Anthropic向Elon购买算力并承诺向Google投入2000亿美元｜Cerebras IPO｜Ramp以400亿美元估值融资

Are there really enough developers in all of the solar system to keep Anthropic on the unprecedented growth path we have this year?

Kicking us off this week, Anthropic partners with SpaceX. Plus, they commit $200 billion to Google over 5 years. Cerebras’ IPO is 20 times oversubscribed. Ramp eyes a $40 billion valuation in its new fundraise. And finally, AppLovin, HubSpot, and Cloudflare post buying numbers, but the Street doesn’t necessarily share the optimism. What is going on?

Rory O'Driscoll

There are categories of software where, if they don’t have a reason to exist in an agentic world, they will go into a terminal state of decay. If you’re not accelerating, you’re going to be destroyed, right? And at a minimum, you’ve got to raise guidance. ZoomInfo’s growth was stolen from it by Clay and friends, and it’s a brutal case study.

Give me 30% growth, give me profits, give me a story that’s got some future in it, and I’ll get you back to 5 times. I’ll never give you 20 times again. I won’t fall for that one this time. There you go. Ain’t capitalism great?

Ready to go?

We are back. I am so looking forward to this one. As always, we’re going to start with this week in Anthropic. I think I’m going to be really unfair, actually, and just go off on one, because Rory loves it when we go spontaneous.

Anthropic has come out in the last 12 to 18 hours and said, “No. All sales of secondaries and SPVs need to be approved by the board.” That basically brings into question the legitimacy of the transactions that we’ve seen and will be able to get out of moving forward. This is a very big deal, actually, and we’re seeing it impact their price in secondary markets significantly—$200 billion to $400 billion, reportedly. How do we analyze this as an announcement from Anthropic?

Rory O'Driscoll

It makes sense. Not surprisingly, companies like to keep control of their cap table, and they’d probably lost a little bit of control.

Stepping back, the SPV word is a little misleading. There are a lot of different ways shares change hands, and a lot of different ways SPVs are used, so let’s distinguish them.

One is sometimes an SPV is when a venture firm has a big allocation and can’t take it all up. It forms a special purpose vehicle, and that company invests directly in Anthropic, maybe raises money from the venture firm’s LPs, and the SPV appears on the cap table. Nothing wrong, nothing to see here.

But that’s not what we’re talking about here. What we’re talking about here is secondary sales of Anthropic shares—say, outright sales, which we’ll come to in a second—or the transfer of economic value. In other words, I own shares of Anthropic. I’m an early employee. For whatever reason, I couldn’t access the company’s structured tender offer, and I decide to do something outside of that.

The first thing I do is try to sell directly to Jason. He’s willing to buy, I’m willing to sell, but Anthropic says, “We’re not going to effect that transfer. We have rights to approve transfers, and we’re just not approving this,” right? I don’t know what they’ve been doing in the past, but that’s what they say they’re going to do going forward. In other words, they have the right to say yes or no, which makes sense. It’s not unusual.

But the real thing that’s going on—and this is where it gets interesting—is that I could also say to Jason, “Hey, Jason, I can’t sell you my Anthropic shares because they won’t let me, but I’ll contract with you that whatever I get for my Anthropic shares later, I will transfer that value to you.” I will structure a document that says, “I owe you contractually all the money that ever comes from that share,” right?

Now Jason thinks he doesn’t have the shares, but he kind of has the economic right to the shares, so he’s money-good. There are 2 problems with that.

The first is that the Anthropic documents probably say, “Rory, not only can you not sell your shares, you can’t transfer beneficial ownership of those shares to Jason.” So I may not be able to do this. Anthropic could say, “Hey, we don’t agree to that,” and that’s fine.

But the funny thing is, Jason and Rory might still have their contract, because I promised Jason something. Just because Anthropic says they don’t agree to it doesn’t mean Jason and Rory can’t contract.

The tricky thing for poor Jason is that, 2 years later, when we go public, he didn’t have any structure. He can’t go back to Anthropic and say, “My shares are in the cap table. I’m good. I can just sell.” All he has is a commitment from Rory—and Rory turned out to be a liar and a cheat, and he never owned those shares, or he owned them but sold them to someone else as well.

It’s just going to get really messy at that level as individuals start trying to enforce contracts—not against Anthropic, because they’re not in the loop, but against other investors who sold. So there are going to be a lot of losses at that level.

I think Anthropic is just trying to distance itself from that and is probably also worried that, if there have been a lot of transfers, you get into all sorts of equitable remedies. In other words, the court starts saying, “Hey, dude, you knew this was going on forever. You kind of acquiesced to it, so maybe you are involved in this mess.” They don’t want to be involved in this mess.

It totally makes sense as you get ready for an IPO. You’re like, “I’ve been a little sloppy here. Time to tighten up.” The real thing is that we’ll see the second-order impact.

Jason Lemkin

To me, maybe I’m missing something, but this story seems like a nothingburger made up on social media. Let me tell you why.

First of all, all companies have had this for a long time. Most of our portfolio companies have the same provisions in them. This has gotten more and more locked down over my investing career. It used to be that lawyers wouldn’t put this stuff in by default, and then maybe 4 or 5 years ago, every set of charter documents said no transfer without the board’s permission, just like Anthropic and OpenAI.

Nothing new. Second, Anthropic last year warned Memo, one of its best investors, and others: enough with the SPVs. They said publicly, “Enough with the damn SPVs. This isn’t new. They said, ‘Stop the SPVs. If you want to invest, you have to invest directly.’” They were clear on this last year.

I think this is Anthropic just saying, “This is still happening.” They named a bunch of entities in there, which you never see. They named about 5 hedge funds or something and said these guys are bad actors.

I think the interesting thing to me is when folks don’t listen. I don’t think they listened to the company last year. So then you have to go public on the megaphone and make sure everyone hears: “Hey, these are bad actors, and don’t do it,” because people were so greedy that they took the risk anyway.

The triple-layered, quadruple-layered, 20, 20, 20 SPVs—this is the hottest share of the century.

We’re going to discuss the next big deal of the day, which is Anthropic’s deal with Elon. We’ve said before how difficult Dario’s job is in terms of forecasting capex requirements. Rory, do you want to provide some context on the deal with Elon and what it basically means?

Rory O'Driscoll

It just shows: “Needs must when the devil drives,” as they say.

Elon, in the past, has said horrible things about Anthropic: they’re evil, they’re woke, they’re anti-white, they’re anti-Chinese. That was as recently as 3 months ago. Suddenly, he wakes up one morning with excess capacity, and Dario wakes up one morning with a need for capacity. Elon ends up in a trial with OpenAI, so it’s kind of “the enemy of my enemy is my friend.” Here we are, right?

It’s a totally sensible deal. It says a lot about both markets. It says this is a market consolidating. Even though they will deny it until the day they die, what this is is xAI, SpaceX, and Grok basically saying, “We’re not going to be a leading-edge model contender right now.”

Grok is not growing like OpenAI and Anthropic are, and they’re going to effectively switch from being a net demander—a net buyer of capex, because they’re trying to build Grok—to being a net seller, because they’re not going to be able to build Grok. I believe the data center was 11% utilized, right?

This is the market consolidating. The stronger players have the capital to buy more capex. In this context, Grok is a weaker player, and they very wisely are opting to stop, for now at least, the dream. Instead, they’ll take probably $4 million or $5 million a year, which is a big slug, right? Above revenue. Depending on the estimates, $3 billion to $5 billion a year of revenue.

For context, SpaceX’s total revenue runway is around $20 billion. So this is like a 15% revenue lift from something that, a couple of months back, looked like a money pit.

The combination of this and Cursor, I think, has taken xAI from a $250 billion valuation—which is what they paid for it in SpaceX stock, which is astonishing—from a kind of “What are you doing?” to “Okay, I get it.” It’s at least notionally pro forma profitable and has some kind of existence, right?

But it’s not a competitor anymore to Anthropic and OpenAI, and trying to pretend it is is different. You’ve got your own little CoreWeave. Congratulations.

We talked about this a couple of weeks ago, and I said something—I usually don’t use these kinds of lines—like this wouldn’t be hard to imagine, because Samsung does it with Apple.

They compete and sell them components. You're right. We talked about this, and they were ridiculous. No, they would sell it to each other, right? Actually, the most obvious candidate now is SpaceX because they've got to have different BUs, right? So you've got—how can I even keep track?—you've got the rocket guys, the Starlink guys, and now you've got xAI, which is two things mashed together, right? And they each have their own P&Ls, BUs, and responsibilities.

If I'm stressed and I'm running this xAI-Twitter thing and I can get another $4 to $5 billion into Rory, it doesn't just make sense for the IPO; it makes sense for me. It just selfishly makes sense for me. So I don't think the deal is as surprising.

It's a reminder to keep meeting with your partners and your competitors. This is a classic SaaStr post: always meet with the CEOs of your competitors. You never know where it's going to lead. It's never a bad idea to have lunch once or twice a year with your competitors. It's never a bad idea because this is one of those deals.

And 2, the more interesting thing to me is, my God, this changes so quickly. Just a couple of weeks ago, it was like, well, Anthropic can't launch models because it doesn't have enough capacity. And now Anthropic—I hate the 5D, 8D chess metaphor—but now it's figuring out a way to have more capacity than OpenAI. It's pretty epic to just hoover up anything that's available on planet Earth and possibly orbiting soon enough.

But it will hoover up everything available, right? CoreWeave, xAI, anything. It'll probably buy capacity from OpenAI if Sam lets them, for some reason.

Jason Lemkin

You know, capitalism works, and assets should get reallocated to the person who can create the most value from them. Right now, Anthropic can turn that CapEx into the most amount of money the quickest, and Grok could not. So I agree.

The person who is in charge of making the P&L work for the data centers is sure glad to get an extra $3 to $4 billion to make his math work and not get fired by Elon. Pretty happy.

Jason Lemkin

It would be fun, by the way, to watch the SpaceX roadshow because someone's going to have to pro forma it out. Well, this is what we owned a year ago. Then, late last year, we bought xAI. So bear with me. We dropped all this stuff in here. That's what it is today, and we have a whole quarter of combined revenue.

Oh, and by the way, we then sold all that capacity, and that should hit next quarter. So you've got to pro forma that in, and then the quarter after that, we're going to drop in Grok, which hopefully Colossus tool will be online then, and you've got to pro forma that in.

Basically, Mr. IPO investor, you're buying a $15 billion run-rate company today, and it'll be a $23 billion run-rate company in 2 quarters with 2 totally different businesses on top. That's why the bankers are going to earn a couple of hundred million dollars, but it won't just be an extend-the-model-and-grow-20%-Q-on-Q kind of analysis here.

They also did a deal committing $200 billion to Google over, I believe, a 5-year period. To your point, Jason, on Samsung and selling into partners: ultimate sign of a circular economy, ultimate sign of Google's superiority in that positioning, owning both Gemini and TPUs, and now you have them selling to Anthropic. Anything of note there?

Jason Lemkin

To some extent, it plays like Google. It's like, okay, obviously we want to win. We want Gemini to win. We want to beat everybody everywhere, but we're not strong everywhere. We're not. But let the best model win.

We're going to win, but we're okay with winning with internal competition by selling capacity to Anthropic as well. It's okay, at their scale, to let the best buyer of these different things win. As long as—I mean, they don't have infinite capacity—but it's not necessarily a bad way to keep everybody on your toes, to have a little bit of the competition inside of you. It's not necessarily a bad thing.

It is interesting, though. I think it's a couple of things, but one is, I think now the Anthropic revenue commitment is about 40% of Google's total future backlog, right? So it underlines quite how heavily dependent the hyperscalers are on these 2 privately held companies, which are effectively providing all those revenues.

And you are, Jason—reminder, everyone—Google has Gemini, which, in theory, is a direct competitor of OpenAI and Anthropic. And this is Google—one part of Google—giving Anthropic the compute they need to grow. And I think, in one sense, you're in a win-win situation.

I'm sure at the margin you'd prefer Google to be the winner of the model company, because I think—and I could be wrong—I think in the end the value will accrete mainly to the model providers and everyone down the stack that's selling to them, even though they're all making out like bandits today, starting with the memory guys all the way up to the hyperscalers. Over time, that's not the obviously differentiated place. And I could be wrong about that.

Maybe CapEx and the ability to invest hundreds of billions of dollars in a data center is, in fact, the moat itself. But over the long term, if I'm Google, I'm like, I'm happy that I'm doing $200 billion of revenue with Anthropic, no more than Microsoft is happy they're doing $200 billion of revenue with OpenAI.

But deep in your heart, you should be saying to yourself, “God, I really wish Gemini was so busy that they needed $200 billion of compute.” And in Microsoft's case, “I really wish I even had a model that was worth a damn, which I don't.” Right? Because all you're doing is enabling, with your balance sheet, the 2 most exciting next-generation tech companies who are going to draft on your balance sheet air cover and become huge.

Jason Lemkin

It's for sure that—I mean, there's a trade-off. You're enabling your competitor, right? It was interesting: The Wall Street Journal today published the market shares in the enterprise for OpenAI, Claude, Gemini, and Grok. Grok is a rounding error, going to the prior conversation, right? It's not making any progress.

Everything is so multimodal that Gemini went from 27% to 40% market share, I think, in the last 9 months, and Claude went from 21% to 48%, okay? And obviously, OpenAI actually only went down a little bit. They don't sum to 100 because you're multimodal, right?

But if Gemini's gone from 27% to 40%, and Claude, as we know, has gone from 21% to 48%, now Google has both pieces. It's got its own winner going to 40%, 50%, and 60% of the enterprise, and it's got a large share of this other leader, 48%. There are worse ways to solve for revenue growth than having both the fastest-growing players—yourself and your competitor. You get a piece of each.

Jason, I was so intrigued to hear your thoughts on this Goldman piece, where they essentially summarize, saying agents will push token consumption up 24x by 2030. Again, our job is to invest on the back of this and invest in companies that provide these services.

When you heard that, and living as you do with the company structure that you do today, do you agree with that? Do you think that is enough? Do you think it's underplaying it, overplaying it? How do you respond to that?

Jason Lemkin

Well, I wish I had the fluency in numbers that Rory has, but 24x—I know, I highlighted this first—it sounds way too low.

Agreed.

Jason Lemkin

The theme of a lot of the rest of the year is parallel agents. Now, we don't need parallel agents in everything. We don't need 100 SDRs hitting up our 1,000 potential customers every minute. There are definitely plenty of workflows that do not need 10, 20, or 100 parallel agents.

You don't need a thousand flights going to the Bahamas for your vacation. But for workflows that can benefit from parallel agents, it's just kicking off inside of these LLMs and their models. And so, not only is this 24x—what if we have 10 agents? That's 250x, right? And more.

We also forget how underpenetrated enterprises are. It's so early outside of tech, right? So I don't see why it's not 250x, but I have to put it on a better spreadsheet. I think we're underestimating the potential impact of parallel agents.

Still, most of us live in a sequential world. We fire up something—Claude, Claude Code, ChatGPT; it doesn't matter. And we kind of have a human interface where we're doing things sequentially because that's how our brains work, but it's also how the LLMs have worked.

Now that they can natively run parallel agents, we just get these superpowers we didn't have a couple of months ago. And what these parallel agents are doing, if you haven't seen them in action, is that they're even better because, for things like coding, they'll go out and do 10 different versions of the same feature, of the same iteration.

Then the LLM will decide which is the best of the 10. It can present you with the 2 or 3 best options, and you can approve it. So why build a feature once if you can build it 10 times? Have the LLM decide which of the 2 or 3 expressions of the feature is the best, and then you pick the best combination of the expression, right?

You can see hints of it in image generation. You use some image generation; sometimes you get 4 images, right? But what if you were in real time blending the best of all of these with much more complicated workflows?

So 24x just sounds conservative. I think that's why Anthropic's right to buy every TPU, GPU, every Cerebras chip, 12-in chip they can buy.

Yeah, I think the token count is always misleading, because I've tried to do these numbers, Jason, and it's so hard. I saw the summary of the Goldman report; I haven't read the detail, and I really want to, because I was trying to think through the same stuff myself.

Every 18 months, the raw performance of the chip gets roughly 3x faster. Then, on top of that, other optimizations in how they run LLMs, how they do quantization, and all the other clever stuff that you can read about and try to understand gets you another factor. So, you're probably roughly 10x-ing the number of tokens per unit of money every couple of years.

If you weren't using more tokens, I would expect total revenue to be going way down. It used to be that if you only needed 1 million tokens and it cost you $5, and the prices are down 10x, then for the same number of tokens at the same efficiency, it would be way, way cheaper.

But what you have on the other side is that the systems are getting better, so you use more tokens to get better results. What you see, as Jason said, is that the token count to support a chat 2 or 3 years ago, when you were interacting with an LLM, was whatever X is. It's 10x that now to do some kind of simple co-work analysis, and it's 10x that again if you're doing coding. It's 10x that again if you're doing parallel work and moving toward where these agents are going.

You're just going to see the cost per token go way down, and the value and use of those tokens go way up. The interesting thing is that you're dealing with 2 numbers, one on each side, both moving an order of magnitude every 18 months. Trying to forecast where the net multiplicative effect of that comes out is hard. I wouldn't fool myself into saying that you can be wildly accurate on that. You can get the rough direction, but it's hard to say, "I know exactly how this is coming out."

Jason Lemkin

There's a growing counterargument here. When you talk to some of the best CTOs and engineering leaders, they say that we don't need as many tokens as we think. If you talk to people who didn't just get their team going in the last quarter or so on Claude 4.7, but have been deep in this for a while and have gone through multiple releases with their whole team being AI-pilled, there's a theme that we actually don't need this much code.

We don't need this many lines of code. It's too much, and we cannot process all of it. This is just a lot of token-maxing that's unnecessary to deliver what the end customer needs. We're all learning, and we're all excited about these tools. They work, and people are running Claude Code and Codex 8–10 hours a day, but it's not necessary, guys. You're producing too much code that's never going to be committed to production, and you're wasting your energy.

There's the whole token-maxing at Amazon, where people are pretending to work because they have quotas. But this is different. This is some of the smartest people saying, "The folks that need this much Claude coding are the mediocre web heads, the web developers who don't know what they're doing. These are people a little bit better than Lemkin."

How do you think about that with respect to Mike Cannon-Brookes at Atlassian saying, "Our demand for new software, new products, and new features is infinite, so we will continuously need a labor supply of great developers. We will continuously need more tokens because the demand for new technology and new products is infinite"?

Jason Lemkin

But the question is, can the average Atlassian engineer effectively consume $10,000 or $20,000 a month of tokens and be productive? There's a growing micro-backlash that the best developers and engineers don't necessarily need $20,000 a month, and that this is going to fade.

We're not going to go back to handcrafting code, but we don't need to be running Claude Code 10 hours a day. This is a bad way to ship enterprise-grade software. It's a great way to ship hacks and proofs of concept and impress my boss, but do we really need this many lines of code a day? Do we really need it?

Hang on. There's a lot to unpack. I'm a little tentative because it's more my infrastructure colleagues who are doing this, but I've been talking to them about wrestling with this question. I want to pick this apart, Jason, because there's a lot to get into.

One is this whole idea of monitoring people's token consumption and what impact that has. There's an economic concept called Goodhart's law, from an LSE professor, which basically says that whenever you monitor a variable, you actually change the causal relationship of that variable. Which is a way of saying that if you tell people, "I'm going to monitor your token production and how much you use," you will, in fact, distort the result.

That's what Jason is hinting at. I think we saw it at either Amazon or Meta, where employees were internally just burning tokens on stupid tasks, which is easy to do, just to make sure they, quote unquote, made their quota.

To assess the market for these companies at a high level, you do have to have some mental model of how much LLM spend—which, for tokens, is a tricky but meaningful proxy—is going to be relative to salary. Does the average developer spend 2% of their salary dollars on tokens? Is it 5%? Is it 10%? It's a huge number, and it's a very important number.

The odd thing is that the more these big companies target their employees based on it, the more likely the employees are to distort the outcome. That's the first big-picture comment, and I agree that there's a ton of that going on.

To some extent, when people find that out—and I think they will—I think there is going to be a push to get more control over cost, because there has to be. If you think back to December, when Anthropic was at a $9 billion run rate, very few of those CIOs had in their budget, "Oh, by the way, you're going to spend 10x that next year." It's a big sum of money.

Someone's going to have to find $50 billion or $60 billion of budget across U.S. corporates, and that's real money. What that means is that there's going to start being some kind of pressure on where this money is being spent, even if they still want to pursue this AI-max concept. So, I agree with you, Jason. On that, I agree.

The thing I don't know enough about, and I'm just going to ask you because you said it, is this backlash: do the best engineers not need as many tokens? I hear you, but I have 2 questions on that, and they're questions I ask my infrastructure people, where I don't have the answer.

The first is that I've heard them articulate a perspective that actually says it's the opposite. The very best engineers can, in fact, manage these tools because they have the conceptual vision of what they're trying to build. They can actually be more productive with these tools. The less productive you are, the more yield loss you're getting, in the sense that you're using tokens that aren't turning into effective code.

Jason Lemkin

Yeah, but I think both are right. They really are becoming 100x engineers, so let them use whatever they want. Give them all the tools in the world.

The concern is that there's a kind of snarky tone from web developers and others—folks 1 or 2 steps above me, or the mediocre people on your team—who just aren't that good. They're web developers, and they're consuming massive amounts of tokens for relatively low productivity gains.

It's not all performative, like the Amazon thing. Some of it is an attempt to keep up, but they need so many tokens to contribute so little value. It's 100,000 lines of code that some folks made up on Twitter. Do you really need 100,000 lines of code to run a blog? Maybe you don't.

Jason, you're right. Which is why the question I'm always asking the rest of my team—and I don't know the answer to—is, what is the objective? How do you think about measuring this?

You're right: lines of code is a dumb measure, because this stuff grinds out lines of code. You need some kind of conceptual effective lines of code, and I haven't found anyone who has it.

In fact, we just surveyed, I think, 30 of our VPs of engineering to try to understand what's going on in terms of spend. They're all spending a lot, and they all think they're going to spend more. But you're right, I didn't get clarity on the heuristic for success, and there has to be one over time.

Jason Lemkin

What's hard is knowing what the ratio of web developers and token-trashers is to 10x or 100x engineers. It just goes to the point that I'm not smart enough to answer the question.

It goes back to the question we asked before: are there really enough developers in the entire solar system to keep Anthropic on the unprecedented growth path we've had this year? Probably. But the counterargument is that these web developers are trashing tokens, and we're going to see this play itself out.

We're not going to, a year from now, waste tokens on mediocre web developers playing with stuff. We're going to clamp down on it because it's a huge waste. We're going to give the S-tier guys all they want—the 100x engineers—but that's always been true.

Probably there is. Also, Anthropic released 10 financial agent templates, killing a load of YC companies in the process of doing them. In a couple of weeks, they're scheduled to come out with a legal product that will challenge Harvey and Legora, apparently in a very meaningful way. So there probably aren't enough developers to satisfy the insatiable market cap increase, but there are when they take legal, and then when they move into financial analysis, financial modeling, and everything in between.

Jason Lemkin

It might be. And look, I don't want to go too far. We're still trying to see—we talked on the show about Claude design, whether it was a killer, right? It isn't a killer yet, right? It's hard to predict some of this stuff.

Listen, there have been many YC and other startups that have been destroyed by a Claude feature. I've invested in 1 or 2. Being able to innovate faster than Anthropic is tough, okay? It's pretty effing tough. This is not a classic slow-company, slow-pace company, right?

Having said that, if I'm a lawyer billing $2,000 an hour and doing this, I don't want to take a risk. It's not that I'm not going to ask Anthropic and ChatGPT my questions; I'm going to ask questions too, in addition to Harvey or Legora or other tools, right? But I don't know. I don't want to take any risk that this doesn't have the level of domain investment in anything that is close to regulated or has—

I mean, you can't—people already got the memo from The Wall Street Journal. You can't submit briefs to a court with hallucinations in them. Everyone gets that that's a problem now.

The question, though, is: are these guys going to take away everyone's horizontal and vertical businesses? I don't know. I think what Jason said is that the model companies have a lot to do: building their models, building broad horizontal harnesses, and building products like CoCounsel, right? It's not clear to me that they'll be able to have the time and the focus to do all these specific verticals, nor should they.

Rory O'Driscoll

I agree, but I think there are also very cool verticals, like customer support, legal, financial modeling, and accounting, where they are mega and very clearly winnable.

I don't think there's any chance that Anthropic is going to do applications in CX. Actually, I could be wrong. I will bet you a lot of money they're not going to go all the way in legal, either. That's why Design was interesting: it was an application, okay?

What does Harvey cost on average? $150,000 a year per law firm? I don't want to roll that into the $200 a month I'm paying for Claude if there's any risk. It's just not worth it. I need a solution. I need it to do everything. I need it to integrate with DocuSign. I need it to prepare the brief properly. I need it to review it a different way.

It's just not worth it. This is not a career where saving a few pennies is worth it, outside of the low end of the market. Maybe the ambulance chasers do use Claude, right? That's fine. But I don't. I may be wrong next week, but I'm not confident that these verticals are going to put the resources into building an application.

That's the thing. They're not going to build a CX application. They're not going to rebuild Decagon, Sierra, Fin, or Gorgias, or any of these others. It's not that they won't build chunks of it. It's not that they won't take away pieces. It's not that OpenAI hasn't taken away pieces of ElevenLabs or other pieces of other companies. But we haven't seen them commit to building applications. That's different.

If the LLM can express what an application does, that's where the YC companies get killed, because all of a sudden you don't need an application. It just works in the prompt, right? That will kill 1,000 startups. It already has. It's killed 1 or 2 of mine. One is on its third version because it was super innovative a year ago, and then it just became built into Claude today.

Rory O'Driscoll

Step back here. This is not the first time you've faced this question. If you look at every single platform, there was a dominant compute-level provider, and the million-dollar question is how much of the app layer they take over, right?

Let's just do 2 obvious ones. Microsoft in the '90s: they dominated. They were the operating system, so they dominated broad horizontal application software for the consumer and the individual knowledge worker—the Office suite. I think Co-worker could be like the Office suite, right? They also dominated networking at the infrastructure layer, just connecting things.

But there were hundreds of application software companies on top of that—Siebel, Vantive, Scopus, Clarify, Baan, and SAP—that said, "We run on top of Microsoft. We build specifically for this vertical or this horizontal use case." Microsoft tried. They bought Great Plains. They never really made it happen.

Fast-forward a decade later—hang on—Amazon, right? Same kind of question: does AWS eat everything? I know lots of brilliant investors who passed on Snowflake, which wasn't even an app; it was an infrastructure-level player, because they said, "Oh my God, Microsoft and Amazon Redshift are just going to eat their lunch. It's not going to be a thing." And it turned out it was a $50 billion thing, right?

So my point is, you have to have some approach to thinking about this, but you have to face this question every compute revolution, and this is just the latest turn of the crank, right? To me, the default is the Microsoft outcome, which is broad horizontal—the equivalent of broad horizontal compute is now broad horizontal intelligence. That's going to be provided maybe not by the monopoly like Microsoft, but by the oligopoly of Anthropic, OpenAI, and maybe Gemini.

Microsoft will probably regret until their dying day that they let this slip and let Claude Cowork take their lunch. But that's their problem, not mine. At the app layer, I'm kind of with Jason. I think all of these individual apps—especially, and this is a key point, especially if you take legal—are different.

If all you're going to do is mark up a document for an individual user, I think there's an argument that an individual user might get a skill from Claude, and they'll be fine, right? But most of these companies are selling coordinated workflows across an enterprise, and I think once you get to that point, I'm with Jason.

I think Legora and Harvey, if you're selling to AmLaw, and GC AI, which is ours, if you're selling to corporates, are going to want both the relationship and the ability to customize it to what you want. There's just going to be a whole bunch of work that's better done by a focused firm. I think history's on our side. That's the way it shakes out.

I would just offer 2 thoughts, maybe. One, this is more OpenAI than Anthropic. Sam did hire Fidji to be the CEO of Applications, and there is no longer a CEO of Applications. Now she's CEO of AGI. So that's really walking it back.

You could say it's a little bit different, but just think about hiring a CEO of Applications and saying, "You know what? That's not a business we want to be in." Anthropic doesn't seem to be investing 100 or 500 people in design, right? Which they could. That could change next week, but right now neither of them seems to see applications.

Paradigms do shift. On Rory's point about the low-end redlining or whatever, that could get more powerful. I'll give you a different example. It's not the same, but I don't think in a year we're going to need any traditional marketing automation software. It's too dated. It doesn't work for agents.

There are categories of software where, if they don't have a reason to exist in an agentic world, they will go into a terminal state of decay. Agents don't need HubSpot, Marketo, Salesforce, or any of these marketing automation tools because they have no need to hand-compose an email in a third-party template. I'm not saying that would happen in legal, but you could see that if they don't keep up and the paradigm changes, you can become obsolete over time.

Rory O'Driscoll

There are 2 separate dynamics. One is: does new software get eaten by new models? In other words, how will Legora be eaten by Anthropic? You're referencing something different, which is: does old software get eaten by agents—either Claude agents or new agents?

New software can get old, too.

Rory O'Driscoll

Yeah, so you're conflating the two, but yes, I agree, it can. I just think they're different questions. I'm just trying to distinguish them.

I just think the rate of decay might accelerate in the agentic era. That's the connection I didn't make. It used to take a decade; now it could be 18 months. If you talk to senior folks at Lovable and Replika, you will hear they're well aware of this. They do not want to be Claude'd out of existence.

They think about it every day: they have to stay ahead of it because they're closer to the core, right? I just think stuff gets old much faster than it used to. The evolutionary pressure from a model that's underneath you—an intelligent model that's underneath you, a company building that intelligent model—to not enter the market is much more powerful than the pressure of an operating system or a compute system like AWS.

Jason Lemkin

Every day that you wake up as Lovable or Replit, the model provider underneath you is doing more. You said you can’t—if you could be behind by a year competing with Redshift, you can’t be behind by a week competing with these guys.

I’m going to bring some semblance of structure to this. Jason, you said HubSpot. We had some public-market activity: HubSpot crashed 20% despite decent and consistent growth. We had AppLovin at a $7 billion run rate; its stock crashed. And then Cloudflare beats and lays off 20% of its base. Which one do you want to pick on first there, guys?

Jason Lemkin

Maybe it’s not as exciting as the Anthropic wars. I just thought the slight contrast that was interesting was Monday versus HubSpot. They’re both decelerating. They come out with a quarter, and Monday trades up after being maybe the most beaten-down stock out there. It trades up, I think, 20%, and HubSpot is down 18%. What’s the difference?

They’re actually not much further than each other on their agentic journey, which is early. Monday’s in production and HubSpot’s sort of in, but they’re pretty early. All Monday did, that I can tell, was for the next quarter, they’re still decelerating, but at least they raised the guidance for real. HubSpot couldn’t do it. HubSpot lowered its guidance.

So, if you’re not accelerating, you’re going to be destroyed. At a minimum, you’ve got to raise guidance. Even if you’re not accelerating, like Monday, you’ve got to at least raise your guidance to get some breath, because at least it says, when Monday says, “Hey, we’re raising our guidance next quarter,” we’re not going to zero. We’re not being destroyed by AI if you’re raising your guidance.

But I think even though the SaaS apocalypse is behind us, for many, there is still a worry that the terminal value is zero. I don’t think that fear has gone away. If you keep decelerating while budget is accelerating everywhere around you—it’s not that everyone’s decelerating. The budget is accelerating and you are decelerating. Those aren’t 2 good lines to cross over. They’re not crossing over each other. But at least it’s nice to see some bounces off the hard deck.

Rory O'Driscoll

I agree. I think it was a super interesting quarter. Lots of different people reported and did different things. Some of them cut expenses. Some of them had decent quarters. And the stocks—we had all sorts of weird movements.

Cloudflare had a good, really good quarter. I think it was a mid-30s quarter. I meant to say, yes, it laid off a bunch of its base. The stock went down. You had HubSpot, a decent quarter, for the record here. You had decent growth; the stock went down. You had Monday earlier on, a little bit better. Bill.com, where I used to be on the board, unfortunately also had to lay off a bunch of people. The stock bounced up because of a buyback. So, you look at all this and go, “What’s going on?”

I think there are really 2 things, and Jason’s done a really good job of articulating the important one, which is: Where are you structurally? Where’s your business structurally? Are you getting better or worse in your business? Is AI messing with your head? Is AI—are you accelerating? Are you on that, as Jason said, thinking about Monday and HubSpot? Are you on the agentic journey?

One of the things that happened, interestingly enough, definitely to Cloudflare and possibly to HubSpot, is a little bit of the, “Oh, my God, you’re cutting 20% of your costs,” which had been perceived as a net positive. Now there’s a little bit of, “Oh, Cloudflare, you’re a good company—WTF, what’s going on?” A little bit of uncertainty. What’s going on with that?

But if the first big bucket is that whole question of what’s going on in your business, then separately, it’s going to sound really Captain Obvious, as Jason would say: The second half of this is you just have to look at price.

In other words, if you have a modeled quarter and your stock is already at the bottom, you might get a bounce. If you have a tricky story and your stock is like Cloudflare or AppLovin—and those 2 were among the 2 highest-valued companies—if you’re going down 30% or 40%, 30% in the case of Cloudflare, but you have a story that’s even a little bit messy, people are like, “I might be okay with this at 4 times; I ain’t okay with this at 15 times.”

Sometimes you forget that there’s both the strategic journey and then how price is dealing with that. I think what you saw in the case of AppLovin and Cloudflare is that high-priced stocks at the start of a paradigm shift, no matter how amazing the quarter is, are just vulnerable to disruption. Whereas when you’re trading like Monday—poor Monday was at under 2 times revenues with a bunch of cash—you’re like, “Pretty much anything you do other than burn the company down, the stock goes up.” Excuse my language.

Jason Lemkin

It was basically priced at nothing because at one point it was basically almost 1.5 times cash.

Bill Gurley

It’s basically just saying, you know, any positive momentum will get rewarded. It’s, as I say, very much Captain Obvious, but price is the vector, and we forget about this often on the private side because you’re not really dealing with price on a day-to-day basis. But on the public side, that’s how it works.

Jason Lemkin

I think the most brutal one, and I love Henry, but the most brutal one that maybe was underdiscussed was ZoomInfo. The reason I bring it up is ZoomInfo also tanked when they were growing 1%, and I think they guided to negative revenue growth going forward.

ZoomInfo just won the pre-AI game of sales intelligence: data on your customers, even just basic stuff—emails, phone numbers. And then agentic—really, Clay and others, they’re in some ways pre-4.5. They’re not epic products in some ways. We use Clay, we love it, but I wouldn’t say it’s epic. I would just say it’s LLM-infused. It’s AI-infused.

But that’s just a 1-to-1 loss of dollars. What’s Clay doing? $200 million, $300 million, right? Something like that. It’s hard to say. So, ZoomInfo’s doing $1 billion-something. That’s AI taking all of ZoomInfo’s growth away from it. All of it.

I only bring it up as a tough case study of how, even if you’re treading water, you can turn around and these AI companies—they’re not all Harvey and the Goras. These kids can just take all your growth from you. I think ZoomInfo’s growth was stolen from it by Clay and friends, and it’s a brutal case study. It’s a brutal case study.

Bill Gurley

Agreed. And for the record, I think the Clay story is amazing, and it’s much less, to me, an AI-first story. It’s even simpler than that. They built a waterfall product that allowed them to optimize among multiple different data providers and allowed RevOps to waterfall a bunch of different data providers and pick the best.

What that means is, instead of ZoomInfo being the only game in town, you can compare and contrast 5 or 6 data providers. Data then becomes a commodity, and you realize your business is commodified.

On top of that, they’ve done a great job—much better than anyone else—of building Claygents and having an AI story. They’re really, to some extent, a pre-LLM company, which makes it even more impressive. The funny thing is, it’s a pre-AI company at its core, with an initial value proposition that’s morphed brilliantly.

You’re right: It’s sucked all the value out of the data providers. Those guys need to figure out how to become relevant in an agent-first world pretty damn quick.

What happens to a ZoomInfo growing at 1% a year?

Bill Gurley

Nothing stays growing at 1% a year in the public markets for a long period of time.

Jason Lemkin

It gets bought by PE. It does a take-private.

Bill Gurley

You know, “bought by PE”—I always resist that because it’s such a lazy man’s approach. The implicit assumption is that, no matter what happens, you can get bought by PE. That might not be true going forward.

Though, again, contradicting myself, a friend of mine used to say, “Price clears all markets.” There’s a price at which PE will buy something like that and say, “We’ll do the hard thing for 2 years to fix it.” It just might not be a particularly compelling price.

Jason Lemkin

It’s trading at 1 times revenue with 35% adjusted operating income. It’s trading at 1 times revenue.

Bill Gurley

Yes. Yes, exactly.

Jason Lemkin

1 times revenue with 35% adjusted operating income—that is a classic take-private, if you can put the—I mean, you’d have a hard time finding someone, I think, better than Henry, but that’s the counterargument, right?

On paper, that’s the classic take-private: trading at 1 times revenue, 30% adjusted operating income, reasonably stable, not adding customers, but reasonably—it’s net-net neutral on customer growth.

Bill Gurley

I mean, we just said it 2 minutes ago vis-à-vis Monday. Get your act together for 2 quarters, get a small amount of AI-enabled growth, and you probably won’t ever go back to 30 times or whatever absurd number you’re trading at in ’21. But it doesn’t take a lot to get you to 2 times, and that doubles your stock, to state the banal.

Yeah, they’re not—I mean, the big-picture comment is this.

Rory O'Driscoll

The market is not asking these old-school $500 million-to-$1 billion-revenue SaaS companies to become the next Anthropic and double, treble, and 10× every quarter. They're saying, “Give me 30% growth. Give me profits. Give me a story that's got some future in it, and I'll get you back to 5×. I'll never give you 20× again. I won't fall for that one this time, but I'll give you 5 or 6× on one miss, right? Provided you have the growth, provided you have the Rule of 40 and the profits. That's what you can do.”

Now, when this goes live, Cerebras will be going public. This is one of the most hotly anticipated IPOs of the year. In terms of oversubscription, it's 20× oversubscribed. They've bumped the range from $115–$125; it's now $150–$160. The offering will raise $4.8 billion, valuing the company at $48 billion fully diluted. How do we think this IPO is going to go? We've spoken before about Andrew being fantastic. How's this going to play out, guys?

Bill Gurley

Look, it's going to go great. The fact that they've raised the range—which you only do when you're highly confident, especially by that much—means that you've got a killer IPO on your hands. There's no more new information. It's going to go out and trade amazingly, right?

Now, it's an entirely separate question: how's it going to do 2 years from now? That's a business question, but the technicals—I'm not a banker—

When you say it's going to go out and trade amazingly, do you mean comparably to a Figma, which had an incredible, parabolic pop?

Bill Gurley

No, let's put it this way. I think what the bankers would say in the boardroom, when the pricing committee starts asking them, “Are we leaving money on the table?” is something like this: “We believe that at this price we'll have that nice 20% pop. Everyone will be happy.”

I know we should hate the pop. Thank you, Bill.

Bill Gurley

But they're probably trying to get that perfect IPO. Can it run away from them totally? Can you have a Figma phenomenon where retail piles in? Entirely possible. Nothing excites the mind like some of this AI stuff. It fires the public imagination. There's a dearth of opportunities to play.

It's entirely possible that you have a whole host of retail demand that you can't forecast, and in the short term it runs away from you. But a reminder: when Figma popped to $100, when it priced at $35, I think I said, “I think it's worth $35 or $40.” Now it's actually significantly below that.

So you can't control the weirdness of retail. As we've discussed, it's impossible—I mean, it's impossible to try, and frankly, it's also impossible to let it drive the narrative, right? I really feel for Figma in that their narrative is this: “Oh, you had $100, now you're down 80%.” No, you're not. You were priced at $35, and now you're at $20. It still sucks, but it's not 80%, right?

Same thing here. Who the hell knows? Will it do the normal pop, or will it do something crazy? But fundamentally, it's going to go out priced well and trade to the upside, because otherwise these bankers would have been manifestly incompetent to do that raise, and they're not manifestly incompetent. They're smart dudes.

Jason Lemkin

Well, look, I of course agree with Rory. When they raise this much, there's no question people are going to buy. They're going to buy into the first day, so you should get a pop. Maybe there are examples where the range is raised this high and it doesn't happen historically, but in my limited experience, it's almost built into the system, right? But we just have to see.

Listen, it's a fun one to watch because, on the one hand, demand for inference is infinite, right? It's great. On the one hand, they've got support from OpenAI and Amazon and everyone, which they didn't have when they tried to IPO last time. Right timing, right partners, right-seeming backlog, right commitments.

But it's competitive, right? Everyone's going to buy every solution, and if NVIDIA or Groq is better and they can get the chips, they'll use it. So it's hard to predict when there's this much explosion of demand and when there's also hedging happening. There's hedging for capacity, hedging for performance, right?

I just think this one is impossible to predict, but it's a great derivative play. It's great to IPO before Anthropic and OpenAI IPO, too. It's a great time because I get into that zeitgeist. It's more interesting than CoreWeave, which is a data center, right? This is real technology that's fueling inference, but it's so damn early to really know how it's going to go.

This is not 2 years of massive Cerebras chips used in production in data centers, proving a massive competitive advantage. It's early. So how the hell do we know where it's going to be in 2 years? How the hell do we know? I might take my profits. I just might take my profits.

Eric Vishria

Hard to argue with that. And because when they went to go public, and even now, the historical revenue is very much concentrated on a couple of customers from the UAE—the United Arab Emirates, right? What they're leaning into here going forward is the contract from OpenAI and a less fully fleshed-out contract from Amazon. So you're right, Jason: you are leaning into a future that's not like the past. To that extent, you have a lot of risk going on here.

Yet at the same time, their story—and you read the CEO letter or the founder's letter—is great. What they're selling is speed, right? What they're selling is that their inference can be faster than anyone else's. I love the tagline in the thing: “How much would you have to be paid to have a slower internet?” And you just don't. Once you see speed, you don't like to go back.

One of my companies, Tabula, was mentioned deep in the S-1 because we use them for our inference. We have these AI humans, and you need real-time responsiveness. I think there is a focused market for that kind of real-time, blazingly fast inference that they can maybe have over the medium term. That is the medium-term bet.

If they are the first really reasonable, competitive alternative to NVIDIA's solutions—and NVIDIA is a $5.5 trillion company, just roll with me—being priced at $48 billion, if you take a 5-to-10-year potential view—

Jason Lemkin

Yeah.

That passes the Monday partner meeting test.

Jason Lemkin

Agreed. Now—

Doesn't it? Yeah. Thank you, Jason. I'm glad you—what was that?

Jason Lemkin

10% of the fund in. Cathie and ARK should be into this deal. It makes perfect sense, right? You can't argue with the upside.

Rory O'Driscoll

We're losing you there, but yes, agreed. I think that is the sound bite in a nutshell, Harry. It's like: the other guys are worth $5 trillion. You are one of the only ways that you have an at-bat against them. Are you worth 1% of that? If your probability of making it is 10%, and you get that, your expected value is positive.

But it's still a risky way to make a buck. I totally see how you get there. And, you know, just to step back—

We don't do risky ways to make a buck, Rory, so that's fine.

Eric Vishria

No, I hear you. That's my point. But, yeah, just to say it: great achievement. These guys started, I think, in 2016. It wasn't obvious then.

The one thing I thought on that achievement—sorry, Jason—is the 20% ownership. What is it? Sorry, they have 20% ownership—sorry, Benchmark.

Rory O'Driscoll

No, they don't. You just listen to Twitter, but if you go and actually look up the S-1, they have 8% or 9% ownership, right? I do these things because every VC reads an S-1 the same way you read the front page. You figure out what it does, and then you type on the end, push on the index, and then you go to ownership, shareholder ownership.

Foundation, Benchmark, and Eclipse all have an incredible 8% or 9% ownership. To hold on to that after 8 or 9 years in a wildly capital-intensive business that even in their own S-1 said, “Oh my God, we were early in '21, '22”—I just think it's an amazing achievement. Huge credit to Steve, to Eric, to all those guys. Seriously.

To be fair to Eric, that's a $4.5 billion gain on a $500 million fund.

Rory O'Driscoll

20% to do. No, it's an amazing result.

It is. I got to tell you, though, when I looked at Cerebras—this is 20VC, the show, right?—for the first time in a long time, I was kind of jealous of Foundation.

Because they did the hard work. Steve and the team incubated this company. You saw on Twitter the barbecue, okay? Not only was it not obvious, this isn't even after it gets the kudos in Y Combinator or Foundation socializes the deal. This is what VCs are supposed to do, right?

No one does this in venture. No one goes out and finds this really smart guy, plays tennis with him for a year, works the deal, seeds it, incubates it, and then does it even in a crazy category that didn't totally make sense in 2016, and then, wherever this thing ends up trading, has a $40 billion IPO.

This is what actual venture capital is. My job is to do what Steve and the team did. I'm jealous. I put this in quotes—I'm not literally jealous—but this is the job of early-stage investing. It's what Foundation did.

This is not using Marc Andreessen's brand to muscle into the B round.

Jason Lemkin

Well, without trying to shade the others, watch this: I tend to take things positively. You wanted to be jealous; I'm going to say I'm impressed. Good job, Foundation. Good job, all of them. I'm not going to diss Eclipse; I thought they all did great.

It's what the industry is meant to do. It's the kind of innovation the industry should be supporting. And that applies all the way up the stack to the same thing—the round for metrics, all that. It's exactly what should be happening. I agree. Great credit to Steve. I can't play tennis, so I'm not going to be able to make it.

It might not have been tennis. Maybe I'm making up the story, but it's directionally correct. It might have been another activity like that, right?

Jason Lemkin

Okay, something with less ball. Yeah.

But that's real venture capital, right? That's not leaning into a deal. That's not winning a deal.

Eric Vishria

That's not winning a deal. That's in the founder's note. That's calling a shot on AI in 2016, saying, even in the founder's letter in 2021 or 2022, “Oh my God, we're way too early.” Finding a way to survive. Then seeing the tailwind in 2022 from GPT. Still struggling to get orders. Getting some business from the UAE, and all credit to the guy who said, “I'm going to get on the plane, I'm going to Dubai, and I'm going to sell me some chips.” Right? Whoever that sales guy is, I hope he got a big stock order.

And then surviving, pulling the IPO in 2024 or 2025—I can't remember which—when it just wasn't ready. And then the moment has come. You get the OpenAI commitment, you get the Amazon commitment, and now you can go out on strength. It's a great story. Whatever they make, they earned it.

Jason Lemkin

You know, the other thought I had—so I put “jealous” in air quotes, right? I'm not only jealous in that that's what I should be doing, right? The other thought I had, just at a high level, was—and granted, I only met him on Riverside when Harry and I did a show a while back—but Andrew Feldman's just so good.

Now, I didn't meet him in 2016. Maybe in 2016 it wasn't Captain Obvious; maybe it was, right? But to me, it's also a reminder: we all talk about going long on Twitter and social and betting on great entrepreneurs, but so many more folks quit this week on my LinkedIn and Twitter. He's so good.

This is the kind of—and it's also that if you're not quite as good as Andrew, like, you're really, really good, but you're not as good as him, you would have quit. You would have quit. So the combination is just my learnings: the slight jealousy, in air quotes, of them, and then a reminder that when it's so fun to do a startup today, you've got to be so great to win, right?

And you guys have to be so great to have these massive exits in venture. The founders have to be so great, not just— This is the tough part of venture. Very, very, very good founders aren't going to build this type of outcome.

We have Ramp at a $40 billion valuation, Parker, an alternative fintech company, filing Chapter 7, and Gusto passing a billion. Which one do you want to take?

Eric Vishria

So let's start with the company card. I think the distinction between the two is that Ramp is a broadly horizontal corporate business card. And as we've discussed many times, the actual economics on cards are good, but they're not amazing, right? You get this interchange revenue, but you have to give a lot of it back to the customer, so your contribution margins are only okay.

The only way to make that business better is you've got to add a lot of software and a lot of functionality, and Ramp is doing a truly amazing job of doing that. You start doing ACH payments. They just announced something super interesting yesterday in the market we like, which is kind of agents on top of their system to automate your purchasing, right?

So you, in a midsize business, can now have the Ramp agent go out and try and optimize your spend, reach out to your suppliers, and beat the crap out of them on price. That's a market we like independently, but that's a good add for Ramp.

So, zooming out, the outcome is Ramp was in a broadly horizontal market with a lot of running room to add, and the other guys were in a very constrained market with a lot of margin pressure. To some extent, it's not surprising at all. I don't know. Sorry, Jason, I went off there, but I love those agents. You should check them out. They will pass the Jason Lemkin acceleration test.

Jason Lemkin

Yeah. Certainly, we're going to rebuild our financial stack after SaaStr Annual, and we will move from Brex to Ramp if it is the most agent-friendly. Automating procurement is a huge bonus. We just take the humans out of it. Just have the agents negotiate procurement. We've all had enough of it.

Just like Delve solved SOC 2, I want an instant solution to procurement without this moronic back-and-forth, the games, the politics, the fake contracts that procurement cuts back 10% to get their slice, so you have to overprice the deal. I'm waiting. I'm saying this a little facetiously. This is a problem agents—at least the next generation of agents—could solve.

I'm going to see whether I stay or leave Brex in a month or 2 based on which has the best agents.

Jason Lemkin

Now, the separate issue is, you've just got to put it out there: it's $1 billion in revenue, trading at a $40 billion valuation, and it raised money at $40 billion.

Rory O'Driscoll

That's where I get confused a little bit.

Jason Lemkin

Yeah, and again, it's back to my comment earlier, which is—you know, again, as I say, I apologize for saying something so obvious—you have your discussion on the strategic dynamics of the business, and then separately you have price.

I think Ramp, on the strategic things they're doing, is just amazing. A 40x revenue multiple, when the comp—Brex—traded at 6x, albeit on a lower growth rate, is a pretty healthy valuation.

I do wonder, with a lot of these valuations, whether there is that scrutiny of revenue quality and revenue multiples. I mean, we're all just addicted to growth, right? So we all pay the same multiples almost regardless of what gross margins or anything are today, and maybe it's fine. Would you buy Ramp at $40 billion?

Jason Lemkin

Probably not. I haven't seen the growth rate, which is the only thing that matters. I just think there's a gravitational pull to these businesses. The amazing thing about these fintech businesses is that some of them can be just enormous: Stripe, Revolut, Nubank in Brazil, because you're selling to consumers or SMBs and everyone does this.

Fintech—everyone does finance. Everyone has a payables division. Everyone has a corporate credit card. So they're big-ass businesses, but they trade like there's no magic, kind of AI premium. They trade just like AmEx, adjusted for growth.

Whenever you get wildly far away from a revenue multiple, you really have to be certain that, in 2 or 3 years of further growth, you've grown into it. If you double and double again, maybe the way to think about it is: how many years of growth do you have to get before you're trading at a normal multiple?

If you go 1, 2, 4, 8, it takes probably 2.5 years of growth until you're at the Brex multiple. That's pretty scary. That's the outer edge of terrifying. If you're leaving in a year, you're like, “Yeah, whatever. It's going to double and be fine.” If you're underwriting 2.5 years of doubling to get to the Brex multiple, that's pretty scary.

Rory O'Driscoll

I agree.

Jason Lemkin

I mean, you have the protection of preference, and it's the same investors who did the prior round. So, to some extent, they're probably saying, across the investment, they need the fuel. “I got my return, and it'll all be good.” They're going to make out like bandits here.

I mean, because the big kahuna is 2022 at $5 billion, that looks like a pretty damn good deal now.

The other cool IPO, which isn't on schedule, is Lime announcing that they're preparing to IPO. Do you see those Lime bikes?

Jason Lemkin

I did. I had that feeling of, “Oh my God, they're alive.”

Oh my God, I mean, if you come to London, they're alive and they dominate large parts of the city. But that is a hard business that's been through its turnaround of the day. Incredible journey there, important to say.

Shout-out of the week for me: Lime announcing its IPO. Good, healthy business.

Jason Lemkin

I mean, obviously. Yeah, I will be interested to see the numbers. And again, it's back to: good on you, entrepreneur. Well done.

Totally agree there. Boys, should we do some more? Actually, one final thing before Musk versus Altman. We have to do it. Come on, it's that moment of the week.

Brockman says Musk wanted a for-profit. We had Ilya come out today and say that he's, I think, worth $7 billion. What do we need to know in the Musk versus Altman trial of the century?

Jason Lemkin

I mean, first of all, we're going to know a lot more than we need to know. In the sense of—as is the nature of these trials—a whole bunch of stuff that's marginally extraneous will come out, just because that's the nature of the beast.

In the end, I remind you here, it's not a jury—the jury's advisory, the judge decides. You saw some really nice profiles.

Rory O'Driscoll

I was actually checking out the judge—Gonzalez. I can’t remember her first name. She seems super tough and hard-nosed. She’s driving the case here; she’s making the decision.

All this noise will just fritter away. It’ll be fun for the headlines, but she’ll make a decision on the legal issues. My gut continues to be that, even though everyone will look crappy, OpenAI gets to escape with their deal intact. That would be my gut.

Jason, help us out. SaaStr this year: who will be the best speaker? Who would you put money on? You do reviews of your audience. Who’s going to be the most popular speaker?

Jason Lemkin

I don’t think there’s any popular speaker. We barely have any speakers this year, Harry, because I think podcasting has kind of destroyed the whole need for a speaker.

We have a lot of workshops. We have people coding, showing you how to build things, but we don’t need any speakers. Why would I go see Andrew when he was on 20VC and was better from here? So we have no firesides and no speakers.

It’s going to be great, but when I have Amjad from Replit, who was on 20VC, we’re going to walk through my agents, what they are, why they built them, and why they work. I’m going to have Tyrell here, who is the father—the Blade Runner father—of my agents. But we’re not talking about companies, because podcasts are better, right?

They’re way better. I’m going to see you—

Jason Lemkin

We don’t have any speakers. We don’t have any speakers this year. We’re going to have everyone showing how they built the agents. We’re going to have Rubrik demoing their agents, and we’re going to have Andrew from Klaviyo, the CEO, demoing his agents.

Everyone’s going to demo what they built, why they built it, and do an AMA. That’s stuff you don’t get on an average podcast. We’ll see, but we’ll never do firesides unless I’m forced to. I mean, I’ll do it for Sam Altman, but otherwise I’m not going to do it. Firesides are dead and speakers are dead. There’s just no point when podcasts are better.

For what it’s worth, that’s super insightful.

The minute you say it, you always know when something’s insightful: when someone says it and you go, “Yeah, I hadn’t thought of that before, but you’re absolutely right.”

The minute you said that, I thought, “Why?” I’m going to go over and see you doing a little thing. Why would you schlep over to see 10 back-to-back speakers say the same thing they’ve said on a podcast when you can listen to them while you’re working out?

I’d be super interested—I think I’m conflicted—I’d be super interested to hear, Jason, you and I talk about your agents, agent security, and all that kind of stuff.

Jason Lemkin

Yeah, and how one of them willed itself into existence. We didn’t even try to build an agent. Our agent, Annie, willed itself into existence. How does that happen? That’s pretty cool, right?

Rory, can I just ask you something? Sorry. Should I be buying Micron or SK Hynix? I feel like I’m super late to the game, and I don’t want to rock up to the party at 11, but I’m also like, “Shit.”

Rory O'Driscoll

On the one hand, you’re right. You kind of go, “Oh my God, it’s gone up 5× in the last year. How can it be right?” But I hate that thinking, because the correct thinking is to say, “But then you still look at it relative to earnings, and they’re still relatively cheap,” right?

So really what you’re saying is, on an earnings basis, you start trying to say to yourself: how long does the capex boom last, and how long before they double the number of fabs that make DRAM and commoditize it? Do you think there’s more oomph in the stock? Because once those 2 things happen—once the boom starts to slow down, just when the extra capacity comes online—the combination is brutal to the downside, and you see that over and over again.

I was actually just looking at SanDisk and the DRAM guys because we’re going to discuss this in the podcast, but you, of course, ignore the agenda. As late as 2022–2023, all those guys were in the crapper because you had that post-COVID period. Remember COVID? Everyone bought a laptop because they were working from home, and then after COVID, everyone didn’t buy a laptop because they didn’t need another damn laptop.

All those stocks went way down in 2022–2023, and the last 2 years—and 1 year in particular—have been amazing. So I don’t know. It’s not a good enough reason to say, “I won’t buy them because they’ve already gone up 5×,” because you’ve got to look at the pricing still. But I don’t have a developed opinion on it.

You have to decide those 2 things: the duration of the capex boom relative to the speed at which they can build X more fabs, because in the end, they always do build more fabs. That’s one of the things about human nature. When you start making 50–60% net margins on a ton of money, the temptation to build a fab just becomes huge.

So I’m sure Samsung and SK Hynix, even as we speak, are digging holes in the ground. Then Anthropic starts to cut orders in 2028, and there you go. Ain’t capitalism great?

Are you long Navitas?

Rory O'Driscoll

No, I’m not long. I’ll short Navitas. You asked me a fact-based question: have I chosen to put my money in? No, I have not put my money in, but that’s different. You can’t buy every stock you talk about, Harry.

Fine. The important round, which was very popular rage bait but real: MrBeast basically posted saying that the sacrifice of mental health was essentially required to have the level of success he’s had. The commitment and willingness to suffer for long periods of time is what separates those who are successful from those who aren’t—paraphrasing, but very close.

I agreed with that, and I said I 100% would not have achieved what I have without sacrificing large parts of my health and commitment. I got a lot of pushback. Do you think that is rage bait, or do you think that is real when you look at the $20-plus founders? Can you have that success without sacrificing mental health?

Rory O'Driscoll

It’s very hard to have that success without sacrifice, right? Real, meaningful sacrifice: time and alternative uses of your life. Sometimes, in the case of brutal competition, people sacrifice family life. Loved ones and marriages end up in divorce.

It’s really hard to do something intensely, and most of these things require real intensity, right? So I do think that part’s probably true. However, I think at the point in time when you’re getting into mental health, you probably owe it to yourself to try and find some way not to tilt over to the point of making bad decisions.

I find myself, when I get to that point of being wholly stressed and spread thin, I don’t make good decisions. So you actually owe it to yourself at that point to pull back a little. You actually aren’t that useful when you’re on tilt, Harry, all right?

I don’t know if that makes it rage bait or not bait. Maybe “true but not worth rage” would probably be my advice. Don’t rage.

Jason Lemkin

I’ll give you a different learning. This was one of the first posts I ever wrote, right when I got out of Adobe, and it took me a while to realize this.

When I sold my first startup, I sold it for $50 million after 12½ months, which today would be more money—there’s been inflation. It was great. It was so hard, my first startup. I was building implantable batteries from nanomaterials. It had never been done before.

We had customer concentration. I had to close $6 million. My VCs pulled my term sheet. I had to do payroll myself. I had to take a full-recourse loan against my house to get the round done. It was just everything that happened—everything that could happen.

I would do day trips across the globe. I would fly to any airport in the world, do a meeting in the airport, and fly back the same day. Unsustainable stuff. But we bounced back after a week after we sold the company, right? It was 12½ months. I was given a 2-week package to stay, and I was okay in a month, right?

The second time, it was 5 years, which is not long now. But for me, what I realized after the next one was that my brain was permanently rewired. I was no longer the same human being.

The level of intensity—from almost going bankrupt multiple times, to dealing with those issues, to saving the deals, to going through the GFC, and having to turn a $100K customer into a $500K customer so we could survive and make payroll when everyone wanted to cut the deals in the GFC—going through all that drama, I could not go back. My brain would not allow me to go back.

So, yeah, there’s a rage-bait element, but some of it is doing the 996 and working all the hours you’re talking about, Harry. But I think what founders who have been doing it for more than 5 years understand—the 4-to-5-year mark, I think, is the breakpoint—and you have been, Harry, is that you’re changed.

It’s not as simple as mental health. A vacation doesn’t do it anymore. It’s not enough to go for a run on the beach. It’s not enough to start buying watches or even to buy a yacht. It’s not enough.

You’re a different person, and if you want to win, you have to commit to being a different person. You’re not going to be that happy-go-lucky person who got into YC, got your $2 million, and thought it was great and really fun. You told your friends and went to the hackathons. It seemed really hard, but you know what? That first year, it’s just fun and games, okay? You will be a different person.

You can never go back. You often can't even talk with non-founders for real anymore. Founders stick together. They're in WhatsApp groups. They're in chats. You're all changed.

It's not just a peer group. You're not the same people. You're not the same person when I met you, Harry. And so I think that's the meta issue. Seeing a shrink is great. I'm all for mental health, right? But it's not going to change the fact that you're changed after 4 to 5 years.

I don't think it happens in a year. As hard as that first year was, man, I was back to runs and cruising, and I was never going to do another startup again. I'd made a couple of million. That was enough. Life was good. Checked the box. Owned a house.

One of my favorite quotes ever on a show is Daniel Dines from UiPath. He said, “A lot of people think they want to be me, but I promise you, when the lights go out at the end of the day, it's very lonely in my head.”

And then ask the next one: Are you even the same person you were when you started this journey? He's going to say, “No, I'm not. I've been rewired”—well, your brain has been rewired, though. The brain has been rewired.

I watched your Apple event one with that guy. So effing intense, right? So good, right? But one of the things I thought when I watched that one was, his brain's been rewired. The things he's saying make total sense to me after my journey, right? But they don't make sense to most people. That was the point of your tweet.

What he was saying, most people wouldn't get it because their brains have not been rewired for the level of intensity it takes to succeed. It's not just the hours; it's the intensity that is like nothing else. It is like nothing else.

Jason Lemkin

Yeah. And every time you cut a corner, every time you invest in a founder who's really nice and really hardworking and a really good guy, you never make any real money.

That intensity—okay, whole question: Do you see that intensity in your portfolio?

Jason Lemkin

In your winners, you do. Any normal human will sell for 50 million after a year, or 200 million, or 1 billion. Any normal human will sell. It is idiotic not to sell.

You don't want to do ServiceNow; you want to do Grok, as we talked about. It just makes no sense to be Daniel Dines after 20 years. So it's all self-selecting. You've got to be Daniel or crazier to do this journey. There was a brief period where you could do it 35 hours a week in late 2020 and early 2021, but it's gone now, right?

What he's basically saying—and I'm violently agreeing with you—is that if only intensity is required to be successful, it's no surprise that all successful people are intense. And the part of that that resonated with me, Jason, you're right, is that good companies get offers along the way, and if you're not intense, you'll take them.

Jason Lemkin

That's why I tell all founders to take the offer. I tell all founders to take it. It's the opposite of the VCs: Take it.

It is an intensity test, and if they push back and say, “No, 1 billion is not enough,” great. Then go for it. But if they're not sure, you may not make it anyway. You may not have that level of intensity to do it.

So take the 1 billion, or the 50 million, or the 200 million, as long as it's 3 times or more than what you raised, and go enjoy your life, man. Daniel's not happy, and the AppLovin event guy is not happy.

Rory O'Driscoll

I just want to say one thing: I find myself violently agreeing with you on the rewired-brain and intensity side, right? I do still think that you have to find a way in that maelstrom to keep perspective.

I see it myself, and I see it in CEOs. When you're working so hard and you don't have any way to clear your head, you actually can become ineffective and a weaker decision-maker.

Jason Lemkin

Adam does surfing now.

Rory O'Driscoll

Yeah, actually, I saw an interesting talk by Admiral Stavridis. I'm going to pronounce his name incorrectly. The guy was commander-in-chief of NATO. He was speaking at a Cambridge Associates event recently.

Someone asked him the obvious: the 3 lessons for life as a senior leader. They were expecting some kind of “hire good people”—the usual clichés. His first comment was, “You've got to stay healthy,” because if you're not at some level vaguely healthy and functioning—in terms of sleep, in terms of fitness—you just won't be able to cope with the pressure and the intensity.

It was a super interesting comment from a guy who, remember, has been commander-in-chief of NATO. My point is merely that when you disappear into your own head to the point where you're not making good decisions, you will not maximize value for yourself.

I'm sure there are times when you're just not thinking straight and making bad decisions. Just make sure that you have whatever coping mechanisms you need to avoid doing that. It's a little like—I don't drink much now, but the day after you've had a lot of alcohol, when you're slightly hungover, you make horrible decisions because you're all jittery. And that's one of the many reasons why I don't drink now.

The reason I love rage bait when it's real is that 2 out of 2 of you guys have both agreed that it's real, not rage bait. I'm like, yeah, I feel vindicated.

Jason Lemkin

No, both times I've actually agreed it's real and it's rage bait because it's something that annoys everyone, perhaps because there's an element of truth in it.

Rory O'Driscoll

Yeah, right.

I know the work-from-home Friday brigade were pissed off with me about work-from-home Friday.

Jason Lemkin

None of them have figured out, Harry, that you do it because you enjoy the rage. You know, that's what the problem is here.

Okay. Right, boys. Rock and roll, Jason. I got to run, baby.