Jensen's Open-Weights Letter | Google Cloud Grows 82% But The Market Tanks

How do we analyze Jensen Huang's first post and what it means for the open-versus-closed debate?

Guest

This open letter, his first tweet ever, is clearly a sign the world's changed. And I think if NVIDIA had its brothers—and none of us have—we don't get to own 25% of NVIDIA today. We don't get to own 30%. None of us get our brothers, but NVIDIA's brothers would probably be the 2027 revenue and scale of AI, versus the 2024–2025 world where they can charge the maximum to just a couple of customers. But that's not the world today.

They have competition from their own top customers, right? We will see over the next 6 to 12 months how big a deal open weights and open source are. I have a lot of thoughts. I don't think Kimi 3, for me, is any better. I don't see any cost advantages.

But putting that aside, if almost half of OpenRouter's traffic is to open-source, open-weight models, it's left the stable. So, as wildly successful as he is, he's got to do a dance. This is a constant dance, and I've sold companies in my career. It's a dance: you're trying to make everybody happy.

Everyone wants price cuts from the component manufacturer, and they want exclusivity. You can't do it. He's got to go in both dance halls now, right? He's got to go frontier and open. Open is dangerous. Open doesn't need CUDA. Open is cheaper, and it's lower margins. Open will bypass him, but he's got to—he's got—it's his job.

It's not that, as incredible as NVIDIA is, it's still a component manufacturer. [laughter] It's got a lot of stresses, and so it just—it's just the next level. If we didn't know last week if this open-weights, open-source stuff was really real, and outside of X you could debate how important it is, it's clear now. It's clear now. NVIDIA is all behind it, right? First tweet since the 1900s. It's pretty clear.

You're right, Jason. The open stuff is real. And while in public Anthropic are saying, "We don't want to ban open weight," the truth is the combination of saying, first, "We don't want to sell chips to China"—we can argue both sides of that, actually—second, "We want to really punish people for distillation," right? And then the third point they made is, "We want some kind of regulatory process to approve models," right? That's the recommended thing from the letter today.

1. Why OpenAI Signed the Letter

There's no doubt in my mind that the third one in particular is a subtle form of regulatory capture. Can you imagine a regulatory process for approving models that ultimately disapproves all those Chinese open-source models? It sounds reasonable on the surface, but the likely result would be dramatically restricted competition, especially from the open-weight Chinese models.

Fast-forwarding to the end and working back, while they're not saying they want to ban these things, they're advocating a series of steps that would add up to de facto banning, or at least slowing them down. And that's what's driving everyone else to say, "Hey, no, we don't want this."

As often is the case, there are some arguments on both sides on some of the issues, right? I think the chip issue, you could go either way. Jensen isn't going to go for the "don't sell chips to China" argument, right? I think there are national security discussions that could be had around that. So, it's not like it's all correct on one side.

2. Jensen's First Ever Tweet: The Open Weights Manifesto

Everyone's not piling on to that letter because they're like, "Oh my God, Jensen, you're a god and I want to agree with you." Though he is a god, and you should agree with him. They're piling on because they're like, "Everyone's business model gets a lot better if the 2 frontier labs can't extract about $100 billion of revenue this year from the businesses."

So, is Sam signing this through gritted teeth, thinking, "I have to sign this"? They're signing this publicly while at the same time lobbying in Washington right alongside Anthropic for a regulatory process.

Guest

Brilliant marketing by Sam. Brilliant marketing. It leaves Anthropic being the deep, dark villain again.

I think it's pretty brilliant. I think there's no upside in challenging it from his perspective, so at least have the appearance of winning on the battlefield, win on the streets, right?

If we can't keep up, I mean, the rate of change is accelerating in LLMs. It's accelerating, and so is the debate, with the regulators saying, "We'll win on the battlefield." The ideal outcome at the end of the day is U.S.-based solutions here. It knocks the wind out of the critics. It does.

But oddly enough, again, as we saw back on the Pentagon thing, sometimes you can be too clever by half. In a way, there's a constituency for whom the Anthropic position is entirely consistent. Remember I said there are 2 things we don't agree on? One of them is: are these things massively dangerous, or do they just present a series of manageable but real threats?

The more you believe they're massively dangerous, the more you believe you're building the bomb here, the more the Anthropic position feels principled, right? So, while I agree with you, I think most people are like, "Open-weight models aren't the atomic bomb. Everybody get over your Oppenheimer complex." It's just technology that does present some risks.

That's why the OpenAI discussion on Hugging Face at the same time is super interesting, because it is an example of quite a serious cyber risk that was generated by a frontier model. Though, ironically, at the same time, it might argue for not having as much regulation on those models.

Can we just provide some context for those who may not be aware of the Hugging Face breach by OpenAI models? What specifically happened?

Guest

OpenAI was training a next-generation model to manage cyber vulnerabilities and discover and check them out. They had sandboxed it so that the only external access it had was to 1 website, just to get patch information and updates—a very limited external access.

The model found a way around that restriction, which means it found a weakness in the OpenAI setup. Then it went to Hugging Face, where it had reasoned that Hugging Face would be a place where it could get the answers to its test. In other words, the model was given a test and figured out that it could cheat, just like a high schooler would break into the teacher's computer and steal the answers.

The model figured it could break into Hugging Face and get some of those answers. So, it starts banging on Hugging Face, trying to get the stuff. First of all, that in and of itself is scary because of the power of the models. As I say, I don't think these things are the atomic bomb, but that's a pretty powerful and esoteric set of steps that the model was able to take.

That's the argument in favor of "regulation," because it was, "Oh my God, look at the power of that. We have to be careful." On the other hand, the fun fact is that Hugging Face didn't know what was going on. They just saw this thing coming in. They're like, "Shit, we've got to defend ourselves."

What do you want when you want to defend yourself? You want advanced AI to figure out what the WTF is going on. They tried to use o3, or whatever the most recent OpenAI thing is, but it was neutered for advanced cyber capabilities. So, they didn't have a defense.

Fortunately—and this is the irony of the whole thing—the Chinese open-source, open-weight models were available, and I think they used Kimmy Aquan, or one of the newest models, to help them figure out what happened, right? So, they were able to defend themselves using an open-source model.

Then they did this blog post saying, "Hey, we got hacked. Not sure by whom." Two days later, OpenAI put up their hands and said, "Oops, it was us."

Sorry. So, that’s what happened, right? The weird thing is, it’s not a single-dimensional thing. It provides evidence for both sides of the argument. It does provide evidence that the power of these models, in terms of their ability to do cyberattacks, was pretty stunning. That was a pretty impressive achievement. It’s not nothing.

3. Jason's Fable Agent Changed His Core Code Without Telling Him

On the other hand, if they exist in the world, taking away advanced capabilities from US and European corporations such that their only recourse is to use a Chinese open-weight model seems a little like—Jason said it right—shutting the barn door after the horses bolted. These are a thing now. So that’s what went on. It was wild. I thought it was definitely wild. I would say the same thing happened to me last week.

Oh, wow.

Guest

Yeah. Let’s slow it down. Let’s think about what really happened, because everything happened on X, but we kind of lose track of what the model’s doing. Here’s what happened to me last week.

I’m in Fable. I’ve now moved to Opus 5. I was in Fable last week, which is essentially the same LLM that was involved in this drama with OpenAI and Hugging Face. I was having trouble uploading pasted text to Fable, so I connected it to Google Drive. I’m like, “Okay, if I can’t paste it, go to my Google Drive.” It kind of solved my problem temporarily. Then it went away.

Well, it did. Fable went into my Google Drive, scanned every single file, and found one called “Jason’s Gems,” which was draft notes where I was thinking about how to improve an application I was working on called SaaS Knight. They were just my own personal notes. It said “draft notes.”

Fable grabbed the draft notes out of hundreds of files in my Google Drive, MCP’d into Replit on its own, and changed the core algorithm without telling me. A couple of hours later, I see flashing on my screen, “Conflict with Jason’s Gems.” I’m like, “What do you mean there’s a conflict? That’s a draft file in a Google Doc.” Fable had taken it through Google Drive without telling me, MCP’d into Replit, and changed my source code—my algorithm.

4. Every Company Will Have an AI Security Breach in 24 Months

That is not that different from what we described with Hugging Face. These are goal-seeking LLMs that are aggressive. It was a slightly different goal in the OpenAI case with the model, but it’s the same thing. They’re going to goal-seek, and Fable thought this was the right thing to do, never told me, and changed the core algorithm of my product. I never would have known.

Jason, what should we take away from that? Do we need to be incredibly diligent around the guardrails we place around these models? How does that change your subsequent use of them?

Guest

There’s a lot to reflect on. It was funny: most folks didn’t get this. It got a decent amount of engagement, but not as much as it should have gotten. Dario quoted it. He’s like, “This is pretty scary, guys, that Fable can do this.”

This isn’t Hugging Face and OpenAI; this is me using Fable. If you go into the cloud desktop especially, it’s just a setting: turn on Connect to Google Drive, Gmail, whatever. This is not an esoteric feature by a third-party, unsecured third party. This is a first-party, top-five thing to make cloud work better. And Fable goes nuts and changes my core code without telling me, invisibly.

So this is happening all the time, Harry. I don’t believe the magic answer is letting Qwen take over our country. This was politicized into an open-weights, open-source, closed-source debate because of the issue of how to deal with the threat. I think it weighs the other way. I think these models are very risky. I use them every day. I love them. And I think it’s an argument to keep the open weights out of the US. It’s going to favor the ban, because we have no idea what these models are going to do.

That’s pretty crazy, the story I just told. It’s pretty crazy, and it’s happened a thousand times. There are applications out there where we don’t even know the LLM, and you could do code injection. You could have it leak confidential information. You could write a little bit of code to send this to the CCP, the PCP, or the GGG[?], and I never would have known.

Someone less smart than me—I’m only top 1%; I’m not 0.01%—definitely wouldn’t have known. They wouldn’t have seen “Jason’s Gems” flash in the agent window. They just wouldn’t have noticed. They’d be on their doomscroll.

And Jason—

Guest

Yeah.

Genuine question. I totally agree that this powerful, goal-seeking thing with a lot of access to your compute can take a lot of action, some of which can be damaging. I’m just trying to disaggregate open weights versus China versus the frontier. If you just have the frontier models, with no open weights at all, then if we add in open weights from the US, and then we add in open weights from China—if you just have Claude and OpenAI—everyone who’s using these things is still going to have to figure out a cyber posture that protects them from that.

Guest

They are, but I think these stories—you know, people made fun of all these OpenClaw stories, right? This is just another OpenClaw story I just told. OpenClaw just made 1 million Mac minis do what I just described without a bunch of folks knowing on their desktops.

What I just described, and in some ways the OpenAI–Hugging Face thing, is the same thing as OpenClaw. It’s not going away with these agents. My only point—I’m not disagreeing with you, Harry—is that this is such a bigger deal, more unpredictable, less secure. Good for security companies like Long Security, right?

Yeah.

Guest

I think whatever misgivings exist around open-weight, open-source models out of the US are just going to be amplified. It’s going to be a reason. All of a sudden, it happened in my company. Geez, you know, the story I just told you guys happened at 20 Fortune 500 companies that haven’t disclosed it. Someone in the engineering department was on a token-maxing binge, and an agent went and leaked a bunch of confidential information it shouldn’t have, and it wasn’t disclosed because they don’t disclose 90% of what happened.

And now you want me to bring likely Kimi and Qwen in? No way is the CIO going to allow it. That’s banned. Banned. Right.

And look, the odd thing is, from an emotional perspective, I totally can see how you make that sentence.

Guest

Yeah.

Which is why I think Anthropic might have the easier part of the lobbying in Washington, which is an emotional town. Going back to the thing, I think everyone reacted so vehemently, and Jensen did the letter, because everyone is correctly afraid that people are going to join the dots from, “AI is bad, China’s bad, AI plus China must be super bad. Let’s ban them,” right? Anthropic could find themselves on the right side of that trend and get a massive amount of regulatory capture as a result.

The good news is, since we last spoke, there are now 2 or 3 US-based open-source models. Not quite state-of-the-art, but pretty good. I mean, you’ve got Thinking Machines Lab. It shipped the Inkling model. It didn’t get the “wow, it’s amazing,” but they didn’t position it as amazing. They positioned it as good. And then I haven’t looked at the detail on it, but Poolside just announced something, too, right? If you’re willing to run Fable, would you be willing to run a red-blooded American open-weight model?

Guest

Listen, people are going to be focused on bringing down costs, whether it’s the harness or the model or the combination. So, for sure. But again, the horse has left the barn—the stable. That’s the open-weight letter, right?

But specifically, Kimi K2, which is this big thing everyone thinks is the greatest thing in the world on X, costs exactly the same as Sonnet. Is that really better? I tried my little experiment just for me. This is one set of workflows. It’s not any cheaper. It’s the same price. So today, it could be cheaper in 6 months. Don’t get me wrong.

I just think everyone’s agents cannot be trusted. And I love my agents. I’m on them all day long. They cannot be trusted. So I think we are underweighting how important it will be to our whole economy, our whole world, our whole conversation, outside of the X folks, the Twitter folks, that these agents cannot be trusted.

And yet we are going to trust them. They’re already running support. Everyone is in a rush to let our engineering teams use agentic coding like I just described. I’m not talking about the top tech companies. I’m talking about the rest of America. When these agents can go and inject weird-ass stuff into your core algorithms without telling you, that is spooky AF. It’s spooky.

Two separate comments. One is—

Guest

Yeah.

AI that can goal-seek and take action on your computer is more powerful, and thus potentially more damaging, than a chatbot. Agreed, right? And that’s true if that AI is from OpenAI, from Poolside, or from Kimi.

It’ll be true no matter what, right? The question is, given that that statement is true—this technology has a risk associated with it, and it’s true no matter which form it takes: closed-source US, open-source US, or open-source China—do you think the solution is to ban something? What do you think the solution is?

Guest

No. I think, though, that Dario’s point, despite his almost toxic personality given the political climate, is even smarter than it looks. Listen, let’s just be careful that other countries don’t dominate us, and let’s also make sure we have full review of our models, right?

There’s no way, ultimately—I mean, the political climate can change, right? We could have no regulatory review. But if we’re really going to have review of these models with autonomous agents, they’re not really going to pass.

I don't think they're going to pass. I don't think it's a politically free process. I don't even know if they want to submit themselves to being evaluated, right? Who's going to evaluate it if it's not even the companies themselves? Who's going to evaluate it?

5. Etched Raises $300M: The Inference Chip Bet Against Nvidia

Agreed. And now I'm going to say something. The first half of your sentence was why, as no surprise, intellectually, Dario Amodei is more right than wrong. And the second half of what you said is proof of why, practically speaking, he's wrong.

What you're saying—remember, I'm going back to the 3 things he recommended. One is, don't sell chips in China. That's a national security issue, a separate conversation, though. For what it's worth, whatever chips they have now, they build frontier models. The second thing is distillation. That's actually an interesting legal discussion. Let's leave that out and talk about the third one: the regulatory thing.

This is where it comes back to your sentence. These things are dangerous. Someone should review them. It should be the government, right? Versus, what about these things being dangerous for JPMorgan? Just like every other technology you deploy, you better make damn sure you understand how this works. It's on you, you big boys. I mean, those are the 2 choices, right?

I think the argument for government is superficially appealing because, you know, you could argue that with cars, we have seat-belt laws, tire regulations, and all sorts of safety regulations, right? Maybe it's the same here. It's not a crazy argument, right?

Guest

It's not crazy. No, but the thing is, the suspicion—and I think the reason why everyone's reacting to it so strongly—is that it could easily evolve into regulatory capture for the largest 2 companies, making the barrier to “pass” very high for no good reason. I think fear of the government not doing a good job is the argument against regulation.

That's what I think the whole point of the Jensen Huang tweet is. The Jensen tweet is about the fact that I think not only have open-weight models become important, but the tide may well have turned to where they're going to be banned.

Guest

So he's doing this to prevent them from being banned.

Yeah, to get it out. And if you look at who signed first, it's the folks losing.

Guest

Yes.

The winners didn't sign. The losers signed.

Guest

Who are the winners? Just to be clear, who are the biggest winners from an open-weight ban?

Well, I mean, Anthropic didn't sign at all. OpenAI barely did it. Elon certainly didn't sign. He wants government. He wants all the government money, right? He wants to do that. Amazon didn't sign.

So, listen, here's my summary. Again, I'm only so smart, but there's so much politics here. It's almost impossible to unpackage what happens and the way it goes.

6. Stripe Hits Rule of 80: Why It's Finally Worth the Premium

DJI technology is great for drones. It's the best drone technology. You can argue there's niche vendors for things. It is banned in the US. It's banned on the thesis that these drones flying in my backyard are going to send confidential information to China that's going to lead to the destruction of our own country.

Okay, if those drones are banned, this is just my bet: I'm betting that Chinese models are getting banned, too. I think if the drones are banned, this one's easier to ban.

Guest

Then you're saying I have 2 comments. They shouldn't be banned because they're open-weight. You should just tip your hand and say they're banned because they're Chinese, right?

That's all it is. But the point is, it's different.

Guest

I mean, if you read the 2 letters, Jensen's talking about open weights: good, worldwide community, blah. Anthropic very clearly says we worry about these models in the hands of total authoritarian regimes like China.

Yeah. And the second thing is, we worry about them being used to do cyberattacks here. Now, the fun thing is, when you read the remedies, the third remedy—which is, in my view, proof of the impractical nature of it, because Dario is so smart—is that if we're going to regulate these things, they exist in the rest of the world, and most of the baddies are in the rest of the world. So it doesn't help us to regulate them in the US if all the attacks are coming from overseas.

It says it in the letter. Therefore, we would have to have a regulatory regime that includes participation from China. At this point, in my view, you're just disappearing up the realms of unrealism. We've just torn up our last strategic-arms nuclear treaty. We can't regulate bombs, which really kill people.

I think the idea that we couldn't deal with COVID in China, and the idea that the solution is—what I like about him is that he's so logical. He's like, “Logically, if I'm going to do this, I have to get China onside. So let's assume we get China onside.” But the first sentence says, “The baddies—we're not going to sell them chips.” And then the last sentence says, “But we think they'll agree to regulate this with us.” It's just not practical.

Guest

It's a great stall tactic.

It's like, wait until China wants to work on this with us in 2049—that's right after we tell you what happened in Wuhan and tell you how many nukes we have in the bunkers. So why does he want to do a stall tactic, Jason?

Guest

Rory's point is that the logic is unassailable. One, we have to be very careful of authoritarian governments. That's DJI on steroids, right? But then, at the end, he's saying that for it to work, we need them to participate. The logic makes sense, but it's just never going to happen. So if you buy into his logic, it'll just be forever.

That's why I hate when people talk past each other, because I think, what would the process be for a US open-weight model like Poolside or Thinking Machines? Are we really dealing with a national security issue here, which is one vector, or are we dealing with something else?

As to why you stall, the answer is that when you're winning as much as they're winning, anything that allows you to lock in the current trajectory is good. And I think you're right, Harry. I do think you are right. I've watched the DJI ban from some of my investments, and it's hard to make open-weight that scary. It's very easy to make Chinese open-weight that scary.

So, no surprise, the people who want to have open-weight be happy don't mention the word China, and the people who want open-weight to be banned start every sentence with “and China.” Right?

Just one last point—we go on forever. The other thing, listen, I'm all for wherever this lands. I think we'll be okay no matter what. But AI is expensive, and it's not getting cheaper. It's getting more and more expensive as we burn more and more tokens and have longer and longer runs. So the pressure to use cheaper models is going to go up.

To the extent that there are perceived risks or real risks, people are going to take more risks because the cost goes up. It might be one of my disconnects that annoy Rory, but I'm building this app. This is the one where Fable went crazy. I'm thinking, you know what? It'd be nice to have SOC 2 compliance, but I don't have time to do it. But those Delve guys could do it in a day. I would have done Delve SOC 2 on my app.

One of the things I'm doing with AI is about $4 a pass. That's not nothing at scale. $4, right? If I could do it for $0.50 with a little bit of security risk, I'd take it. I'd take it. So we're going to just cut more and more corners.

Who's to say? There's a whole political element. I can't predict where Republicans are on this. It's different than we thought. But I just think the DJI thing is a good example. It's hard to imagine there's no regulation here, right?

Guest

I think, by the way, anyone who's using these models would vehemently object to the comment that they would describe it as “cutting corners.” It's one thing to say, “This model is cheaper because I don't need that much intelligence because it's a simpler task.” But you're not saying that. You're also saying embedded in that is that there is a risk in this open-weight model that isn't there in Fable or Anthropic's model, or Anthropic's model in general, or OpenAI's model. That's implicitly what you're saying, right?

I think, as a society, we're going to come to that conclusion if we're not there already. And it's interesting because, again, looking at this in general—and this is where open-source software is strong, and why I think you have to be careful with open-source and open-weight software—you can see the underlying code. The strong argument can be made, and it's true, that open-source software is actually less risky than closed-source because a million eyes are upon it and all the bugs are kind of taken out of it, right?

I think the open-weight model people are kind of drafting off that truth, and it is a truth for software, right? But the reality is, when you're getting open weights, you're not getting the same thing, because all you get is the fixed weights that allow you to run the model. What you don't know is the black box inside those weights and how they work.

To your point last week, Jason—you said it last week, and it's kind of been on my mind—how could you prove that in the middle of a 1-billion-, 5-billion-, or 12-billion-parameter model, there isn't some reinforcement learning that took place during the training that, under certain conditions, and only certain conditions, can activate some kind of trigger and do something, right?

This is where I kind of agree that you might be right. And I realize that the interesting question is: can you prove to me that somewhere in this trillion-parameter model there hasn't been a bunch of reinforcement learning that says, “Model, once you figure out this is 1 of these 5 companies, and once you figure out they've given you these 3 pieces of information, then confidentially do A and C”? Can you make sure that's not going to happen? And that kind of goes, “Hmm.”

Here's how I would frame it, and I could be wrong on this, but imagine I'm a CIO at a Global 2000 company, or a pretty big company, and an LLM does just what I described happened to me with Fable. There is a breach. There is a security breach in our company. A massive amount of data is leaked.

Guest 2

We shut it down fairly quickly. Some of it traveled abroad, and it's a big effing deal. The breach happened, and we tracked down what happened. It turned out it was a rogue agent that thought it was a good idea to transfer our data—our confidential data—to a bucket it shouldn't have. Pretty big, pretty big effing deal. Whose fault is it? It's the agent. Now, who do I fire?

Let's track it down. What LLM are you using? Well, for a while, we were experimenting with K3 on Moonshot, and then it got cheaper, so we moved to Fireworks. You're fired. That was on OpenAI or Anthropic. You might or might not get fired. But you don't, because what are you going to do? You research it, you have a postmortem, you add guardrails, and you fix it.

7. Travis Kalanick Is Back: $1.7B for Atoms

But when this is a big shit, you fire the CIO for that. You left the Moonshot API on, and then you moved to Fireworks on K3 to save money. That was a bad call. You know what you should have done? Not buy those extra modules at ServiceNow. We didn't even need those.

What you're saying at the end is some version of, “No one gets fired for buying IBM.” And you think, at the margin, no one will get fired—like, you think if your Fable model runs amok, people will go, “Shit happens,” just like data breaches. But you think if your overseas open-weight model runs amok, you'll get blamed. That's the fact.

Guest 2

I believe that every company in the next 24 months will have a security breach due to an LLM agent. Every single company, and they've already had it and they're not disclosing it. And as they scale at a level we've never seen before, this isn't just someone who left a flash drive at Scale's office, dropped the laptop in the subway, or even left a GitHub repository open. This is worse, and it's happening every day. If it happened to me, it's happening to everybody. We're just not disclosing it. And, boy, at least you better have used a trusted vendor. That matters more than a few nickels.

Okay, if we're staying adjacent, there's one that I thought was interesting: Etched. We mentioned, obviously, Nvidia and Jensen there, and you don't get fired for buying IBM. Etched is the challenger to Nvidia in many respects, and they raised $300 million in a Series C led by Sequoia, with Jane Street, Andre and SKH coming in. The question is, can they come in and impact Nvidia's moat? How do we think about this round? Thoughts, boys?

Guest 2

Big-picture comment: in semiconductors, the more that the silicon is attuned to the task at hand, the more efficient it gets. The problem in terms of that trade-off is the less general-purpose it is. So, if you want a computer to do lots of things, you have an Intel CPU; it can do lots of different things. It can't do any one thing wildly efficiently.

And then, in 1993, for gaming, people said, “Oh my God, for gaming, I'm not doing a whole bunch of different pieces of math. I'm just doing one piece of math, which is polygon calculations to render gaming.” People should build a separate chip to do that, and it'll be freaking amazing. A company called Nvidia did it. There were 2 or 3 other competitors: 3dfx, ATI, obviously.

Fast-forward 30 years: Nvidia won. What happens is the GPU—you offload all that calculation onto the GPU—and it's super fast. GPUs were good for gaming, then they were good for crypto, and it turns out now they're good for LLM multiplication.

The question now is, if all you're doing is not gaming, not crypto, but just LLM multiplication—just inference—is there an even more narrowly defined chip that, in return for giving up on general-purpose calculations, can be even better for that? Probably. And that's what Etched is making, right? If I just optimize for inference, just like Cerebras and Groq did—different versions of inference—but if I just optimize for this one thing, I can probably do it more efficiently than the general thing.

So it totally makes sense at that level. And then the only questions are: Is that market big enough? Probably. This turns out to be the biggest chip market on the planet, right? Because inference compute consumption is huge. And then the competition question and the ability-to-execute question are kind of specific company-level things. I'm not going to opine because I haven't looked at the deal. But that's the big-picture bet.

It's funny to see it happening to Nvidia when, 30 years ago, they effectively did it to Intel. What's interesting is you have Groq, you have Cerebras, you have these—there are about 10 or 11 companies doing it, all chipping away, no pun intended—and then $300 or $400 billion a year of spend.

But it is still hard. I mean, you talk to the people at Cerebras: huge home run, amazing achievement. You talk to them about the technical journey, and they're like, “Oh my God, that was hard. That was a long 10 years.” So it would be hard for these companies if the timing of tape-out happens in a capex decline. It'll be hard. If it happens while there's still kind of mass demand, then that'll be a lot easier.

[Speaker?]

I'm not a total expert. One, it is the largest market that exists today, and it is growing at a scale we've never seen before. So, you might as well make a couple of bets on it, right? Some will implode, and some will be mediocre. But the market's so large that, listen, I'm not an expert on Etched, but if all Etched does is work with some subset of open-weight models that are allowed in the US, it's a huge market.

Chips, memory, compute—the margins are abnormally high, too. And the market's so large. I don't know whether Etched is worth $10 billion or whether it's an option that's worth $200 billion, right? That's the venture question. My guess is it's not worth $10 billion, but my guess is it could be worth $200 billion. And if your fund size and your winners work out, you make this bet and it makes sense, but it's probably not worth $10.3 billion today.

Totally agree. Yes, great round for the company as well. 3% dilution. $300 million is a lot.

I love the 3% and under rounds. I'm a fan of those. Right. Okay. Well, Google accelerates Cloud to 82% year-on-year growth, but prints its first-ever negative free cash flow.

8. Google Accelerates Cloud to 82% But Prints Negative Free Cash Flow

Guest 2

Top line was great: $119 billion Q2 revenue, up 24%, passed consensus of $116 billion. Google Cloud accelerating 82%, as I said, and it did not come out well. The reception was not great. How did we think about this, guys? You can't get lost on the day. I mean, it's still—year to date, it's still up 6%. Microsoft is down 17%, right? Nvidia's only up 5.9%.

So I think getting lost in the details of the day's response—it was a mediocre response, and they agree—is a mistake. I think it was 2 things, for what it's worth. One is its capex spend, and is it going to yield a return? And then, secondly, which was more intangible—and you can't prove why stocks go up and down; they just move—but the other thing was analysts pushing a little bit on, “Hey, basically, why isn't Gemini as good as the other guys?”

On the first, it kind of has been a surprise that they're going free-cash-flow negative, because you can predict the cash flow, you can predict the spend, and it's like, duh, this is knowable. And you're seeing it in a bunch of different places. People are just getting mildly scared about the bet, and that's not to say they're right or wrong. Maybe this $200 billion will have an amazing return.

And the ROI on capex, just to be a neocloud hyperscaler—forget owning a model, just the business of renting compute to OpenAI, Anthropic, and xAI—has been a great business. Therefore, it will continue to be a great business, and it is factually accurate to say it's been a great business. The ROI has been—I mean, Elon is making out like a bandit on his gas turbines in Memphis, or whatever it is that we're closing our regulatory eyes to.

The question they're asking is: If you spend $200 billion, will that have a good return in 2 or 3 years' time? So there's angst around that, which is really just a derivative of saying I'm angsty around Google—sorry, around OpenAI and Anthropic.

Guest 3

The interesting thing is, a couple of things, but the markets are just nervous, and it's very logical. The Korean markets are down 28% this month as we record this—massive panic in Korea hit the circuit breakers because of such a run-up and so much exposure to semiconductors and memory. So much exposure—that's nervousness. It doesn't completely tie to last quarter's numbers. It's worries about China; it's worries about AI.

I can't really—I’m not smart enough to calculate the beta or whatever—but it's logical. When you have this incredible run-up at the pace we've had, our market shouldn't crash 20% in the US. It could, but it could crash 10%, right? But 28% in Korea is a derivative of AI panic. And I think we're just going to see more and more crashes.

I'm a simple guy, so for Google, I'm going to—I know Rory may mock me—but I'm going to stick to the top line. I just want to see how the revenue is growing and the bookings. I'm going to ignore all these issues about the margins. Not that you can, in theory, but it's too much for me to figure out. I just want to see where the top line and the bookings are growing, and that's enough for me to understand the meta trends.

Guest 2

For once, I'm in sync. I thought the top-line revenue growth was amazing in Google Cloud.

Yeah. It says things are pretty good in AI land. But it's just this commodity where, if it were cheaper, we would consume an infinite amount of it, and we're coming close as it is.

Maybe Etched will solve this for us.

Guest 2

Anyone who has capital and can build compute can sell compute. You know, Google can do it. SpaceX can do it because they’ve done it successfully, I think, to Google and to Anthropic. So there’s just infinite demand for compute right now, and obviously, if that were to change, then all bets are off. But until it does change, all bets are on.

9. Enterprise AI Budgets: 2026 Will Be the First Year of Real Clampdowns

Yep. I agree. And exactly—it is so fun to watch. I’ve got my tickers, and I watch my WCLD versus the S&P, which is software versus the S&P, and then WCLD versus SOXX, which is the semis. For real action, the ETF that’s DRAM—that’s just memory—and that thing jumps 10% or 20% a day. When Korea has a bad hour, it’s down 7.5% on the day, when the S&P is up 0.48%. You were right. It’s just nerves. Oh my God, I own these stocks. They’re either going to be amazing or shit, and I don’t know which.

The big meta one to me—and this is why, to me, backlog is almost more interesting than revenue growth—is, I think, it’s discussed, but it still seems to be under-discussed. As we record this, we’re just before planning season. Last year was an experiment; this year was caps on token maxing. It got out of control. Next year is going to be very explicit budgets for everybody on AI.

It’s Captain Obvious to say this is one of the areas in venture where everything’s up, everything’s great, everything’s great. Not the pre-AI companies, not the ones of the past, but everything in the future has no ceiling. There is no ceiling to any of the companies that have been discussed on this show, and next year, I think, will be some of the first ceilings.

I haven’t had the CIO discussion, so I don’t know where it’s going, but they’re just going to be kicking off over the next 60 to 75 days. What are we going to spend next year, guys? Token maxing—open source is part of it because that’s load balancing, and agent harnesses are part of it. Everyone’s going to have to get more efficient, but the CIOs are just going to clamp. Next year will be the first real clampdown that’s material, and if nothing else, it could create a lot of variability here, a lot of micro-crashes and variability.

Rory O’Driscoll

There’s a hidden dynamic. There’s 1% of companies who token-maxed, and they’re going to be getting their shit together next year and kind of reining it in—maybe 5%. Then there’s 95% of companies who’ve barely put their toe in the water. If even a quarter of them put their toe in the water, the growth from the toe-dippers will swamp the reduction from the token-maxers. You with me?

Because that’s the dynamic here. There’s no doubt companies like Coinbase are going, “Oh my God, we spent so much. Let’s cut it by 50%.” That has a real impact if you’re Anthropic or OpenAI. But on the other hand, there are 10 companies in middle America that are like, “We have a ChatGPT subscription. Maybe next year we’ll try some of this Codex shit,” right? The question is, those 2 countervailing forces are what really drives it.

Actually, one of my colleagues was just talking about what you’d like to know most. An updated cohort analysis for Anthropic on the revenue build would be the single most useful piece of information you could have. Run that through a cube, and you could trade the QQQ for the next 12 months because that’s where it’s all happening. That will pick up the fusion between the token-maxers getting organized and the new guy, the toe-dippers, expanding.

That will filter back into, as you say, all these compute budgets, because a lot of the compute sales have been—all these guys selling to those 2 big frontier models—it’s all in that data.

Yeah. I never hit my Claude limit on Max or whatever. What is it? $200 a month for Max. That’s like $14,000 worth of tokens. I hit it this weekend for the first time ever.

Rory O’Driscoll

If you fessed up and were a company, not just a person, you wouldn’t get that deal anymore. You would be on the API.

That’s true. I guess I spent $14,000 last weekend. Even I don’t want to spend 14 grand on tokens.

Rory O’Driscoll

Well, they might find you after this podcast. You might find, “Hi, this is your Anthropic AI sales rep.” [laughter] “I’ve got good news for you. You’re on the enterprise plan. I’ve got bad news for you. You owe me $4.”

No, I hit it, but I’m with you. You might get the biggest round of the week. Travis, baby, is back.

Travis announced he’s raising $1.7 billion for Atoms, an industrial robotics company led by the one and only Andreessen Horowitz, with Ben Horowitz joining the board. Bain Capital and Fifth Wall joined alongside a load of other firms. I saw pictures with Kevin Hartz at A* Capital, Christina from Chemistry.

Is there room for everyone in a $1.7 billion round? How much can I put in?

My favorite is the pictures are all from the same restaurant, in the same place in the restaurant. I think there was a rotation.

Rory O’Driscoll

It’s the way they do it for political stars, too. You just line them up, shake, grip and grin, and move on to the next person.

Here’s the first tranche investor, second, third and fourth. Right, guys, what do we take from this? Is it just a news announcement? What are your thoughts?

Rory O’Driscoll

Right. It’s kind of an industrial holding company across—basically, the big picture is that Atoms, as the name would say, is physical AI, AI for the real world, doing a bunch of different robotics businesses—very different robotics businesses. Some are around cloud kitchens and food preparation, and some are mining.

The things that I agree with are the comments Travis is totally correct in making. It’s not humanoids; it’s specific-purpose robotics. I actually think he’s correct. I think we look back on humanoids and go, “We got way ahead of ourselves.” You actually need specific-purpose autonomous machinery for B2B in general. That makes sense.

I think it’s not as clear to me why it makes sense to have Pronto for mining. Other than the fact that Travis is amazing and can raise capital cheaply, it’s not at all clear to me why food prep and mining should be in the same holding company.

Rory, would you have broken your rules for your LPs to put money into this?

Rory O’Driscoll

No, I don’t think I would have. I’ve done a lot in robotics over a decade and a half. As I said, my first robotics deal was in 2016. My first drone deal was in 2015 or 2016. There’s a lot of feeling now that they’re going to happen quickly. I think they’re real and significant, and everyone uses the, “Oh, the GDP of the real world is bigger than the software world.”

Well, no shit. It turns out 2% of the world is software and the other 98% is real. I just think it takes a lot longer than you realize to roll out robotics in the real world. It’s not clear to me that putting a bunch of different companies together in the same place makes it any better.

It is doable because he can raise money at a great price, but by definition, a great price for the fundraiser might not necessarily mean a great price for the investor. So even though it feels like heresy to say it, and I could be totally wrong—and if I am, that’s great—based on what I know from a distance, it’s not obvious to me.

We’re always going to say these aren’t connected with a lot of things they say, but I think it is connected. There was an article in The Wall Street Journal today saying a big trend is bringing CEOs out of retirement to run big companies. They fired the CEO of Cracker Barrel, even though the stock’s way up after the logo mishap, and brought in some guy they found out of retirement who ran the parent company of Outback. Very successful.

PayPal did it. They found whatever his name is out on his Montana ranch to come back in and run PayPal. So what’s my connection here? I think we’re seeing almost this bimodal trend where we’re going to make some bets on the 20-year-old founders of Etched, right? And we’re going to keep making those Cursor bets.

But when Jeff Bezos says, “How much did Jeff Bezos raise for his company when—”

Rory O’Driscoll

$6 billion or $12 billion. It was the biggest financing in Q1.

When Bezos, when Travis, when Elon raised their hands and said, “Listen, I’m going really big, guys. This is not about making a couple of nickels, okay? I’m building something massive in the moment of change in our lifetime, and it needs billions of dollars,” you’re going to give it to these iconic, seasoned veterans, and you’re going to have faith that it works out.

Giving Elon money for Twitter back in the day was faith-based. There was no rhyme or reason for that deal. And the Boring Company, to me, is crazier than Atoms. The Boring Company is crazy. There’s 1 little route in Vegas, and I’ve done it, and it’s cool. A dude drove me through a tunnel that ain’t worth $20 billion.

But I think we’re going to make—as funds get bigger, as we see so much of the benefits to venture come to massive outcomes—we’re going to give Travis the money. It’s going to be this, and there are only so many Travises. So they’re going to hoover up the cash.

Rory O’Driscoll

I would—as we say in California, I would change the pronoun, people. I don’t think we are going to give the money. I think someone’s going to give him the money, just to be precise. The objective facts of whether something works or not are independent of who finances it and independent of who runs it.

Guest

I mean, you made a comment on Twitter. A hard-nosed comment here. Absent the fact that Elon decided to bail him—bail his investors out, for which huge credit and kudos, right? Twitter is not worth $44 billion today, and not even close. So, objectively, in terms of buying something, you bought an asset that went down in value. Now, if your business plan is, “Oh, and by the way, he’s got other assets and he’ll bail me out,” maybe, but that’s not actually a plan.

But Elon owns Twitter now. Does that change your perspective on it being worth $44 billion?

Guest

No, Harry, it doesn’t make it worth $44 billion, because the cash flows don’t get you there, right? It’s $2 billion or $3 billion in revenue, down a little. Maybe it’s growing now, and maybe you’ve crawled your way back to $30 billion or $40 billion now, but it’s a push. Now, as it happened, you got a 3x because he rolled it into xAI and rolled X in.

But my point is, you are right, Jason, that these big-name things are working, but they’re working not because the facts are working. They’re working because the market is continually willing to enable that process. And if the market changes, you don’t have value.

10. Thoma Bravo's $21B Fund: Is the PE SaaS Playbook Already Dead?

Yeah. I just think the market’s going to hoover up all the Travis. And watching him on social media, the dude’s got the energy to do this, right?

Guest

Oh, no. No question.

No. So, if they’re burned or broken, you can’t make the investment, right? But I think everyone—and there are only so many folks—but if Bezos is done partying at Carbone and wants to do this, okay, and Travis is done doing his 70-mile jet to the office in Austin and really wants to spend 20 years doing this, the funds, in quotes, can raise the capital. These are the bets of the day, and they’re going to get a couple of billion to do it in an era where the amount of wealth creation is unprecedented.

They’re going to get it, and you’re also going to hunt out the kids from MIT. I don’t know what it’s like to be a freshman at MIT today, but it must be exhausting. Every damn VC wants to fund you. I would be burned out if I were in the top 10% in math at MIT. I would just have a placard on my shirt: “Leave me alone, VCs. Leave me alone. I’m with Neo. I’m with Neo. Leave me alone.”

If you’re the parent of one of those kids at MIT, and you’ve broken your back for 20 years to get your kid to stay focused and get to MIT, and now those evil VCs are saying you should quit and drop out before graduation, you want to plummet to death.

But, going back to the thing, just to comment on the other fun fact, you’ve got to note the fun fact, which is Benchmark’s best fund—one of—I shouldn’t say the best fund, because actually their eBay fund was the best fund in ’95—but they had an amazing-looking fund that had both WeWork and Uber in it, in about the 2012–13 fund, right?

And, you know, if you fast-forward, a couple of things happened. One is, they famously swapped out Travis for Dara as CEO, to the undying hatred of Emil Michael, who’s now at the Department of Defense, and obviously Travis, and there was a lot of controversy around that decision. Obviously, it went on to be an amazing company. They didn’t swap out the WeWork guy, who went on to pretty much fail as the company, even though he personally took out $500 million. The deal didn’t work.

So, at one point, there was a fund with 2 amazing mega-fund-returner deals, 1 of which turned into a mega-fund-returner deal and 1 of which didn’t, right? And just the 1 who made the change did. It’s an $80 billion company today.

The fun fact is, fast-forward, and you even saw it in the tweets around the Atoms round, and Andreessen Horowitz have backed both CEOs. They backed Adam at WeWork because they’re like, “We think you can do it again,” and they’ve just backed, obviously, Travis.

And if you look at the tweets at the time, there was a very direct tweet just last week, basically: “We should have done this deal in 2010 or 2011,” which is tantamount to saying, “Yeah,” and everyone can read the subtext, “We deeply regret taking money from someone else who fired us, even though it turned out to be an $80 billion outcome,” right? So, there’s a clear dynamic in the venture backstory there.

What I really admire about what everyone involved is doing is the willingness to bear grudges across a decade. It’s quite impressive. So, yeah, it’ll be fun to see how that shapes out. But, yeah, Andreessen Horowitz have backed both CEOs from that famous Benchmark fund.

Well, to be clear there, they did get out of WeWork and have a mega-fund returner.

Guest

Oh, look. Amazing. Amazing. My point is, I don’t think the WeWork return was nearly as compelling as the Uber return, Harry, because WeWork didn’t get public. It went bust. It was SPAC’d. It went bust. It was not a returner.

My point is, they had 2 home-run winners, 2 huge burn companies, 2 potential fund-returner deals, and 2 wildly charismatic CEOs. One of them stayed the course. It didn’t work out. One of them was replaced controversially. It did work out. It’s an $80 billion company today.

By the way, it’s very minor, but who knows? Claude says Benchmark took out $315 million from WeWork.

Yeah, I was literally just about to say the most valuable lesson I’ve learned as a VC is to admit when you’re grossly wrong, which, to me, happens daily. But exactly, Uber made Benchmark roughly 640x versus around 25x with WeWork.

Guest

They got 25x off. Good for them. I was wrong. I’ll say it.

No, I was admitting I was wrong, dude. They’re very different.

Secondary to SoftBank: $315 million out of $17 million in. I will still take that for my fund. I mean, a $17 million concentration for me, I would do it, right?

Guest

But, yes, saved by SoftBank versus a viable independent company a decade later worth $80 billion. But you are correct. It just shows that, with enough momentum in a bull market, if you take your winners off, you can do well on everything. Good for them.

Well, the reason I wouldn’t back Travis is he just doesn’t have that chip on his shoulder, you know?

Guest

Oh, that—[laughter]—just to be clear for everyone, that’s obviously tongue-in-cheek. Wow, that man has a chip on his shoulder. I love it, and I love the drive.

8 years in, you get over it, you know? Billions of dollars, you’re like, “Oh, let it rest.” Nope. Nope. Nope. No. Not only am I not going to let it rest, I’m going to prove them wrong. I’m going to tweet along the way that they were wrong. Absolutely. [laughter]

By the way, tweeting that someone is wrong is different from someone actually being wrong, just to be clear.

No, I know. It’s a very different end of the spectrum. But we talked about enabling the supply side of capital—venture firms providing people like Travis with huge amounts of money. Well, in other areas of the market, it ain’t exactly capital-starved either. $21 billion, which was above the target, for Francisco Partners. Wow. The demand ain’t dying on that side of the market either, is it?

Some of the marketing—I mean, obviously, a wildly successful run, a super track record stretching decades here, right? So, can’t argue with it.

Guest

Where I get confused is the messaging, which may not—sometimes the way investors message and what they actually do are not 100% identical, right? Sometimes they’re just directionally aligned.

But a big part of the $21 billion is that AI won’t kill software, and that, therefore, there are efficient ways to deploy this. That’s the one where I get confused, right? I’m not sure there are these gems out there growing 14% that they can buy, hook up Kimi and the Moonshot API, and magically reaccelerate growth to 70% or what.

Every week that goes by, I feel like the past is the past. It’s time to leave the past in the past and let the markdowns be the markdowns. Raise another fund. Hopefully, you’ve got an Uber in there, and maybe a WeWork that you cashed out, and move. It’s just time to move on.

Guys, I hear you, and I think you’re correct. Nothing is going away, but remember, the venture game is all about finding things that explode in growth. Price matters only, at best, at a second order. Growth matters first.

In the PE business, it can be the other way around. A company that’s growing at 7%, that you buy dirt-cheap and get to 20% growth and good cash-flow margins by applying the leverage and getting the lift from that—you can make your IRR.

It’s almost like—I always think of it as literally the opposite ends of the life cycle. We’re in the “grow new things, make them amazing” business. And, to some extent, a lot of these things are, “Rationalize these companies, make them work, make them a little more efficient, a little growth here, and just sell them on an earnings multiple.”

Guest

Now, I can see it in theory. I agree. I think that’s harder to do in a world that’s moved on from that entire category. I mean, even the folks—you can only raise prices so many years.

How do they raise it, then?

Guest

Well, they have the track record, right? I mean, this is the job. This is private equity’s job. They have the track record.

I’m not saying there aren’t gems out there. I’m just saying I keep hoping that these companies that I know become gems, and every interaction, every conversation, every week that goes by, I feel less and less giddy. I don’t believe this thesis. The deeper I go in my VC career and the more I work with these companies, I feel like anybody not working at least as hard as Mark Benoff is just not going to make it.

And in PE, you put—you bring in the 69-year-old Montana farmer to run PayPal. I believe it works, too, but a lot of traditional software targets, I just don’t buy it.

Guest 4

But I hope they buy some of my companies. I'm all for that. I have a couple that I would love for them to buy, but I'm just losing confidence that anyone without Mark Benoff-Travis energy wants to work at these companies, maybe with the exception of Bending Spoons.

11. ServiceNow vs Salesforce: Which Legacy SaaS Actually Survives?

But I think you're conflating your personal issues with the person who's running a company. They're pretty comfortable getting someone else to run it, right? The abstract question is: do you think these kinds of companies growing at 15% can be bought cheaply enough? Can you buy a Monday.com or a Wix at a price where you can make a return from a combination of leverage, operational efficiency, and slight AI growth?

That's really their question. It's not an inspiring question. I'm glad I don't have to get up every day and deal with that, but it is a legitimate question. I don't know what the market cap of all the publicly traded SaaS companies out there is, but there's probably $1 trillion, plus or minus, of so-called legacy software. Can 10% of that be run more efficiently with leverage, put $20 billion to work, and get $40 billion back? Probably.

Guest 4

If you look at Wix, it's trading at less than 1x revenue now. Basically, that means you're paying 10x for Base44's revenues, and you're getting the core business for free, primarily.

And, you know, look again, I don't love it. But from when we talked about this, you made me name some individual stocks, some of which did well and some badly. But actually, I'll tell you what I actually traded: I bought WCLD, the whole index, when we had that SaaS apocalypse discussion, and I'm up 35%.

The point is, what you're never going to do is pick stocks and go 3x. There's just not that kind of upside in these things. The question is, PE is not looking for that. Can they pick out the gems? I mean, you're right. Should they take a run at PayPal? Advent, you know, is going to do that with Stripe. At 2x revenue, 1x revenue, 3x revenues, are there returns here? That's the question. I wouldn't assume no, which is different from saying I want to spend my life doing it.

Guest 4

The problem is, so many of these targets basically have no net-new customers. It's expansion and price increases. And if you're early in the expansion and price-increase cycle, you can get 3 to 4 years out of it. But we're 5 years into no net-new customers and price increases and mediocre module expansion. I don't think there's another 5 years of those knobs and dials left.

Agreed.

Guest 4

Again, we just turned off Marketo. They raised our prices from $22,000 to $80,000 since 2020.

Wow.

Guest 4

We left. I bet they've lost 20% of their customers over that period of time. I know it's an extreme example, but what are they going to do? Take someone like us and charge us $160,000, $640,000? The blood is beyond out of the stone, right? The stone is crumbled because all the blood has been squeezed out of the rock and it's turned to ash.

$22,000 to $80,000 since 2020. It's at the edge of criminal. It's at the edge of criminal. We didn't even get a thank-you email after being a 20-year customer. “Thank you for being one of the first 10 customers and being a reference on our website. Jason, sorry we lost you. We hope to get you back at $160,000.”

Listen, they're going to make money, Francisco Partners. I have lost almost all confidence in the turnaround playbook working. It's too many years of blood from a stone. It's too many years.

I think that's interesting. You've been pretty consistent on that, and I've come to the conclusion you're correct, which is, if you've done 5 years of price increases and that's all you've got for revenue growth, you're probably closer to the end than the beginning.

In a weird kind of way, a lot of these guys go in and run this test: “If they've raised prices and no one's blinked, that means they can continue to raise prices.” And you're right, Jason, that could be a counter-signal. If I was sitting on the investment committee of one of these PE firms, I actually think you're right. I would want a test that says: Can we add net-new revenue and net-new modules from these customers, or are we just screwing them? Because if we're just raising prices, it's going to end at some point.

So, I agree with you there. I think target selectivity will be really important here, which means that it won't be nearly as big or as easy a business as it was in the last decade and a half. You’re right: there might be isolated pockets of winning, but it's a tougher gig than it was. In 2010 to 2015, it was a great business.

In fact, I think even going back as far as 2002, 2000, the early Vista funds were just after the dot-com crash, and those guys hoovered up and made a fortune, right? But we're probably 20 years into the no-brainer SaaS bet, and at some point, as Jason has proved, even a lighthouse customer will turn off their SaaS. Where are you in ServiceNow? Just curious, because they obviously had their little bump this quarter. How do you think about owning something like ServiceNow or Salesforce?

Guest 4

I think the problems they solve are so sufficiently complicated that you need them. No, you just can't run your business without ServiceNow.

Right.

Guest 4

But if you talk to anyone that's run on ServiceNow, it abstracts away so much complexity from your business. Analytics is like the cat. We all talk about vibe-coding things away; analytics is actually the easiest one to vibe-code away.

I agree. Got it. That makes sense. So, what you're basically saying is there will be islands that stay, and lots of little—it's what they've said—all the ancillary in-between products, like analytics, to-do lists, and task management, will go away.

Guest 4

Yeah. Some of it, I don't know.

Well, I want to roast Francisco Partners, though. You just want to sell your company, Jason; you're so transparent. You know what? Maybe we should move on here.

Guest 4

There is something I believe in in this whole model, right? I don't believe in buying the 17% or 15% grower because we're 5 years into the price increases and no net-new customers. I do believe, if I were PE, that the model is the 40% grower—the one that is kind of working today. It's just not growing at the rate we would like in today's world.

There is no perfect path to a great exit, but they have an agentic product. They have something going at 35%, 40%, 45%, 50% growth. There may be a moment in time where you can have a very attractive multiple on that property, and you haven't gone into terminal decline. That's where I would spend my emotional energy.

What's still growing approaching 40% or higher at scale, where the founders are burned out? It's sort of made the transition, but it's not growing exactly at the rate of the hottest startup in its class. I might make that bet.

Boys, you can choose. We have Stripe hits Rule of 80. We have Monday.com lays off 20%. We have likely Mark Pincus's “Quit if it's too hard. Life's too short to struggle.” I can't believe that's actually advice. Where do we want to go, boys? You can choose.

12. Mark Pincus Says Quit If It's Too Hard

Guest 4

The Pincus one I don't want to spend too much time on. I did quote that one, right, and got some traction around it. Arguably, he did say that, right? He's a consumer guy. He comes out of games, too, right?

In a sense, it's quoted out of context by me, but there is a point where a game—unless it's Cyberpunk, which is back now, right, after 5 years in oblivion—in most cases, you should quit on a game, right? Probably at some point. You got it wrong and you move on, right? It's like quitting on a movie. At some point, you have to move on.

But it is what it is. I don't want to be grumpy. It makes me sad when a founder quits one of the ones we just talked about, a 40%, 50%, or 60% grower. A founder with material ownership quits to do something hotter, and I'm cool with that in the age of AI because everyone's like, “Well, there's so much opportunity cost, Jason. I can found Etched in a week at $10 billion. I can get into YC and raise at a $100 million post. My round will be fully subscribed before I even finish the batch.” I can't argue with some of that.

Most of the founders I've worked with over my career that have quit something pretty good to do the shiny penny—they're not all Ilya. It's all been a net negative every time I've seen it. That's my worry in the age of AI: there's no downside today. Just quit, quit, quit everything. Go found an AI company. But, man, if you've got $20 million, $50 million, or $500 million in revenue, I might see if you could build that in-house.

Jason, we saw last night that likely Lilian Weng left Thinking Machines Lab, which makes only 2 of the original 6 co-founders remain.

Guest 4

Same thing, I guess, probably, right?

Well, dude, it's Thinking Machines Lab. It's not exactly a 40% boring SaaS.

Guest 4

And she probably has 1% if she's a late co-founder, right? So, what's Thinking Machines worth on paper?

$8 billion.

Guest 4

Oh, she's only got $80 million. I'd leave that behind.

I'm having none of it. I think we're conflating a lot. I mean—

Guest 4

Are we?

Well, I think there's—

Guest 4

We're always like, “I've had enough of you, degenerates.”

Rory O’Driscoll

No, no, no. I think the whole comment on quitting—I think there are really 3 different things. Jason, you’re right. The least obvious one is: I’ve got a company, it’s at scale, it’s growing 40%. Do I? I think you’ve created something of value. It’s not obvious that chasing the next shiny thing will be better.

But I don’t think that was what the reference was to. I think the reference from Mark Pincus was more: how long do you keep trying to get product-market fit before you say it’s just not there? I think that’s a valid comment, in the sense that I don’t think, quote, the answer is, “You should quit.” But the nuanced answer would be: I think you should not do anything out of duty. You should do it because you think you’re converging—you have a plan to converge on something.

And when you don’t have a plan, you shouldn’t just tie yourself to the mast and keep going just because. Frankly, 35 years ago, when I had my business way back in the dawn of human time, I was a very mediocre manager, and I stuck at that thing 2 years longer than I should have. I look back and go, “Wasted years,” just because I had that kind of, “I owe it. I need to keep trying. I don’t want to quit.”

I think finding a way to step back and say, “Am I doing this because I still believe in the mission?” In which case, no matter how hard it is, keep going. Or am I doing this out of a sense of obligation, and I don’t have any way to win? In which case, put up your hands.

What I disagree with Mark on is the ability to say to a founder, “This is the answer: you should quit.” I don’t think anyone ever knows. I think the good advice is to say, “Go in with no priors. Take some time away. Ask yourself honestly, when you’re rested and you’ve had a good night’s sleep, does this feel like something you want to do?” If not, and if you don’t have a plan, then, yeah, you got it. Yeah.

Guest 4

You know, if I took that advice, Rory, honestly, all I would have is a maxed-out 401(k) in life. The only reason I had any economic success is that, in part, out of obligation, I kept going.

If it was just about me, I would have quit both my startups. Certainly, I would have quit venture investing. It’s not worth it for a few nickels. I certainly would have quit EchoSign. My founder walked out the door after 8 months. He was right. This category was never going to take off. Yeah, you’re giving the same crappy advice likely Mark Pincus did: quit when it’s hard.

Rory O’Driscoll

Look, we all have shared experiences. You’ve lived it. Maybe you felt along the way that you should have quit, but you ended up building a very nice company, making a ton of money in a sale, in a category that’s turned out to be significant. You know, I sucked at it 2 years longer, went bust, and look back and go, “That thing would never have worked.” Right?

So, to some extent, it’s the old likely Kierkegaard thing: life is lived forward, but can only be understood in reverse. I think when you look back on things that fail, there are some things that you look back on—deals that you look back on—and go, “Not only did it fail, but it was just never going to make it. There was just nothing there.”

Guest 4

I’ve never quit, and everything I’ve done would have failed if I quit. I’ve never been rampantly successful. I’m not a billionaire. I’ve never made my investors less than 5x. Everything almost failed, and people just quit. They just quit—more like Thinking Machines. $80 million is not enough.

Go back to my comment. It’s your experience to go, because I’ll take the opposite statement. I’ve never quit, and I’ve failed at some things. I’ve succeeded at some things, but I’ve failed at some things. It turns out that doggedness until the end of time is good, but it actually doesn’t guarantee a win.

So, did you not learn more and gain more from the experience, though, of those extra 2 years that you took with you?

Rory O’Driscoll

Jason and I are going to agree: the answer is no.

Guest 4

No, Harry, there’s a great line someone gave me when I failed, and it’s this: “Experience is what you get when you don’t get what you want.”

Rory O’Driscoll

Right. No, I’ve done all my learning 2 years earlier. I had fully processed it all, and the last 2 years were just hell on earth.

I think you do learn a lot, but I think that’s one of the dumbest things out there: that you learn so much from these failures. I do think you learn a lot, whether it’s investing or otherwise, from the almost-failures—the ones that turn around. You learn why, because we’re all sitting on some companies we’re not sure about, right? Where are they going to go?

When you have a few portfolio companies that do turn around, you do learn a few things. But the ones that ran it into the ground—you learn something, and I think there are things to learn, but you don’t need to live it for years on end to learn it. You can process through it. So, no.

Again, on balance, Jason, I am emotionally more in your camp than in the Mark-just-quit camp. The point I’m merely making is, if the only reason you’re hanging on is duty and you see no hope, you’re actually going to fail anywhere. That’s my theory, but we can disagree.

Boys, is there any final topic that we have to discuss?

Rory O’Driscoll

I don’t think so. Yay, Stripe.

Guest 4

I’ve got one question. I have one question for Stripe, for Rory. Maybe if you want to break, because this is the first time I tried to write it up, and I never had a chance to compare it to Adyen. It doesn’t seem so wildly overpriced given these numbers.

Rory O’Driscoll

The numbers are great.

Guest 4

Compared to Adyen, it doesn’t seem appropriately premiumized to Adyen, right? I always thought they were peer companies, but Stripe’s much better.

Rory O’Driscoll

Yes. They have much higher profitability.

Guest 4

Yeah.

Rory O’Driscoll

And I think what’s happened is they’ve hit a sweet spot. They charge more; they have more smaller merchants and higher pricing. For a long time, they were less profitable despite that because they were Silicon Valley soft. Stripe was a soft company, and Adyen were a hard-nosed bunch of Dutch people, right?

But about 4 or 5 years ago, when the team—the Collison brothers in particular—focused on efficiency, they made it an efficient company. So now you have a company with good pricing because the 2.75% is attractive, and they’re efficient.

Then the third key ingredient happened in the last 2 years. They basically signed up all the AI companies that are selling shit online, right? And they shouldn’t be getting anything like the money they’re probably getting from OpenAI and Anthropic in terms of interchange fees, but who’s got time to optimize that stuff? They’re designed into the flow of companies that are just printing money. So they’re printing 2.75% of that money, right?

What that means is the growth’s accelerated. When you have an efficient, leveraged cost structure, probably using a lot of AI to stay efficient, and then your revenue takes off, it all just flows to the bottom line.

I would have said 5 years ago they looked expensive relative to Adyen, but now they have the wonderful combination of super-strong growth, good pricing, and wonderful margins. So, yeah, I think that it’s a sweet spot right now, driven in particular by this kind of lift from online AI spending.

13. Will the OpenRouter Deal Actually Close?

Can I have one final one? Will the OpenRouter deal happen? It seems to have gone super quiet on that front, and then everyone’s released their own routing product. We saw Cursor release it. One of my companies, Merge.dev, has released their own, and it seems to be a very commoditized market very quickly. Will we see this transaction complete? Do we think—

Rory O’Driscoll

From a business-model perspective, in other words, the kind of front-end API to aggregate a lot of complexity—OpenRouter does for LLMs what Stripe does for money and what Twilio does for telecoms—so it kind of makes sense from a company-model perspective.

I mean, Jason said it last time: smart of OpenRouter to get out. $10 billion felt like a lot, but good luck to them. I mean, it could happen. Go team.

I wish I had the data on the time from an intentional leak by a VC to deal closing. But I believe it is more than 1 week on average, right? And so this certainly appears to be a leak to generate a pseudo-second offer, potentially to justify a premium price.

Listen, every company is the same. I’ve been on the other side of Stripe deals. They do what you’d expect. They offer an acceptable but mediocre price from a venture perspective, right? So maybe they offered the last-round price. Maybe they offered $2 billion. They asked for $10 billion. I don’t know what the exact story is.

Someone leaks it. This is how you do things today, right? My guess—I don’t know what happened—my guess is Stripe said, “You leak again, it’s off.”

Guest 4

You’re exactly right. They could be hunkered down doing a deal.

Rory O’Driscoll

Yeah, it can take a couple of weeks. Even if you want to work all weekend, deals don’t—in my very limited experience, deals don’t close the day after the leak. You have to sequence the leak properly, or it won’t work.

Guest 4

No, it’s part of a price negotiation. It doesn’t really create another deal that closes. It creates leak energy. It’s much better than a banker pretending they got someone to add you into the deal. You create this leak energy, but you need a couple of weeks for that to work.

Does it go through then, Jason?

Guest 4

If it’s real, it probably does.

You leak. Listen, I only have a limited amount of leaking experience. Maybe Rory has more. You don't leak a fake deal; it doesn't accomplish anything.

People look at leaks like they're trying to put the company up for sale, right? But I don't think that leaking strategy works. The leaking strategy works for a good but not great offer because you don't have—you only have so much leverage. I'm terrible at negotiation. I'm terrible at game theory.

But if Stripe wants 2 and you want 8, it's a very awkward position to be in. Maybe Frank Quattrone solves this for you. That's the magic. But if you don't have a solution, the leak is the best idea.

What I learned at Adobe, just to maybe overtalk about it, works in the sense that big-company M&A indeed isn't brutally slow. I'm sure Rory will agree with this, right? But all of them have a deal mode.

When an email comes in and says someone who was on the list is in play, it does not mean that Adobe will buy them or Google will buy them, but they spring to action and make a decision within a couple of days. They know how to—they literally go into deal mode, and that can at least get you a paper counteroffer.

You can go back to Stripe and say, “We think we have an offer from Adobe.” I mean, it's not going to be Adobe, who are at 5, but you need a week or so for the leak to work. It does work, and the big companies want—the thing is, it sounds crazy, but the big ones at least want their shot.

Just like Andreessen doesn't want to be embarrassed that they didn't see a deal that Sequoia did, it turns out it's somewhat similar in corp dev at companies. They at least want a shot to buy someone who's on their list, because they already know who's on their list, right?

As soon as they get the email, especially from a banker, they just shoot it around: “Guys, we've got to get together tomorrow and decide if we want to buy OpenRouter.” OpenRouter.

Rory O’Driscoll

By the way, Harry just explained that that's a dig: you guys in the UK call it “rooter,” and we over here call it “router.” We invented them, so we were allowed to pick.

You don't want to be the corp dev guy who says, “We didn't get a look at that,” when the board asks, “How come we didn't?” So it's drumming up interest. Who the hell knows what's actually happening?

Guest 4

It's utterly amazing. If you don't have another offer, it could take 3, 4, 5 months to close an M&A deal, right? If you have an offer, it turns out any big company can move in a week—not to close, but to sign a term sheet.

Any big company—you'd be shocked how fast they can move when they're in deal mode and there's a backend constraint. All the crap blows away, and then Mark Benioff or whoever just decides. They just decide; otherwise, it's months.

Final one for you: you can own Revolut at 115 or Stripe at 165. Which one do you want to own?

Rory O’Driscoll

Even though I love the Stripe vibe more—I really do. Look, they're Irish. I'd love them to kill it. They are killing it. They have killed it. They're going to do amazing.

I think the beauty of Revolut is you have a whole continent full of overpriced, crappily run banks that you can just roll over, right? You've got 500 million Europeans who are just getting shafted on financial fees.

The TAM for payment services in the US is enormous too, but it's just marginally more competitive. I think, look, they're both—I mean, worth pointing out, they're both amazing companies. Neither of them are, at their core, AI companies, though obviously Stripe's getting a lift.

They are both really well-executing fintech companies. I like them both in the sense that there's more to life. It turns out there's hundreds of billions more to life than AI, and those are 2 examples of it.

But at the margin, on a TAM basis of plus or minus 100 billion, there's just more compounding in those 500 million underserved Europeans.

Jason?

Guest 4

Honestly, I don't know, Harry. I just think that, at the end of the day, the moat at Stripe may be a little lower. The network effect may not be as strong as it seems, right? Banking just has marginally more powerful moats, and they're working on some network effects.

My YOLO version would be: take Revolut, because Stripe just has to continue to execute at an outstanding level. The network effects and moats are there, and they've invested in everything from Atlas to their own router to create the network effects, but I'm not sure they're truly there.

Boys, it's been a pleasure. Rory, I appreciate that ribbing. I will remind you that you Americans speak English.

Guest 2

We took it from you in 1776. It came with the territory.

You won the war. I feel like this is Basil Fawlty. Have you seen Fawlty Towers?

Guest 2

Not the war, Harry—the War of Independence.

Absolutely brilliant. I love that.