:57So, you know, we're just seeing acceleration
00:59in all of our business.
01:00Like, I really mean it.
01:01Like, we see acceleration in our core business,
01:04and we're like, why is it?
01:05Why are we seeing acceleration in the business?
01:07We slice it by Europe, by United States, by Latin America,
01:10Asia, everything accelerated.
01:11We slice it by different clouds.
01:13No matter how we look at it, it's just, you know,
01:15the business is growing.
01:16So that requires you to invest, you know, to keep up.
01:19So we're investing in the core business.
01:21So that's the data platform, the security, and all of that.
01:24So that's the huge investments going in there.
01:27On the AI side, I think that the biggest, and maybe we'll get
01:30into it later, is that, you know, I think the biggest opportunities,
01:32I think AGI is here.
01:34So I think we don't need more intelligence.
01:36We need context for the AI.
01:38The AIs without context are useless.
01:40So that's our Genie product.
01:41So, you know, there's a big research team that just does research
01:44on reinforcement learning and advanced AI techniques to make that better.
01:47How many researchers will you add?
01:49You know, I mean, we're hiring hundreds, and they're very expensive.
01:52You know, so it's almost like thousands.
01:56It feels like it.
01:58Like, you know, other employees that you would otherwise hire.
02:01And then, you know, they need GPUs.
02:03You know, so it's, there's a lot of investment goes into that.
02:06So we're very excited about that.
02:07Happy to talk about that.
02:08That's the thing I'm most excited about.
02:10And then we're getting into new markets.
02:12Because what's happening, basically, is that the thing that people call
02:15SaaS apocalypse, you know, people interpret that in extremes.
02:19But really what's happening is that the barriers to entry for other
02:22people to come in and compete in any business in software is just lowered.
02:26So we're getting into new markets as well.
02:28Talk about what that looks like.
02:30What new markets?
02:31Where?
02:31Yeah, I mean, we launched lake watch, which brings us into the
02:34Security market.
02:35Yep.
02:35And the security market is itself going through a big
02:38Revolution.
02:39Because what's happening is that, I mean, we all heard it with mythos.
02:42But actually, it started before mythos.
02:44Yes.
02:45You know, it's already the case that the number one attacker,
02:49According to hacker one's top list, is, you know, this person
02:53Called expo.
02:54But if you look behind it, it's actually all automated agents.
02:56And it's been dominating charts.
02:58Even before mythos, they found, you know, they used openai's
03:02O3 model to hack the kernel of linux.
03:04So, like, this has been going on.
03:05So there's an onslaught of agents coming, attacking these
03:08Businesses.
03:09So cyber is a big risk.
03:10And the existing businesses that do security protection, they're
03:14Sort of in this legacy model where you have to ingest certain data.
03:18It's very expensive.
03:19They're not going to be able to keep up with the agents.
03:20Right.
03:20So, you know, we got into that market.
03:23Now, maybe we would have gone into that market anyway.
03:26But with ai, it's compressing almost 10x the time it takes for us
03:30To build that.
03:31It's making it a lot easier for you to come in and try to disrupt.
03:35Yeah.
03:36Okay.
03:36Yeah.
03:36For multiple reasons, right?
03:37Writing that software is ten times faster now, thanks to ai.
03:41So we can just produce the competing software very quickly.
03:44Yeah.
03:44We could not have done that five years ago if it wasn't for ai.
03:46So that's like one.
03:47But the second is that the ai threats themselves have changed the landscape.
03:51So the way the incumbents were actually approaching security, you need to change those assumptions.
03:56So that also always opens up an entry point for the existing.
03:59And then thirdly, it's not, you know, specific to this time.
04:02We don't have a security business.
04:03So we don't have an innovative dilemma.
04:05We don't have to protect some revenue.
04:07Yeah.
04:07We can just, you know, launch something and we can, you know, it's sort of your
04:10Margin is my opportunity, as bezos used to say.
04:12Do you have a target for the number of people?
04:15Because you're going to have to hire researchers, engineers.
04:18You're going to have to build up your capability in that area.
04:20You can't just rely on agents and mythos.
04:23No, we've staffed up the whole team.
04:25How big do you envision that becoming?
04:28I mean, unfortunately, I think it's going to be in the hundreds.
04:30But, you know, the way you want to operate in this ai era is actually a very small team
04:33that can move very fast.
04:34It's actually preferable.
04:35So actually, you know, so if it was a thousand, it would be a terrible answer.
04:39That would be, we would go nowhere.
04:41You know, I think it's going to get to the hundreds.
04:43But you want to keep it small and fast and nimble and execute quickly.
04:46Because in this ai era, big teams slow you down.
04:50Because it requires coordination between lots of people.
04:53And that then becomes the actually, you know, slowest or lowest common denominator for progress.
04:59You want to unleash the agents and let them go fast.
05:01You've talked very eloquently about how the idea that many businesses over-indexed early on on chatbots.
05:08We sort of went a little nuts on that.
05:10Yeah.
05:12You've obviously thrown your, you've thrown yourself your resources behind this age of agentic AI.
05:22Curious about how we protect against the downsides.
05:25Harvard business school put out some research that talked about how agentic AI, particularly when we humanize it,
05:32Can result in shifting accountability away from individuals, reducing review quality, eroding trust.
05:38They talk about brain fry.
05:40I mean, it's a little bit of a doomsday scenario, but how do we protect against that?
05:45How do we, how do we avoid the risk of over-indexing on agentic AI?
05:49Buy our new security product Lakewatch.
05:51Just kidding.
05:52Just kidding.
05:53Just kidding.
05:53But actually only half kidding.
05:56So, you know, these things are now merging, right?
05:59Yeah.
05:59So we were a data and AI company.
06:00We were focused on data and AI.
06:01We were not doing security before.
06:02And people would tell us like, oh, make sure that no one breaks into the data and make sure that
06:05the data and the AI is.
06:07But now people are saying, hey, how do we know that the agents are also not doing something weird?
06:11So actually whole markets are collapsing.
06:13So us getting into security is not just happenstance.
06:16That market is kind of collapsing with the data and AI market that we used to be in itself.
06:20So all those things that they say is absolutely true.
06:22So, you know, you unleashed agents.
06:25Well, should agents act on behalf of you?
06:27How do you know that they're not going to do something that they're not supposed to do?
06:30So, you know, I think that ultimately this is why I think actually humans are going to continue to be
06:33in the loop for a very long time.
06:35You know, a lot of people talk about like, you know, AI disruption is going to, you know, displace jobs,
06:3920% unemployment.
06:40And so on.
06:40I actually think it's not at all the case.
06:42I think that we're going to, we need to be firmly in the loop because we need accountability to your
06:46point.
06:46So we're going to ask, you know, if there's a, you go to the doctor and they say that, oh,
06:50you know, we have to amputate your arm.
06:51You want to know who, who, like, if that was a mistake, who approved it?
06:55Oh, some agents.
06:55You're, you know, sorry.
06:57They're like, oh, I lost my arm.
06:58Like, that's not cool.
06:59So, you know, so I do think that there needs to be like, okay, well, it was actually signed by
07:03this doctor.
07:04Did they look through the results?
07:05Did they look through the traces?
07:06Did they understand what was going on?
07:07Why did they sign off on this?
07:09Same thing with our kids.
07:10We send them to school.
07:10We don't like, oh, you know, they were trained in the wrong stuff.
07:12And now your kid just went crazy because some agent thought of that.
07:15Okay.
07:15I'm okay with that.
07:16No, we want to know who was the teacher.
07:17Who was the person in the room?
07:18Why didn't I oversee this?
07:20Why didn't they?
07:20So I think we're going to have humans in the loop and we need to have this.
07:23And it comes also oftentimes down to the legal issue of, you know, who took the legal risk here.
07:29So this is going to continue.
07:30One way to think about this is just to look at the past.
07:33Yeah.
07:33You know, we could solve self-driving kind of airplanes a long time ago.
07:37You know?
07:37But if I told you all that I'm going to give you like a huge discount, 90% discount, and,
07:41you know, you can go with Ali's airline, we just have, you know, one slightly drunk pilot
07:46that only has like 50 hours of, you know, flying experience, but it's really cheap, none of you
07:50would get on it.
07:51You would say, oh, that's terrifying.
07:52But, you know, we already have these autopilots in the planes.
07:55So why do we still want humans in the plane?
07:56Because we still want someone to sit there and we don't want to just delegate everything
08:00to the, you know, to the sort of bots to take over.
08:03To the bots.
08:03So, yeah.
08:04I think that's one of the focus areas with our security product.
08:07Yeah.
08:07And I don't think it's solved, by the way.
08:09If you, I was joking when I said just buy our product and it will solve itself.
08:11This is going to be a journey we go through together as humanity, as AI comes and permeates
08:16society.
08:17Yeah.
08:17Just like when internet came and there was all these, you know, Nigeria letters and fraud
08:20and all of those kind of things, we learned how to deal with it and we figured out how
08:24to handle payments.
08:25You know, I was a kid in college and they were saying, no one's going to ever swipe a credit
08:28card online because, you know, that could just disappear.
08:31Now we only shop things using, so we figure out techniques to deal with it.
08:34We'll have to go through that together.
08:36I want to talk about numbers a little bit.
08:39Bloomberg being Bloomberg in a Bloomberg.
08:42You talked in February about your AR being at about 5.4 billion for Q4.
08:48We're well past Q1.
08:50Can you update that for us and give us a sense of kind of where the AR is now?
08:54Yeah.
08:55You know, my finances would kill me if I do that live.
08:57But, you know, there's no quiet period when you're a private company and you have no
09:02plans to IPO, as you told my colleagues earlier.
09:04This is what I keep telling my finance team.
09:06But, you know, they still give me a hard time after these meetings.
09:09So, let's just say that's significant acceleration that we never would have thought we would have.
09:14Acceleration from the 65% YOY increase.
09:18So, it was faster than 65% in Q1 over the year earlier.
09:23Yes.
09:23I can reveal that much.
09:24And you already got me into trouble.
09:25And I can tell you that.
09:27You're welcome.
09:27Yeah.
09:28Yeah.
09:28And also, I can say this.
09:29It's also when you look at different product lines.
09:31Right.
09:31Or by different regions.
09:33Or, you know, we have many of the AI companies that are our customers.
09:35Like, OpenAI has shared that they're, you know, a customer of Databricks.
09:38They use Databricks to, you know, 800 million users to make sure that they're, you know, nothing
09:42crazy is going on.
09:43You know, they use us for that.
09:45So, it's back to the kind of safety thing that we discussed.
09:48But, you know, you say, okay, well, what if we remove some of the...
09:50Maybe it's just a few big AI companies that are using us.
09:53Anthropic, you know, use Databricks.
09:55If we remove those, still the same thing.
09:57Acceleration in the whole business.
10:00So, why is that?
10:01Why is the business accelerating so much?
10:03I think one of the main reasons, and this is maybe one thing that
10:06People are now sort of thinking through, which is the agents are
10:10Now starting to behave, you know, they're starting to actually act
10:13Through, throughout the ecosystem.
10:15So, a month and a half ago, i shared that we've now seen more
10:19Queries come to Databricks by agents than we see by humans.
10:23So, it's like this tipping point where, you know, it was 100%
10:26Humans, i would say, three years ago.
10:29And now we already have over 50%.
10:32Database is launched on Databricks.
10:35It's over 81% now agents.
10:37And these agents are just way faster.
10:39So, they just launch lots of stuff.
10:40We're consumption-based pricing model.
10:42You know, our revenue goes up when that happens.
10:44So, that's one of the main drivers behind this.
10:47You have a lot of visibility into which LLMs, engineers,
10:53Programmers are using.
10:54You talked in the past about how people, engineers, will just switch
11:00LLMs very, very quickly.
11:02And they're becoming better.
11:04And we're creating more tools that enable them to do that.
11:06And we're able to measure, kind of, well, who's using which.
11:10What have you observed over the last six months?
11:12What has surprised you in terms of LLM use?
11:16And who's Ascendant?
11:18Which ones are Ascendant right now?
11:20Well, I mean, you know, without picking, they're all partners and customers of ours.
11:24So, I have to be careful.
11:24You keep getting me into trouble here.
11:27That's my job.
11:28Yeah.
11:29Great job.
11:30But, you know, I would just say this, that the biggest shocker for me is that it used to
11:35be that, you know, GPT-4 comes out.
11:37It could dominate for six months or so on.
11:39And then, you know, just not long ago, the saying was that, you know, frontier models only
11:43last three months, a quarter.
11:45So, even if you have the world's best AI model, it'll only last, you know, a quarter.
11:50Okay.
11:50That's kind of surprising.
11:51So, the most shocking thing now is that that number is not one month.
11:54Like, you get the right, for one month, like, you know, maybe I should say, like, when Gemini
11:59came out November last year from Google, it was like everybody was blown away.
12:02Everybody switched to Gemini, you know?
12:04But that's, right now, people are not talking about Gemini, right?
12:07Who are they talking about?
12:08Well, I mean, like, you know, people are, okay, you're very good at your job.
12:13Look, people are very excited about Codex, but people are also very excited about Cloud,
12:16you know?
12:17And I think Gemini 3.5, when it comes, we're going to be very excited about that, too.
12:19Here's the second.
12:21So, number one was this shrinking time of how long a model.
12:24In some sense, this question even won't matter soon.
12:26You know, what was the hottest model yesterday?
12:28What is it today?
12:29Well, who cares?
12:30Tomorrow is a new day.
12:31So that, but the fact that this shrunk was interesting.
12:34The second thing that's really been a big surprise for me is that it used to be
12:39that the developers are changing models because they want the latest model.
12:42The models just speak English or any other natural language.
12:44So it's very easy to switch between them.
12:45You don't need to worry about, you know, oh, I'm changing everything.
12:48No, it's just, you know, it's a probabilistic thing.
12:50Every time the answer is different anyway.
12:51So you might as well just pick the cheaper, smarter thing that came
12:53Out just yesterday.
12:54But this was developers.
12:56Now we're seeing that corporations, i.T.
12:59Departments, leadership is freaking out about the cost.
13:02Costs.
13:03You know?
13:03And they're saying this is insane.
13:04Because, you know, when you see these companies, you know, get
13:07$30 billion of revenue just in one quarter out of nowhere, well,
13:10someone's paying for that.
13:11So on the other side, they're freaking out.
13:13That, you know, how do we put cost controls in place?
13:15Right.
13:15So one thing that has a new trend that has emerged.
13:18So we put this thing out.
13:20It's called the unity ai gateway, which lets you basically, it's
13:23Kind of a, we did it for security reasons.
13:25Because back to our security.
13:26We said, look, you need to control which ai's your company is using.
13:29Yes.
13:29That was our purpose.
13:30Like, you know, to make sure they're not using some model you
13:32Don't want.
13:32To track the data.
13:33Did they swipe a credit card?
13:34But what the main thing people are using it for is actually, you
13:38Know, can i put a budget on it?
13:40And the thing they're doing is they're using more and more open
13:43Source models.
13:43Right.
13:44And what they're doing is they're using this new paradigm that's
13:46Called the advisor model.
13:47So you use a small, not so intelligent, very cheap, very fast
13:51Open source model.
13:53And then you give it a tool.
13:54Because these ai's are very good at using tools.
13:56You know, they can go type stuff and they can go into word
13:58Processor.
13:58The tool you give them is access to a smart big model with a big
14:02Brain.
14:02And you say, hey, when you get stuck, use the smarter model.
14:04And this way, you know, people are on this increase of
14:07Exponential increase in costs.
14:09And they're able to completely flatten it.
14:12Even though usage is increasing, you know, cost is
14:14Flattening.
14:15To the extent that cost is an issue and to the extent that
14:18You have visibility into the popularity of chinese-based
14:20Models, which on a per-use basis are relatively cheaper than
14:26Those in the u.s.
14:27Are you seeing people gravitate toward that, especially as
14:30Cost as they become more price sensitive?
14:32People don't want to talk about it.
14:33Okay.
14:34But yes, absolutely.
14:36You know.
14:37It's chinese models.
14:38Open source models are absolutely dominating.
14:40You know.
14:41And
14:41When people have the choice to use them.
14:43Yeah.
14:44It's like, hey, would you want to pay 30 million dollars or
14:461 million dollars?
14:47Well, you have to use this chinese model.
14:49Okay, fine.
14:49You know.
14:50It's like, you know.
14:51It might be like, oh, I don't want to use a chinese model
14:52And we have concerns and so on.
14:53It's like, well, it's going to cost 30 million versus 1 million.
14:56It's like, okay.
14:56I take that.
14:57What does that say about the efforts that we have
14:59That the u.s. has led globally to sort of hamstring chinese ability to advance in artificial
15:06Intelligence?
15:07Yeah.
15:07Is that effective?
15:08Has it been effective?
15:09Well, I mean, open source is just this force.
15:12Right.
15:12You can't fight it.
15:13You know, it's just and these are like you put the weights out there and the whole world
15:16Start using it.
15:16By the way, it starts reappearing.
15:19So, for instance, cursor put out the model composer.
15:24Well, it became clear that they actually trained composer on top of a chinese model.
15:27So composer would not have been able to be trained without kimi model that came out of
15:33China.
15:33So there's also this happening.
15:35So then some of our customers, like, we do not want to use a chinese model.
15:37We're okay using composer.
15:39You know.
15:40It's like, okay.
15:40Here you go.
15:42So, yeah.
15:43So I do think that, you know, in the end of the day, it's kind of like, you
15:46know, money talks, you know.
15:49Sure.
15:50Speaking of money, earlier today you were on Bloomberg Television and they stole my thunder
15:55by asking you the IPO question.
15:58And you said something that was maybe a little controversial about this being a terrible
16:04year for an IPO.
16:05Don't tell Elon or Dario or Sam.
16:10What is, when is a good year?
16:12Yeah.
16:13When is a good time?
16:13What are the ingredients needed for you to say, this is a great time?
16:18Yeah.
16:18Look, I just think that there's, if you have, like, big dislocations happening and there's
16:23big things happening, you know, election year, there's, you know, there's macroeconomic
16:28indicators that, you know, could be great, could not be great.
16:31We don't know right now.
16:32You know, there is uncertainty, you know, with energy and other things going on in the world.
16:38And then there's just mega IPOs.
16:40And no one's ever raised that much money that any one of them, the amount that they're raising
16:44in the IPO has never been raised ever by any company before.
16:48So, you know, I think it's better to wait and see.
16:50And, you know, and maybe I'm wrong, actually, and it'll be a great year, but I don't lose
16:53anything by waiting.
16:54You can go later.
16:55And, you know, go at a time where, you know, you just have more stability.
16:58Like, as everyone knows, businesses prefer predictability and no big surprises.
17:03No surprises is good.
17:04Do you think that you'll need to go out and raise again before an IPO given 2026?
17:11You seem to be putting that off the list.
17:13We might raise again.
17:14I think that, you know, there's going to be much more interest after, you know, these three
17:18gigantic private companies go public.
17:20You know, the way it works is that private capital people who, you know, have to invest
17:23in private companies, they can't deploy it in public stocks, right?
17:26So I think there's just going to be a bigger supply in the private markets for, you know, companies
17:31like Databricks, Stripe, others.
17:33So, you know, I don't think there'll be a shortage of capital for us.
17:36So, you know, I don't think there'll be a shortage of capital for us.
Comments