20VC：按比例认购权之战：Valar Atomics 起诉 Day One Ventures｜Meta 凭 Muse Wonder 重返 AI 前沿｜NVIDIA 瞄准 300亿美元 Reflection 交易，Jev 以 750亿美元估值融资 8.7亿美元

I think Zuck's had a lonely stretch. There's a material chance Meta has just a monster AI product on its hands.

We roughly generated 20, plus or minus, decacorns per decade for the last 5 decades. And we generated—what did you say?—42 decacorns in less than a full year of this decade.

I think everyone without massive scale is living a worse SaaSpocalypse than in March or April. I don't think we should be breathing any sighs of relief.

This is 20VC with me, Harry Stebbings. Now, Jason Lemkin and Rory O'Driscoll: the biggest news in tech. What is on the agenda today? My word, the salaciousness. We have Valor Atomics suing its own pre-seed investor, Day 1 Ventures, and the pro rata debate that ensued on Twitter. Then we have Meta, Zuck, and Alex Wang. The bromance is real. They're crushing it with Muse Wonder, their new model, just shy of Opus. And then big tech goes shopping: Amazon in talks to buy Descartes, and NVIDIA circling Reflection for a whopping $30 billion. This and so much more later today. But before we dive into the show today, every great software company eventually runs into money. Uber had to move it. Shopify had to hold it. Airbnb had to settle it. Money movement stopped being a fintech problem. It became a software problem. And increasingly, God, it's an AI problem, too. Zero Hash is the infrastructure that makes global instant money movement seamless. One API integration for stablecoins, digital assets, and modern payment rails so builders can stay builders. Zero Hash powers some of the world's largest enterprises and financial institutions, including Calshi, Stripe, Morgan Stanley, Gusto, and Interactive Brokers. If you're thinking about stablecoins, digital assets, and the future of money, it's time to talk to the team at Zero Hash. Visit zerohash.com/20vc to learn more. While Zero Hash powers your crypto infrastructure, Deel helps you build a global team. We get it. Global IT can be a headache. New hire in Tokyo, but the laptop, well, it's stuck in London. Someone left the company last week. Wait, is their access already revoked? Deel IT handles hardware, software, and access to across 130 countries from one system. You have the right to be impressed. Brands like Eleven Labs are already managing their global IT at scale with Deel. It's one system, one source of truth, with zero vendor gaps or manual workarounds. So build your global team with Deel. Visit deel.com/20vc and start expanding your business today. While Deel builds the team, Framer builds the site. Framer is a complete website platform, not just a builder, so teams can launch and keep improving their sites in one place. Framer is the AI website builder that helps creators, teams, and businesses ship production-ready sites faster than ever while getting every detail right. Prompt, inspect, edit, and publish in one place at a whole new pace. Agents and humans work in tandem. Agents bring speed and scale. You bring taste, judgment, and control. The work lands on the canvas and stays editable. Build custom code components, manage CMS content, optimize SEO, and audit for issues all in one place. Enterprise-grade hosting, security, and 99.99% uptime SLAs. Trusted by leading brands like Perplexity and Miro. Learn how you can get more out of your site from a Framer specialist, or get started building for free today at framer.com/20vc for 30% off a Framer Pro annual plan. That's framer.com/20vc. Rules and restrictions may apply.

You have now arrived at your destination.

Boys, it is my favorite 90 minutes of the week. We have some big stories. I love the salaciousness of this market. I remember when it was quite boring. I look back. We're going to start off with the news of the day: Valar Atomics sues its seed investor over pro rata.

1. Valor Atomics Fights Pro Rata

The backdrop here is that Day 1 Ventures put about $1 million into the seed and $15 million into the A. They hold about 5%, and the fight is over August's billion-dollar Series B at $6 billion, led by Sequoia. Basically, the CEO asked every major investor to take zero Series B allocation. But there's a spring-back clause, which basically meant that Day 1 wouldn't relinquish this and put up a bit of a fuss when everyone else didn't. How should we think about this, and is pro rata outdated in modern venturing?

And just to remind everyone, pro rata simply means that if you want 5% of the company, you can get 5% of any round, so you can, quote, “protect your ownership.” Typically, all major investors get that. It's also typical that a majority of the investors can waive it for everyone.

So someone can say, “Hey, if most of us don't want to do this because we want to do a different thing for the company that's right for the company, normally you should just be able to waive it, and then those rights go away for everyone, so it cancels the problem entirely.”

But then you mentioned that the second thing is you have these weird spring-back rights, which are something like: if we waive our pro rata rights, and then one of the investors who had pro rata rights gets to put some money in, then you can say, “Hey, you waived, but then you took care of yourself specially. That's kind of unfair. It's not really a fair vote, so I get my spring-back rights, too.”

And that's what I think Day One was asserting at one point in time, and there was actually another level to it. But those are the 2 big rights.

As usual in these disputes, no one's 100% right here. It's rare in life. There are disputes where someone's 100% in the wrong, and even when you look back, you might say, “You know what? Maybe I contributed just a little bit to that.”

Here's my read of the situation, for what it's worth. Here's the more interesting thing: what happened. Listen, we can only read the tweets, right? I love the fact that immediately everyone was all over it, saying that Day One was suing its portfolio company, which was completely wrong, right? It was quite the opposite. They may have hired Quinn Emanuel, now a notorious firm, but it's Valar that wants a declaratory judgment to move on. They just want an answer.

All of a sudden, X decided that Day One had sued its portfolio company, which wasn't true. The other thing is that, if I read what Masha said—who's my co-investor and owner; I don't know her that well, but I have observed her behavior, and she's a good investor—it sounds like, to Rory's point, what really happened was Sequoia came in very aggressively and said, “We're going to make you a great offer. We're going to make you a fast offer. It's clean, it's simple. And just like Sequoia and Index and others do, everyone has to waive their pro rata so that we can get our ownership target.”

I can say, as a seed investor, this happens all the time. It happens all the time. The big funds, in fact, put it in their term sheet without even checking it. Their lawyers immediately put in that my pro rata is waived in every single round.

So it happened here. It probably annoyed Day One because, even though they're a $450 million total fund, they have more funds, whether it's an SPV or whatever. This annoyed them. They wanted to do the pro rata. So be it.

Then Valar said, “Oh, we're actually letting one of our investors invest because they're strategic.” They're strategic. This might piss me off, too. What is strategic? First of all, if they're strategic, they're already on the cap table. You got the strategy. You already got the strategy you need.

If they're a new investor, if someone wants to invest and help this nuclear startup go further, and that's truly strategic, let them the hell in, right? But this is someone already on the cap table who wants to own more, which I get. That probably means they're adding more nominal value than your early investor, which I get, but this is the whole point: you're not supposed to play favorites.

Having said all of this, I've observed versions of this many times. I've observed VCs suing their portfolio companies for doing the wrong thing. When founders have decided to do this, you just gotta get with the program.

At the end of the day, when the founders are in the wrong here—and the big funds often convince founders to do things that are a little bit wrong—you just gotta get with the program. It's not your company.

And, to be clear, are you saying, “That feels unfair, but I accept. I'm not going to waste time with Mr. Quinn and Mr. Emanuel. I'm just going to go with it”?

Listen, if you feel so wronged—and I'll tell you why I think Day 1 feels wronged, if you want—maybe you do send a letter. Maybe you don't hire Quinn Emanuel. Maybe you hire Stebbings and O'Driscoll because they're not quite as intimidating.

Maybe once in a while you send a letter, which you shouldn't, but you do. Then you gotta let it go. You just gotta get in line and move on. It's not your company. It is not your cap table.

These aren't just assets. I mean, the founders have to work at this company for 20 years, right? That's the uncomfortable part. But the fact that they bridged the company when no one else would, led the rounds when no one else would, and then were discarded later—that can hurt.

It's happened to me several times, and it hurts as a human being. It does hurt as a human. I didn't have $100 million lined up to do the Valor round, don't get me wrong, but I've had 3 or 4 portfolio companies where, with a small fund, I had to bridge or do an A, B, or C, which I don't even have the resources to do. And then sometimes the founders just forget it even happened, and that would bother me.

I actually agree with everything you said there, Jason. The challenge is that you are seeing people like Ali Partovi, who’s a fantastic investor, who’s made a lot of money for his investors, and who’s brilliant.

Yeah, I nominated him last year for Investor of the Year, right?

Yeah, they come out and say, “We waive our rights, and we don’t think it’s right.” Many great investors support him. There is a moving tide toward pro rata rights not being included.

Yeah, but you know why he said that? It’s because his fund doesn’t need them. His fund doesn’t need them.

Yes. You then have Gary Tan’s comment on how pro rata rights are what made his IRR and, I think, Initialized one of his funds on Coinbase, right? Look, if your model is all seed and no follow-on, then you can waive your pro rata rights no matter what.

The interesting question is, will the megafund be waiving its pro rata rights? In other words, is Sequoia coming into this round saying, “I’m not going to take anything on the follow-on round”?

To me, I always think anything that a majority can control in a cap table that is heterogeneous, where there are lots of people, should be roughly fair. If a majority can waive pro rata rights if it’s good for the company, typically 2 or 3 of the VC investors can get you there, and it should all be fine. That would be my assumption most of the time.

I do think, for the seed guys, I can see why it sticks in the throat of a seed investor who takes all these risks. Most of the deals don’t work out, and only a few of them work out well, and they’re going to want to grab some of that when the guys coming in later say, “Oh, thanks for everything, and now you should waive.” I totally see that.

So is it complete bullshit that we have to let the big platforms come in with their term sheets that automatically waive our rights?

Look, there are obviously many types of rounds, but I would say there are roughly 2 later-stage rounds. There is Sequoia, Index, or whomever that wants to take all of it. They have the resources, the capability, the fund size, and they’ve been very aggressive for 20 years.

These terms are not new. We take the whole round, everyone waives. This is not a new thing. Some funds do that.

Sometimes you’ll find other investors that are growth-stage funds whose target is 5% or 7%, and all of a sudden everyone can play in the round. It really seems like there are 2 types, and founders get really pressured to go along with the program.

It’s like, look, the Sequoia brand is super powerful, right? It has been since the 1870s. You’re treated with love, you’re paraded in the office, and these days you’re given a very generous valuation and set of terms. You’re given help, and you feel great, and I just ask one thing.

“We have a target, Harry, and we just need to own 15% of the company, so I just ask that you let us do the whole thing.” 9 times out of 10, founders agree to it without thinking. Maybe that’s the right thing to do, maybe it’s not.

But there’s another category of investors for whom that’s not their target, right? There are plenty of growth funds for whom it isn’t. That’s the way the sharp elbows are in the industry. This is not an industry full of soft, cushy elbows, and this is a messy, messy way for it to spill out, right?

So, Jason, you’re on the side of Day One.

I think everyone’s wrong here.

Yeah.

I think it went too far. I think the CEO handled it poorly and said he wished they’d never been an investor when they clearly bailed him out multiple times. That’s not cool. Hiring Quinn Emanuel, a very aggressive law firm, is a step too far. They both did the wrong thing.

Yeah, I agree. Play it out here. Now, down in the weeds, what are you actually going to win here? Whenever you start an argument—and I do this a lot too—you forget to ask yourself, “What are you going to win?”

First of all, you’ve waived your pro rata rights. So you want 5%, but you’re not going to get 5% of the round because someone’s waived the rights. Once that’s happened, it’s all up for grabs.

The spring-back is only proportional to how much the other person who got to come in on spring-back rights puts in. In other words, if the strategic got to put in $5 million, then your ratio is relative to that, so it might not be a whole ton of money anyway, right?

You’re suing for the right to get $10 million or $15 million in at $6 billion pre-money when you already own 5% of the company, which is worth $300 million. What are you doing here? Even if it’s $100 million, right? You have your win.

Picking this fight is a mistake. Whenever you pick a fight with a founder—and occasionally you do it when you’re not thinking—it’s just not worth it. I’m with Jason: both people can be intellectually wrong but pragmatically right.

Pragmatically, this is a whine. Say, “I’m hurt,” and then move on. We’ve left out the whole Russia-Epstein dimension. There’s a whole bunch of noise around that that’s very Twitter-centric about Day 1, and I know nothing about it. I’m going to acknowledge it, and I’m sure at some point it’ll be used in the defense.

I think, forget that. Leaving that aside, I think the pragmatic thing is, what are you fighting for relative to what you already have? Why are you doing this?

That’s just my life lesson. If you are, quote-unquote, “wronged” by a founder as a VC, raise it once and let it go. Raise it once and let it go.

I made this mistake once years ago. Harry doesn’t even know: he wrote a small check into a company he didn’t know, and a very large fund came in. This is one of the rare cases where I wanted my pro rata. It was important to me. I don’t really care after the second check, but I needed to get my second check in.

I love this guy, Harry Stebbings, but I get an email that Harry’s invested in the round and I’m cut back to zero. I’m like, “I was the only investor in this company for years. I bridged you, I helped you. I’ve been the only one. And I love Harry, but I know he has this fabulous podcast. Why is Harry getting an allocation and I get nothing?”

This is a true story. When I said it that way—and I shouldn’t have quite said it the way I did—it was so obviously wrong that Harry got his allocation that it really bent the relationship with the founder. It didn’t break it, but it bent it, right? There was no good answer to why Harry gets his allocation and I don’t.

Then I learned from that: don’t even say it, because the decision’s been made. It doesn’t matter if it’s right or wrong. Let it go. It doesn’t matter.

But Jason, I felt a real bond with you as a result of that. You sacrificed—

You didn’t even know it happened—

—yourself—

—until today.

—for my allocation.

Say that, Harry. Say it.

This was 5 or 6 years ago. You never even knew until today.

I have no idea what company you’re even talking about right now. I hope it’s a good one.

It’s a good one. Yeah, but—

Oh, good. Then fantastic. Great.

But you have to let it go, right?

Jason, I thought that was incredibly nuanced commentary, by the way, from you on that. I was very impressed.

The interesting thing is that everyone’s so biased on this issue. I’ve never seen more crap, more selfish takes on Twitter that are seen as generous. You see people saying, “I’m only a pure seed investor. I don’t need pro rata,” right? Then you see other folks who are like, “This is the most important right under the constitution of venture capital investing.”

If you peel the layer back, everyone’s biased in the answer they give. You just can’t get a straight answer on this, right?

The only thoughtful answer I ever heard, the last time this issue came up a couple of years ago, was what Keith Rabois wrote, which I thought was a simple answer. He said, “You could get rid of pro rata, but it’s core to our business model as larger funds.” Maybe it was at Founders Fund at the time, right?

He said, “I just have to pay a much lower price than, worst case. If I’m investing early with an expectation that I can deploy a lot later, I’m comfortable overpaying in the early round for my pro rata. I’d have to pay half the price I’m paying otherwise. Are you comfortable with that, founders?”

And the crickets. No one said, “I’m willing to raise it at half the price for no pro rata.”

Yes. You can value an option and deduct it from the price. You’re exactly right.

It has a value, right?

And it is worth pointing out, as I said, and you said it, Jason: all the people—even the megafunds who are coming in saying, “You guys should waive your rights”—ain’t fixing to waive theirs. Which should be a clue about the real value thereof, right?

You waive, but I ain’t waving mine.

2. Meta Returns To The Frontier

Moving swiftly on, I love the bromance and partnership that is Zuck and Alex Wang. I just find it uplifting in so many ways. They’re cool to see together.

I like how they match. That’s my favorite part: the outfits match.

Did you see Zuck drinking the Guinness in the cheeky pint? Then he pulls away, and it’s clear that he’s never tried Guinness before and he hates it. But Meta is back at the frontier with a new model, Muse Wonder, just shy of Opus 5.5 on the independent index.

Let's start there. I don't want to add too much. How do we think about this, guys? Another meaningful movement from Meta into the model layer. Even more pros for Zuck?

It's a zero.

It's a zero.

Yeah, because there's no API. We can't even use it. It's only for internal use by Meta. This is the most performative eval I've seen to date in a never-ending spiral of performative evals.

I'm not saying they're not nominally true. I'm not saying that you don't want to celebrate the team or congratulate the team. You do want to do it. Clearly, the team has done good work at some level, right? For the show, I'm like, “Okay, look, I'll hook it up.” Oh, it's on—you can't use it. You can use it in Muse.

What's the point of a model that allegedly has parity for the same price? It's not even cheaper, and you can't use it. This is the height of performative evals.

It's available in Muse Code. It's on an API as well. I didn't check the API. Are you saying it's not up yet?

You can't put it into your own app, no.

Okay, got it. That's a fair comment from a business-impact perspective, but nonetheless, you have to give him credit for getting, at the risk of sounding like someone we all know, back in the arena. They were competitive early on with open weight. Then they faded way back to the back of the pack, and now they're up front again, notably with closed-source models.

So we've decided that if we're going to spend $100 billion, we at least want to be able to charge for the damn thing. That's the takeaway. It'll be interesting to see if anything comes of the enterprise business, but at the very least, the Muse instinct war will now be powered by a super model model. But yeah, that's the takeaway. And $12 billion got you more than you thought a year ago.

Yeah, and on the body-language thing, it got him a right-hand man.

Yeah.

And great tweets.

That's what I saw in the cheeky pint thing. I think Zuck's had a lonely stretch.

Aw.

I think it's been a tough job. No, no, it's hard being a founder-CEO.

No, I know.

It's been lonely, and I think it's been lonely since the beginning. It's been lonely before the IPO. I think he had Sheryl Sandberg back in the day. He's had others, but clearly, it isn't easy to pay someone $10 billion and have them stick around and work that hard. It's easier said than done, and it's even harder to find someone you can truly trust. That body language is key to making this work, right?

I know he's got all the other folks, the Nat Freemans and the Daniel Grosses and all these guys, but it's hard to get a right-hand person. It's hard. So I think that's the untold story. To Rory's point, this may have been a great deal for Scale. It may turn out—

We almost laughed at it when it happened—not literally, right?—but it may turn out to be the next WhatsApp or Instagram. The more interesting thing is, I wasn't that impressed with the model I can't use. It's hard for me to get too excited, right?

The fact that it's just going to make the Muse app better, though, is super interesting. The world changes so fast, but I think there's a material chance Meta has just a monster AI product on its hands. The amount of investment into it is so high. The capabilities are so high. The autonomy is so high. And now you're going to make it 10 or 50 times more powerful. I mean, this could do better than the AR/VR goggle stuff.

The amount of Muse and Instant competitors right now is pretty astonishing. Okay, big tech goes shopping, baby. We have 2 rumors that are circulating. Obviously, we don't deal in rumors, so I'd love to start with the first one.

3. NVIDIA Goes Model Shopping

Reflection, for those that don't know, is a fantastic team building open-source models in the US, backed by Sequoia. I think the last round was at $25 billion or so, and the rumor is that NVIDIA is buying them for $30 billion. This is on the back of NVIDIA buying Poolside for $12 billion, which is in the open-model game as well. Guys, how do we analyze Jensen further as he goes into the open-model space?

Yeah, and by “further,” just to be clear, you're reminding us that they already bought Poolside. They already have their existing Nemotron model. I think the answer is, as we've discussed several times, if you're in the chip business and some of your biggest customers are closed-source model companies, zooming out, you've got to think about it this way: if you've got a pile of money that's going to go into that collective pool, it's in your interest that more of it stays with you and the chips, and less of it stays with the model companies, right?

By definition, you love open-source models and only one chip provider, just like on the other side of the table, the model companies love closed-source models and everyone having their own chip provider. So we're friends. We're the best of friends. We lend each other $100 billion or so, but we're watching you closely, and we want to be able to fight back if you get margin-greedy. It all makes sense.

Do you think it'll go through, Rory?

When you've got $5 trillion, if you want to buy something for $30 billion, you can buy something for $30 billion. Whether or not there'll be that whole pushback after the Groq thing—will the structure of an acqui-hire work now, or will there be pushback? Who knows?

The world's pretty busy right now. It's not like these next 2 months are going to be idle time for the US government, so who knows?

Rory, if you're on the board, would you say sell or would you say reject?

I'd hit the bid because they're amazing companies, but the standalone business is a different question. If you have an open-weight model, you're relying a lot on monetization through inference. Are you in that?

I've always thought open weight is really exciting, but how does it work monetarily? If you're getting these kinds of excess prices, the value to the acquirer is very high, and the market cap of the acquirer is very high. This might be a moment in time that doesn't come back.

It's also, honestly, I've been on the other side of it. It's very stressful to say no to NVIDIA, knowing that they're going to directly compete with you and promote your competitor. We're doing this. We're investing $50 billion, $100 billion in our own open-weights model, US-based and open. We're going to continue to do it. This isn't a one-off, and you have a chance to take $7 million or $9 million, but if you don't, we're going to be a bigger and better competitor.

It's not simple. You can say no, but it is a little scary. It is a little scary when you say no and they're going to buy other people and get bigger and bigger. It doesn't change the game, but as a founder, it certainly weighs on you when you get that “build by our partner,” sort of “threaten with a smile” message, right?

Yeah, so I would take it. Unless you're confident, you've got to have the winning hand, and we don't know where it's going to go. Salesforce announced they've got their own models that are built on Nemotron and built with NVIDIA. You've got to predict where this is going to be in 18 months with NVIDIA putting all of its software engineers and all of its team behind it, right?

This is a software company, not just a hardware company. You could imagine that the moat here between software, hardware, and models is so powerful that in 12 or 18 months, why would you want to use any other sort of open-weight model other than NVIDIA's? Today, it doesn't seem logical, but I could imagine it being the case in 18 months. It's such a perfect combination.

Does it make Thinking Machines more or less attractive at $44 billion, which is the round that they're raising now?

Do you get pro rata rights in the round? Do we have a sense? Do we know?

That's a very good question. Let me just check my—

Otherwise, I'm lowering my price—

…Quinn Emanuel.

…substantially, to Keith's point.

As long as there's either a larger acquirer to pay more or an obvious, clear business model. There are examples on both sides of the table. There are companies that were last and made a ton after all the early guys sold, and there are companies where the early guys made a ton of money and then everyone after that was sorry they're still in it.

It's all so situation-dependent; it's useless to have an ironclad rule. You just have to know what the game here is, and again, I go back to this: I believe Thinking Machines has grown nicely in revenue, but you've got to have a lot of clarity on revenue and revenue visibility to make that call.

Now we're just going to see. It's interesting: we're recording on Tuesday, publishing on Thursday, and on Wednesday, Anthropic is doing its investor meeting. I assume the Q3 revenue numbers will leak within 24 hours. It's not like there aren't big revenues attached to these businesses. They're the biggest revenues ever seen, but the only question is, how will it check out? How will it check out for one level down, like Thinking Machines?

The sort of Poolside “why we're selling” memo was that we just couldn't get enough compute. We just couldn't afford it. We could not keep up with the arms race. It's not as simple as $44 billion versus whatever. Can you even get the compute to be competitive? I don't know.

You need someone who's even closer to it than I am, but my gut is no. My gut is you can't get enough compute to be competitive. Everything's being taken, right? Even Anthropic's reaching out to Meta, saying, “Hey, Meta, can we get some of your compute?” Maybe they'll give it to Thinking Machines, I don't know, but that war has just begun. You can't even get memory, let alone compute. It's not as simple as software, right? So I don't know.

4. Jev Enters Investment Committee

It's time for investment committee. Jason, on the agenda for today, we have Jev and TypeSafe, the parent company, raising an $870 million Series A at a $7.5 billion valuation weeks after launching Jev. Andreessen led, Sequoia participating—one of the largest Series A rounds ever. Jason, will we be participating in this round at $7.5 billion?

Well, listen, obviously we're lucky to get into the second close because there's plenty of money that wants to get in at $10 billion, $12 billion, even $15 billion today. We can't even get our full allocation, so I recommend we put in the maximum amount we can out of the fund because it's already valued more than this, right?

That valuation is several weeks stale, and Jev's usage has quadrupled since Andreessen actually shook hands on the deal. So we have no choice. We have to be all in and get every share we can. I'm asking that all the earlier investors waive their pro rata. That's my only requirement, but we want to put in every share we can.

Are we worried about competition? Are we worried that it appears Cloudflare already has a competitive product? Are we worried that OpenAI appears to have a competitive product? Are we worried that Anthropic could even give this away for free? Are we worried that Amazon has suggested they're building their own version of Jev? We are, but growth like this is the point of venture, boys, and if we don't do this, we should just stay home.

We're raising new funds every 12 to 13 months. These are massive funds. We have to deploy them into winners. And to be clear, we have years of data on this: not every winner wins, but you can only win by being in the winners. This is a winner. Will it be a winner in a year? I don't know. It is one of the massive breakout winners today.

We're still getting to invest at $7.5 billion. We can probably do a paper markup to $15 billion and tell our LPs we're geniuses. I say we buy every share, block the pro rata, and do as much founder secondary as we can.

The terrible thing about that parody is it's accurate as hell.

Totally. It's not a parody at all. It makes you realize, yes, you can do this job in 3-minute sound bites.

Can you paint the upside for me on Jev? How does this become a wildly successful, massive company? Where's the $500 billion vision?

I'll tell you, though, just at a tactical level, at a builder level, right? No models are particularly—going back to the evals and the meta whatever, the Muse Turbo—none of these hit the 90s or 100. All these models make many errors and have many issues, okay? They're not reliable. I could tell you endless issues from this week.

Jev is a classifier, and we're now in Classifier 2.0, and it works. Maybe it gets 70% to 80% of carefully prompted matches right. Will the Cloudflare team keep putting as much effort in as a standalone team? Will Amazon, OpenRouter, and all these people try as hard, or will these be niche products with not enough revenue for them to have time to invest in, while these guys at Jev make classifiers do something they couldn't do before: get incredibly high resolution rates?

Then they can expand it. Right now, you can ask relatively complicated questions, but you only get binary outputs. That's why it works so well, right? Rory or Harry, who's my better investor? Rory, Harry, Rory.

They're going to get better and better and better, and I don't know that even if the Cloudflare and Amazon teams can catch up today, when it's easy—it's easy to build what OpenRouter was 6 months ago today—they're going to maintain that investment.

I think as VCs, we have to make the stressful investment that a core team that's very strong out of OpenAI and doing nothing else will pull ahead from the seeming parity you can achieve today. I'm actually willing to make that bet because I'm using Jev a lot, and I can already see that if it never changed, maybe I'd swap it out, right? But if it gets better than the rest, there's no way I would change.

Even just somewhat better on its evals, you would never use anything else. So I would still make that bet today: this team will do nothing else. And I don't think the inference costs are going to destroy them. I think the 750, this, this is a low inference model. I don't know the training costs.

For real, I think it gets you further than it gets you to Thinking Machines. It might get you 2 or 3 years of runway. So I'm cautiously bullish that a focused team, rather than 4 smart kids at Cloudflare, can maintain its lead and get to accuracy levels that, to summarize, were never possible before. That's the unlock.

Jev is super-fast and super-cheap, okay? And it is accurate enough to use for maybe 10% of your AI use cases. If they make it hyper-accurate, it changes the game.

And that's the theory. Then, just to get back to the numbers, if it's 20% of your total use—we said this, I think, 2 weeks ago—if total AI frontier inference is a $200 billion market, which it better be in a year or 2 or we're all in trouble, then 20% of that is $40 billion.

And if the price of that goes down by 80%, you still get $8 billion. If they get half of it, that's a $4 billion or $5 billion revenue company, which is probably worth $50 billion. That's the case. Every one of those sentences is credible. I think the biggest question is the other competition, but to Jason's point, the business is there.

What worries me in every justification for a lot of these prices is, “Wow, the liability is perhaps only $1 billion, and it's such a strategic asset with a good team that the downside protection is they'll just get bought by a great acquirer.” I hear this every single time. Maybe that's true, but not everything has downside protection through being acquired, does it?

I mean, it does until it doesn't, and that's meant to be glib. But in a world where you've got 12 companies above $1 trillion, all of them are acquisitive, all of them are in a wild, tearing hurry to build companies where time is money, and especially if the legal window is open, then there's going to be a lot of acquisitions, and the logic is true.

If the market slows down, then all this stuff will go in a bad place at the same time. So it'll be true until it's not.

I think it's overstated. I will say today it feels like a liquidation preference in Jev is much more valuable than 1,000 liquidation preferences in 1,000 2021 unicorns. Many of those will end up being worthless.

So we overstate the value of the liquidation preference to get the deal through the committee, but they have real value today, right? I'd rather have this liquidation preference than quite a few from 2021.

5. Decacorn Creation Hits New Highs

Okay, you said 2021 there, Jason. In 2021, there were 34 decacorns. We now have a record 42 startups that have become decacorns in 2026.

As they say in The Simpsons cartoon, that's 2026 so far, right? With a quarter to go. There were 34 decacorns in 2021, right? In other words, it's more decacorns than the last time things went horribly wrong.

That's the first obvious description. If you look back on those decacorns, Miro was a decacorn in 2021. Airtable was a decacorn. I don't think they'll be the same, but it's just worth pointing out.

I think the funnier statistic is this: there are, in the public markets, a grand total of 98 decacorns ever. If you take the private and public decacorns as of the start of this year, we roughly generated 20, plus or minus, decacorns per decade for the last 5 decades. And we generated—what did you say?—42 decacorns in less than a full year of this decade.

Now, obviously, market caps continually go up. All those things point in your favor. But it is stunning to think of 20 to 30 decacorns per decade, and then 42 decacorns in less than a year.

And there are only 98 public companies above $10 billion—I just looked it up—another 30 above $100 billion. So 130 companies in total worth $10 billion or more since the dawn of American capitalism in tech, and a further 42 this year.

Scratching the obvious itch there, you're basically saying that we're in a bubble.

I don't know if I'm saying we're in a bubble. You're trying to make me say I'm in a bubble. I just think that, well, we got the preference, as Jason points out, so we're all fine.

What you can say for certain is the production rate of decacorns is higher compared to anything we've seen before, including 5 decades of data in the public markets. It definitely feels like either this is the most economically transformative event since the Industrial Revolution, or we're getting a wee bit ahead of ourselves. You pick.

Which would you pick, Rory?

Maybe the better answer is this. I'm actually going to go back to the preference answer, right? It's kind of funny because you have to remember that all these private-company valuations—the valuation is set by someone who does have a liquidation preference.

Once you have a liquidation preference and you believe it exists, you can say to yourself: unlike common stock, which trades every day, it’s 1× plus option value. The thing right now is that option values have high value because there’s a lot of volatility. Apparently, 3 companies in this decade will become worth $1 trillion. Many will become $100 billion. Even if it’s 10, it’s rational to say there’s some nonzero probability this company becomes freaking amazing, and my downside is 1×.

To make it concrete, I don’t think all these stocks would trade at $10 billion or more if they all traded on the public markets.

That’s not giving me an answer.

I think it is.

Well, we’re all betting this is true, right? Menlo had some math when they raised their last fund. It was, I think, a $3 billion fund to deploy over the next 12 months, and they’re betting they’ll have at least 4 $25 billion-plus exits to make the fund, right? That’s just the math. It’s not complicated. They can do that like clockwork. That’s the bet.

The AI spend and the recent exits of the last several months justify it. Historical evidence makes it suspect. So are you living in the moment, or are you living as a student of history? I just don’t think in venture there’s a lot of upside in being a student of history, even though we have one starring on this show, which I’m grateful for. But I don’t know if it gets you the huge carry checks.

In fact, as I looked at this data that one of my colleagues was working on for something else, I realized what you need to do is say to yourself, “Okay, if you extrapolate, what level of growth are you assuming to make this math work?” Because, yes, every year—every decade—it looks bigger than the last decade, with the exception of the 1970s, because stocks go up.

But you made a comment about 4 $25 billion-plus outcomes, and in a world—

Per fund.

Per fund, yeah. There are only 130 companies above $10 billion ever, so there are probably only 70 above $25 billion. You’ve got to get 4 new ones. It’s doable, but you can’t miss a lot here, dude.

But that’s what Andreessen got so right. People mocked Andreessen in the early days for overpaying and doing this wrong, saying, “We have to be in all the winners.” They just did the math. Yes, you have to be in all the winners. Even a lot of the winners won’t make you money because you’ll overpay, but the super-winners will make up for all of it.

It’s basic. All the yahoos mocked them, and they were completely right. The price of overpaying for a winner is being in the super-winner.

6. Isomorphic Bets On Drug Discovery

Speaking of super-winners, do you want to come back to the investment committee? We can do a DeepMind. Demis Hassabis, the legend that is Demis Hassabis, raised at about $10 billion, give or take, from Thrive matter of months ago. Now Isomorphic Labs, his company, is in talks to raise at a $40 billion to $50 billion valuation. Are we going to do this?

Generational talent is generational pricing today. As much as I believe in the Jev teams and the TypeSafe teams, he’s going to pull together a team like we’ve never seen since the dawn of mankind in artificial intelligence. Maybe he fails. Maybe he’s had enough of it. Maybe he’s enjoyed life in London too much, with too many posh clubs. But I think I’ve seen the fire.

It’s a scary bet. We have our liquidation preference, and no one will be able to assemble a set of talent like this. This is the next and last chance to cherry-pick the best of the best of the best, and this is the job.

The interesting thing about this company will be that, unlike the vast bulk of the revenues that come from foundation models, where it’s intelligence automation and the value capture is replacing labor, here the value capture is not replacing labor. It’s actually making scientific discoveries.

When you discover a molecule, automating the $500 million of science spend that you pay in science wages isn’t really going to cut it. What you really want to do is, when you create the $200 billion drug, get 10% of that on a royalty basis. As the models devour all the knowledge-worker dollars available to them in terms of direct replacement, at some point they’re going to have to monetize by creating whole new things with their intelligence.

It’s a lot more credible today, given the amount of math they’ve done recently, which, as one mathematician correctly pointed out, doesn’t actually get you paid a whole ton for solving Stokes-Navier, whatever it is. But if you can discover the next GLP-1 drug, that’s a winner, because those things make money.

It will be interesting to see. I’ve said in the past that it’s hard to get the drug companies to pay meaningful royalties when they still have to do Phase 1, Phase 2, and Phase 3, and many of these companies end up being drug companies down the line. It’ll be interesting to see what happens here. But it’s a big pile of money if you can extract the value, not just from a labor percentage.

No, I think we’re all going to be on 10 GLP-1-like drugs in a few years, and I think we never really got the full story on Anthropic’s wet lab. The potential here is even greater than is let on. I think the potential for drug discovery is probably massive.

We’re going from weirdos microdosing to finding out we’re all going to have a set of drugs that dramatically extend longevity and quality of life. They can radically accelerate it, own the IP, and have other people do the hard work. Maybe AI really can reboot the entire industry fairly rapidly. Of course, not all of it—the trials and the rest—but if you could get a GLP-1 out 5 years faster, that would be huge.

Do you actually buy that? I don’t mean that cynically.

The drug stuff?

Yeah. What caused you to change your mindset?

Only very recently did I buy it. I thought it was wackos, microdosers, and people selling crap. But the medical evidence for GLP-1s is so profound. It’s so profound for heart disease and so many other things, and I think our ability to tailor drugs to extend longevity is the massive market.

It’s better than inference. If you talk to any physicians at the cutting edge now, they’re all saying to people, “Your hip or whatever? Just don’t hurt the other one. Just get there.” In another couple of years, the amount of longevity we’re going to be able to achieve will be so much higher.

The wackos and the microdosers of San Francisco made me skeptical, but the overwhelming evidence of the benefits—that almost everyone should be on a GLP-1, right?—and the fact that we’re going to get better and better at figuring all this stuff out, I mean, we’re just walking sets of cancer and heart disease and other things that, having material impacts, will radically extend longevity. Radically extend longevity.

There’s a whole sequence of steps before you even have adequately good molecule identification with some understanding of how it works, and then you still have all the different biological phases. So I do believe it’ll be harder to capture value than you think, but I’m really glad they’re trying.

7. Robotics Data Finds Its Market

Guys, I want to talk about one that may be a little bit earlier than usual. We had Maca [?], which is the Scale or the Surge for robotics data—physical-world data that can most often be implemented in robotics—raise $60 million in a Series B from Sequoia.

The reason I wanted to bring it up is that we have MicroAGI [?] in Europe, which has posted a raise to $100 million very quickly. There’s Maca, which is well over $100 million in ARR. There’s another provider whose name I’m forgetting, AxoFT or AxoSoft [?] or whatever it is, that’s again over $100 million. We just see the sheer volume of data providers over $100 million. What should we read, infer, or analyze from this?

You can infer from data in general that, at Anthropic and OpenAI, I spend a large amount of money on training data, as do the other 3 wannabe foundation models, and there’s an entire industry that exists to take care of that.

It’s an interesting business because you only have 5 or 6 large competitors, and they all obviously want to go on and sell to enterprises, but the vast, vast bulk of the dollars come from a relatively small number of enterprises. Those companies are growing very quickly because the foundation model companies want data, they want it now, and they’re willing to spend big to get it.

They’ve been able to go from zero to hundreds, and in the case of some, billions, very quickly. The interesting thing in the foundation model case is that there’s clearly an end revenue source that’s paying for that. In the case of robotics data, it all makes sense. It’s the same analogy.

There are a bunch of really interesting companies doing various forms of models for robotics. It’s obviously early in its deployment life cycle, so you’ve got the early, burgeoning spend. You’ve got the $10 million and $100 million deals, and you can get that in venture dollars from the robotics companies and, yes, foundation model companies too, because they want to be in robotics.

I think what’s going to scale that to $1 billion or $2 billion is when the deployments happen, and that’s still unproven in terms of timing.

With the humanoids, I saw a post from GXO, a logistics company that one of my companies deals with. One of the guys there sent it to me, and the summary of the post was: “42 POCs of humanoids, zero deployed.” There was also some thread about the BMW relationship with Figure where, again, we talked a big game, but the humanoids aren’t out there.

I do think there’s a deployment issue in robotics, which will filter back upstream to ultimately spending on the data business until the unlock happens at the deployment level. But for now, we’re definitely making hay while the sun is shining.

Are we doing M&A from our investment viewpoint?

Well, you have to do the second half of the Sequoia round at the regulatory 20x step-up, or whatever they deem to be appropriate, so no.

Well, it just—I’m not smart enough to really know. I do think it’s an interesting play in the sense that the current level of interest in robotics is beyond an all-time high. They all need training data, right? You can be skeptical of some of these folks, but the demand for high-quality training data—even if it’s just paying people to wear sensors and do work—is off the charts.

I met a founder at YC Demo Day. He wasn’t in that class of YC; he was just there, and I knew him from Faster stuff. He tilted his company into doing a very niche version of training data, and he did $30 million in 60 days. He said the problem is that they just don’t want to do all the stuff he’s asked them to do. He’s like, “We have $300 million of demand. It just takes so many specialized humans in so many countries to do this niche product. We're turning away 270 out of 30 'cause we don't wanna do it.”

If you could nail that 2-by-2 of being able to provide the data labeling, the humans, everything, you have the right one. Maybe I’d rather do this than invest in a bunch of robots that are going to throw Aunt Bessie out the door or smash their fist through my kitchen when they’re trying to grab the pots. I’m certainly skeptical of the residential ones. Maybe this is the safest place to put our money.

Oh my God, I would way rather do this than any robotics company.

But the absurdity of that is that these things aren’t going to exist unless the robotics companies are successful, so I disagree.

No, no, no. I’m saying I don’t know which robotics company will be successful. I don’t know what form factor or what vertical, but I believe that robotics will be. So I want to be in the infrastructure that allows any of them to be.

But pushing back on that, that’s like saying I want to do Scale AI, which is a, or Merco, which are great companies, instead of Anthropic. If I had a choice, I’d do Anthropic. If you can pick the winner, pick the winner, right? The model companies are, by definition, bigger than the companies that sell to the model companies.

Sorry, I didn’t quite understand that analogy. It’s a bit like saying, “I want to be in Scale, not Legora, because I don’t know who’s going to win the application layer, but I do know that models are going to need to get better.” That’s a better comparable, which would be understandable.

I understand. What you’re saying implicitly is that you believe each robotics company has its own foundation model, and that the companies we’re in, like Generalist AI and Skild AI, are the generalists. If Generalist AI and Skild AI become the dominant foundation models for robotics, then it is equivalent to my foundation-model comment. But if every robotics company has to build its own specific model, then yes, it becomes a very fragmented market at the model layer.

I don’t know. It’s definitely one of the areas where I have a fair number of robotics investments and the least amount of clarity.

8. Anthropic Starts The Price War

Okay, boys. There’s a lot else. Guide me in terms of where you would like to go. We can talk about Anthropic cutting the price of Haiku by 90% as the price war with OpenAI further intensifies. I’m relatively intrigued to hear your thoughts on that, given how intensely you use them.

I know we’ve talked a lot about this kind of stuff on the show. I will say it’s under-discussed. These models—Haiku 5.5 and Sonnet 5.5—they’re great and, from my perspective, they’re the first thing that’s actually been deflationary with LLMs and AI.

Nothing else has been deflationary because, no matter how much you talk about token prices falling, we’ve been consuming so many more tokens that we’re not getting the benefit of it. None of our portfolio companies are seeing their COGS fall in AI, are they? Not really. We’ve all consumed so many more tokens and used more and more of the state-of-the-art frontier models that we haven’t seen the inflationary benefits.

These, to me, are the first ones that are deflationary. Since the beginning, since before the show, there’s always been the N-minus-1 model, which isn’t quite the frontier one. You get last year’s Cadillac and fins for this year’s shiny new one. It was sort of worth it.

But now Haiku 5.5—which is a joke, I mean, developers would laugh at you if you used Haiku 5.5—is almost as good for a lot of use cases as Sonnet, and it’s a tenth or a 20th of the price. It’s shockingly cheap. Haiku 5.5 is like the first reasoning model, and you can dial up how much reasoning you want, so you can almost not care about what it costs like Jev. We haven't had this before, at least not one that worked for me. We haven’t had a reasoning-level model where you didn’t have to stress every token and every dollar, because it is a tenth the price of what Sonnet 4.5 was.

So I think it’s disruptive. Sonnet 5.5 is interesting as an aside. It seems, at least in the testing that I’ve done, almost as good as the previous Opus models. But what’s interesting is that it’s only about 10% cheaper than the previous Sonnet, while it’s twice as fast. These evals don’t talk enough about speed. This is why a lot of these things, like Grok and others, are useless: They’re too slow or they time out.

Speed is a big deal. If you can have your chat go twice as fast, that matters. Sonnet 5.5 is twice as fast, 10% cheaper—not that much cheaper—and better. Speed is also deflationary because you can do more and more quickly.

For me, these two, plus Jevons, are the first time we’ve seen deflation. You’re going to have your Jevons paradox, so we’re going to use even more, of course. But this is the first time you could literally go in, swap out models, have much more speed, and pay a lot less, like in a week.

It’s exciting, and I think it’s most exciting for B2B stuff, where we’re all still making these trade-offs for a lot of B2B apps. We’re asking: How good can we make the resolution? How well can we resolve this issue? How well can we do this workflow? If it’s twice as fast and, in some cases, a tenth the price, we’d better see a lot of results at the next board meeting. No more excuses, boys. It’s the first one I’ve seen at a practical level that’s deflationary.

I think the next interesting thing will be Jevons, or price elasticity. Price elasticity simply says: For a proportional reduction in prices, does the net dollar spend go up or down? The bet they’re making, obviously, is that it’s highly elastic, that it goes down in price and you end up spending more. We’ll see. That’s obviously the bet.

You would not want to be cutting prices in a demand-inelastic product, because then you’d have the classic problem of high fixed costs, low demand expansion, and lower prices, which is terribly, horribly bad. So the bet here, as you see these price cuts, is that this stuff is very elastic in terms of demand.

I assume we’ll see it in the S-1 soon enough. All the accounting for these LLMs is going to be interesting, right? But I assume that when you’re using Haiku or Sonnet—when you’re using essentially the N-minus-1 or N model—the only COGS is going to be inference, right?

You can basically write off the training cost because you can assign that all to the frontier models, to Fable and Opus, right? Assuming that’s the case, it’s sort of a win-win all around. I believe that Anthropic’s inference margins are quite high. But I believe their real-world gross margins are still pretty crappy because of training costs.

I think it’ll be interesting to see how they break this out, but I think it’s exciting. We’ll just use more. There’s going to be a lot of stuff where it’s like taking a 20-minute shower when you don’t have to pay for the hot water. You never get out.

Just as a reminder—and I agree that we will, and I agree with that assessment—when you cut prices by 90%, “we’re going to use more” doesn’t cut it. We have to use 10 times more. Otherwise, demand elasticity is less than 1 and revenue goes down.

I do think we’ll use 10 times as much, because what’s going to happen is, instead of sampling something, you’ll just do it all.

That was what my colleague said about using Jevons. You’ll use it in every row of the database.

Everything. And the key is speed, Matt. We don’t discuss speed as much.

Jev is so fast, right? And, um, Haiku 5.5 is actually slower than the one before but it's not even comparable. Sonnet's twice as fast. You can just use more of it 'cause it's faster. If it's slower it doesn't, it doesn't really help. You can only use so much in parallel but yeah, I agree with your part. You just use it for everything.

What does this mean for open source in U.S. enterprises? Does this make it inherently less attractive, where you're like, “Hell, I can get Haiku and Anthropic for 90% less”? That really reduces the attractiveness of open source.

That's why I thought—I said on the—I don't know whether it was Dev or Jack, but I said you could imagine it would all flip around next year, and then the whole OpenRouter thing would change. People are not loyal. No one wants to use, right or wrong, at least China-based open-weight models. No one in the enterprise wants to use them.

So when you have a domestic solution, ideally from a vendor that's already been qualified and trusted, I'm not saying it will happen, because there's a lot of smart people working at DeepSeek et al., right? I mean, some of the best of the best. But you could imagine this being a moment in time, and 12 months from now, they're niche models that are used by some companies we invest in, but really in the enterprise are almost nonexistent.

Again, I think Salesforce making Fin rip it out before the deal closed is telling. It's a moment in time. Maybe that won't happen, but some people don't want these China open-weight models. They just don't want them.

But they'll have Reflection.

Yeah, they may. That's why OpenRouter could—maybe if NVIDIA buys Reflection, then Reflection plus Poolside plus everything ends up being, you know, 50% of the usage and 15% of the revenue. It wouldn't surprise me at all, right?

Jason, what else tickled your fancy from this schedule?

9. Bending Spoons Cuts Airtable

You know, I almost don't want to bring it up, but Bending Spoons laying off basically all of Airtable. Not a surprise, right? But, man, I wrote this when it happened, and no one cared. I'm like, “Guys, everyone's talking about the money and the ARR, revenue, and the deal.” And I'm like, “You realize almost everyone's losing their job?”

I didn't see anyone care or give a rat's arse about the fact that everyone's going to lose their job. Everyone lost their job at Vimeo, everyone lost their job at Evernote, everyone's going to lose their job at Miro, and I barely saw a tweet about anyone giving a rat's ass. I'm not even judging, because this is capitalism, but, man, when you sell your company to Bending Spoons, you just have to realize your entire team's going to be gone.

Do founders owe anything to their teams? Do boards of directors owe anything to their teams, or do they just owe their best returns to Cedars-Sinai, right? Do you have any moral obligation to sell to someone who's going to lay off 80% of your team within weeks? They are stakeholders, especially if you're cash-flow positive and growing, like Miro and Airtable, right? They both have hundreds of millions in the bank and are growing. Do they matter? Do they matter legally? Do they matter ethically? Do they matter at all, or is this just part of the startup game?

I wouldn't do it. There's no way I would sell my company and have 80% of my team laid off. I just would never do it.

No, no, but just pause, Jason. Let's just stay on that. You have a fiduciary responsibility to your board and to your shareholders if you're Miro, say, and that is the best and only offer for you. What do you do?

I think Rory can help remind me. I do think legally you have an obligation to your shareholders above your employees, right? I believe that. I'm mixing up my Delaware law. I'm pretty sure that's true, right, legally?

Yeah. For a public company, it is the only obligation you have, and they're extraordinarily clear on that. Even if in your mind you're thinking about something else, you never say it in a public company board because it just lays itself open to shareholders. Your sole job when you enter into an acquisition period is to maximize value for shareholders. So yes, that's there. I actually don't know what the rules are in private companies.

But they're private, right?

Yeah.

So, Jason, would you not sell if you're Miro, which is a private company?

No. The people who work for me so hard, I wouldn't let them all be laid off the next week. No, I wouldn't do it. Life's too short.

Pushing only on that, I hear you, and obviously, if you have another alternative that, even if it's going to make you less money, you might well choose to sell to them. The question is, if you're locked into a forever march where, over the next 2 years, a lot of them are going to leave anyway, right? Because when you get on the forever march, as you know, retention is the problem, especially for the best people.

There's a little part of me that says accepting that, having that kind of talent repurposed, is actually just better for everyone. Go into the Bending Spoons deal, but cut your deal on your employee economics up front so everyone knows what they're getting into. I do think that the death march will get old pretty quickly.

You'll try and do the noble thing for your employees, and then over the next 2 years they'll all peel off anyway, and you'll be left with an even more empty husk that's now worth 1 times revenue. So I hear you, and I'm normally the softie in this group, but sometimes resolution is worth it, even for the employees.

It's complicated.

It's complicated.

I just don't think that there's an acquirer you can be more sure will lay off the vast majority of your team than Bending Spoons.

No, the mission is clear.

You don't know what Vista's going to do, you don't know what Thoma Bravo's going to do, you don't know what Francisco Partners is going to do, but you know what the Bending guys are going to do. They're going to axe almost your entire team, right? To an extreme level I've never seen before, right? Maybe Constellation does this—I don't know—but I've never seen it before.

For what it's worth, I think Bending Spoons has a bigger centralized component. They all are very efficient. I think Constellation's argument is everybody runs on their own, but we don't spend a ton of money in sales and marketing, and you have to have 40% EBITDA. I don't care how you do it.

I think Bending Spoons' argument is all of the above, plus all your G&A is centralized, all your people are centralized, so you just need even less. So it's variations on a theme of lean and mean, and even leaner and even meaner.

I'll tell you the other one, just on the theme—and Harry, you can cut me off if you want—but I was thinking about this. I'm very sensitive to this. As we had on the Dev podcast, I'm confused about what loyalty means in the AI age. I've got many flaws. Loyalty isn't one of them.

Cloudflare is an interesting one because a lot of folks are doing little layoffs. Netflix, 5%; HubSpot, 7%. Cloudflare, earlier this year, was crushing it and had a great year the whole year, even during the SaaSpocalypse. Matthew just came in and said, “I'm laying off 20% of the team.” He did it in 1 day. They beat the quarter, they've reaccelerated growth even more, and the stock price is up, like, 80% this year.

It's not Bending Spoons. He didn't decide to gut Cloudflare, but he made the right call. He's like, “We're not failing. We're an AI beneficiary, but we are too slow. We're wasting too much effort on folks that aren't going to get us there. I'm going to do a 20% cut, not because we're struggling, but because we're trying to grow even faster.” And they've reaccelerated.

I think Bending Spoons, it will—I don't know, but some of these are just too small. That's why when I see these layoffs, as much as I care, I'm like, “They're too small.” Cloudflare did 20%, and it worked.

The hard thing for Bending Spoons now is it actually has to find large enough companies that it can buy. There's lots of smaller companies, but it needs to find large enough, revenue-based companies that it can actually buy. That was what was really interesting when I spoke to them. But yes, I mean—

Yeah, it's 3% organic growth, so if you want to grow 80% like they are, you have to buy big stuff, right?

Honestly, Jason, the churn on these companies when they buy them, though—I mean, the numbers are just—phew. You have to increase prices to the max just to keep these businesses going.

While we were talking, it may be uninteresting to you, but I might have been a user of this company in the past. Bumble has announced that it will be taken private. Rory, Bumble—if you don't know, you probably don't, being a happily married man for so many years—is a dating app where young men can find young women, and I think there's a professional network as well, but obviously, it's a dating app.

They are being taken private by Ravi Gupta and his new firm, Ithaca Holdings, for $1.3 billion, I believe it is. It's a pretty cool bit of news—a holding company taking Bumble private.

Yeah, its market cap was, like, $500 million before the deal, right?

Yeah.

Yeah.

Good premium, man. VCs have always hated it, right? The churn, the problems with customer lifetime value—they've always hated it, right?

Yeah, it's funny, because, you know, given the percentage of people who meet on the apps now is 50% to 60%.

But if you think about it, there's a business that provides the most basic human need, which is marriage, et cetera—dating, marriage. Literally, the thing that people care about the most. Just look at the number of pop songs about it versus everything else, and all you get is a few lousy billion bucks. I mean, you're solving the needs of half the Western world, and all you get is a couple billion bucks. It's a funny world.

Listen, I could be wrong. I'll give you the bull case on Bumble, just as a builder. This may be wrong. Having built a matchmaking app for our SaaStr community and thought through this in AI, you can do so much more with AI than you could before. I haven't used Bumble, no criticism, but I can only assume the matches are, at best, mediocre. They're visual, and you have a fairly small amount of data to grab from.

If Bumble has massive historic data, you can get other data and use AI to make dating better. If you have a data science team that can do this better than others, everything's competitive, but I could imagine the outcomes being much better with AI—much better with deep analytics across lots of data. This is Captain Obvious. Obviously, it's not new to the team, but I would imagine there's a bull case where someone could actually make much better matches. Charge more for it. Raise prices because it's much better.

Instead of going on 20 horrible dates or getting a bunch of creeps, you have 5 great ones. You know, the old eHarmony idea, but it actually works. Anchor high and quintuple or 10x the value of this business. That would be the bull case to me. But they're not the only guys probably thinking about it in the space: how to use more data and AI to make better matches.

And also, actually, how to use synthetic versions of yourself—AI avatars that represent you—and then you can go out and meet 100 different people without actually spending any time meeting anyone. What a wonderfully antisocial way to date.

Yeah, but it might work. But it might work.

In the end, I do believe the object of dating is, at some point, to make contact in the physical world, as I understand the mission. So—

Oh, Rory—

For a while, but at some point.

That sounds terrifying. Oof. No, we sit scrolling, which is why we have such a good business.

Yeah, but AI should get to know you. If AI can reduce the friction, the wasted time on dating, and the wasted time and money, the ROI is really high.

I totally agree.

What's a good date cost? $200? I put $150 on Bumble AI, and I'm already $50 ahead.

Jason, any final ones? Rory's not going to give me a suggestion.

10. The SaaSpocalypse Still Spreads

This one is barely worth talking about, but you had it on the list. It's just worth reflecting on as we're in Q4 of the year. Do I have this right? Yeah, we're in Q4. Reflecting on the SaaSpocalypse for a moment, I really don't even want to talk about it in 2027 because 2021 is so far behind us.

I think the SaaSpocalypse is still here, and I think about this a lot. What I mean is, the SaaSpocalypse earlier this year was in part a short-driven thesis that we would vibe-code our way out of all these applications. The market feared that. Growth for many leaders slowed at the same time, which seemed to self-reinforce the fact that we'd vibe-code our way out of all these applications, and we reached all-time lows in March or April, something like that.

The markets roared back. The Nasdaq's up 22% this year. I have a SaaStr AI Index. You can see at SaaStr, of those, I think it's up 28%, or WCLD's up 25%. Nominally, the SaaSpocalypse was a joke, and Marc Benioff made this point at the low: this, too, will pass. Nominally, it did. That fear passed. But here's what I think about this: it passed for those well above $1 billion in revenue.

Agreed.

I think it only passed for those with massive scale, and I think everyone without massive scale is living through a worse SaaSpocalypse than in March or April, because no one wants to buy software. I think we need a new term, but everyone's breathing a sigh of relief. I don't think we should be breathing any sighs of relief.

Agreed, because if you think about it, all that happened is that it's a mature, consolidating industry now. And what happens in mature, consolidating industries is that the market leaders have scale, they have market cap, and they have leverage, and they're just going to start acquiring companies. You're seeing Benioff doing it. You're going to see companies acquiring, and then the ones they don't acquire, you can grind out.

Because your idea is, if you're already at $40 billion, you're money good. You've got cash flow; you're good. If you're at $100 million and trying to get to $300 million in a SaaS product, it's really hard today.

Yeah, and I wonder if, when we started the show, it seemed like at $500 million, you could IPO. I'm not sure, after Orin [?] and friends, that's even true today in the SaaSpocalypse. I don't know that $500 million is durable or safe any longer. You might need $1.5 billion or $2 billion of ARR as a software company to have a safe IPO, and that, outside of Jev and friends, has a lot of implications that the SaaSpocalypse is worse than we thought.

$500 million isn't good enough anymore. Now it's got to be $1 billion to $2 billion in ARR for these companies to go public, and the math gets harder and harder. I don't want to talk about it forever because it's a new world, but the only thing that is safe is scale.

And credit to the few who are able to transition pre-AI plays into post-AI plays, from the pre-AI play to the post-AI play, like Intercom did. One of our businesses last week—I'm not allowed to say how much it sold for, but it sold for a very large amount of money. Not billions, but hundreds and hundreds of millions. TechWolf to SAP.

It was a pre-AI enterprise software company in the HR space of all spaces, and it was able to transition into an AI-first company and sold to SAP for a big amount of money. It was the biggest ever for Belgium, which isn't particularly saying much, but still big. And so I think there is still room—

But I don't think, in all fairness—and congrats to the TechWolf guys—I remember, actually, I forwarded that to you in 2020. But I just didn't get the HR thing. Or even Fin. Intercom and the team are great; I'm a minority shareholder, and Rory's a shareholder, but that's not true scale. That's not north of $1 billion in recurring revenue.

Those deals are great for the founders, and they're fine for the VCs, but I don't think they countermand this point that the apocalypse is as bad as it ever was. Kudos to Intercom for building an AI-native reboot of their company to 9 figures in revenue, but that's very different from building to $1 billion while growing 50%, and so I don't think it counteracts that.

These exits—Menlo's point is we need three $25 billion exits per fund. TechWolf and Fin don't count.

Well, not for Menlo, but for my fund, they return 24%.

Yeah, it made a difference. I mean, I think there are 2 separate points, and I think you should disaggregate them. Who does the $500 million exit count for, and who doesn't? If you're a megafund, it doesn't count. If you're a normal-sized fund, you really can make good money there. It's not to be sneered at when the check hits.

But the second comment, I think, is the more interesting one you're implying, Jason: post-SaaSpocalypse, there just aren't going to be as many of them. Because let's just say there are 200 or 300 of these similarly sized SaaS companies. Big Tech can only buy so many at a time because it's a distraction, so they get hoovered up at the rate of 10 a year, and that's an infinite amount of time.

That's the real issue: because there's no probability of getting critical mass and scale and an exit, you're just destined to be one of hundreds of midsize companies trying to compound your way to something that matters. You very much have to proactively get AI-native, try to drive to an outcome, and try to be important to some other company. Or be happy running your company for the next decade and taking a dividend.

You could cynically say the same is true in AI apps. There are way too many apps relative to the likely number of half-a-billion- or billion-dollar outcomes, but in an expanding market, all is forgiven. Product categories look like company categories. Everybody expands just like SaaS did 20 years ago. It'll all be great until you hit the tipping point again.

11. Flock Safety Faces Security Fallout

Can I ask, guys? I don't want to be Mr. Negative or finish on a negative, but Flock Safety had some problems with privacy backlash previously. We touched on it previously in an episode. Now they're cutting 270 people, 18% of staff. Is this a company in real trouble?

I saw a tweet that said, “I prefer the crime to the surveillance.” Personally, I'd prefer the surveillance to the crime. I'd prefer the surveillance to be actively controlled. We talked about that a while back.

The problem here is a bunch of renegade, off-limits use case for law enforcement weren't being used in the right way—or weren't even being used for law enforcement at all, but to check up on people's personal lives—and that has now damaged the brand.

I won't say irrevocably, but it's a pretty tough problem. You wouldn't want to be Flock right now trying to position this because whenever red and blue America both hate you, you've got a tough thing going. Again, I defend the other side. I do think the crime deferral is really worth it.

I think the improvements in crime in San Francisco and New York are just awesome, and it really expands the thing. So I'm not a hater, but it always sucks when the public has you—when everyone has you like that. I don't know. I'm sure it bottoms out, but it's a tough—not just a comms problem, it's a tough business problem too.

But I'll tell you my concern, which is that this worries me with a lot of startups: I don't know that they have a strong enough engineering team, and I think they cut too many corners. The security issues they've had are unacceptable. They've had cameras just streaming unsecured across the internet.

They were using dated versions of Google on old Android tablets to save money that were no longer supported or hadn't had security patches in the last year or year and a half. This is unacceptable from a company where security has to be at its core and paramount, and it's happened multiple times.

Interesting, because I was more focused on the unauthorized use, but you're saying there's also been access. That sucks, because that's on them.

I'm saying they cut corners. All startups cut corners; it's how you get off the ground. It's something we don't—we don't like to talk about too much at board meetings. But when you are ultimately a security company like Flock, maybe the police departments don't understand the corners you cut, and maybe the VCs don't want to probe, but there are certain companies where you can't cut too many corners, and I worry Flock cut too many.

It just didn't build safe hardware or software. And if you're not investing in the hardware and the software, what about the rest of the team? What about other vulnerabilities up and down the stack? It's culturally consistent, and so I don't want to pick on them because I think they've done a lot of social good, but I wonder if they cut too many corners and now they're reaping what comes from it.

They stored images unencrypted. The cameras sent credentials in clear text. Pressing the button on the back of the camera would open up a Wi-Fi network that any attacker could get into. The police interface didn't require any authentication. This is pretty bad stuff, right? Security patches stopped in June 2018 for many of its cameras.

Ouch. Those are not good facts, I will admit.

Not good facts. And is it ever just that? It's always more when you pull on the string, right?

I mean, it's a very valid worry, to be fair. And actually, I wasn't aware of the security vulnerabilities there. But on a theme that Rory loves, Jensen Huang's just announced that he's going to invest $2 billion more into European AI companies. So go Europe, hey, Rory.

Go, go, go Europe. Yeah, that's about the right ratio.

How much is that? How much money is that for Nvidia?

It's about a coffee, maybe a coffee and a croissant.

$2 billion.

I think the cash flow is $150–160 billion. Yeah, so it's 1% of cash flow. Don't spend it all at once, kids.

I did debate bringing it up because I wasn't so confident about the rebuttal there.

Jensen is just going to spend that money wherever he can to fragment this market. Okay.

Well, probably also Harry is much closer to this than I am. I'm not an expert, but anything you can spend to deflect any anti-Americanism is probably worth the money up to a certain point, right? Keep your head down. Don't be associated with any anti-Americanism. A couple billion here and there to mitigate it is probably a great investment, right?

You know what, boys? No one is talking about what they should be talking about: the downfall of France. I mean, Jesus Christ.

Yeah.

You talk about a failed ecosystem in Paris.

Yeah.

It's—

It's really good of them to step up and make the UK look great.

I know. Do you know what? Genuinely, yes. Thank you so much, France, for making us look like—

Yeah, no, totally, yeah.

—the peaceful, orderly society.

I did enjoy—I don't know why we're getting onto it—but I did enjoy the fact that the Greeks now borrow at a lower interest rate than France or Germany. It's just pleasing after what we all went through a decade and a half ago. There you go.

Dude, Italian politics resembles stability.

I know.

That is a worrying state of Europe.

Yeah, right, totally. Yes.

Yeah.

Yes.