20VC：SpaceX 会以1.5万亿美元IPO吗｜Cursor 会终结 Figma 吗｜Lightspeed 募资90亿美元｜OpenAI：获 Disney 10亿美元、新任 CRO 及 App Store 第一名｜Oracle 与 Broadcom 遭重击：现在是买入时机吗？

Boys, it is great to be back. Jason, you look very smart today. Thank you for joining us from the beach house.

Jason Lemkin

You're welcome.

You've got to have one. You've got to have one. Rory and I are in the office, but we're back. What can I say?

Jason Lemkin

Well, based on your tweet, I've given up because I'm not in OpenAI or Anthropic, so I've decided to call it a day for the rest of the year, I think.

Listen, you're a smart dude. What's the point otherwise, right?

Jason Lemkin

What's the point? Yeah.

1. Lightspeed Raises Multistage Capital

Well, what's the point indeed if you don't have mega funds? Lightspeed raises $9 billion across 6 funds, point number 1. I did backward math on it to understand how that's split up. It's about $2 billion for venture and early stage, and then $7 billion across other vehicles, mostly growth. So the $9 billion is a bit misleading. The question becomes, to my tweet: if you're not playing the big game, do you really matter?

Jason Lemkin

Kudos to Lightspeed, right? They're playing the game on the field. It's bad for seed VCs—I put “bad” in air quotes, right? Because whether it's $2 billion or $9 billion, and you've got to slice these funds up, to Harry's point, to really understand what's going on, it's not all $9 billion for seed.

It really means you don't care what you pay for seed. It just doesn't matter, and you work for speed. Let's do the math. This is why we have $20 million or $30 million pre-seed rounds, because it just doesn't matter at that scale, does it? You just have to get into a $100 billion outcome. I'm not saying it's bad; it just continues to contribute to the barbell, the barbell side of venture.

Rory O'Driscoll

That's such a funny answer, Jason, because it's the classic human thing: the world is ending, but what does it mean for me? Lightspeed raised $9 billion, but Jason's first comment is, “What does it mean for my business?” It's just a good reminder that everything is personal.

Thinking about it, to state the obvious, they earned it. If you're evaluating a big, multistage manager as an LP in 2025, you probably want to see them do 2 things. You want to see their early fund have wins 8 or 9 years in, maybe 10 or 12 years in. In the last 2 years, they had Rubrik last year and Navan this year. An early-stage deal where they were seed or Series A went the distance.

The second thing you want—because you're also not just going to give them half a billion for early stage—is to see whether they're picking and concentrating in the late-stage deals. Obviously, in the last 12 to 24 months, Lightspeed put, by all accounts, $1 billion into 2 rounds of Anthropic, and that feels pretty smart right now.

If you zoom out a million miles, they did the 2 things that a multistage manager has to do. They had great early-stage companies that they built over 10 or 12 years, and they stuffed a ton of money into the hot late-stage deals. Tick, tick, insert $9 billion. It all makes sense.

If you want to make this bet as an LP, that's the kind of fund you'd be looking at. Those are the success criteria to give someone $9 billion.

Jason Lemkin

To raise $9 billion, you can actually make the math barely work on paper with the exits, but you've got to be in so many huge ones, right? You have to have—I’d even forgotten Lightspeed was in a few of the IPOs we have. Otherwise, the math doesn't pencil out.

But then, I just released a show on Monday with David George, and Databricks 7X'd a $1 billion fund for them, while Coinbase 5X'd it.

Rory O'Driscoll

Yeah.

That fund is 15X on a $1 billion fund.

Rory O'Driscoll

Yes.

Jason Lemkin

Yeah, seed's for suckers.

Rory O'Driscoll

If you're in one of the—I think, in Coinbase's case—3 or 4 largest exits, and in Databricks, what looks like one of the 4 or 5 largest upcoming exits, then most math works, right? I'm shocked to discover that if you buy the largest market-cap company on the planet, you probably make money if you buy in early, right? Really, provided you execute and get into those deals, it can work.

This is why I'm always so surprised by the LPs' unwavering appetite for early-stage managers in San Francisco between $50 million and $100 million in fund size. Because, to your point, Jason, what Lightspeed can pay at seed is completely irrelevant. They don't give a shit. $30 million, $40 million, $50 million—it doesn't matter. It's an entry ticket for them to do the Series A, the Series B, and the Series C.

I'm just consistently surprised by LP appetite for pure seed plays given, to me, the destruction of seed economics by multistage.

Jason Lemkin

I think that's a myth. I don't actually see that appetite. I see that appetite having faded since 2021, and I do not see a resurgence in finding tiny new managers. I don't see it. They want to find a 20VC or a Neo or whatever, but I don't think it's easy for emerging managers.

Rory O'Driscoll

There's no doubt that the prevalence of this kind of money must make it slightly harder, at the margin, for everyone further down the food chain to make money. You are competing with someone who does have the ability—and the desire—to invest a lot of money in the very best companies.

Actually, you cited Netskope, and Navan has done an amazing job for Lightspeed. One of the most interesting things is that the aggregate return on one of them—I can't remember which—was only a 6X or a 7X.

Rory O'Driscoll

But the real insight was that they got $200 million plus to work. That's the game they're playing with their $9 billion, and it's a great game, and they do it bloody well, right? But your point, you're right, Howie. I do think, I can't quantify it, and I'm not a seed investor, but there's no doubt that the dynamics of a multi-stage firm mean that if they choose to, they can swamp a seed business to some extent and just write it off as marketing, right?

Rory O'Driscoll

I totally agree with that. And use it as an acquisition to get into the later-stage rounds, as we've discussed. And that's where Dragoneer raising a $4.3 billion venture fund comes in. So there's more than $13 billion across those 2. But have we ever seen late stage as competitive as this?

Competitive is an interesting word. The better question might be, how do you feel about the capital versus the potential return? And even though there's a lot of capital now, it's competitive. One of the attractive things about now, and the reason this money is flowing, is that there are a lot of amazing late-stage companies that look like they're growing very strongly. There are places to put that capital.

2021 felt like that, but it turned out to be treacherous. A lot of those companies—we'll talk about it later—the growth just attenuated and indeed went backwards. So it was a very competitive time in '21 to be in late-stage growth capital, and it turned out to be a very dangerous time. It is competitive today, but I suppose one of the, quote, blessings of having OpenAI and Anthropic in the market is that they can soak up $60 billion of your late-stage dollars and just keep on moving. So there are places to put that money. I don't know if it'll be as treacherously competitive today as it was in '21. We'll see. The wall of money keeps on climbing up.

2. Private Markets Drive Venture

Jason Lemkin

Maybe it's not directly to the point, but Rory made this point when we started this pod, and it's become true in spades just not that long later: all these leaders not IPO-ing is the greatest gift to venture in our lifetimes. The greatest gift to venture capital. The fact that you can flood these top 20 companies with venture capital—well, maybe it's not venture capital, right? Maybe it's a fusion of— but it doesn't matter—the fact that the VCs are able to keep this for themselves—

Yes.

Jason Lemkin

Of course Lightspeed should raise $9 billion, because the public markets aren't getting this. And when people used to say that the retail investors were getting ripped off, I used to scoff because most IPOs don't do well, right? But this is a supercycle where growth is the big beneficiary of this supercycle. It wasn't true of other supercycles, not of the SaaS era and others. And if you're not playing that game, you're losing, to Harry's point. That's the real game. It's not just being an OpenAI; it's playing the growth supercycle bet today. That's the winning play.

Rory O'Driscoll

Totally. But there's a combination of reasons for no IPOs, but no doubt one of them was vague consumer protection post-2000. You're right. The good news is the consumer's been protected from a whole load of bad deals where you can lose 1X your money, and the bad news is they've left the entire compounding of Databricks, of SpaceX, of Anthropic, of OpenAI on the table.

If SpaceX goes public north of $1 trillion, if OpenAI goes public at $600–$800 billion, all that value's been taken in the private arena. And you're right, it's been great for, in particular, the late-stage firms who've been able to get early-stage venture economics on masses of money and, to all intents and purposes, put it to work fairly profitably.

Jason Lemkin

Yeah, it's a different time, but if you just compare Tesla and SpaceX, just for fun, right? Tesla had to IPO. It was a different time. It really barely had any revenue, but it was the same guy running them. It IPO'd at $1.7 billion, which seemed very expensive—$1.7 billion. The sister company will IPO at $1.7 trillion, or $1 trillion. I mean—

Rory O'Driscoll

Jason, you're exactly right.

Jason Lemkin

That is the difference in time.

It's 1,000 times more. No, it's a stunning difference.

Jason Lemkin

And that all went to VCs, or Elon or others. None of that went to the private investors—to the public investors, right?

No. Say what you will about Elon, you're exactly right. Anyone who chose after 2010 could have a 70% compound IRR for 15 years. Exactly. That product was not available for SpaceX and Databricks either.

Jason Lemkin

And he chose. It's the same founder, right? Different situation, right? I'm sure he wouldn't have taken Tesla public if he had any other choice a long time ago, but he kept SpaceX private.

Rory O'Driscoll

You're exactly right, which gets to my point. I think the primary reason you stay—you go public or stay private—is relative cost of capital. He'd had that difficult private round in Tesla where he had to, frankly, save the company from some fairly predatory VC behavior, as he at least recounts it. I wasn't in the room, but it sounds convincing.

At that point, he's like, “The cost of capital from these guys, the VCs, is too damn high. Let me go public.” And that's worked for him, obviously, right? Now the cost of capital in the public market feels higher than the private, so everyone's staying here, and we'll see how that plays out.

3. OpenAI Expands Its Empire

Speaking of capital-ingestion machines, the biggest of all right now is OpenAI and ChatGPT. As always, this could be called This Week in OpenAI. But there was a lot that happened.

Jason Lemkin

Yeah.

ChatGPT was the most-downloaded app in the US. Disney investing $1 billion into OpenAI. And then Denise Dessa leaves CEO of Slack to join as CRO. I just want to break them up. The one that I would love to start—

Jason Lemkin

And OpenAI having to give up its moat to compete.

Rory O'Driscoll

Wow.

I mean, this is almost one a day. I mean, but yeah, put—

Jason Lemkin

Yeah.

Sam has got a busy calendar, doesn't he?

Rory O'Driscoll

I think we can assume that.

Jason Lemkin

Yeah. Plus all his other companies.

Can we start on Disney investing a billion dollars into OpenAI?

Rory O'Driscoll

I think it's one of the least interesting. First of all, $1 billion is neither here nor there. My understanding is also that it's a cross-licensing deal. We'll get $1 billion in equity, and we'll give you money, and we'll get money back from you as a license on the characters that we give you. So it's very round-trippy.

They're leaning into allowing OpenAI, as an image generator, to use the Disney content to generate images. And the interesting thing is that, simultaneously, they sent a cease-and-desist letter to Google because of unlicensed use of the content. So I don't think it's a huge thing at all.

I think it's fairly experimental for Disney to say, “Okay, we better embrace this new thing. Let's see what happens. We're effectively getting $1 billion in equity in return for allowing these guys to play with our characters.” I thought that was, in the scheme of things, interesting, but a no-op.

Jason Lemkin

I thought it was a little more interesting as a content creator, which is just that we're entering the next age beyond just ripping everybody's content off.

Yeah.

Jason Lemkin

And so I think what Disney is saying—it's a 3-year deal—and they're saying, “Look, in the next era, here's the template. And yeah, we're gonna...” This is like when all the content creators take stuff offline on YouTube or cable.

Yeah.

Jason Lemkin

It's a negotiation. So here's the deal. The good news is OpenAI is the leading consumer player in the space. You're going to get the leading IP in the world. You're going to get Disney. And here are the economics. Now there's a template, just like there's a template for ESPN or Disney with YouTube.

Yeah.

Jason Lemkin

You may actually have to pay more, because usually the first ones that go in get a slightly better deal, right? Everyone else is going to pay worse, and we're going to ratchet up the terms to use our IP. And then in 3 years, we're going to raise the rates again.

It is an interesting resurgence of the value of IP in the age of AI, when the first phase was just ripping everybody's content off, and it was great for all of us as consumers, but it may be the revenge of IP.

It will be interesting to see, in 3 years, whether existing, quote-unquote, old-media IP is worth a lot in this new age, whether that's what consumers want to use, and whether it will be worth ChatGPT's or any image-generation software company's while licensing that IP. Will there be an economic return on it versus UGC-type content?

But yes, first of all, does it feel good for Disney? And then secondly, does it yield an economic return for the model providers? Do you get any extra return from having that content?

Jason Lemkin

Disney's a big deal in IP, and I think in 3 years no one's going to be working because of AI. I know this sounds facetious. We'll be spending all our time at Disneyland, because no one's electively working and the jobs are gone. We're going to watch Disney and live in Disneyland, right?

I liked what Bob Iger said, even though I'm not sure what it means, which is that creativity is the new productivity. There's no long tail. What we said 3 weeks ago doesn't matter, right? It's the constant creativity and creation of top-tier assets like 20VC that matter.

Rory, I understand you're saying it's not the most interesting, but to our point last week and before, on Benioff talking about the commoditization of models and the ease of switching, with IP lock-in—hopefully, I'm sure Sam is thinking like this—this is a core element that would retain consumers in a way that other people aren't thinking about.

Rory O'Driscoll

The big-picture point is that I think this makes Benioff wrong in this. It's not a commodity if 800 million people use it and actively go and download it on iOS. How much extra? Would it be 880 million if they have Mickey Mouse? Will it be 820? I don't know. Will that 800 stick?

But I think the big-picture comment is that, in 2025, the most downloaded app on Apple was ChatGPT. If you look over the last 10 years, it's interesting to see what has been the winner. It's been 2 years of TikTok, 2 years of Temu, and 1 year of Zoom. Guess which year, everybody? You know, 2020—the year you realized you needed Zoom. Going back, some of the winners were social media apps.

Rory, do you think they will retain the consumer over Gemini as the consumer front end?

Rory O'Driscoll

My gut would be yes, because they're all in on making it happen, and Google obviously has a lot of other ways to push Gemini, but it would be a very uninformed opinion.

Jason Lemkin

I think, to me, that's an interesting horse-race question. For now, the more interesting question is that OpenAI's mobile app growth has declined because it has to decline because we've run out of humans on planet Earth. So it'll be interesting to see whether it Robinhoods—how well it becomes a meta-app, right? Robinhood is on fire even though its new customers are only growing at 8%.

OpenAI's going to have to do that. That's the whole point of bringing in a head of apps and all of that, I guess. Otherwise, it will inherently stagnate around a billion or something. They're at 1.2 billion. So it'll just be fun to watch whether a mega-app works for ChatGPT or not.

Rory O'Driscoll

I don't know if I buy that.

Jason Lemkin

You have to buy it, because if you look at the numbers, growth is down to single digits. It is empirically true on any source. The growth is down to single digits on every measure.

Rory O'Driscoll

Well, let's take that. Okay, I buy that sentence. I buy that the facts are true. As Senator Moynihan said, we're all entitled to our own opinion, but we all have to have the same facts. So I'm giving you the facts, Jason.

The question is, do they go from roughly 800 million—do they do what Robinhood did and stay at 800 and just sell them more shit, or do they do what Meta did and find ways to go to 4 billion out of 5 billion, pretty much every active human on the planet? That is a big-picture question. I'm not sure everyone on the planet wants to do complex AI lookups.

Jason Lemkin

And if you have no ads, it's tough too. They have cheap versions in India—

Rory O'Driscoll

True.

Jason Lemkin

Otherwise, I don't know that they'll be as big as Meta in terms of footprint unless they want to go all in on free.

Rory O'Driscoll

Yes, you're right. The reason I push on the Robinhood thing is that I think in fintech—actually, kind of financial, in financial in general—the movie is always the same: acquire customers and then cross-sell them up the wazoo, right? Which is what Robinhood is doing.

The young, gambling-addicted financial sector, they're going to give him any product he wants—he or she wants. The interesting question is, I don't know what the cross-sell would be for a consumer on ChatGPT. I suppose the only thing you can do is drive up the percentage of the free users that opt for the conversion to the $20-a-month plan.

Jason Lemkin

Well, there's shopping, and ads are the classic one, right? Everyone's tried that—Instagram for e-commerce ads.

Rory O'Driscoll

Yeah.

Jason Lemkin

I'm not smart enough, but I know that growth has slowed. There's no—

Rory O'Driscoll

Yeah.

Jason Lemkin

—debate. It'll be interesting to watch. There's Google versus Gemini armchair quarterbacking, and then maybe this is one simple reason Anthropic is better: it doesn't have the same headwinds of already having 800 million users. Consumer gets you there faster, but maybe Anthropic wins the bigger prize because the enterprise—

Rory O'Driscoll

Interesting.

Jason Lemkin

—and the back end are just—they've just gotten going.

Rory O'Driscoll

I don't think it is the case, but to take your point, there is nothing as terrifying as a high-growth bet that slows down. What happens is you go from being valued on growth to being valued on cash flow, right? You really would not want that to happen while you're still private.

I don't think it is. I'm not sounding the alarm on OpenAI. But the one big risk of the staying-private-longer bet is that at some point, someone is left holding the bag, and we're seeing a lot of it in the class of '21. I'm holding the bag on this thing that I paid 20 or 30 times revenues for because it was growing at 100%, and now it's growing at 8%. It's 5 years later, and I can barely clear the last-round price.

That would be very bad for a lot of folks if OpenAI's growth slowed. There's no indication it has. User growth maybe has slowed, to your point, Jason, but they appear to be still finding ways to monetize. But if that were to taper off, it would be a world of pain.

4. Cliff Vesting Disappears

What about the cliff vesting ending entirely, emblematic of the hiring wars that we're seeing? How do you think about that?

Rory O'Driscoll

Maybe just an explanation for everyone: typically, when startups hire someone, they universally give stock options, and the typical format is 4-year vesting, but with a 1-year cliff. In other words, if you leave within the first 12 months, you get nothing. At the end of the 12th month, you catch up on a full year's vesting, and then you vest ratably over the remaining 3 years.

The idea is you hire people, you go through—there's a lot of change. They leave, they don't work out early on. Do you really want to accumulate a whole lot of extra shareholders for 3 months' vesting? When 3 months' vesting was worth $10,000, you could see that.

But clearly what's happened here is market pressure has said to them—hires are saying, "I don't want to be here 11 months, have you whack me, and maybe at the current rate I'm vesting on $2 million or $3 million. I want to know I'm going to get that." So I think it is a sign of the times, a sign of the extraordinary sums of money you're dealing with, where even an individual contributor coming in says, "One-fourth—just under 25% of my total vesting package—is real money."

"If I'm getting $10 million over 4 years, it's $2 million if it's slightly less than 12 months." People will probably say, "For $2 million, I'm going to push back on vesting." And they've clearly decided to give in.

Jason Lemkin

It is. The one thing it took me a Captain Obvious beat to get is that it obviously makes leaving easier, right? It makes leaving easier because if you're leaving somewhere, you've hit your cliff.

So on the one hand, it seems like a dumb idea, like you're creating mercenaries, and perhaps you are. You're asking someone who's 18 months in to leave for somebody else and wait 12 months to make a dollar. That can be a tough sell in an age of plenty, right? So they probably also had to make so many exceptions that it stops mattering.

Rory O'Driscoll

You're exactly right. I'm sure they had a gazillion exceptions, and at some point the VP of HR said, "Guys, I just can't be dealing with this. Let's just accept this is the market today."

5. AI Infrastructure Gets Tested

Listen, Oracle—oh my God. Oracle shares plunged 15% on Friday. Disappointing earnings. They've plunged 45% from September highs, 14% down in a week. They've spent $12 billion in quarterly CapEx, higher than the $8.2 billion that was expected. The bulk of it is going to data centers dedicated to OpenAI. Guys, this is above my pay grade. What's going on?

Rory O'Driscoll

One of the reasons I like to admit I'm wrong is because it then allows me to do an "I told you so" when I'm right. Play the tape back. I know what I got wrong this year in our conversations. This one, at the time when they had that 30% pop, it was absurd, and it's just been unwound.

The pop was because, "Oh my God, you sign all this revenue," and everyone gets really excited about the RPO. That was 40, maybe 50—I don't know—60 or 90 days ago, probably. Last quarter's announcement. And now everyone's saying, "Oh my God, to meet this revenue obligation, you're going to have to incur a whole lot of expenses."

Well, shock horror. You just opted to enter a very capital-intensive physical data-center-building business to service 1 or 2 super-large customers, principally OpenAI. It's a tough business, and now I think people are internalizing that.

The stock is—it probably is now below where it was. In fact, I know it is. It's about, I think, 15% below where it was just before they announced all the quote-unquote good news. To me, my real sentiment here was, I knew it, and I should have done those shorts. I should have bought those puts.

In retrospect, 90 days ago we were right. It's a sugar high, a total sugar high on a huge contract with someone who may or may not be able to afford to pay for it—that's OpenAI. And even if they can afford to pay for it, it's not a great business because it's capital-intensive and not nearly as good as your existing free-cash-flow-positive business.

So you'd say that it's returned to normal. It's not going to rebound from here.

Rory O'Driscoll

No.

Jason Lemkin

That's not what I think.

Rory O'Driscoll

No. Oh, you do? Go for it.

Jason Lemkin

I think it sounds odd, but let's bring Oracle and CoreWeave in together. I think they are very interesting, and if you look at the leaders, they're the weak guys. They don't really own anything themselves. They're at high risk of margin compression. So the market's going to have jitters between now and an even bigger AI future. We've talked about it. It's going to have ups and downs. It's going to have bumps, and it makes sense.

Oracle is down 46% since September, right? CoreWeave is down 60% from its high in July. It makes sense those should have the strongest—

Rory O'Driscoll

Totally.

Jason Lemkin

They should see the biggest impact from bumps, and there's no reason, actually. Listen, I'm not saying that these are even the best companies we've ever had, but I don't think there's any reason they won't rebound as the overall trends continue. I think these are just jitters. The overall trend, I think, is that we're still anti-gravity here.

Rory O'Driscoll

You're good. I agree.

Jason Lemkin

Right?

Rory O'Driscoll

I agree with your characterization. You're right. So, to take your point, Jason, if you think the CapEx cycle has 2 more years of strong legs, then you could see these guys rebound. And if you think we may have found an equilibrium—there's not much more increase up from here; in other words, the marginal investment rate is going to go down from here—then you're right: You would be scared of owning them.

That's exactly it. It's an easy way to figure out what's going on at the margin in the AI space, and at the margin is where the money's made.

This is my question. Is this a micro-market hiccup, a small undulation that we pass over, or is it a canary in the coal mine that's foreshadowing something much bigger to come next year?

Rory O'Driscoll

It doesn't have to be something bigger to come. I always think the markets are like this person always interrogating you for truth. The markets are always trying to find out what's true: Should you really be spending this money, right?

You could argue right now the market—6 months ago, the market—was saying everyone should be spending money on AI, and that's great. Now the market's saying, "Gemini, Google, you keep going. You've got a plan. Microsoft, hmm, I'd like to see more of a plan, but yeah, you've got this thing. OpenAI, we're good for it. We'll give you $40 billion. Meta, not so much in love. And then CoreWeave and Oracle, you guys, it's not clear to me why you're doing it."

So it's doing a good job of filtering through from the overall mega-trend, which is amazing, to which of the guys have a good plan in this space and which of the guys do not. So there is a world, to answer your question, Harry, where what happens next year is Google still continues to invest in more CapEx than before, which increases overall demand. But some of these marginal players maybe can't be as aggressive as they've been. Neither the end of the world nor a rebound is a totally implausible scenario.

That's why Jason and I have bigger social media brands than you, my friend, because we're great at binary statements.

Rory O'Driscoll

Yes.

You're like the third option.

Jason Lemkin

In the middle.

Rory O'Driscoll

You're so right. The world is full of the middle, and no one wants to hear it. By the way, that might explain a lot of our problems in wider society with social media, but we'll come back to that another time.

6. Broadcom Reckons With Margins

Broadcom was another one. It lost $300 billion in market cap in 48 hours. Investors are concerned that a $21 billion order from Anthropic will drag down margins because of higher costs in the chips business. Is this a reasonable concern? An order from Anthropic seems like a fairly secure asset to back, and a $300 billion loss in market cap seems like an exaggeration.

Rory O'Driscoll

I don't think it's unreasonable in the following sense. One, Broadcom's an amazing story. Just an amazing story. You look at what that guy's achieved over the last 10 years, it's stunning. Its current market cap is, plus or minus, $1.6 trillion.

As a reminder, the first trillion-dollar market-cap company was Apple in 2018, and now Broadcom—a company 90% of people couldn't even name or tell you what it does—has a $1.6 trillion market cap even after this correction. I think it's trading at high teens in terms of its sales multiple, so it's not like it's cheap.

It totally makes sense. It's actually similar to the OpenAI–Oracle discussion. Why is Anthropic buying chips from Broadcom? The answer is that they don't want to pay 75% gross margins to NVIDIA. And so they designed this other chip and they say, "Hey, dude, I'll buy this chip from you, but I'm not going to give you quite as much money as I'm giving NVIDIA, because if I'm going to pay full retail, I might as well go buy the designer brand."

The whole point is this is meant to be coals here, dude. You're meant to get lower gross margins. It just makes sense that it's a really good business. They'll make good coin at it. But it's not going to have quite the same margins and defensibility that you'd expect NVIDIA to have, where they're imposing their architecture on their customers.

Broadcom is being kind of a made-to-order business. They're saying, "Mr. Customer, Mr. Anthropic, tell me what kind of chip you want, and we'll make it, we'll design it," et cetera. But it's not the same as marketing a branded product like the NVIDIA Blackwell or whatever it is.

I think, again, it's the same thing. The market's just got a little bit ahead of themselves. When you type in the revenue number, you get all excited. When you type in the EPS number, you get a little less excited, and it's just this process of discovery. It's just stunning that you can drop $300 billion in a single day and still be worth $1.6 trillion. Let me tell you what a real crash is like. When you're still trading at a high-teens sales multiple, it's not like everything went cheap, Harry. It's just slightly less expensive.

Jason Lemkin

It is interesting that you lose so much over guiding 100 basis points lower. What's really interesting—and I'm not smart enough to fully predict this—is who gets a pass on gross margins and who doesn't? If you're Broadcom or someone, you're not getting any pass. Even though you're getting massive AI spend, we're very worried.

Rory O'Driscoll

Yeah.

Jason Lemkin

All we care about is that we're worried you're going to lose insane profitability in the semiconductor industry due to AI. Oracle got this great pass until it didn't. CoreWeave gets, apparently, an entire pass. OpenAI does; Meta doesn't. I can't keep up with who gets the gross-margin pass. All I know is we learned in Palantir that some folks deserve the pass, and then we're going to find out some don't deserve a gross-margin pass. I don't know.

7. The Valuation Warning Arrives

Rory, you said it's not a crash, and it can get much worse, and it can go much lower. Markets are at the same P/E peaks as 2000 and 2021. Apollo predicted zero public-equity returns in the coming years. How did you read that, Rory?

Rory O'Driscoll

Sure. I think you have to be very precise in what Apollo said. They said the predicted 10-year return is zero. This is a piece of work that's fairly well understood, and it's actually a very important piece. It's probably one of the things I look at most. Vanguard sends it to you. They're really good about that.

They basically show a correlation between entry P/E and subsequent 10-year return, and they show 3 graphs. They show your 1-year return. There's little or no correlation. In other words, when you buy at a high price, it can still go higher. The correlation gets stronger for 5 years, and it's strongest at 10.

In other words, if you buy at a high price, I can't tell you how you'll do next year. Maybe the stocks will keep going up, and you'll feel smart. But what I can tell you for sure is the probability of making money over 10 years is very correlated to your entry price, and that's the graph that Apollo showed.

Because what they're doing—and it's what all the sensible investment houses and asset managers do—is they don't make 1-year predictions, because you can look like an idiot, and it's hard to know. I mean, it's not just that you look like an idiot. It's actually empirically hard to know. But you can say with a high level of certainty that if you buy at a P/E that's 50%, 60%, 70% higher than the long-term average, your forecast return will be significantly lower.

Going back to '99 and 2000, you see the same thing. You could have said things were expensive in '96, and Greenspan said they were expensive. That's when he gave his "irrational exuberance" speech. But shit kept going up for 3 more years. You can't predict the short term.

But what was true is that by the time you were piling in in '99, it took 10 years for the overall stock market to get back. So you did go zero for 10 years, and even more impressively—and this is a really amazing fact—Cisco, the darling of '99, just got back this week to its '99 stock price.

Rory O'Driscoll

So, in other words, if you buy the hyper-expensive company at just the wrong time, it takes 25 years to earn it back. That's a compelling statistic, and I think it speaks to the same thing as the Apollo comment. Now, the interesting thing is: what do you do with that information? I know over the next 10 years you're screwed, but it might go up next year. Do you stay in? Do you go out and risk having a '96 moment where you leave three more years?

I think you actually end up just looking at your overall asset allocation. But I don't think you go binary on it, because you do have that thing: it's not predictive in the short term, but it is a warning sign. There's a reason that nice Mr. Buffett has piled up $300 billion in cash, because he reads these data too, and he understands them better than most.

The meta point to make that kind of goes back to our business, because we're not public market investors, is when you look at all these private valuations that look attractive relative to the public—and they do—you have to say to yourself, they look attractive relative to a public market at an all-time high. A more normalized public market might well leave some of these private valuations somewhat high and dry. Now, growth might save you.

In a nice story, if the Apollo graph is correct and you just get zero return for the next 10 years, that's not traumatic. That's not the way life and markets work. What tends to happen is, at one point in the next 2 or 3 years, things drop 30%, and then you crawl back slowly over the rest of the decade. If that happens, then some of these private valuations that are comped off that could feel lofty.

8. Cursor Challenges Figma

We mentioned Broadcom's dip, which was on the back of many things, one of them being Anthropic. Cursor announced they're launching an AI coding tool for designers. It's a UI inside the Cursor browser that lets you tweak web apps and drag and drop CSS. It's really the first move up the stack, so to speak, for them. Jason, is this the first credible threat to Figma's position, where design decisions start?

Jason Lemkin

You mean Cursor doing it, right? Not Anthropic?

Yeah, sorry, Cursor.

Jason Lemkin

Well, maybe. I think all of us need to be hyper-aware in '26 and '27 of the massive convergence of categories. It's not just the old days, when 7 years later Datadog would decide to compete with PagerDuty, which certainly hurt them. That's how we grew up: every couple of years would go by and, all of a sudden, Brex is competing with its partners rather than partnering. That's the old school.

AI is creating convergence, where the same products can do more things. I'll give you an example, and then I'll talk about Cursor, because you asked. It's already happened in e-commerce, which is that marketing, sales, and support have already converged. Just this last week, Andrew Bilecki, the CEO of Klaviyo, brought on the former co-CEO of Workday to run most of Klaviyo, which is at $1.3 billion, growing 30%, because the entire world of e-commerce software has changed.

So he needs to get back to product full time, because in 2026 and 2027, there will be no such thing as the marketing software that got Klaviyo public. Marketing, sales, and support have already converged to one agent, so it's already happening there. It's actually surprising it's taken a full 12 months to happen in coding. That's a lot of time, because if you've built any apps like I have, the disconnect between design and the output is the most jarring thing when you get good at something.

When you get good at Replit, like I am, or Cursor or whatever, you can build such cool stuff and you're like, "Man, the design—it all looks like friggin' AI Claude artifacts." They all look like Claude. I can find a vibe-coded site, and 30% of the last YC class looked vibe-coded to me. I could see the Claude artifacts all over their homepage. So it kind of breaks your heart. The fact that it honestly took this long is a surprise.

Should Cursor own that, or should Figma own that, or should it be someone new? They're all going to converge. You shouldn't have different tools for design, prototyping, and production. The agents are just too good. To put it differently, what I've learned is we all want to talk to the same agent: designers, product people, engineers, DevOps.

In an ideal world, there's this meta-agent where we can all collaborate and work together as one company, rather than all being on 11 different AIs. There's a lot of fracturing in AI. I don't know who will win, and it's probably mean to say Figma feels behind, but it is how it feels as we record. Figma is tiptoeing into vibe coding, just like Canva and others, and there is some disruption risk that the tiptoe is too slow.

Rory O'Driscoll

I think that's awesome, Jason. I want to do the coding-Figma thing, but also the bigger comment on wanting a single agent for everything. That's just a huge insight, because it's only become obvious to me in the last few months as I've been talking to companies.

No matter how hard you try, you try and translate your prior experience into this. You have a sales SaaS company, so now you have an AI sales SaaS company. You have a SaaS marketing company, so now is there an AI marketing company? I think you're right. In a lot of these processes, the reason you have these separate siloed companies is that the humans were siloed. There was a salesperson and a marketing person.

But if the AI is doing everything, let's just take the customer journey. You can have a single AI agent within your company dealing with your customers as you're prospecting them, as you're selling to them, and as you're supporting them after they onboard. I've seen some companies doing that, and I'm like, "Wow, that's a powerful idea."

Because if you think about it, one of the shittiest things about dealing with any company is you start off with sales and build this rapport. They tell you what you want and seem to know exactly what you need, and then you just get transitioned to a totally different person and start again. It's like you, as the customer, are being put through that.

Jason Lemkin

It's not okay in the age of AI.

Rory O'Driscoll

Yeah. I'm thinking of a couple of deals I saw within the last week where what you expressed suddenly goes, "You're exactly right": the single view of the customer on the customer side. I just think that's a really powerful 2026 theme.

I don't even know what it means in terms of what kind of apps are built. I'm thinking of sales and marketing at this point, and we'll come to Figma and Cursor in a second, but I just want to say you nailed it there.

Rory, can I ask: how does that shape your thoughts when you look at investing in support tools, as you have done in the past?

Rory O'Driscoll

I think all these companies are going to expand their footprint. Going back to the thing, the good news is that there's a huge return—an ROI—on AI, and I do believe there is. The bad news is I think to grab that return over time, you're going to have to be more expansive and aggressive, and you're going to be going into adjacencies much more than in the past.

It's only obvious to me now that I've started to think about it: if you were selling software to automate work, then you probably sold to each department that did that work. But if you're selling AI software to automate an outcome, just sell to the customer—the person who wants the outcome—and they'll be like, "Yeah, I'll take all that."

So I think you're right. We have an investment in customer support, and we have a number of investments in the AI sales stack. I just think over the next 1 to 3 years, there's going to be mass convergence.

Just to jump back to Jason's example, in the same way, the old line is that in any company, you're either selling stuff or making stuff. We just talked about the selling-stuff people, but then Jason was talking about the making-stuff people. Among the making-stuff people, between design and production, I think what he's saying there, too, is that instead of having a separate design function and a coding function, you could argue that all comes together over time. Is that right, Jason?

Jason Lemkin

I don't think we're going to eliminate designers. I think we're going to have designers. I just don't think we're going to have multiple platforms. It's better if they're the same. It's so much faster and so much more efficient.

I've invested in a small startup called Alloy App that sort of bridges it. It lets you vibe-code your existing product and change it. It's early. There's already insane demand, but that's just the first step.

Everyone loves Figma. Every designer—I'm sure some are grouchy—but I think it's up there with Klaviyo.

Rory O'Driscoll

Yes.

Jason Lemkin

For everyone in e-commerce, their favorite app is Klaviyo because it just gets you more customers. Klaviyo has brought in a co-CEO because their whole business model has changed. It's going to be there in software in a year. It just doesn't make sense to have to wait days for designs to change, then have them only sort of work in my codebase, and then have to integrate them.

All these guys have added a design mode with Gemini 3 Pro or whatever it is. They all added a design mode. It's pretty good, but it's still recycled stuff. You can see a hint of it, where now you can make a somewhat beautiful website while you vibe-code. 2026 is going to be 50× better.

Jason, you're so well informed on this. If we take that view of the collapsing of it all into one platform and one location, and if that is the outcome, if I pressed you on who the winner of that will be, who will that be? Will that be Cursor, Claude Code, or Figma?

Jason Lemkin

Honestly, and I know this sounds Captain Obvious, I honestly think it's going to be who wants it the most.

You almost have to work so hard. You have to 12, 12, 9. As hard as we're working at Cursor, and it's been great, we might have to work even harder. But maybe Figma works even harder. I know this sounds silly, but everyone can copy each other in weeks now, not in months, years, or quarters. It's who really, really wills it into existence.

I don't think we can sit back as VCs and, even with our great king-making checkbooks, fully control the outcome. I think the ones that work, that produce 10 times more output, are honestly the ones that are going to win—the leaders. I don't think we can predict. It's easy to bet in favor of Cursor over Figma because Figma took, what, a decade to get to $1 billion? And Cursor took, you know, a year. So if we're momentum bettors, we have to bet Cursor, right? Even though there are reasons to bet against it, right? They aren't designers. They don't have the base. They don't have the customers. But if I had to pick, I would bet on that rapidity today.

Is this the interim step to Cursor moving down further—or up, whichever way you want to take it—into consumers and into Lovable and Replit's zone?

Jason Lemkin

It could be. I think the only reason Cursor didn't build Replit or Lovable is that it wasn't worth their time, because they're in an even bigger, better market for the moment.

Rory O'Driscoll

Agreed.

Jason Lemkin

I honestly think it was just a distraction. As big as that market is, as exciting as those guys together are, they'll get to almost $500 million in a year. Cursor got there in 9 months. So you're taking your eye off the ball to invest in a high-churn, friction-full space when they probably have 160% effective NRR at a revenue level, because nobody leaves Cursor, right? Why would you invest in a smaller, high-churn space when you have insane retention and you're growing even faster? And you want to have a team of employees in the triple digits.

For software teams building professional, enterprise-grade software, the budget is just logically larger than for demo apps or prosumer apps. So maybe there are more individual users of Lovable, or Replit is kind of in the middle, but the reality is you're accessing the big-dollar spend of every software company and every enterprise company when you're Cursor.

I don't think the Figma market and the Cursor market fully converge. The equivalent now of a Figma design is, in fact, someone using Replit or Lovable to do a direct mock-up, an interactive mock-up, versus just a design mock-up. But clearly, among them all, there's this kind of smooshing together, and that's really significant because what were separate markets are all going to be competing with each other. They're all going to be competing with each other because the prize is just so big.

And I thought, Jason, you did a nice blog post. I don't know if you did it or your AI machine did it, but making the point, citing the Menlo work, which is really nice, that 50% to 60% of all the end-user spend on AI right now is coding and coding-related.

Rory O'Driscoll

Yep.

Jason Lemkin

This is the big kahuna. Everything else, even customer support, all the other stuff is... Everything else to a rounding error is 45%. Software- and coding-related stuff is 55% of all enterprise end-user spend. This is the epicenter of the enterprise AI revolution right now.

Rory O'Driscoll

55% of AI spend is in coding. What will it be in 3 years?

Jason Lemkin

You hope the other categories are just behind, in a sense, right? But maybe it's intrinsic to the extreme value in a very large category of software.

Rory O'Driscoll

Yeah, I don't know, but how about this answer? I don't know the answer to that question. I can give you a quick answer to an adjacent question, which is fun: the total spend is kind of $15 billion or $16 billion on apps, and another $15 billion, plus or minus, on enterprise infrastructure, which I don't fully understand as a category, but leave that by the by.

So the big aha, again, is that end users in the enterprises are spending about $15 billion or $16 billion on AI, per the Menlo data, and the people who make AI are spending $400 billion on, quote, “making AI.” If that's $16 billion at 3 or 4 times last year, it's got to 3 or 4 times a bunch more times before it can cover the nut on the CapEx spend.

To me, that's the most important thing. I'm less clear on what the mix will be, but somehow enterprises have to find not $15 billion but $150 billion of budget. Otherwise, the people investing $150 billion in CapEx are going to have a sad day when their CapEx is greater than the revenue line. So I think the overall growth is the key question.

9. AI Starts Maiming Incumbents

Jason Lemkin

Could I just go back to one point on Cursor versus Figma? I do think it's so important to founders, investors, and execs. I think one risk is that Cursor completely displaces Figma. We don't need it. That's unlikely for a lot of reasons.

Rory O'Driscoll

Yeah.

Jason Lemkin

I think the bigger risk for so many vendors is that it maims Figma.

Rory O'Driscoll

What does that mean, Jason?

Jason Lemkin

The old customers don't leave, okay? HubSpot and Box, and even Anthropic using it, don't leave. They renew. They don't buy as many seats, right, because the team is also using Cursor. But of course, they don't churn. The logo retention remains good, but NRR drifts down, and new customers—the next generation, the kids from YC—defer that purchase because they're doing enough in Cursor.

I'm seeing this across my older portfolio, that folks are maimed. The existing retention is good, but the new guys are just taking enough of the new budget that your growth materially decreases. I think this is a risk for almost everyone that's established: you just get maimed. You don't get killed. You get maimed. And Cursor could easily maim this market. It's so big.

Rory O'Driscoll

I think of Monty Python and the Holy Grail, where it's just a flesh wound. Sorry. Monday.

Jason Lemkin

It's a deep one, though.

Rory O'Driscoll

It's a deep one.

Jason Lemkin

It's the one that never quite heals. And you sit in the board meeting and you're like, “Well, we had 15,000 customers last quarter, Rory. Now we have 15,200. Hooray. Kudos.”

Rory O'Driscoll

Low growth is miserable. Agreed.

Jason Lemkin

But it's this maiming that I think people aren't being honest about. A lot of founders aren't being honest about how they're being maimed by AI leaders. They're being maimed. They're not being crushed. The crushing is the narrative we talk about.

Rory O'Driscoll

Jason, who else is being maimed?

Jason Lemkin

Atlassian is slowing down, right? GitLab is arguably being maimed. Many public companies are being maimed. You've got to be Datadog and have so many leading products in market almost not to be maimed.

It looked like Mongo was being maimed until they fought back, right? It really looked like they were being maimed by all the new Postgres and other competitors. They're back, although maybe they're maimed in the sense that they should be growing faster given the growth of AI.

Mongo should be this great Broadcom-like AI beneficiary, and they are, but they're not growing as quickly. They have so many competitors now. There are so many Supabases and others taking pockets of market share away from them that you might not even see some of the maiming, right? Maybe Mongo should be growing 50%. Why shouldn't it be? It was the leading platform, and the explosion of apps that we're building today is unprecedented. Why isn't Mongo growing 50%?

Rory O'Driscoll

The way you see it is, you see those CIO surveys where they rank their priorities, and what happens is, as AI has gone up in the zeitgeist, it's gone up the priority list. Something that was number 3 goes to number 6, and then it doesn't get funded. That's part of the SaaS slowdown that we've seen across the board in the public companies. You're not selling the new, new thing.

I'm kind of with you, Jason. Even at the infrastructure level of the GitHubs and things like that, you have to co-attach to where the budget is. If you're selling infrastructure but you're selling it to the people who aren't doing AI, then you're going to be dealing with a slow-growth secular story. If you're selling that infrastructure to Anthropic and OpenAI and the JPMorgan AI initiative, you're not going to be an AI company, but you're going to have some of that growth rate.

Jason Lemkin

Maybe the most maimed that I can think of—it may not be fair—is UiPath.

Rory O'Driscoll

Ooh. Yeah.

Jason Lemkin

Until Daniel came back, we missed the AI wave. But it's not that simple. UiPath is not hemorrhaging customers. It's back to 16% growth. The stock is up 27% this year. Go Daniel, right?

It just got maimed because RPA, in part, got replaced with agents and AI. It got maimed. It got maimed.

Rory O'Driscoll

Yes. I think that's a great example, genuinely, Jason, because it's kind of one of those moments where you'd do it in a different way if you were starting now to automate, and you'd start with an agent. You just wouldn't start by doing the thing that they do.

What do you do if you're Daniel? Daniel's one of my closest friends. What do you do if you're him?

Jason Lemkin

Well, he has stabilized the ship. Growth has come back a bit. He basically checked out, like a lot of folks did, when times were easy, when products didn't change. He called it a day.

And then how long did that last? 8 months, until he had to kick out the CEO he brought in? It wasn't even a year.

Rory O'Driscoll

I think what you have to do is say to yourself, there's a way you do automations in 2025 that are agentic.

UiPath's automations were very deterministic and brittle. You do exactly these 3 steps in exactly this order, then it works, and you can automate away humans. Now you can have a 10-step automation where you have some decisions. The good news is there's a bunch of companies from Y Combinator that are doing this kind of stuff, and you should buy or build enough to just take the pain and insert yourself into relevance.

I'm sure he's on it. This is a very smart man who knows what he has to do. But as we've discussed before, the hard thing isn't intellectually knowing what you have to do. The hard thing is just driving it through an organization that will invariably say, "Our stuff already does that."

We've looked at a lot of the companies that are the next-generation companies. You do the research and you do the reference calls, and the UiPath team will say they have something like this. They do, but in the eyes of the customer, it's not perceived as the new thing, and you've got to change that perception. It just requires a supreme act of CEO will.

I didn't like the founder mode cliché, but I've come to the conclusion that it's what you're dealing with here. You've just got to say, "We are refounding the company in the age of AI. We are not going to lose any of these AI-first deals, and we're going to make it happen."

Jason Lemkin

You know, in a way, though, I agree with all that. I think Daniel, as hard a job as it is, has an easier job than some. Let me just step back to the numbers. Here's what I would do: $1.8 billion in ARR, but he's got 98% GRR and 107% NRR.

Mathematically, one problem is his NRR has fallen from the 140% peak at the IPO because people aren't buying that much more from UiPath. But he's got 98% GRR—98% retention. He has enough time, as one of the greatest B2B founders out there, to build the agentic products that $2 billion of his customers want to buy, and they aren't leaving.

It's the same opportunity Marc Benioff has. I would argue Marc, with his megaphone, is actually, like an Aaron Levie, doing a better job publicly of bridging the gap, but they have the same job. We have extremely high revenue retention. People aren't leaving. No matter what anybody says, we're in the first inning for AI B2B.

I can tell you why we're in the first inning. It's brutal, but you have time. You've got 6, 12, 18, 24 months to roll out high-ROI agentic products, and you just have to get enough of your $2 billion base to buy them, and you're back to 30% growth. He goes from 107% NRR to 100%.

You know, Databricks has 150% NRR at $5 billion ARR. I know it's easy to say, but that's the job. Make those happy customers buy more of your agentic product. It doesn't have to be all of them tomorrow. It has to be more of them each quarter, and you're back in the game.

Rory O'Driscoll

Isn't it as simple as the Alex Rampell quote, which I love: "Can the incumbent acquire innovation before the startup acquires distribution?" Exactly this point.

Jason Lemkin

Yes, but you don't have infinite time. But I think everyone has time with 98% GRR in 2026 and 2027. You don't have infinite time. Get rid of the CEO you brought in so you could relax. Go into Sergey Brin mode, but you do have time. The game is not over.

Rory O'Driscoll

To take that quote, I think it's a good quote, and to some extent it is that simple. But actually, I'm going to find myself surprisingly agreeing again with Jason on the human factor. The hard part of doing it is: can big co acquire innovation faster than new co can acquire distribution? Let's take that as a construct.

That's actually not the issue, because they can acquire innovation if they can push it through. The really hard part is that there are going to be about 2 or 3 years where you're lifting the growth rate from 9% to 11%, and the stock doesn't give a shit. Then 11% to 13%, and the stock gives a shit. And then you're at 15%, and it doesn't give a shit. Then the activists whine.

The problem is the physics. We talked about this with Salesforce. If I have a zero-revenue company and I can go to $100 million, that looks amazing. If I have a $2 billion revenue company and I sell them $200 million of the new thing, I'm twice as big as this little sexy startup, but I've only got 10% growth, so I'm valued at 5 times revenue.

You've got to push that Sisyphean rock up the hill for 4 or 5 years, keep everyone motivated, and just accept that it's a grind, right? That actually is the hard part, and that's why I admire the CEOs. I admire Benioff for turning up and keeping doing it. I admire Dines. I admire Aaron enormously for that.

They're just saying, "We're not going to roll over and die, because we can be relevant, and we're just going to do what it takes." But the point I'm trying to make, more than that, is there's not a moment of, "I've invented the magic thing. We're cool too." No, the market will say, "No, you're not cool. You're a $2 billion boring old company."

You've got to make yourself cool, because 30% growth—that's what's cool. It's just that journey.

30% growth, that's what's cool. I love it. Daniel is also one of the biggest warrior CEOs—the Romanian cockroach founder.

Yeah.

Rory O'Driscoll

Just relentless. I mean, Daniel is just extraordinary. The man survived on a dollar a day in Romania for years.

You mean he was on drugs? That's a Romanian pre-'89 joke.

Yeah, a pre-'89 joke.

10. Boom Bets on Hard Tech

Listen, you said about attaching to budget and to AI budget. There are 2 that you could pick on for that from this week: Boom Supersonic, the plane builder that gets an order from Crusoe and is raising $300 million in the back of this to fund it and to open up this new line of business, or Harness, which raised $240 million at $5.5 billion to automate AI's after-code gap. Which one do you want to take? Because both of them are...

Rory O'Driscoll

Harry, anyone who chooses between a supersonic plane and a software infrastructure provider and chooses anyone other than option A has no soul. Of course you've got to talk about the supersonic plane. For God's sake, man, this isn't even a choice.

Just for context for everyone, Boom is an awesome company. They are building a supersonic airplane from scratch, and they're designing both the plane and the engines. This is a hard, hard, hard task. Even for existing plane manufacturers, typically Boeing builds planes, and GE and Rolls-Royce build engines. So they're taking on the full enchilada.

I love it. It's a Y Combinator company. They've got a prototype out with less than 100 headcount. Everything in your heart wants that to work, A, because it would be great to have supersonic flight other than from Concorde, and B, just because it's an awesome entrepreneurial story.

When I saw this, I will admit, I was like, "Huh." But then you do a little research, and I'd forgotten this, but Rolls-Royce—it's not as crazy as it looks. What Boom said is they're also going to sell their engines to data centers for power generation, right? When you say it like that, it's like, "Huh, you were making planes and now you're selling turbines?"

But in fact, all the airline engine manufacturers—GE in the United States and Rolls-Royce in Derby in the UK; I used to run a manufacturing company near there—they all do the same thing. It turns out the bulk of what it takes to build an airplane engine, other than the last bit of propulsion, is very similar to what it takes to build a generator.

My guess is—this is, Jason, Captain Obvious—it's a lot easier to take an engine and plop it on the ground and have it generate electricity than it is to put the same damn thing in a plane and have it generate propulsion. So it was actually not crazy at all.

I don't know if it's defensive because the plane has a huge regulatory hurdle, or if it's offensive because they can make more money there now. But genuine comment: I wish them literally all the best of luck.

Jason Lemkin

Well, hold on. Boom comes out of YC, a cool company. It quickly raises at an arguably fake billion-dollar round from airlines with no revenue. In December 2024, a year ago, it crashes. Down round, maybe a $500 million valuation, but it might be worse when you think about the cram-down and the effects. That's a tough moment. AI booms, and now you're back up to $1.5 billion.

Rory O'Driscoll

Yeah.

Jason Lemkin

In 2021, $1 billion, no revenue. It crashed to half of that or less at the end of last year—a massive cram-down round—and then back to $1.5 billion. If I was an investor, I might still have whiplash, but it's great. Who knows how it looks with all that change? I'm not sure. Crazy story. There are so many of those. We all have them. We all have the couple of ones in our portfolio that actually got a big AI lift this year. Didn't know it would be Boom, but it was. Boom, boom.

Rory O'Driscoll

But you must admit, it's one thing to get an AI lift because your cool little network-monitoring tool is selling to Anthropic. It's quite another thing to walk into the factory and say, "Guys, strip those engines off those planes, slap them on a trailer, and make them generate electricity for a data center down in Texas." I just love the hard engineering of it all.

I would argue, I don't lump Boom into the reacceleration. I think the real, interesting distinction is that Boom, in its old and new incarnation, is an ultra-high-risk, very ambitious company, versus the software-centric, known-business-model reacceleration. I wouldn't lump Boom in with some SaaS company that's reaccelerating to 30%. They're just such dramatic risk profiles.

I think the interesting question, and I'm going to tie it into the forthcoming biggest IPO ever, is: how do you think as a growth investor about ambitious, hard-engineering projects like Boom, where, as we've seen, the risk of it going horribly wrong and not building a plane is quite high, and maybe you get saved by building an engine?

So you go, “Oh my God, that is way more risky than anything in even AI/LLM land.” But on the other hand, we’re about to cover the big story of 2026, which is possibly the largest IPO in history: a rocket company. I think the aha is, when those hard tech problems work, and when you solve the technology problem, and if you pick the right problem, you have a wildly compelling business. But as the Boom plane part of the story makes clear, they’re damn hard problems.

Jason Lemkin

Well, they’ve sold zero of either. They’ve sold no planes and no jet turbines for AI power. So it’s a big bet.

Rory O'Driscoll

It is a big bet. Three failed launches in 2007 and 2008—that’s probably what SpaceX looked like, too. And if it works, you’re a genius.

11. SpaceX Tests the IPO Market

Speaking of the trillion-dollar outcome and the big IPO for next year, we chatted about it. We chatted about the $800 billion secondary. Following that, pretty quickly, rumors emerged around the $1.5 trillion IPO.

Rory O'Driscoll

It’s funny, the exact chronology last week. We recorded on a Tuesday. We talked about $800 billion being a high price for that secondary round, and I was like, “Ooh, that feels like a high price based on the numbers.” I’d looked at the numbers and the growth rate and some of the acceleration going on, and then I thought, “Ooh, $800 billion feels high.”

Then on Wednesday, before we even released the damn podcast, we saw the leak that said they plan to go public at $1.5 trillion. So I’m sure everyone piled into the $800 billion round thinking, “Oh my God, it’s cheap.” You can be wrong by $800 billion in a day.

So I’ve obviously been thinking about that a lot. How do you get your head around that? The company’s doing $15–16 billion this year, 2025. Starlink is the growth driver. Growth rate was down a little this year over last year. Maybe they do the early to mid-20s next year. So you’re talking 78x 2026 revenues.

What’s the thought process? I’m genuinely thinking, “How do you talk constructively about that?” What I realized is you have to factor in what I’m now going to refer to as the EOV, the Elon Option Value.

Because on all his public companies, you just run the math. You value what it’s worth as a business with the TAM, take it all into account, and then you look at the difference between that and what it’s trading at. That’s basically the Elon Option Value, right?

I mean, if you look at Tesla, it’s trading north of $1 trillion. It’s roughly $100-something billion in revenue. Earnings are down. You kind of apply a normal multiple, maybe you get $300 billion. The rest is the EOV.

It will be the same thing with SpaceX. And he’s earned that EOV, because he pulled the Starlink rabbit out of the hat. It was a rocket company, and then it became a communications company.

So it is credible to say, in his case, as in almost no one else’s case, this is one of the few people on the planet who literally might find you another trillion-dollar market that you weren’t in, that wasn’t in the original plan. So roll the dice.

If that premium ever evaporates—oh, dear God, if he were ever to die, oh my God, the stock gap would be something horrific.

What I recognize is you can’t run the numbers on SpaceX and come up with $1.5 trillion. You just can’t. But what you can say is, someone who’s founded, let’s be clear, the most successful car company ever, and let’s not forget, is also the founder of OpenAI, which looks like the most successful AI company ever, and the most successful rocket company ever.

You know, we said last week, once you’re lucky, twice you’re good. Three times, you’re freaking amazing. So I’ve just let go. I’ve let go of valuation. I’m like, I can’t figure out how it’s worth $1.5 trillion, but you just apply the EOV on top of the 10x multiple, and away you get. It’s awesome.

Jason Lemkin

You have to manifest it. That’s the key you’re missing. This is what he’s doing. Even Harry’s manifested almost $1 billion under management. You can laugh, but if you have enough behind you, if you have enough magic and proof points—

Totally.

Jason Lemkin

—you manifest. I think he’s manifesting a $1.5 trillion company, and I don’t know if all the kids can manifest it, but he’s doing it.

I can teach him. I’ve learned a lot.

Jason Lemkin

Yeah. Teach him how to manifest it.

I’m manifesting a beach house in Laguna Beach, boy.

Jason Lemkin

You’re a manifester. Harry, you are—people don’t get it—you are an S-tier manifester.

I mean, then Elon would definitely be S-squared, S-cubed, S to the power of 10. I mean, it’s just genuine.

Jason Lemkin

You just have this other thing up. Yeah, it’s just not even playing the same game.

Rory O'Driscoll

You look at the numbers, and you go, $15 billion, plus or minus $2–5 billion of space revenues, which is wonderful but capped. You’ve got these customers. There’s a certain amount of volume—they already have 90% of the volume.

Then you start estimating, well, how big can Starlink be? Dear God, let it be on every United flight soon. If anyone from United is listening, I actively avoid your planes because they don’t have Starlink enough, so please fix that.

So they get all that. They get all the world broadband. You still struggle to get to anywhere close to $1.5 trillion. And then you just have to go, you’re buying a share in probably the only person since the demise of Steve Jobs to literally not do it once, but do it 3 times.

I mean, Jobs did it with Apple, Pixar, and then Apple again, right? Elon’s done it with—oh, I mean, you forget the small one like PayPal. And then on top of that, Tesla, SpaceX as a rocket company, SpaceX as a communications company, then, of course, OpenAI as the leading AI company.

He did SolarCity as well, didn’t he?

Rory O'Driscoll

He did, but that didn’t matter as much. I was just a simple, humble $2 or $3 billion outcome. Nothing among friends.

So, yeah, even rounding out the little ones, you’re just left with it’s the most impressive entrepreneurial record possibly of the last 30 or 40 years, with, as I say, the exception of Steve Jobs. You lean in.

Jason Lemkin

No, the only thing is, I can’t get the math to work either, right? But the one thing is, I do think Elon’s all over the place, unlike Sam Altman, where every single thing I think Sam says, as off-the-cuff as it looks, is very thoughtful.

But I think Elon talking about space-based data centers is his next big play, potentially. I think there’s a reason he’s communicating this well ahead of the IPO. I think this is one of the big plays. I mean, it’s an incredible business as it already is. I have 3 Starlinks. I don’t know what I’m paying. A lot, right? And it’s great.

But if all of that—all that Oracle, CoreWeave, and Nebius revenue we were talking about at the beginning of the conversation—goes up, up in the sky, deep, far in the sky, that is a lot of money, and he’s the only one that can do it. He’s the only one that can build data centers in space.

And I don’t think he’s joking. And it’s easier than going to Mars, or at least it’s easier than getting back from Mars. It is not easier—going to Mars is not a huge challenge. It’s getting back. So we might be underestimating the data centers in space. It’s possible.

Would it be a successful IPO at $1.5 trillion?

Rory O'Driscoll

That’s an interesting question. For the buyers or the sellers? It will be the most successful IPO in human history for the sellers. I just think what Founders Fund are going to record on this is going to boggle the imagination.

Will it be a good stock to buy at $1.5 trillion? Is that what you’re asking, Harry?

I’m saying, will there be investor demand at $1.5 trillion, enough to satisfy it, and then a stabilized period where it doesn’t tank afterward?

Rory O'Driscoll

Gotcha. I don’t know, because you’re asking me to assess something that’s not within the bounds of logic. I don’t understand the Tesla market cap relative to the financials or even the TAM. I don’t think people get that when running the numbers.

So you’re trying to assess the nonquantifiable. I don’t know how you can come to an informed conclusion. I don’t understand the mindset of someone who would buy based on the intangible at that kind of price.

I think as a rational buyer, you look at the upside-downside risk and you get very nervous. But I think the same has been true of Tesla for the last 5 years. It’s been up and down. The financial performance has been fairly mediocre for 4 years, but the stock has stayed up.

So you’re asking me to speculate on the propensity of the marginal Elon believer to buy the stock at $1.5 trillion, and I have no way of assessing that. I think it’ll be a tricky one to get done. The bankers are pitching this week. I would not envy them their task.

Jason Lemkin

He does seem to have an ability that maybe no one other than Bezos in his prime has, which is, “Give me 5 years.” All of Elon’s dates are wrong. Everything is full self-driving. Everything’s behind, from the Model S to the Model 3 to Starlink, and everyone gives him another 4 to 5 years.

The day traders are going to enjoy this stock, so it’s just hard to predict when everyone gives you another 4 to 5 years. They’re really not trading on today’s revenue or today’s anything. It’s a gift, but he’s very good at that. He’s very good at overpredicting, but ultimately delivering.

Rory O'Driscoll

You’re exactly right, Jason. It is a daunting task. Are you going to raise $30 billion, which is only 2% dilution, which means it’s still widely traded, or are you going to do the typical IPO at 8% dilution?

So that would be—I mean, let’s do the math here—about $120 billion, right? Is there $120 billion of raw risk capital that says, “What I really need in my portfolio in 2026 is some 70x run-rate revenue space investments,” right?

Some poor banker is going to have to get on his PowerPoint, get on his private jet, and start flying around the world and saying, “How much of this do you want?” Now, I think in this market, it…

I don't know how to assess who that investor is. Do the index funds buy it? Do the sovereign wealth funds buy it? I think there'll be a lot of consumer and retail appetite. I just don't know how that all comes together.

It will be fun to watch because it will probably be the most challenging IPO story if it's anything like that valuation. The other thing you might see is the valuation getting walked back to the merely outrageous. The most impressive thing is the entrepreneurial oomph. I heard Peter Thiel speak about this years ago.

It's like when you have a big enough vision, you do get a buy on the little shit. And Elon, by having that big vision, I think has been able to paper over the cracks that would've killed many a lesser man. Good luck to him. It's great. In the end, I like my Starlink, too.

Jason Lemkin

Here's one scenario, for what it's worth. Google anchors him with $10 billion. They're already a 10% shareholder.

Rory O'Driscoll

In for a penny, in for a pound.

Jason Lemkin

They may have—maybe it's a Gemini partnership in space that's part of it. They get first access to the TPUs in space. So they put in $10 billion. It's good timing with AI.

As soon as that happens, we're running out of space in the IPO, and the banks run around. Google's in for $10 billion, Fidelity's in for $2 billion, and all of a sudden you start to panic that you're not gonna get your shares.

Rory O'Driscoll

If you get a call in 10 minutes from Goldman Sachs or Morgan Stanley, you'll be signed up as a banker by Friday. I love it. You're exactly right. It's a narrative—maybe the best company. It's a narrative story, and Jason's just shown how easily a narrative can change.

It's not a story like a PE-backed public IPO where it's like, at 8 times ARR, you can get 6× coverage, and at 10 times ARR, you can only get 3, so the deal gets done at 8. This is a totally different thing, and this is in tune with the zeitgeist. It's all narrative all the time. And you're right, that would be a very clever way to do it.

Jason, that was fantastic.

Rory O'Driscoll

Yeah. He should be a banker. He's exactly right.

Jason Lemkin

Well, Google almost bailed out Tesla at the last hour when it almost went bankrupt. It's already happened, and Sergey’s back running Google. It's just $10 billion.

I can tell you, we do work closely with a lot of the folks at Google Cloud with SaaStr. They are feeling strong. We work with a lot of folks in marketing and product. I've never seen a team feel more energized in the entire decade I've worked with them.

They are feeling that they are winning, and so why not put $10 billion to get my TPUs in space? It's a good deal.

Rory O'Driscoll

Fuck, it's nothing. It's a couple percentage points of market cap. Not crazy.

They already have engineers. They don't need to spend it on that. Jensen's doing buybacks, so maybe he can stick a couple of billion in, and there you are.

Jason Lemkin

Maybe NVIDIA could put in $5 billion. If they do $15 billion together—Google and NVIDIA—and $2 billion from Fidelity, there's no room for retail. We're sold out. The round is sold out.

Rory O'Driscoll

The genuine comment and the genuine learning is, as I said, the power of a big story. At one level, there's a little part of me—the boring, curmudgeon numbers guy—that screams, “This is all madness. You're all insane.”

And then there's another little part of you that just gets inspired by the fact that we're sending rockets to Mars, for God's sake, and we're the telecom company to connect the whole world. This has gotta matter.

Jason Lemkin

Just because you brought up Founders Fund owning 10%, it's just a story people forget. Peter Thiel fired Elon Musk.

Rory O'Driscoll

Yes.

Jason Lemkin

He fired him. But he treated him like a human being, fully vested all his stock, took nothing out of his pocket, thanked him for it, and told him that was the way it was gonna be. He treated him pretty damn well, and so when SpaceX was gonna die, he went to Founders Fund and Peter Thiel gave him the money in an hour.

It's a reminder in today's age of extreme greed to be kind. He wasn't just nice to him. He accelerated all of his stock and made sure that it wasn't about money. He was appreciated. He walked him out the door.

I don't see this happening often enough. I've seen 2 CEOs at the end of this year just quit their startups that I invested in. Be kinder. Go the extra yard like Peter did with Elon. It may, just may, pay off. Owning 10% of a $1.5 trillion IPO, even for a billionaire, it might pay off.

Rory O'Driscoll

Yeah. Reserving the right to say I'm not sure how much of it was kindness versus the fact that Elon was also a big PayPal stockholder, I think the meta point you're making is right.

Two things are wildly impressive. One is the density of talent at PayPal—it blows the mind. It's like you have Thiel, you have Sachs, et cetera. You have Reid Hoffman. It's just amazing. You have Elon.

And then the second thing is, you're right. They clearly had their bumps, and the fact that they were able to be rational business people is about to pay off to the tune of $100 billion. It's a great story. It's kind of one of those wow moments.

Jason Lemkin

Mm.

Rory O'Driscoll

And good for them. It's like Harry tweeted today something about people who aren't in the very biggest deals feel totally irrelevant, and you correctly got a whole bunch of pushback on it, Harry, because it was a bit obnoxious.

There's no doubt that in a world where one investment might return $50 to $100 billion, everyone else has to feel just a teeny, tiny tad irrelevant compared to that return for Founders Fund. And good on them. It's a 20-year compounder at 50% plus or minus.

Okay, we're gonna do a would-you-rather, Rory. You love this. You've got Figma at $17 billion or Cursor at $29 billion.

Rory O'Driscoll

Cursor.

Would you rather... Jason?

Jason Lemkin

I'm gonna take Figma only because I don't believe Cursor's future is as short. I think Cursor is magical. I think it's a gift to humanity for folks that have built software.

If we keep doing this podcast long enough, we're gonna see I'm right about something else, which is that everything's much less stable than we thought this year. The categories, the Harveys, the Cursors—these are incredible companies. I wish I'd invested in all of them.

Don't get me wrong. We're gonna look back in 24 months and see there was a lot less stability in these so-called leaders than we thought. It's not that they're gonna go to zero, but we are so early in AI that I'm gonna take Figma on this one. But it's a tough one. It's a tough one.

Rory O'Driscoll

No, I don't feel the need to argue. I think that's the beauty of money. You just make your bet, and then you don't have to justify your position with English. In the end, you're right or wrong on the numbers.

Yeah, Jason. He doesn't need to argue. God.

Jason Lemkin

I like to argue.

Let's shit on Harry's tweet and then just not argue.

Jason Lemkin

I was a little argumentative last week. I hope it doesn't happen again. But yeah, keep going.

Oh, dear.

Jason Lemkin

What's the next one?

OpenAI at $500 billion, Anthropic at $360 billion, or Google at $2 trillion.

Jason Lemkin

I have to take Google. The reason is just the level of confidence—not arrogance, confidence—in where they're going that I see across the team we work with. I know it sounds small, but it's not.

When I see everything's going right and the team feels it and the team knows it, how often does that happen at this scale? I don't think everyone at Meta is cheering that they're crushing it, right? Risk-adjusted, take Google. Wouldn't have been true a year ago. Gotta take it.

Rory O'Driscoll

Okay. I don't know my answer yet, but actually, the point is, a year ago it should've been time to take Google because the stock's up 60%. You may now be on the wrong side of that trade. You should've taken the quick pop.

Jason Lemkin

I'm not a trader.

Rory O'Driscoll

I know.

Jason Lemkin

True or not.

Rory O'Driscoll

They make north of $100 billion a year. So you're buying in at 20 times, plus or minus, EBITDA. It clocks in north of $100 billion of profit a year.

Everything else was a positive when we talked about it 9 months ago, and the risk was erosion of Search. All the positives have kind of gotten better, and the erosion of Search hasn't happened. So, yeah, you do have to give it consideration.

Ugh, this is hard. At $170 billion, I'd have taken Anthropic. You're giving me the future round on Anthropic. Has it happened at $300 billion? But the future round at OpenAI—when would that be? Is there another round coming on OpenAI?

They're continuously raising, but if you were to do live pricing on them, Anthropic's would be $360 billion, unless you were getting grandfathered into a prior round. The real live price is $360 billion.

Rory O'Driscoll

At $170 billion, I'd take Anthropic all day, every day, because I think they're being way more sensible than OpenAI, and I think they will get public. At $300 billion, you might be fully valued.

I think the difference between Anthropic and OpenAI is OpenAI has more ambition, perhaps, but it's more likely to just get caught in the middle with commitments it can't meet. Whereas I think Anthropic is very sensibly and boringly converging on profitability, will go public, and will be a very nice public company.

So of the two, I'd take Anthropic. I'm processing on Google, and it just feels late. You could argue they're all almost public already.

Come on, Rory.

Rory O'Driscoll

Shut your face. I think I'd go Anthropic. I think I'd go one of the pure plays, even though I do agree with Google, I think I'd go one of the pure plays.

Jason Lemkin

Cloudflare, Snowflake.

Cloudflare, Snowflake.

Rory O'Driscoll

I like to think before I comment, so call me strange.

Don't worry, we edit anyway. It's all good.

Rory O'Driscoll

Yeah, yeah, yeah. Get rid of that one.

Boys, thank you so much. You continuously make me more and more questioning of my own questions. I've done this for 10 to 11 years, Rory. I never questioned my own interviewing ability until I started doing this with you, and now I'm deeply forensic on the quality of my questions.

Rory O'Driscoll

I'm just a dick. What can I tell you?

No.

Rory O'Driscoll

No, but it's great. No.