The Twenty Minute VCLeo Aschenbrenner's Situational Awareness Blows Up | Moonshot AI Raises $3.5B at $35B
EVERY SPOKEN WORD
80 min read · 16,393 words- 0:00 – 1:44
Intro
- NANikesh Arora
The long term average intelligence is gonna be free, and the average intelligence will get smarter. Land, permits, energy, the, this is the thing that is gonna get priced for the next three to five years, and-
- HSHarry Stebbings
Hold up. Nikesh Arora joins us in the studio? Oh, yeah, baby. Palo Alto Networks CEO, $280 billion company for this incredible session today, and we discuss Leo Aschenbrenner's situational awareness imploding, sad face, Airtable being acquired by Bending Spoons for $1.285 billion, sad face again. They were worth $11 billion before. Anthropic's model breaching three companies. Oh, God, when does the security problems end? That and so much more in this incredible conversation.
- RORory O’Driscoll
Absolutely right on trend, absolutely wrong on portfolio construction. It's almost like it was inevitable. If you bought in in April, May, or June, you've been wiped. We did a hedge fund, but it appears it wasn't hedged, [laughs] right? And we lost all our money in a week.
- NANikesh Arora
I think it's a bit of a gold rush moment. I think every consumer app will get rewritten in the next five to 10 years.
- RORory O’Driscoll
In the face of insatiable demand, all things are possible.
- JLJason Lemkin
You know, Palantir can do it. They came back from 15% growth four years ago. Why can't you do it, kids? Work harder. [upbeat rock music]
- HSHarry Stebbings
Team, it is so good to be back, and we have the one and only Nikesh joining us. Nikesh, thank you so much for agreeing to join Rory and me and Jason.
- NANikesh Arora
I am, uh, a little apprehensive watching Jason and Rory there, uh, in their full glory, so let's see how this sort of plays out for-
- HSHarry Stebbings
Yeah, we, we, we have Professor O'Driscoll in the corner.
- RORory O’Driscoll
Oish.
- 1:44 – 15:27
Airtable Sold to Bending Spoons for $1.285B
- HSHarry Stebbings
[laughs] But we're gonna start on the news of the day, and the news of the day is Airtable, one of the big names from the last decade, has been bought by Rory? European-based-
- RORory O’Driscoll
Bending Spoons
- HSHarry Stebbings
... Bending Spoons. The company was doing $485 million ARR, growing 20% year on year, ultimately at a $1.285 billion, um, acquisition price. It's not-
- RORory O’Driscoll
Enterprise value
- HSHarry Stebbings
... it's not the outcome that everyone quite wanted or expected, [laughs] but it's what we have today. Rory, why don't I hand over to you first? I'm sure you've got some perspective.
- RORory O’Driscoll
Well, you know, we're all gonna do the Airtable side of the analysis, but just worth pointing out the Bending Spoons side. Um, I'm buying stuff at 2.8 while I'm trading in the market at naught to 10 times revenues. They're gonna do this all day, every day. And I think, as we said a few weeks ago, one of the big advantages is they're in the game with capital and a traded currency to hoover up a whole bunch of this stuff. So on their side of the table, totally get it. You know, obviously on the Airtable side, you know, there are a lot of comments around is this giving up? Is this a reflection of where SaaS is? If we weren't anchoring off the 11 billion, it would be a great price. If you said someone set up a company 10 years ago, grew it to 450 million in revenues and sold for 2 billion plus or minus, they'd be like, "That's an amazing outcome," right? But of course we all anchor off the 11 billion 2021 price and it feels like a lowball. But I think it's, it's still a great value creation achievement. And you know, we have a talented entrepreneur on the show with us, and we need to start with that. It's a great outcome.
- JLJason Lemkin
You know, can I ask Nikesh one particular question on it, Harry, if it's okay? 'Cause-
- RORory O’Driscoll
I was gonna ask you too, Jason
- JLJason Lemkin
... I have a lot of interesting thoughts on this deal, whether Airtable matters in the age of agents and AI. But I mean, Nikesh is also one of the best deal makers out there. The shocker to me with Airtable wasn't the price, 'cause I think it's low market. The shocker to me is no one else stepped up. No PE firm, no Thoma Bravo, no Vista. It's 20% at 50 mil- 500 million with some AI dust going on it. Um, there... Do you think there was even another offer? That's the, the... I just assumed someone would outbid them.
- NANikesh Arora
And you're asking me that because I'm like-
- JLJason Lemkin
You're the deal m-
- NANikesh Arora
... rule number one private equity guy?
- JLJason Lemkin
You're a deal maker par, par excellence. [laughs]
- NANikesh Arora
[laughs]
- JLJason Lemkin
We just happen to have the k- a king deal maker on the show.
- NANikesh Arora
I look, I, you know, I, I met Howie a few times. He's a great guy. He's done, built a great business. I think, uh, I don't know, to Rory's point, there's a bit of founder fatigue here. I think, uh, he's been through a lot of sort of ups and downs in terms of sort of internally and in the market, but he's got a good product. I think the broader question, which Jason hits on right, is, you know, what is going on in the SaaS marketplace? You know, you see disloca- uh, is this a pricing dislocation or is it a fundamental change in the long-term growth rate that people expect out of SaaS? If it's a fundamental change in the long-term growth rate people expect out of SaaS, then the multiples are right. Now, you know, and that's I think where the market is grappling with this. I think the people you mentioned, uh, Jason, the, the PE guys, they might have a full roster of stuff they'd like to sell to Bending Spoons as opposed to they'd like to buy against Bending Spoons.
- RORory O’Driscoll
[laughs]
- NANikesh Arora
So I think they, they-
- RORory O’Driscoll
Totally
- NANikesh Arora
... might be caught in the sort of demand and supply problem right now. They, you know, they have a lot of inventory.
- JLJason Lemkin
It just worries me. I'll let you... We just, I, I... 'Cause you see more deals than we do, uh, on the, on the acquirer side, right? I mean, we, we see it on the, on the target side. It's just-
- NANikesh Arora
Well, we-
- JLJason Lemkin
When you look at Francisco Partners-
- NANikesh Arora
We stick to unprofitable long-term growth comp- businesses. [laughs]
- JLJason Lemkin
Yeah. Say, say that again. Sorry?
- NANikesh Arora
So we stick to long-term unprofitable, currently unprofitable-
- RORory O’Driscoll
[laughs]
- NANikesh Arora
... long-term growth businesses. But yeah, no, I see what you mean.
- JLJason Lemkin
No, I get it, and, and that might be the answer. I just, with F- I mean, when Francisco Partners raised 22 billion to do deals sort of like this, right? Uh, you know, Thoma Bravo was like, "We're looking for AI-infused B2B companies." You know, we could pick at Airtable, but it, it did do that. It did infuse AI workflows in others, and it, and I don't know that it's growing, but it did that, right? 20% at 500 million in AI workflows isn't nothing. I just would have, for all the founders out there looking to be picked up, this one would have seemed to me to be just above the fold. Like, the, the... They should have leaned in on this one, not the one at growing 8% and shrinking because it was destroyed by AI. And it's cash flow positive, and it has plenty of cash. So all the boxes you check, right, are sort of there as an attractive target, and yet no one, no one outbid them.
- NANikesh Arora
Well, hey look, Bending Spoons did buy them, so clearly somebody saw value there. Uh, not everybody's seen it. But Jason, I want to go back to something you said around AI infusion. I'm a little wary about this AI infusion stuff, and this is what I talk to my team every day from an operator perspective. I say to them, you know, "Are we Mercedes? We're trying to sprinkle a little bit of AI in our car and say I have a little bit of AI. [clears throat] Are we Tesla? Are we making sure that our car will drive, you know, the next 10 exits by itself and might have to grab the steering wheel once in a while? Or are we building a Waymo?" And the question back to you is, did Airtable do a bit of a Mercedes action, little Tesla action, or a little bit of a Waymo action? 'Cause my biggest fear is a bunch of people are out there in their garages getting funded by Harry and Rory, and they're gonna build Waymos of the future, and we will be busy putting lipstick on the pig.
- 15:27 – 21:10
Leo Aschenbrenner's Situational Awareness Blows Up
- HSHarry Stebbings
[laughs] really rode the wave, so to speak. He did it with 4X leverage. Um, and in the past week it, it kind of came crashing down, and then Ken Griffin and Citadel bought his public book for a reported $16 billion. Ken has made out like a bandit, reportedly making about $3 billion [laughs] on the back of it in a very short amount of time. Um, how do we think about this? He was the wonder kid of the AI wave.
- RORory O’Driscoll
I mean, I think he was absolutely right on the tre- I mean, absolutely right on the trend, and remains to date right on the trend. In other words, the data just last week about CapEx absolutely supports his memo. So conceptually, right on the trend, and then absolutely wrong on portfolio construction. If you accumulate a portfolio of high volatility stocks with 4X leverage, the math makes it clear your probability of getting wiped out once is just very high. This is as simple as that. Absolutely right on trend, absolutely wrong on portfolio construction. It's almost like it was inevitable.
- HSHarry Stebbings
I'm really sorry. How do your investors let you get to that place?
- RORory O’Driscoll
Because you made them 10X last year, and you probably don't question anything. [laughs] Uh, he did amazing, you know? And which of us really, let's ask ourselves honestly, when someone makes you a 10X, do you sit there... Is your first response, "Yeah, but what can go wrong?" Or is like, "Ooh, can I put in more money?"
- JLJason Lemkin
[laughs]
- RORory O’Driscoll
You know? That's what happened. And, and my wife said when I was talking about it, she said, "I don't wanna see any schadenfreude." She was like, "You know, there's a lot of schadenfreude, laugh when the poor guy..." I feel sorry for him. It was a tough call to have to go through that, just to put it out there on a human level. I mean, he was clearly wrong on the bet, but that was a brutal week.
- HSHarry Stebbings
Will he be okay? Uh, it says that he's still managing both the private and the public, but he's got his Anthropic position, and then other people are like, "Oh, no, no, no, lawsuits are coming and it, it's not gonna be okay." Is he gonna be okay?
- RORory O’Driscoll
I think he'll be fine. I, I promise you, he's not gonna be caught at the same place again. [laughs] So, you know, good news, he's, he's learned a lesson and he's gonna live, he's gonna survive to sort of live to-
- HSHarry Stebbings
Another day.
- JLJason Lemkin
Odd, I suppose-
- RORory O’Driscoll
Yeah.
- JLJason Lemkin
But in a way- I mean, Harry, it's your job to mentor some of these younger kids like, like Leo.
- RORory O’Driscoll
[laughs]
- JLJason Lemkin
So I think maybe you c- you could step in. What, he's 26 or something like that?
- HSHarry Stebbings
He was 25. Dude, uh-
- JLJason Lemkin
25, yeah. So I think you've got... It's time for you to, to t- become the elder statesman in the industry and start mentoring him on leverage, when to lever up to 4X, when not to.
- HSHarry Stebbings
[laughs]
- JLJason Lemkin
Uh, I, I mean, in all seriousness, I mean, I think someone smarter-
- RORory O’Driscoll
Harry's busy with the, the-
- JLJason Lemkin
I've read it
- RORory O’Driscoll
... drooping plants and, and, you know, to-
- HSHarry Stebbings
Yes
- RORory O’Driscoll
... expanding homes with, you know, multiple people living in the same house. Don't, don't bother him.
- JLJason Lemkin
I assume his LPs or his investors knew this was a highly levered fund, right? Um, am I wrong, Rory? I mean, if they know it's 4X levered, then they know there is black swan, uh, issues, uh, when there's short squeezes and others, and I don't think the, uh, the, his investors should cry if they knew how, how it was playing. Um, I mean, I, my limited experience as an LP in funds with leverage, not quite this much, is you, you, you know it's not free.
- NANikesh Arora
I have a feeling, I, I have a feeling Zelpy's didn't lose any money. If you are up 440% and you go down-
- RORory O’Driscoll
Yeah, from day one
- NANikesh Arora
... from $45B, you're back to where you started. So I think it's fine.
- RORory O’Driscoll
The interesting thing about that, not quite, 'cause actually this is where I think it could get a little hard. It all depends on timing, right? 'Cause th- they hedge funds are weird. If you came in early, you made a ton of money, and then you lost two-thirds of what you made, and you still made money, right? Brutal comment, if you came in in the last six months, you might have been wiped 80%, right? 'Cause hedge funds, unlike, unlike pe- venture funds, people come in at different times at different bases.
- NANikesh Arora
True.
- RORory O’Driscoll
So I think the real... And I think, I thought even perhaps Jane Street had put in some money recently, but a bunch of people had put in money, and if you bought in in April, May, or June, you've been wiped. And then to your comment on, those pe- I mean, the, the litiga- I mean, I think fundamentally, yes, he will be fine, and there's a lot... Larry Fink, who, you know, founded and runs BlackRock, had a blow up early in his career. There's lots of people who had blowups early in his career. Uh, Nikesh has worked for one of the most aggressive risk-taking human beings on the planet at SoftBank. He's seen ups and he's seen downs. So you can survive-
- 21:10 – 30:18
Anthropic's AI Models Breach Three Companies as Cyber Threat Accelerates
- HSHarry Stebbings
of OpenRouter on the show on, on Friday, Rory. So there we go.
- RORory O’Driscoll
Cool.
- HSHarry Stebbings
Uh, I'm excited for this next topic, 'cause with Nikesh, I think we've got the most prescient person. Anthropic's models-
- RORory O’Driscoll
Yes
- HSHarry Stebbings
... breach three companies too. This is obviously on the back of the OpenAI and the Hugging Face debacle. Really? Anthropic breaching three models also? Uh, is this just, like, the most epic beginning of, like, a bull run in security?
- NANikesh Arora
Like, first and foremost, this is all flex, right? They all wanna tell you how good their models are, how powerful they are. So it's kind of bizarre, because normally, if you end up breaching somebody's infrastructure, it's not a good thing. But we're all saying, "Look, look at these models. They're so powerful." So fine. Granted, they're all very powerful. Uh, I think the first things our friends at Anthropic and OpenAI should have done, [clears throat] which I've told them, is point your models at your own sandbox to make sure your sandbox doesn't have any zero-day vulnerabilities, and make sure your sandbox is your first catch as a flag, flag exercise. But they decided to give a target to say, "Go out in the wild, persist, take as long as you want, and go capture a flag." So fine. You know, we have these models which have these capabilities. I think the, the challenge we have from a cybersecurity perspective is we're filing vulnerabilities which would take us days, months to find. Uh, the average time to patch a vulnerability, a zero-day vulnerability found in the wild is fifty-five days. Just think about that. These things are finding vulnerabilities in split seconds and then turning around and building an attack on the back of that. So I think the, the fundamental speed at which cyberattacks will happen and need to be defended changes. And this is good for us. It's kinda like, you know, the sound of revenue. But I think from a more fundamental perspective, y- I was thinking this capability is gonna show up in six months. I think I said that with you, Harry, and it showed up in four months. And I think in two, three months from now, open source have distilled all these capabilities, and we'll find open source models out there, which you can fine-tune if you're an attacker to actually do this on an autos basis. So it's gonna change the game.
- RORory O’Driscoll
How are your enterprise customers reacting? Because this feels to me like the mother of all of... I mean, security sells on fear, and this is terrifying. So what, what are you seeing in the enterprise customer base-
- NANikesh Arora
[laughs]
- RORory O’Driscoll
... when this is knowable?
- NANikesh Arora
[laughs] Sure. Uh, look, the good news is the, the flex that Anthropic did with Mithas has every CEO talking about Mithas. I spent eight years trying to get CEOs to talk about cybersecurity, couldn't get them to do it, and Dario did it in one fell swoop. [laughs] So this is good news. He's got everybody, like, all hot and heavy about Mithas and the capabilities of Anthropic and how these models are gonna go attack your infrastructure. I've never had so many CEOs call their CIOs and say, "Are we ready? What's gonna happen to us?" Well, the answer is you're not. Because what being ready means is that I have no vulnerabilities, either in my code, any vendor that I've got deployed in my infrastructure, any open source I'm using. That is fundamentally not true. Now, we, we found 14,000 vulnerabilities in open source in the last 14 weeks, testing open source packets. So there's a bunch of stuff that's being used out there. So, A, every company has vulnerabilities. They gotta figure out a way to patch them. B, these models will figure out misconfigurations. If you've left the door open, if you've got a device configured wrong, if you've got a piece of software configured wrong, and there's tons of that out there. Not every IT person building infrastructure or configuring infrastructure is a genius. There's misconfigurations. All these things need to go away. So at the base, as the face of it, a lot of organizations are gonna have to go fix a bunch of these vulnerabilities and misconfigurations. On the flip side, even if you've fixed most of these, the bad guy has just gotta be right once. So he's gonna find one, or she's gonna find one, to get into your infrastructure. The question is, what is your time to detect and respond in that circumstance? The average time to detect and respond is four days. How are you gonna get it down to a minute? So it's not a fear problem, it's a capability problem. It's an infrastructure readiness problem, and now it's come to bear, so it's time to pay your taxes.
- JLJason Lemkin
Yeah, I mean, that's the last sentence. I mean, you're right, we can argue fear versus thing. But what you're basically saying is the security infrastructure that you had a year ago is wholly unfit for purpose in the next year, and you, Mr. Enterprise Buyer, are gonna be buying a whole load more stuff. Are you gonna be the weakest link when these capabilities are everywhere?
- NANikesh Arora
Yeah. You said it so well, Rory.
- JLJason Lemkin
I mean, I'm-- It just feels so good. [laughs] I don't mean to be melodramatic.
- NANikesh Arora
I think you should be on a podcast talking about how people need to buy more cybersecurity.
- JLJason Lemkin
I'm just gonna buy the stocks, but...
- NANikesh Arora
[laughs]
- HSHarry Stebbings
Like, hang on, Nikesh, can we get him some swag? I mean, Jesus, he's like, basically-
- NANikesh Arora
That's my role. We don't want swag
- HSHarry Stebbings
... come on, Jason. Give, give, give Nikesh a hard question. You're, you're the one global-
- JLJason Lemkin
Hold on. Can I wanna ask Nikesh as my-- I wanna ask him as my cybersecurity therapist. I had two Fable issues-
- NANikesh Arora
Yeah
- JLJason Lemkin
... and they're internal security, but I'd love to get your... And you can make fun of me for this. Like, I, I've got-
- NANikesh Arora
No, no
- JLJason Lemkin
... I pretend to have a thick skin. I don't, but I, I, I, I love criticism.
- NANikesh Arora
I already figured I don't have a thick skin the first 30 seconds of this conversation.
- JLJason Lemkin
Okay, good.
- NANikesh Arora
So-
- JLJason Lemkin
So I'm building an app for the Saastr community, it's called Saastr Connect, to help with recruiting. The, that-- details don't really matter, but it's, it's the biggest thing I've built, right, myself in this era. And, um, and I'm-- I-- A-among other things, I've got a Google Doc, it's called Jason's Gems. It's my ideas on how to improve it. It's just ideas. They're just scratch notes, okay? No one's seen them, that's not ready, just a side doc I keep. So the other day, I, I went into Claude and I just turned on the Google Drive connector, since it's one of the three primary connectors. This is not esoteric. This is not third party. And I never knew, it went in, scanned all my docs, found Jason's Gems, found the ideas. Fable then went and changed my code and my algorithm without telling me. Never got a notice, never was told, never was in a change log, never was anywhere. I only found out later when the agent flashed conflict with Jason's Gems when I was trying to fix something else. I mean, I'm not saying it's terrifying, but how can organizations deal with this fact when an LLM will go out and change your core code, your core corporate OS, without even telling you? What if you have a thousand employees doing this? And this is Goo-- this is just... Oh, and, and the, the, the hack to-- pa-part I didn't tell you is, the way it did it is it MCP'd in. So I had, I had Google Drive to Claude to Fable to MCP, and so it was able to do with whatever it wanted, probably thinking it should implement Jason's Gems, but it shouldn't have, and it never asked me, and it, and it never told me it did it. Is this scary? Is it not scary? Is this Orwellian? What, what happened to me?
- NANikesh Arora
It's the Wild West.
- JLJason Lemkin
[laughs]
- 30:18 – 35:15
How Palo Alto Networks Defends Against Unknown Bad Actors at 19PB/Day
- NANikesh Arora
bad. Every cyberattack happens because-
- JLJason Lemkin
You didn't know
- NANikesh Arora
... you didn't know it was bad until it got into your infrastructure. So the question becomes, if you know it's a known bad, you stop it at the perimeter. If it gets through, how quickly can you find it and stop it before it creates harm or damage? So from a cyber perspective, you wanna be in the perimeter business. You wanna be on as many perimeter endpoints in the world as you can, because that becomes a sustaining business. The more perimeters I'm on, the longer my tenure for my business is. So I'm on endpoints, I'm on, you know, devices, I'm on servers, I'm on firewalls. I'm protecting the perimeter for multiple infrastructure set of components in the world. That's good. That's kind of good. Now, the question is- How quickly I find known bad will change using AI, right? There's a concept of data classification. You had to write static rules. Guess what? An LLM can suss it out much faster from a content perspective. You know, we've, we track every malicious website in the world. Can AI tell me it's a malicious website much faster? Yes, it can. So the, the sort of the ingredients of my s- perimeter security will change using AI. The act of stopping things in line will still be needed. So when people tell me, "Oh, open AI is gonna eat my lunch," or, "MIT Technology Review is gonna eat my lunch," guess what? There are no perimeter security scenario, which means I still need to block the bad guy. They need to be the ingredient in my product. They're not gonna take me out of business. 'Cause people have all kinds of infrastructure on the perimeter. The other part is, if you wanna suss out all the bad stuff in the infrastructure and find out the bad actor, guess what? Imagine collecting all the enterprise data-
- HSHarry Stebbings
Yeah
- NANikesh Arora
... and running an LLM right on it. And say, "Find me all the abnormalities, find me behavior that you've never seen before." Now, I'm ingesting 19PB of data a day. Think about it. 19PB of data a day of enterprise data to look in it for anomalous behavior. I have machine learning techniques, I have static techniques, I have rules that I look at it. Guess what? I'm gonna throw some LLMs in there just for fun to see what they find. Now, if I can find the unknown bad actor in your infrastructure much faster using LLMs, I can detect it and block it. Right now we run at one minute-
- HSHarry Stebbings
Okay
- NANikesh Arora
... with machine learning. This is a good thing. The only problem is I only have 1,200 customers who've bought and deployed it. I need to get the rest of the world to go buy it and deploy it, so that's the second half of the problem. The third part is there is stuff which is new, which does not have any security guardrails that have been built. Agents. The world is talking about agents. We can have a whole episode, 90 minutes on about what are agents, what is really an agent, how do you give agency, and how do you control an agent? People tell me they have agentified stuff, but then I ask them, "Does it actually have agency?" Like, "What does that mean?" I'm like, "A Waymo has agency. It can drive you into the wall without human intervention. This is a bad problem." But most people haven't actually given agency to their agents, so they're running glorified workflows which are seemingly agentifying things. But when you start giving agency to things, when piece of, of code can decide what happens next, we're gonna have a whole different conversation around how do you secure those agents? How do you build kill switches? How do you intercept them in line? How do you stop them from doing bad things? Just like Jason's agent, which did-
- HSHarry Stebbings
Yeah
- NANikesh Arora
... a bad thing and took Jason's gems, and now the whole world will find out what Jason's gems are. [laughs]
- HSHarry Stebbings
Totally agreed.
- RORory O’Driscoll
It, it's, uh, it's crazy. [laughs]
- HSHarry Stebbings
We said open, we said China. Uh, whether we have comment on it, Moonshot closes three and a half billion at a 35 billion valuation. And if it's free, you're the product. Moonshot's free.
- RORory O’Driscoll
I think one of the comment on the if it's free or the product is, is totally true, especially in the consumer side. The interesting thing here is I'm not sure how it's true. Put it another way, unless you're... W- one of the really interesting things about these open weight models is the impact they're having and the ability to be a drag on price for the US, you know, closed source frontier model companies. It's not as... And in the apps, if, if you're running the inference as well, then I get the model. The mo- You, you get the business model. The model is free and the inference is how you're making money. It's not as clear to me long term if it's possible to continue on a sustaining basis, offer open weight models without monetizing in some way. And we'll, we'll see if what it is when, when people like Reflection and Thinking Machine start, when these models start to happen in the US, it will be interesting to see, you know, what the business model is of which, of, of which an open weight model is a part, right? It's definitely not, "Hey, download it, have a go, and you can do whatever you want wherever you want," right? I mean, look, open source has evolved the business model of support, so it'll be just interesting to see, you know, what version of if it's free or the product emerges for these companies in the medium term.
- NANikesh Arora
I, I'm gonna give, I'm gonna give Harry a soundbite.
- RORory O’Driscoll
Good.
- NANikesh Arora
Average intelligence is gonna be free in the long term, and the average intelligence will keep getting better.
- 35:15 – 44:00
Energy Is the Real Bottleneck: Land, Permits & Compute Over the Next 5 Years
- RORory O’Driscoll
Nice.
- NANikesh Arora
Exceptional intelligence will be paid for.
- HSHarry Stebbings
Can you give me some tone with that, Nikesh? That was all monitor. I want like drama. Come on, you gotta deliver the soundbite.
- NANikesh Arora
I, I, I was watching somebody speak-
- HSHarry Stebbings
Give me more
- NANikesh Arora
... I was watching somebody speak the other day, and they said, "If you whisper loudly into the mic, people lean over and pay more attention." So I'll say it again.
- HSHarry Stebbings
Nice.
- NANikesh Arora
I say in the long term, average intelligence is gonna be free, and the average intelligence will get smarter.
- HSHarry Stebbings
Oh. [claps]
- RORory O’Driscoll
But do you, but do you think we'll rely less on frontier intelligence? We won't need, we won't need it?
- NANikesh Arora
Oh, no. We'll need exceptional intelligence. We need exceptional intelligence to discover the, the cure for cancer. We need exceptional intelligence to send rockets to the moon. We need exceptional intelligence to build a space data center. Those are exceptional intelligence tasks. They are still gonna require exceptionally intelligent people or exceptionally intelligent models, and people will pay for it because the outcome is so spectacular. I don't think you need to pay $6 a million tokens to answer a call saying, "How can I help you? I'm so sorry, your network connection is not working."
- RORory O’Driscoll
Agreed. Yes, customer support will not be using frontier models.
- HSHarry Stebbings
Maybe.
- RORory O’Driscoll
It won't be.
- HSHarry Stebbings
But, but I... My limited... Uh, listen, I, of course you're right over the long term, right? In the short term, customer support is requiring more and more tokens to do more and more sophisticated resolution, right?
- NANikesh Arora
And guess what? Those are the primary candidates that all these open models are going after with open weight fine-tuning saying, "I don't need hallucination, I need..." I, I think the, the part where we're sort of this is what I mean the capability intent gap is, I think there's a lot of work that needs to happen to go from building a frontier model or any model and taking that and making useful the enterprise context. The amount of effort that goes... The problem we have is like, uh... And sorry to go back to the Waymo example because, like, it's kind of, I think it's the most obvious one out there. It do- I, I drove in the first Set up Google self-driving car, I don't know what I was thinking, in 2009 when I used to work there. It was a Lexus with a bunch of cameras. It drove me from San Francisco to San Martin on the highway, and my hands were not on the wheel. And then I took the... It told me at 11:00 PM to take the, the wheel in my hands. I was driving to Corda wall, and I did. And I sort of was more relaxed about it, saying, "Oh, maybe it's just gonna figure it out when I make a wrong turn because it's so smart." Drove me. I was like, "No, dude, this does not drive when it turns off." So that was 2009. It's taken 14 years after that to get one with all the edge cases trained from machine learning perspective for us to rely on that as being the agency that we've given the agency that, to, to that replacement. So I don't believe we're gonna give 100% agency to use cases for some time. And for us to be able to do that, the amount of data collection and, and context we're gonna create is gonna be humongous. Basically, you have to literally take every edge case in customer support, get into your AI brain of your organization so that you can start relying on AI instead of the human. So you're getting 80% right now. You're getting 80% of customer support solved. All the edge cases are waiting to be solved with AI.
- RORory O’Driscoll
Then there is the, for a given app, how much of the value is purely in the model versus all the other thing. And you're right. For something like customer support, we, you know, you're probably paying 10 or 15% of the revenue you're getting for intelligence, and the rest of it is all the other shit it takes to make that intelligence actionable in the context of answering tickets. And one of the things we look at is just super interesting on the app level is the tokens as a percentage of total revenue, and it varies from, you know, the Salesforces, the... We were in Intercom, stuff like that, where it's, you know, some plus or minus 10, 15%. Obviously in coding and things like that it's 70, 80%, which means it's just raw intelligence and a mild harness, and those are just very different.
- NANikesh Arora
I think over the next three, four years, we won't be paying for intelligence, we'll be paying for compute through our nose.
- HSHarry Stebbings
I mean, speaking of paying for compute through our nose, we often get chastised for being too public markets focused or too Anthropic and OpenAI focused. Valar Atomix triples to $6 billion price as Sequoia bets on nuclear for AI. Uh, it's a three-year-old small modular reactor company, um, raised at $2 billion. Now Sequoia leading a round at six. Um, and specifically there's an Nvidia partnership to power AI data centers, which caused a lot of excitement for the company.
- NANikesh Arora
Harry, I met somebody who's got... You know, I was, I was talking to him, and he's in the, the business where they take, you know, chicken feces and turn that into methane and, and produce gas. And I thought it was like, "Oh, it's a cute project he's got running somewhere in the middle of the country." And then he told me that he raised money, billions of dollars, and he's selling the energy to hyperscalers. So anybody who can produce any energy source, it doesn't matter where you are, is right now trading in multiples because land, permits, energy, compute. This is the, this is the thing that is gonna get priced for the next three to five years. And I, I think it's almost like the question will become between Anthropic and OpenAI, who has more access to more compute in the next three to five years, and that's what people are gonna buy. It's very hard to find compute right now. You can also take all the free Chinese models you want. Where are you gonna run them?
- RORory O’Driscoll
Jason?
- JLJason Lemkin
No, no, for sure. I, I... It's just to, to Nikesh's point, I, I actually... I used to be a little bit in advanced energy storage in my first startup, and things like chicken manure didn't used to make sense. The- these, these models actually used to work-
- NANikesh Arora
It's literally chicken shit
- JLJason Lemkin
... So did cow, so did cows. I've even looked at some of these things. But the marg- the margins were so low, the IR was so low, but it... the business worked. Now, it, it, AI is, uh... There, there's such a demand for compute, everything works, including all types of nuclear, like Valar, right? Including chicken manure. You laugh, but like, I remember talking to a manure farmer doing this back in the day, and he's like, "Well, the best, the best we can commit to is 8% annual return if everything goes well." And it's just, you know, it's hard to get 20VC excited for, for, for, for those returns-
- NANikesh Arora
Harry, you never thought-
- JLJason Lemkin
... but maybe it's 80 today
- NANikesh Arora
... you thought you were gonna be talking about that on this show.
- JLJason Lemkin
Me neither. You brought it up.
- NANikesh Arora
[laughs]
- JLJason Lemkin
But, uh, but everything-
- 44:00 – 1:09:07
The $1 Trillion Token Question
- HSHarry Stebbings
and he's Irish
- NANikesh Arora
I, I'm not. I'm just realistic.
- HSHarry Stebbings
I, it says-
- NANikesh Arora
Yes, we have two Europeans here. [laughs]
- HSHarry Stebbings
Fucking unbelievable. Unbelievable. Uh, uh, is this not just another layer of companies which is dependent on, Rory, to your point, OpenAI and Anthropic continuing to go on their charge and hit their number? Like, we've never had an ecosystem that will be so dislocated if OpenAI and Anthropic do not hit their 2027 numbers.
- NANikesh Arora
I, I don't think so. Whether OpenAI or Anthropic hit their 2027 numbers or not is orthogonal to the fact that there is infinite demand for AI at this moment, and that infinite demand needs to be satisfied by compute. Now, whether it's OpenAI that builds the data centers or buys the data centers or pays for them, or somebody else pays for them, there is demand in the market. Look, if you think about what's going on, I still posit 70% of the compute demand for AI is being consumed by consumers who are getting a free ride. So maybe there'll be reallocation. Maybe we're gonna have to give more compute to enterprises over time as they become better monetization capabilities. Or you'll find that eventually the promise of consumer monetization's gonna start showing up. We all talk about why can't an agent book my airline ticket and buy, make me a restaurant reservation, and these are simple use cases. I don't need to go, you know, solve cancer to get that stuff to work. That stuff's gonna work. When that stuff works, there's gonna be monetization opportunities on the consumer side. So I believe that y- y- you know, at first principles, there will be tremendous amounts of compute that will be needed to satisfy the AI use cases both on consumer and enterprise. Which player ends up monetizing them becomes a question for the markets to decide, and that's a timing question, no different than Leo's question. That's a question of who builds the capability and the, the, the, the services. You know, Google was not the first search engine. Uh, okay, I'm, I'm, uh, arguing against and, you know, just at one level, obviously, if you zoom out enough, you're right. But if you zoom back down, and I'm, uh, we notice in these discussions, I'm always the literal-
- HSHarry Stebbings
I, I see where you live, Rory
- NANikesh Arora
... yeah, I, I li- hey, dude, I'm just, I, I, you, you, you, you're running your $280 billion company. I'm just trying to turn $20 million into a bill- uh, $20 million investment into $100 million and call it a day. I'm, I'm a small guy. But the genuine comment is kind of mixing infinite demand for intelligence, and let's assume it's just enterprise now, 'cause I think you are right. The c- the consumer side is super interesting, especially for OpenAI, but let's leave it to a side 'cause it's just, we can only do one thing at a time. I think the question w- where it does matter is right now the assumption is 70, 80% of that demand gets channeled through Anthropic and OpenAI. In other words, 'cause they're 70% of the Google compute backlogs, they're 70% of the Amazon backlog. So in the short term, most... The, the, the market is assuming that OpenAI and Anthropic buy the compute, buy all the stuff that's further down the stack, buy the chips, and resell that, resell that intelligence on a frontier model basis to US enterprises. And if it doesn't happen that way, it's gonna take a... Uh, there's gonna be a pretty big dislocation.
- HSHarry Stebbings
Yes. It's perfectly possible that there is a public market dislocation because the table, the players at the tables might change. And that's great. That's called a buying opportunity, because that doesn't take away the infinite demand. It's extremely possible that perhaps this wonderful company called Moonshot, which we talked about three seconds ago, could be the model of choice, and that somebody's gonna take that compute, which is not gonna be used by frontier LLMs, and put Moonshot on it and sell it to enterprises at 10 cents a dollar on the-- 10 cents on the dollar for tokens.
- NANikesh Arora
Yeah. Moonshot is happy, NVIDIA's happy, enterprise is happy, OpenAI, very, very sad. You're right. That's, that's the dislocation.
- HSHarry Stebbings
But the question becomes, you know, what, where are the mark-- Which ones of these are the markets gonna support, right? Is the market gonna give you infinite capital to be able to build a compute because they believe you're the anointed winner? Or does the market believe that you're running it differently and it wants you to run differently? So I don't think the demand goes away. I think in all these conversations, one variable goes away when we run into this sort of technology shift, you know, infinite bull market. We take execution out of the picture.
- NANikesh Arora
Yep.
- HSHarry Stebbings
Doesn't matter. Every chicken manure company and every nuclear reactor company who says the words in a PowerPoint is gonna get funded by everyone because they assume flaws-- flawless execution. And you look around, and then poor Jason is looking at SaaS companies and saying, "Holy shit, some of them are not ex-execut- not executing as well as the others." So eventually, execution matters. And that's who decide the winners and losers in the market, not the shift of, you know, which intelligence is the best intelligence.
- NANikesh Arora
I mean, the best example of that would be two years ago, OpenAI was first and Anthropic was second, and now Anthropic is first and OpenAI is second.
- HSHarry Stebbings
And Google was written off.
- NANikesh Arora
And Google was written off.
- HSHarry Stebbings
There was a point that Gemini was non-existent. Google was written off. Now suddenly Google has the compute, the cloud sales, and Gemini. Yeah.
- NANikesh Arora
But still not the amazing open frontier model. Still not the coding agent. They're still not the coding agent.
- HSHarry Stebbings
Oh, your customer support agent is gonna be extremely unhappy because he didn't get a chance to answer it using the best model
- RORory O’Driscoll
Just kidding. [laughs]
- JLJason Lemkin
Well, to Nikesh's point, Harry kicked us off by saying, you know, "Well, have we ever had an ecosystem so dependent," right, "on the success of OpenAI Anthropic?" I mean, it is, but maybe to Nikesh's point, um, you know, so much has changed since we started the show, right? When we started the show, it actually seemed like everyone would benefit because average intelligence, or whatever term Nikesh would use, would permeate software, and that would be good enough. That has... Now the fro- the we- this is the revenge of the frontier, right? We may not care in a year what model we... Like, we need frontier models. We need the best, but we may not care who wins. We may not care who wins this battle. We may... This may all blow over, and it all mi- all may b- be about compute, and we may not... We, whoever wins, wins. Whoever wins, I'll plug in.
- NANikesh Arora
Jason, I think the models will get better and better, and the distinction between models will not be enough for you to decide to rip one out because I think the part which we will be build, we are starting to build and we will be building for the next two to five years is context. So think about it for a second. Like, when I run a simple firewall company, you know, a simple, complicated firewall company, you can stick any model you want. The model doesn't know why my customer's infrastructure is down. It does not know. Because my model doesn't know what product my customer is using. My model does not know what operating system it's using. My model does not know what the configuration of the customer is. My pro- model does not know why this happened the last five times with the customer. All that knowledge, all that learning is being captured by me in effectively vector DBs and in context learning systems, and that's what my team is doing. I have more people collecting context than I've ever had. It's kind of like the Waymo thing. I got people planted to think this is a tree. This is why it goes down. So as I built that organizational and sort of context, then I can stick any model I want on it, and the model distinction will not matter because the context will become as important or perhaps more important.
- RORory O’Driscoll
And you're clearly 100% tracking Satya with the kind of Microsoft comments recently on, you know, age companies, and it makes them too. Enterprises need to build their own value, build their own context rather than do it in frontier model, right? And that's-
- NANikesh Arora
Well, what I think it's, it's... Yes. You know, he's saying something different. I, I understand what he's saying. Uh, that's a different comment. Uh, mine is different comment. I think there's three parts to it. There is, there's the model, which is the raw intelligence, let's just call it that.
- RORory O’Driscoll
Yeah.
- NANikesh Arora
There is the context needed to answer your queries or needed to answer your problems, and then there's the context needed to train-
- RORory O’Driscoll
Okay
- NANikesh Arora
... that ecosystem. I'm talking about-
- RORory O’Driscoll
Yeah
- NANikesh Arora
... the context needed to train the ecosystem, which means I've got every customer case that ever happened at Palo Alto getting transcribed. So my model has, knows what is a good answer, what is a bad answer, right?
- 1:09:07 – 1:18:12
Scale AI Hits $1.5B ARR
- HSHarry Stebbings
Go Rory. 13-year journey, amazing outcome. Um, we have Visa cutting 2,600 jobs. Nikesh, you said it's not a jobs problem. Well, CEO of Visa says it's efficiency and shaping the way work gets done, so 2,600 people gone there. Uh, Whatnot raising at $20 billion.
- JLJason Lemkin
Can I, can I ask Rory about DroneDeploy? 'Cause it ties to the beginning of the conversation with deals-
- HSHarry Stebbings
Yeah
- JLJason Lemkin
... Nikesh, right? So that deal, the... What's interesting, so, so DroneDeploy was bought by Procore, right? Great classic software founder, founded by Twohi to, to do, um, software for real estate, dominated it. It had a great run, right? Growth slowed, right? Uh, most importantly, net new customer count sort of stopped growing. Growth slowed to like 17. So they make a big bet. They... And I'm not an expert on DroneDeploy, obviously. Rory is. But they buy a next generation platform, right, to use drones to accelerate this construction industry. And, and structurally, what's interesting, and I find these deals are always really stressful, okay? So Procore, uh, market cap is beaten down. It's gotta come up with $900 million, a lot of it debt, pay 11, 12X while it's trading at four. I find in the old days, I'm not saying that happened here, these deals are stressful, man. They are... It's, it's not, it's not Palo Alto Networks spending 0.01% of its market cap on some smart kids. This is bet the farm at a much higher revenue multiple. Ki- Doesn't have to work, but man, this is the big, uh, the, the big bet, right? And it wasn't cheap. I mean, you'll say it's cheap because you're on the board, right? But Procore's gonna think this is expensive to pay 12X when it's trading at 4X, right? And we started this on deals with Nikesh, and we started this on whether 3X to 4X for Airtable is a lot. Well, Procore is one of the ones basically trading there too. So is this deal, like, super stressful? Did, did you lose hair? Were people shouting and throwing things through the window?
- RORory O’Driscoll
So I'm not gonna speak for the acquirer 'cause I'm not in that side of the room. But genuinely-
- JLJason Lemkin
Yes
- RORory O’Driscoll
... one of the least stressful deals I've ever done because honestly, I would've been happy to continue. This was not a founder tired. I mean, I think actually some of the interesting lessons, there's about two or three interesting lessons here. First of all, when you have capital discipline and, you know, modest fundraisers, you're, you're set up for success, not failure. You know, none of... You know, we always raised below the price we sold at. We didn't raise a ton of money. We were profitable. We just... You know, it was a fine little company growing nicely. And then the second thing is, I think very important, the trend with our friend, not our enemy. I think some of these very basic SaaS companies, you look back and go, there's been a platform shift, and you're on the wrong side of it. When you're software that's enabling drones and robots, you're actually on the side of the future. And in fact, one of the lessons I learned having in- invested 10 years ago is in the physical world, AI takes a lot longer to happen. I mean, when it happens, it's amazing, but it's clearly... You know, I look back 10 years ago, I thought drones would've exploded five years ago. They're really starting to explode now as are robots. So it took a long time. So in fact, we were on the upswing of this feels really good, we're happy to hold, and then obviously we got an offer that made us do different. I don't wanna comment on specifics of the offer. But I think one of the ahas here is building companies is hard, and, you know, by being disciplined, by putting ourselves in a position... Uh, the, the founding team did an amazing job. Three founders all together, all still wildly actively involved. So no, it was genuinely not a stressful thing at all. It's like at the right price, you'll do this deal, at another price you won't. And for what it's worth, from a distance, I think it's interesting, super interesting for the other side too. I think actually market expansion is what you need to do in some of these spaces. You need to say, and, you know, probably Nikesh has done these kind of big strategic where you just say, "I need... You know, my thing is this big. I need to add the next thing my customer wants." And I think at some level, the customer wants not only to be told the accounting of his business project, but also the physical progress of his building project, and that's what things like physical inspection do.
- HSHarry Stebbings
Nikesh, at what, at what percent of market cap does a deal become a, a BFD, a big fucking deal, uh, a core strategic, this needs to work?
- NANikesh Arora
Um, look, every deal needs to work. We're not buying companies because we have money to spare or my shareholders think we should, you know, we should a lot, sort of lie and waste. And, uh, not... I think the hit rate requirement in us is more than a VC. I think in the last eight years we've bought north of 40 companies, and I would say 75% have worked, 25% haven't. Our, our largest deal Was a $28 billion deal, which probably is currently valued at fif- north of $50 billion. That one's, that one's gotta work. That one's kind of career defining move. If you take a company at $28 billion when your market cap is 200, and you spend 14% of your market cap or 16% of market cap and buy something, it better work. Um, now when you make that work, then you can... If the market gives you credit for making deals work. I said that in my earnings call, and they got all freaked out. And, and I'm just saying, you have to make the big ones work. If you don't make the big ones work, then you lose the license to, to run your business.
- RORory O’Driscoll
And that big one was CyberArk, right?
- NANikesh Arora
Yes.
- RORory O’Driscoll
Yeah, that was a great... Got it.
- NANikesh Arora
Rumor has it, agents are gonna be important. If agents are important, they're gonna need identities.
- RORory O’Driscoll
Amazing
- NANikesh Arora
And they need to be treated like privileged identities. So that's our thesis. Sort of simple.
- RORory O’Driscoll
Like all the best deals. One of my partners always said, "If you can't express it in a sentence, it's probably a bad deal. And if you can, it's probably a good one." Got it.
- JLJason Lemkin
As, as you've seen with my examples, you don't know what these agents are gonna do, man. [laughs]
- NANikesh Arora
In your case, you're just gonna restrict agent behavior, Jason. You're just gonna have to use your own brain.
- JLJason Lemkin
But they're so good. But man, they're so good. But it's so powerful.
- RORory O’Driscoll
Jason, to the point when... Actually, it kind of ties back to when Nikesh said, I'm just was reading some stuff last night that really does accord to what Nikesh has said, is what someone made the point, you know, if you can't trust them with what they're doing, you have to be very clear on who they are as an identity and where they're allowed to go. If you've got this, you know, as I said, this m- this kind of AI employee and you're not quite sure what they do, you've just got to bound the systems they can access very tightly. So I actually think... I, I totally get your point, Nikesh, that the, the ability to-
- NANikesh Arora
The only, the only danger, Rory, is if you take it to the extreme, that's called automated workflows, that's deterministic outcomes. If it's deterministic outcomes, we already had that technology for the last 20 years. So the question is, at what point in time do you let an agent think?
- RORory O’Driscoll
Do something. Yes.
- NANikesh Arora
That's, that's the big debate. Rory-
- JLJason Lemkin
But the flip side is it does a really good job... And listen, we have a, we have a, a... We don't have the perfect security profile, to your point, right? But even with what we have, which is probably o- one agent has about 1,000 rules, to your point, right? The rest probably have five, right? Or zero, right? But even with the zero to five, 99% of the time today, right, uh, it's, it's pretty... It's, since, since January, since the model's upgraded, pretty darn good. In the last couple of months, like, really good, right? So it, it's a trade-off, right?
- NANikesh Arora
It just takes one destructive example to sort of make it all unwind, you know? If you give it access to a bank account, let's see what it does. If you have HBO Finance allowed to write checks, I might want to have a conversation.
- RORory O’Driscoll
Speaking of people being wrong, I'm gonna say I was totally wrong on something. Scale AI, the fact that they've continued that business, uh, I would have thought the acquisition left them a husk. But I think it proves one of those rules that you kind of know but you forget, which is when you're in a great market and you have a product that can meet that need, e- even losing your top people, it's all fine. You know, they were selling data, data products to an insatiable demand for data, and I give them huge credit. They kept the thing going.
- NANikesh Arora
Hey, Windsurf sold to Cognition, right?
- RORory O’Driscoll
And they've...
- JLJason Lemkin
Yeah.
- RORory O’Driscoll
Exactly.
Episode duration: 1:18:22
Install uListen for AI-powered chat & search across the full episode — Get Full Transcript
Transcript of episode Q6kDZJ0xdSw
