AMD Buys Fei-Fei Li's World Labs for $8.2B | Meta Poaches MongoDB's CEO | Bessemer Raises $5.75B

We're definitely not investing at the right speed. We are either investing way too fast or way too slow.

Returns are going to be highly skewed. Variance is going to go up with AI and many of you will fail.

So, what do we have on the cards today? Well, Anthropic's draft S-1 leaks. We break that down. AMD, baby, they are in the market to buy. Fei-Fei Li's World Labs sells to AMD for 8.2 billion in stock. And then, ka-ching, Mr. Mark Zuckerberg decides to whack out the checkbook to hire MongoDB's CEO and pinch him for Meta's enterprise division. This and so much more in a blistering episode out today.

There are like 10 companies that can do $10 billion acquisitions and want to.

Money is a signal. Price is a signal. And price is sending a signal. Everybody go right here. It's a talent war in the most insane way right now.

I think open source has reached its maximum as a market share. I think it's going to keep going down.

Wow. What? If we're willing to trust them not to blow up the world, and they've already said they might, and it's only a 10% chance, I think we can trust them with the votes.

Ready to go, boys. I am so excited for the show today. A, we have a surprise guest in Mr. Jackman. So, Jack, thank you for joining us.

That's very sweet, Jack. Thank you. You know, don't worry, Rory will ruthlessly give truth to me today while we go through the news. But we have a lot of news to get through now. Number one, Anthropic's draft S-1 leaks. I think every venture nerd was so excited for this moment. 4.66 billion of 2025 revenue, 8 billion operating loss, 518 billion of compute commitments.

Genuine comment, not a single piece of information in that leak. Not a single piece of useful data other than one minor comment which I'll come back to. Like, the 2025 numbers long since spent, you know, the fact that the—I mean, lots of shitty bad takes. The whole 'oh my god, they're losing 40 billion' when 34 billion of it was non-accounting loss, the rest of it was exactly what you thought: 4.5 billion in revenues, 8 in compute expense, 5 in other expense, blah. The only interesting factoid in that was that two customers did 25% of the revenue, which means someone spent half a billion dollars on Anthropic last year, which is pretty impressive. But the point I'm making is the only thing that counts is not even Q1 and Q2 of this year, because I think everyone knows Q1 and Q2 is Anthropic killed it, right? What we said last Q, the Q3 numbers, which haven't been finalized yet, are the ones that matter.

Kind of flattened out. The Q3 revenue number is 90% of the data required to make a decision on pricing Anthropic, and everything prior to that, in particular 2025, is so leaked at this point that there was no data in the wider thing. It was fun but no data.

This is going to be one of the nice things about these companies going public is that I actually think the public market investors will be a little bit more long-term oriented than the private investors. I think like we are all hand-wringing month to month about what these companies are doing. And in a weird way, I think the public market investors will think a little bit more long-term, a little bit more structurally. You think about both, you know, Anthropic and OpenAI, and it's like each month it's like the narrative is just all over the place and it's so easy to flip like that. I think at least in my conversations with public market investors, they're going to be thinking a little bit more at least quarters, maybe even years. And so, you know, I agree with all of that, but I think people are going to mostly be oriented towards like what is the market structure of these things and hopefully it will be a little bit calmer.

I think that's true. Um, my only thought—and listen, this could be completely wrong—when I read all the press, right? Yeah. Like to Rory's point, we learned nothing, right? And to Jack's point, just getting the public investors excited to hold an epic asset long. But man, it really reads negative, you know, not just the risk factor that there's existential risk to humanity. I mean, that one is a little jarring to see. But what I mean is it may be like the Facebook IPO in that at least on the retail side, the negatives come to the top, you know, and we're all investors, right? So, we're all playing for the upside, aren't we? We're all, you know, why is Instinct worth 10 billion? Well, it's the upside, right? Um, it's just the jaw-dropping losses that, frankly, I don't think the four of us care about, right? And maybe a lot of long public markets don't care about. I just think the retail market, everyone that is anti-AI may see these negatives as a reason to sour. And I don't know if that can really drive the stock down. But I worry that the IPO will be—it's a minor worry—but I worry the IPO will be successful. They'll hit their number, whatever they ask for, it'll be more than oversubscribed, you know, 10x. But then a month or two in, with no real change, we may see a drift below the IPO price just because of the negativity, right? The massive spend, any hiccup, you know, the CDSs at Oracle, some default on data centers, and all of a sudden the negatives overwhelm the life-changing positives. That was my only thought, is we're going to see a lot of stuff like this right after the IPO, much more than we have before.

I almost think the people who are going to be big buyers of the stock are not an overlapping circle with the people who are, you know, dealing with all the kind of fallout from the noise and the negativity. I mean, so I know—I mean, brutal—and let me say it even more directly, right? 70% of America might think data centers suck, but they're not running Fidelity growth, and a small number of highly compensated managers are the buyers of their stock. Right. And retail people excited for the thing. Where I do think you're right, Jason, is it's not so much that directly the dislike of AI kind of impacts the stock directly, but the indirect comment is I do think there was a non-trivial political backlash brewing. I think it was super interesting to see the guy in Florida file basically—I, you know—basically sue to stop this thing, stop OpenAI and Anthropic from building models. There's going to be a whole load of 'you told me in a legal document that this thing was not safe, therefore I'm entitled to believe you, and if it's not safe you should stop,' right? So I do think indirectly you're right, is that because they have to spew out all these negatives just to cover their ass, right? Though I'm not sure who's going to sue you if the end of the world actually happens, you're probably pretty good on the litigation side. Um, but because they have to spew out all this negativity, people are going to be reading that and saying, 'Huh, this is the first S-1 I've ever seen that poses existential risk and where the product lies to me. Maybe I should, as the attorney general for the great state of fill-in-the-blank, sue these guys and say stop.' And I think we're starting to see that. So it will be way more of a public participatory moment. Everyone's going to have a goddamn opinion on this, perhaps in a way that everyone doesn't have an opinion on. You know, when a semiconductor stock goes public, 95% of the people can barely describe what it does. Everyone's going to have an opinion here.

Speaking of topics that everyone has an opinion on. I think everyone seems to have an opinion on Instinct right now and Instinct versus Meta and the future of agents and personal AI assistants. Instinct closes a billion-dollar Series C at 10 billion. Noah Shin, 23-year-old founder, amazing generational talent, raises a billion at 10 billion, takes on Meta and Alex Wang and Zach. What did we think, Jack? I think you've got a check from Benchmark.

This is my meat-and-potatoes view of the big winners today, right, is the ones that are winning today. And this is—and I'm a Meta fan. I haven't used Instinct. I'm hoping to get an invite, right? But I use—I have 22 agents. I'm all in on the agents, right? But the ones that win—this is an open question—are the ones we use all effing day long. So if you look at coding, even me, I'm running—I'm running my coding agents 10 hours a day now, right? I run—and legal, I got wrong in a sense because folks are running Harvey and Lora all like—like Harry's partner. You're like this. She gets home, she's on the sofa running it all effing day long. And there's lots of wins as VCs we're going to make for sporadic use. And this is just a question. I don't know. It's—it's not even a criticism. Will we run Instinct, Meta 8 hours a day, right? If we do, I guarantee it wins, right? Um, but I think we're in—I think this is the second generation. Generation one was OpenClaw. No one knew how to run it, right? Other than destroying, you know, all your security and credit card. This is generation two. And then Gen 3, I think we'll figure out, but I don't know that we're going to do the use. This may change in two weeks, but I don't know if we have an eight-hours-a-day app yet here. I don't know.

I think you described it exactly right, which is it is super early in its cycle. If you think about like when chat first launched, like how much did we use it then versus how much do we all use it now? You know, Lora and Harvey, you know, which are, you know, when they started, if you did the reference calls on their customers when they were at a million of ARR, the customers were like, 'the thing barely works, we don't use it that much.' And now you talk to them and they're like, 'I run my whole life out of it.' And these businesses are at hundreds of millions of ARR growing really fast. And all of that, I think same with coding. Like at first it could do a little bit, now it kind of does everything. In the future, it's going to be like, you know, times 100. So it's an open question on the—on this stuff. And I think you're exactly right. The question will be, does it make it from where we are right now to you can autonomously trust these things to run your life? And if it can, it's big. And if it can't, it's not.

Does it not massively cannibalize the chat usage? I'm—I'm so—I'm so sorry for any awkwardness, Jack. I—I used to be a massive OpenAI. It's kind of like you win both sides, so forgive me for this. I used to be a massive OpenAI user in ChatGPT, and now I just live my life through Instinct.

Yeah. And you see a lot of people cannibalize that chat usage. You know,

I think it does to some extent. I mean, I think there—like on some level we're each only awake for so many hours a day and like we are all on our screens so much at this point that like there are not more—you can't really get us out of the—like parks where we're playing soccer too much anymore. Like it is all coming from other time that we were spending on the internet. So yes, it—it does to some extent. The form factor cannibalizes. Doesn't mean that these products cannibalize the other companies, but it does mean that the form factor takes some amount of the space whether the labs do it themselves or not.

Yeah. Can I be a dick? Isn't this what like OpenClaw was meant to be? And the founder, Pete Stipe, and like—no, but like, dude, they've been watching from the sidelines for 6 months.

Yeah, but AI didn't give everyone an entire virtual CPU in the cloud, right? And storage and everything. It—this is just these guys looked at—they were all inspired by OpenClaw. We all were. They just—Instinct did it. Jack and I have a story and—and the whole Meta team said they were inspired the next hour to build Meta. Meta was built since OpenClaw. It's just OpenClaw didn't make sense for any of us.

To be clear, Harry, what you're saying is not OpenClaw standalone. What you're basically saying is when OpenClaw got subsumed in OpenAI, shouldn't they have done this?

Yeah, 100%. You were best placed to do this. That was the most strategic thing with the person who did OpenClaw.

I actually don't have the sort of full inside scoop here. But OpenAI's Dev Day is, you know, today. And so we'll see, you know, what, you know, their sort of reaction to all these things are. But, you know, also back to Jason's point, it's like there's generations of these things and they keep moving and everybody learns from each other and it happens in like 2 seconds. You know, when Cursor came out, it was—they had like a long window before the labs caught up. Now, you know, you look at the Instinct team's window, it's short. It—these—these cycles are only getting shorter.

First of all, I just got to say in passing, Jack, I got to hand it to you guys. You know, I've known Bench since '95 when they started, right? And when you guys said after 30 years you're going to embrace growth, I thought you'd tiptoe in and I got to hand it to you. No, when we're embracing this thing, we're fucking embracing it. It's 10 billion pre-revenue, big-ass check. No mincy little—little steps here. It's all in on day one. So congratulations. It's like, you know, coming off the—coming off the dry and embracing the booze, you know, big party night, first night.

Genuine comment here. It actually—one of the things that's been funny is in this cycle we've had investments with early-stage risk require super late-stage capital, which is just definitionally a strange time to be playing. But yes, I think—yes, you thought you were raising a growth fund. In fact, you're just expanding the early-stage fund because it just takes bigger checks now. And—and good on you guys for doing it because I thought your logic at the end is basically some version of if this matters it'll matter a lot and if it doesn't matter it won't matter at all. And there you go.

Right. I mean, it's funny. I—I'm just pushing. I was—I was—I listened to the Noah interview and, you know, I—I did the boring stuff and I'm going to make the boring point and then I'm going to kind of—kind of backtrack it from the boring. When you listen to the interview and you talk about monetization and travel and all that. And then you go away and you do the data on, you know, the number of Americans who fly more than one or two times a year, which is pretty small, the fees available from that. The number of Americans who book not just dining but fine dining—it's, you know, it's like 10% or less of total dining experiences are fine dining. You don't need a reservation for McDonald's or Applebee's. So if you start thinking—and he mentioned—if you think of it as travel monetization, you could get pretty angsty about market size here. I mean, you guys were smart enough to do OpenTable 20 years ago—26 years ago now, Jesus—and, um, you know, it's—it's a billion-dollar business today. You'll be pretty bummed if you're only doing a billion dollars in restaurant bookings on Instinct. So there's implicitly some level of, yeah, we'll get all this travel revenue, but there's something more here. And because you guys run the same numbers and you know them, right? Is there a—is there a part of the monetization that's just, you know, finger in the air and says if this matters, fucking good shit will happen? So I mean, back to the point before, which is slightly tongue-in-cheek but slightly totally true, which is like, do you think of—is this a—is this a growth investment because it was billions of dollars of valuation, or is this an early-stage investment because it's like 9 days old?

And both are kind of true. And if you think of it as a growth investment, you say, okay, how do I underwrite this, you know, relative to, you know, public companies that have 2.5 billion of ARR and they're worth 10 billion. And so, you know, you kind of logic that out. If you're thinking of it as an early-stage investment, you say, gosh, this paradigm, we just have no idea. And so I think in our heads it was closer to that where you say, you know, we are so early innings here where there was OpenClaw, there's this—like, you know, between now and Christmas, there's going to be like three more evolutions of the whole thing. It's like you don't even try to guess all of that.

How do you think about sizing it then when you're writing growth checks at early-stage risk material? I think there's—you still, to the extent that you can—I'm a believer that—and I think, you know, this is sort of like something I've kind of picked up from the crew—that you still—you still need enough investments, no matter—no matter how, you know, accurately you think you can pick at the early stage. Let's define this, forget the price, but, you know, early on in a company's life, you just need enough shots on goal to get something that's going to really matter. And so on some level, I would say the sizing is constrained by wanting the fund to have enough chances to get something great. And so that's the constraint. And then like the upper bound is, you know, these rounds are big, you need to matter in the context of the round, and if the valuation's already high, you know, you got to put more dollars in to sort of get some ownership. And so those are kind of the boundaries.

It was a very good, articulate, not quite answer, but—but exactly the right questions. Absolutely. No, I—I think you're exactly right. It's that—

You do some kind of Kelly betting and then you kind of—you know how certain—what's the upside and how certain are you it's going to happen, edge over odds, and then you say, do I have the stones to do full Kelly or do I back off?

Yeah. And I think like sophisticated public markets investors probably spend as much time thinking about their sizing as their actual positions.

I think, you know, we—we obviously don't, um, but, you know, we try to at least give ourselves some boundaries knowing that we're going to, you know, size incorrectly almost every time. And it's probably actually the hardest one to size because you're right, you have high upside which leans in, super high variance which leans out, but still a requirement for meaningful dollars. You're right. The sizing discussion must have been almost everything.

Yeah, and it's like on some level—

I want—I mean, have you—I ask it in the way you can take the—have you disclosed the dollars, total dollars in the deal?

Got it. Yeah. Because the way to think—I'm just thinking aloud in real time to make it something we all could talk about rather than the guys with the big checks—is that how would you think about that bet as a percentage of your fund? If you're typically doing 20 deals, which means 5% on average, right? You know, you look at the capital needs, you say, 'Hey, maybe I need to do 10% of the fund in this.' You look at the risk and you say, 'Maybe I need to do two and a half.' Harry, you always jump on me with these kind of real-time questions. What percentage of your fund would you put in Instinct at 10 billion pre as a percentage of the fund? Because that's the way a Kelly bettor thinks.

Super easy. I'd do 5%. If it's meaningful enough that it returns the fund with the upside, it's got to be 5% obviously is. I think if you can project out where it is. Um, but then on top of that, I think you have real downside protection here. It's only raised like a billion and a half. It's clearly a really phenomenal team. The bet that Microsoft or any of the large providers would buy it, I think, is very clear. So you've basically got a 50x upside on it being the WeChat of the West that we've all hoped for for many years and downside protection on a 1x plus. I think—

By the way, just as I know I'm being a joke, but you do it to me. If you really believed in that risk profile—50x upside, 1x downside—the Kelly bet answer was probably 30% of the fund, just to be clear. But this is what I think founders don't understand, which is like it's nuance. Like for me at my stage of career, if I did a 30% of the fund bet, my LPs would shit the bed.

You, Rory, respectfully, because of the incredible tenure that you have and the many, many years, you know, as you said, you've known Benchmark since 1995, a year before I was born. Um, you have slightly longer relationships. So maybe it—it does differ.

I wouldn't have the guts to do that at all. So I—I wouldn't want to let it ride. No, I think—look, I can see—I would come out—I mean, 5%—if you're doing 20 deals in the fund, 5% is a full-whack deal. And, you know, you're right. Go ahead, chat.

I think the other thing that I was going to say—I, you know—that sort of like in support of kind of Harry's view on this—is we're in a moment in time where a lot of investments are shaped like this, where like relative to other times, I think it is much more dispersion on these investments and things are both really expensive and then the upsides like apparently look very high. And so one of the things that we, you know, think about and I scratch my head about a lot is we're in a moment in time where everything valuations are like extraordinarily high and then traction and, you know, outcomes and all these other things are extraordinarily high. And so, you know, people are like, do you think we're—are we investing, you know, at the right speed? And I'm like, we're definitely not investing at the right speed. We are either investing way too fast or way too slow. But when both supply, you know, when both sides of the equation are this out of whack, the odds of having it right are zero.

I'm really sorry. Every venture investor is in a WhatsApp group right now saying the market is totally fucked. Just to be super clear.

Yeah, but those would be the same—to Jack's point and the schizophrenia—those would be the same investors who get out of the WhatsApp group and then wire another check for 50 million bucks to Neolab, right? So yeah, look at what you do, not at what you say, as Harry Haldeman said about Watergate. But to your point, it's funny. We actually just did our annual meeting and we had Tyler Cowen, the economist, speak and he was great—just great. But one of my partners who was doing the meeting asked just a tangential question on venture: what's going to happen in venture? And he just quietly said, as economists do, you know, returns are going to be highly skewed. Variance is going to go up with AI and many of you will fail. And then just moved on to the next question. Some of you will get rich and many of you will fail. Well, thank you for clarifying that, Tyler. And he was exactly right, Jack. You're right. It's like the highest variance time, which—which—which speaks to even when you're doing these bets with the upside, having enough to do 20 is key. Which is why I don't get the—we've talked about this before—it's why you write about Instinct being an early-stage bet. The price, despite the price, it's not yet at the stage where you can have the concentration discussion. You know, the folks having the concentration discussion are really talking about OpenAI and Anthropic at the last two pre-IPO rounds. But you're right—at—even at 10 billion pre, if it's where it is today, you don't want to put 30% in.

I—I also think like this is not the round that worries me. If you would look at the Instinct cohorts, which I haven't seen, but I imagine the usage is incredible. I imagine the depth of usage is like we've never seen before. What worries me is when you have three rounds in three weeks with no material movement in between and no data suggests there's been anything different. That's what worries me. Like the people who say, 'Oh, 10 billion, 14 people.' Like, it's just like people who said, you know, a billion for what—for Instagram was stupid. Just like, 'Yep, you don't know that smart today.'

16 billion for WhatsApp.

Yeah. Exactly. Um, but you know what I love? I love the fact that I actually had your partner, Ev, on the show and he said, 'Ah, we're going to look back at Andreessen and General Catalyst and Lightspeed.' And he retweeted it, which is so humble and like non-VC-like, where he was like, 'Maybe I was wrong to say that Andreessen won't do a 5x in this vintage because Mr. Martin Casado is making us look bad because the man has had OpenRouter. The man obviously last night had Fei with World Labs—'

—and then he had Cursor. Um, I mean, so for the new story of the day, AMD buys Fei-Fei Li's World Labs for 8.2 billion in stock, at first big lab exit 2.5 years into the company's journey. Let's focus on the entrepreneur first, right? I think it's great for Fei-Fei. End of story. I—I mean, I've only met her once. I thought she was awesome, very humble. I love her book if you read it. Very moving book about being an immigrant to the US. I was myself, so I know it. And kind of came up from nothing. As someone pointed out, it's not an overnight success. Cranked through ImageNet. Kicked off the whole damn thing because the entire AI kind of resurgence was really kicked off when that ImageNet project that she built, when the winner of that was Ilya and two or three other people, I think, in 2012, where their model using kind of neural nets just kicked ass and blew past everyone else. And it was the first hint that this technology, this deep learning technology, was going to blow everyone away. And she kind of set that up, didn't monetize that, you know, worked at Stanford, worked at Google, and then, you know, midlife quit, did this 2.5 years ago, and nailed it. So I think it's just an awesome story. Um, you know, yet another—just a reminder—yet another wildly successful immigrant story coming to the US, making a ton of money. So I'm stoked for her and, you know, good on AMD. So, I mean, we can come to the venture money later, but that—it's just a great story. If anyone deserved to encourage—

—has been restored and it's good, right? Cash the check. Well, just one thing, you know, it's, you know, AMD was like one of your grandpa's investments, right? Uh, you know, he and his buddies from Fairchild or I don't know what the exact—this is an oldie moldy. It's up 279% this year to 1 trillion. So I don't even know whether they're going to continue the 3D world models or they're just going to be their counterpart to Nvidia's model team. But whatever it is, for 8% of that market—8%. Yeah. Getting a world-class team to make sure that 279% run continues—like it's cheap if it's the right team, right? It's a moment in time, to Jack's point, right? If AMD was up 3% this year, they—they ain't going to be spending 8 billion. But this is a run that you've—you know, it's epic, but you got to maintain it, right? They've got to be number two to Nvidia or whatever the goal is. And it's—that—it's a—it's a—it's a hamster wheel that keeps going faster. So I'd certainly invest 8% of my market cap to—to keep the good times going.

And I think it's the first of a few. We've talked about this a couple of times. You look at these investments and candidly, I don't think the World Labs companies—not the World Labs in particular, but the kind of those World Labs-type companies—would on a standalone basis have anything like the trajectory of Anthropic and OpenAI, where there was a very quick path from here's an amazing AI technology to here's chat and that can monetize, and then oh my god, here's coding and that can monetize like a mother, right? I don't think the path would have been as clear, which is why I've been a bit nervous about those deals over the last couple of years. But in retrospect, the thing I think that's now become obvious is I think all the big foundational model companies are probably in the market to acquire—acquire some kind of robotic foundational model story. I think that, as you say, AMD just wants to be relevant to Nvidia. So I think there wouldn't surprise me if there was a ton of acquisition interest. I mean, there's still 100 Neolabs. So you—you got to be in the 10 that win. But I think there will be a bunch of these big-ass acquisitions over the next 6, 12 months if the market continues to hold.

Do you—because I—I—I invest with a brilliant French partner, Paul, who is, you know, incredibly cynical just by nature of being French. Uh, and he wrote this report on 102 Neolabs, $70 billion plus raised. And my question is just like—just how many of them can get acquired when there were 10 to 12? Like, yeah, it'll be okay. 102, we all going to have downside protection. It was very easy when all the sort of Neolabs got started and the trend took off to just say this is total insanity. And, you know, to your point, Harry, we don't yet know what the—like net balance sheet of the whole thing will be, but we are definitely starting to see data points like, you know, World Labs and many others where like there is real interest. And I think part of—part of what's so different now than certainly any time I've ever seen—may, you know, maybe there have been other points in history like this—but there are just a lot of acquirers. Like there—there are like 10 companies that can do $10 billion acquisitions and want to. And that's just so different. And it's much easier than going public and it's quicker and you don't have all of these long cycles and the hand-wringing. And, um, that—that has really changed things. And obviously, you know, like, you know, extremely impressive run by Martin and it's just shown that like, you know, in these infrastructure lab-type companies, there—there's—there's a lot of room to run. So I—I don't—I don't know how it will—obviously any better than anybody else, but I'm definitely—I'm hesitant to be too skeptical at this point given what's happening.

I agree with that. And I think when you deal with the 100—and yes, there might be 10 or 20 acquisitions—but it's the—yeah, this is where you people—they use the word consensus in a negative sense, but there is no doubt that credentialism matters. Like you got Fei-Fei, you know, I could write that press release, you know, originator of the whole AI thing. It's a little like, you know, the quality folks at OpenAI and Anthropic. So I think the other part of this is—and then Neolabs making sure you have the pristine—the pristine tech asset versus the gritty team. A lot of the stuff I do in the apps level, I love my gritty teams, right? I think for something like, you know, building a world model, you need to have proven technical success because that's both what it needs to deliver and frankly what it needs to navigate and be the desired acquisition candidate. I mean, I'm not sure that's not true overall, but it's just something I've observed.

I was talking to one of the best CIOs in the world last night, and he just said to me, the honest takeaway, dude, you really think you can play in venture with less than a billion-dollar fund now? And I candidly, when I look at many of these companies, you—the first round for this and Fei was like 65. Um, you know, I just paused. I had my team say to me the other day, we can't find anything under 100 million. I said, wow, seed prices are expensive. They said, 'No, no, 100 billion round size.'

Pushing back a little, I—I think the weird thing about this tech is, because I've been thinking about this a lot, too, as a $900 million fund size. Um, you've got two kinds of businesses. You've got businesses where you can't ship a product on less than $450 million, like huge Neolab raises. But you've also got to admit the following, which is also true. You've got companies that use that Neolab and that OpenAI and Anthropic technology that are shipping products on relatively little capital. I mean, you know, you guys are in Higgsfield, right? You run into a bunch of these guys who are like, well, we took $10 million but we shipped the product for 3 million bucks and then the customers took off and fuck, we still got 5 million bucks in the bank but we're going to raise 50 anyway because we can. So it's not like everything is hundred million dollars just to, you know, spin up some GPUs. There's a whole bunch of 10 million bucks got us a long way and then maybe we're raising a 100 but for different reasons because we can, right? So it's not kind of uni-consistent. It's almost two different worlds. And it kind of makes sense is the big technology lift up front from the guys raising 10 billion is what's setting everyone else up to do amazing shit on relatively little. Yes, I would say that just—talent costs are so much higher than they've ever been because of a lot of the alternative options for the talent that you can't do a 2 to 3 million round anymore, like 10 to 15.

You know what the challenge to that though, Harry, I agree with the math and Rory made the point last week and I want to hear Jack's thoughts that I think nominal inflation is like 2.5x in venture since 2010. I might be—I might be—maybe sometimes I get my nominal and non-nominal confused. All true. But if you look at the—at the undiscovered gems, okay, I think seed rounds and maybe they're called pre-seed or pre-pre-seed or post-inception, but whatever we call it, there's still—what does it take—three to four folks to get 18 months down the road? It's the same thing. And if you're coming out and you're lucky enough to get a million dollars of free tokens from whatever—a half million from OpenAI or Anthropic—to start your company, you get all these other things, you can still get pretty far for a couple million bucks, especially if you're willing to share—share in a $8,000-a-month apartment in San Francisco. So you—I—it's not that I disagree with the math, but the truth is you can do as much, I think, for 2 to 3 million bucks as you could 10 years ago. And if you don't have folks dying—dying to give you capital outside of demo day, that—that may still be the natural atomic amount of capital for what—I get the terms wrong today—but a seed round, it still should be 2 to 3 million bucks.

Yes, in certain markets—I mean, just at the risk of being a pedant, right? Yes, in certain markets where you can ship a product leveraging everything else, it's a couple hundred million minimum to enter the Neolab space and it's 500 million to enter the semiconductor space. And, you know, I'm just making a point is that I mean, especially, you know, full scope, there's a—the weird thing to Jack's point, there's such a range of different financing things you got to get your head around, right? You know, you want to put a hundred million in a semiconductor company to still pre-tape-out at 2 billion. Hmm, let me think about that. Then tomorrow, let me look at $5 million for a seed round for a software company where maybe they can get a product out the door and consumer revenue on nothing more. The—the characterization that Harry opened up with, I—I mostly agree with, which is basically that the initial rounds have gotten so extreme and there's two reasons. One is that there's like, you know, you all are saying there's these labs where, you know, you can't do anything for less than 200 million and so that's the first round. The other reason is there is a set of founders that are very like in the Silicon Valley network that don't necessarily need to raise, you know, a $50 million seed, but they can. And so they skip the first round. They skip—they—they skip the 6 at 40 or whatever that round used to be, and they just don't raise it. Or maybe they raise $300,000 just to like get a month in, and then they raise the money. And so in many—in many lanes, I think that round has kind of evaporated. And so I think there is a—there is a cohort of the market where traditional seed investing where you're going to write 3 to $6 million checks by, you know, 8 to 15%, where I just think that is fully broken slash just isn't there anymore. All of that said, there—there is another part of the market which, you know, Jason, I remember reading your blogs like way back when I was starting Lattice, you know, 12 years ago, 10 years ago, whatever. And, um, you know, a lot of it was just like, you know, trying to get my head around what the venture math was. And it's, you know, you have a small fund, you buy 10% of a company, you hope it's worth a billion or two, you get 100 to 200 million, you do that a couple times, you know, out of a basket. I actually think that—that—that might still exist, but the way that it exists is not in the things that we're reading about on Twitter and in headlines. It is these rounds where, um, at the moment they for one reason or another can't go do that round. They're either not in the right markets or the shape of the company isn't quite right and then, you know, they somehow pivot or they get more traction than you thought and it just kind of—it just kind of gets bigger, but it happens the slow way. Like we have now gotten accustomed to all of these stories playing out in a year or 3 years. Like, you know, it's like, you know, we're talking about Instinct, it's like, you know, the—these stories play out so quickly, they get big. But even if you take a company, you know, like Lora, which, you know, obviously I know Harry knows well, and, you know, these stories still play out in a short number of years. And we've all gotten used to that. But it, you know, there—there is another part of venture that takes place over 10 to 20 years where a software company just kind of compounds slowly. And so do I think that there's a 3 at 30 round happening today where in 2041 that company is just going to have quietly compounded in the market of police or fire departments or libraries? Like, yeah, I probably do. So it's just—it's just a different—it's just a different part of the market and it has really had a big dispersion.

I just don't think that's an industry. I just think that's an anomaly and we're in an industry and I don't want to bet on a potential pivot 3 years down the road that might lead to a misshapen company that then becomes shaping again. Like, fuck, I can't bet on Slack.

Every time I try and pitch the story to Harry, he me just like that. And I, you know, and I—I want to believe in it because we've made some magnificent bets on companies like that. But I do agree the—the—the table at the moment is 80% the fast-action table and 20% the slow-action, which makes sense because in 2022 there was a discontinuity and everything before that became obsolete. So by definition anything at this stage, four years in, is fast-action by definition because it started in '22. Also if you want to go for that model, your numbers will be crap for quite a long time. And we always forget that we're in an—we're in an opportunity-cost game where people can put money into your Sarah and your Benchmarks of the world. And while you guys post numbers that are just never before seen, I was messaging Chaan last night, Jack, being like, 'Oh god, just stop.' Like the latest fund is just ridiculous. Um, and like we're an opportunity cost. You can be in those—well, you can't be in those funds, but, you know, you can try and be in those funds. And so trying to be smart where no one else is—

—lot agreed, it's—

Yeah, the—the giant sucking sound basically. Money is a signal, price is a signal, and price is sending a signal. Everybody go right here and everyone will go right here because that's the job of price. Mr. Hayek would be happy.

The problem with the quiet compounder, to Jack's point, in my view, as someone who's pitched quiet compounding since 2012, is they're just not—with exceptions—they're just not stable. You have to build so much more software so much more quickly, right? Jack and I are both on the board of Owner, which is, you know, a rocket ship north of 100 million revenue. But look how much effing software they have to build this year, right? How many folks can pull that off? How many folks can raise the capital like this? Like, and even if you—you have 10 times more competitors, right? I—I have another company at 100 million. They put up a competitor slide the last board meeting. I never heard of eight or nine of the ones. I thought we had one competitor. Turns out we have nine, right, at 100 million. So it's not that I don't want to believe. Forget about the fact that the—the it's harder to raise LP capital, which is Harry's point. I—this instability is something that I think people under-reflect on, right? That's the problem. It's just—it's so unstable that it's just—it's so undurable.

Put another way. Yeah. If you're making a compounding play, the quid pro quo should be low risk. And if the world is such that the tech environment is changing so much that you get the compounding, not the hyper-growth, but you get the same level of risk, that by definition is a sub-optimal game. I think the—sort of to me one of the most dominant reasons why I do agree with Harry here is it's just like it's a talent war in the most insane way right now and it—it—it's just these opportunities are just too enticing for great people.

Which is a perfect segue. Perfect segue. Harry, come on. Figure it out. Are you suggesting that Jack teed me up there, Rory, for a slam dunk? I didn't—

No one told me Supabase was doing so well. I'm out of here.

Harry, I didn't see it either. And I also try to be a podcaster, so that was a hard one. Jack, I think we're just a bit slower than Rory. Okay, Professor O'Driscoll, he's very sharp. He's very sharp.

Stop. Stop. Stop. Stop the fuck. We should just—for the viewers, we—our listeners—we should just say, look, what happened here is the CEO, the chief executive officer of a standalone independent public company, MongoDB, a very successful $20 billion market cap public company who just took the job less than nine months ago, got an offer from Meta, hit the bid, moved over to Meta to run their enterprise—Facebook/Meta—to run their enterprise division and fuck, stock dropped 20% in one day. And then, um, Dev stepped back in who'd been the—not the founder, but the CEO from a long last time and is now back running it. But the point to Jack's point was the money can exert such a powerful influence. It was powerful enough to persuade this guy to quit the top job, go work at Facebook because I assumed the offer was just earth-shatteringly compelling, right? And that—and that's your point, Jack, is that when you have the hot stock, when you have the momentum, you can make people offers that just allow you to take whatever talent you want. And this is an example of that. Do—do you think it is the money? I—when you're earning 52—maybe I—I don't earn 52 million bucks a year, but I imagine if you're earning 52 million bucks a year, earning 120 million bucks a year, that's nice. But I think for him, when he looks at this, he's like, holy fuck, I have the chance to impact far greater. I think it was a $52 million package and he was offered a $500 million package by Zuck to run enterprise. He 10x's easy. What does it take?

520 million then done, right. Um, no one—

wants to step down from CEO to be the chief executive enterprise products and tokens officer. No, he's paid 500 million bucks a year.

I'm just kind of agreeing with Jason because you were going to go down some kind of mango not exciting and Meta enterprise will be exciting. I think that's true. But conversely, no one who's been a CEO goes back to not being a CEO. It's just so damn hard, right? So the answer is it must have been just a compelling offer.

Be careful what you say. I'm just trying to not be mean about Nick Clegg or England or David Cameron or Mark Zuckerberg, and it's just too hard. So I'm just going to let it slide. But what should we take from this though, that actually even the role of CEO is one where departure is normal now in face of money? We should take what Jack said. It gets back to the conversation you were making, because he segued off your comment on opportunity cost. The market is sending a signal that the only place to be is in these extraordinary hot AI companies, and it's sending that signal via price, and people are responding to price.

You're right. I tend to focus on the money, but you're right. It's not just that. It is the zeitgeist. I'll give you and Harry that point. Yeah. It's what's good about here even versus the East. You know, I can't remember, was it someone on garden leave either in the East Coast or the UK? The great thing about California, you got none of that. It's just like I go across the street and next day I'm starting the new company. Yes. And that is the highest and best use of talent. What are you doing, Harry?

It's time for the investment committee.

You already did one. You did a great one for Instinct at the top of the hour.

Oh, no, no, no. That wasn't an investment committee. No, no. The investment committee this week is Jev. Jev is in the market to raise at a $10 billion price. Jason, welcome to the room. Can you please present whether we should be doing this deal or not at $10 billion a week after the seed at about $200 million?

Well, of course we should do it. What's the price? 10 billion. Yeah. Yeah. We should put at first thought we should do 5% of the fund, but after the last 20VC, I've decided to recommend up to 30% of the fund. You know, Jev's already 17% of the traffic on OpenRouter, it's 20% of the traffic through Vercel's router. Many people will of course copy it, just like many will copy Instinct and others, but we are reaching the point where as exciting as the labs are, as exciting as Anthropic's pending IPO and OpenAI, these AI costs are unsustainable, and it doesn't matter if Sonnet 5.5 and the latest Opus is cheaper, it is unsustainable to spend these costs 10, 12 hours a day, and the competitive bar just goes up. Jev is a 70th of the price and 100 times faster. And listen, as my good friend Jack Altman says, the pace of change is so fast or exploding so fast. Maybe Jev isn't the winner next year, but this is exactly the kind of bet we have to do. We have 100% downside protection. Someone's going to scoop up this team out of X OpenAI no matter what. 17% market share on OpenRouter. I mean, as much as I advocated Instinct last year, last week, this week, this is my deal. I'm all in on Jev, 30% of the fund. I know it's risky, but I want to get the ownership. And I do believe there's a 50x upside to 500 billion. So deferring to Rory, I say we do well 20 to 30% of the fund.

Jack, I should warn you, for some reason, Harry likes this kind of soundbite and it plays well on the pod. So who... Jack, you know me so well. At the end of the day, I'm a clip monkey.

We're Benchmark. We just go back to the LPs with one email, we have a new fund. It's not a constraint. It's not a constraint.

Yeah, we'll recycle something.

We also recruited Martin to the team. He's going to help. We'll just recycle.

Yes.

Another big round is Modal triples to $15 billion and Base 10 talks at $26 billion. It has turned out that the right trade, you know, like there was a period where you look back, and maybe we're still in that period, but there was certainly a period where you look back and the right answer with investing was just put it all in the labs. Just buy the labs, right? And every round people were like, 'It's expensive.' But the correct answer was just keep buying the labs. And it has now turned out in the last 18 months or whatever, the correct answer was just keep buying inference. You have Modal, Base 10, Fireworks, Together, and it's just all worked. You know, my partner Eric Vishria had this line that I've been kind of quipping a lot lately because it's true, which is that it's all going to work. And he was on the Invest Like the Best podcast and it was like, 'Is it this or that?' And Eric's like, 'Dude, it's all of it.' And it doesn't mean that every company's going to work, not every sector is going to work, but in general, a crazy number of things are working. And inference has been a really great way to get an index bet on everything outside the labs.

Even if it does, even if it goes down by 50% and total consumption goes up by 10x, you still...

Two reasons. Jack hit one of them and there's a second one. They're crystal clear. One is at the end of the day, Anthropic and OpenAI are just deciding what they want to price their non-Max frontier models, and they have many ways to compete directly. There's no reason they can't be as cost-competitive as they want to be. Now, you know, Anthropic's got its numbers if you exclude 7,000 things that has 80% gross margins. Okay, we can make fun of that stuff, but they have the ability to be as competitive as they want to be, right? And Jack's right there, for sure there's a certain point where all that matters is resolving a task. But even that I don't think is quite that simple. But they can price Sonnet 5.5, which just came out. I just tested it. It's only like 20% cheaper. But that's 20% cheaper in one week. They can do 30. They can do 40 if they want. That's reason number one. Reason number two is, boy, you know, people really don't want Anthropic and OpenAI training on their own first-party data. But they also, I mean, I just got back from Dreamforce and I got to tell you, I know there were a lot of blazers there and ill-fitting suits and stuff. No one wants to run on open-source models there. At least China-based, nobody. Not a single person is comfortable with it that I talked to. Okay, not founders aside, but no customer. So I think just those two trends mean we've hit the peak, right? To Jack's point, oh, it's not that open... no, I mean, it's ridiculous to say it. No one at OpenAI or Anthropic is stupid. They can twist the knobs and dials and do what they want and be as cost-competitive. And just, you know, there are arguments for, but I just think this anxiety at the C-suite level is only going to increase. It's only going to increase, and it's only going to increase as security becomes a bigger issue, and it's only going to increase as Astra 51 was pulled back for security concerns. And no one, listen, I know it's fuddy-duddies, but it is the real world in enterprise. No one, I asked them, no one wants to use an open-weight model on the floor that I talked to. Nobody. So I just think the market share has peaked. I'm not saying it's not material.

I get your point. I'm simplifying that the vast majority of these models that we're consuming today are China-origin models, right?

I'm not saying that they're destroyed, to Jack's point. I just think they've peaked. They've peaked in market share. I think it will come down, and I don't even think it's going to maim Modal or Base 10 or others. But I think these market share charts that have seemed crazy this year, right, there's been a lot of sources, but I think there's a lot of structural benefits that OpenAI and Anthropic have here, and they're going to take advantage of in the coming months.

And OpenAI just hit $70 billion, they said, in enterprise. Like, it's a lot of traction there.

Yes. Your compute share probably proxies to your token share. Probably with a little adjustment to your revenue share.