NVIDIA预测营收达1万亿美元：GTC全解读，Anduril拿下200亿美元合同

1. NVIDIA's GTC: What You Need to Know

Guys, I'm so excited for this. We were just talking about where to start. So much news. And Jason, I think you are absolutely right. It's very important to start with GTC, Jensen, and data centers in space. Obviously, Rory said, “No, not going to happen.” But maybe it does. How did we think about the data centers in space and last night's GTC, Jason?

Jason Lemkin

To me, the interesting thing—which you don't even have to watch anything for; you just have to look at the Twitter stream—is the sheer sense of energy, momentum, and confidence there, and the confidence to do things. Not only talk about data centers in space, but launch things like NemoClaw, which is their version of OpenClaw, and launch a partnership with Thinking Machines and others for their own open-source LLMs.

They're just going for it, right? You can just smell the—I mean, obviously, Nvidia is a pretty good stock and a pretty good company at the center of AI, but you can just smell the companies in decline. You can smell the company struggling. You can smell OpenAI struggling right now. You can just smell it. You can smell the code red and the fact that they said yesterday, “We have to concentrate on enterprise. We have to stop side projects.”

Any time you've worked with startups, you see the seasons, right? But, man, this is summer at Nvidia. They're on fire on everything, and all the things where you see—and they're at risk. They're at risk from their customers using TPUs from Google and Amazon, but everything from already integrating Groq to data centers in space, to crossing $1 trillion in bookings, to NemoClaw just feels like a company firing on about 13 cylinders.

I'll come back to your misportrayal of what I said later, Jason, but let's hit the main point, because in the end, it's all about the money, right? There's a super-interesting counterpoint here. On the one hand, you have an extraordinarily aggressive number: $1 trillion in revenue. On the other hand, the stock moved less than 1%. Why? It was already priced in.

Rory O’Driscoll

Already priced in. If you start parsing it out, I think what happened is as simple as this. If you look at the forecast, the company did $215 billion in revenue last fiscal year, up from $130 billion, I think, the prior year. The forecast is in the mid-$300 billions next year.

The sound bite was half a trillion dollars of demand over the next 2 years. That was a sound bite a year ago. Now the sound bite is $1 trillion in demand, but if you listen carefully, it includes 2027. Lo and behold, if you go and check the analyst forecast for 2027, it's in the mid-$400 billions. So it turns out if you add $500 billion and $400 billion, and then you apply a salesman's roundup, you get to $1 trillion.

Another way of translating that wonderful, amazing number of $1 trillion is that the analyst forecasts for the next 2 or 3 years look roughly right. So once the analysts process that, the stock said, “Yep, nothing—no new information here.”

Which is just fascinating, because what it says is that this is a company—by the way, for context—that 5 years ago was doing $20 billion a year. Last year it did $215 billion, 10× growth over 4 years. It had just under 60% growth last year, is forecasting 60% growth this year, and is attenuating down to 20% or 30% growth in a couple of years.

At one level, it's not insane relative to—if past performance is predictive. If you have a company that grows 10×, saying it's only going to grow 20%, 30%, or 40% over the next couple of years doesn't seem unreasonable. But what it does imply is a massive level of CapEx continuing for the next 4 or 5 years. That's fundamentally the bet.

That's the big statement that came out here: We think this level of CapEx investment is going to keep going for the next 4 or 5 years. We're affirming our estimates, and none of my customers are going to blink on our spend. That's the takeaway.

Jason, I constantly go back to something that you said, which is that we're going to see inference running 24 hours a day for a larger and increasing number of the working population. If we think about that, in 2030, will Nvidia be a $10 trillion company?

Rory O’Driscoll

I think $10 trillion in revenue is more interesting, because he's just predicted $1 trillion, right?

Again, we've got to be precise.

Rory O’Driscoll

Yeah, no, it's a cumulative trillion. How long does it take to do a cumulative $10 trillion? That's an order of magnitude. That's at the limit, I think, of general, non-inference-level human ability to grok—it’s about an order of magnitude.

So I think in 5 years it will have a cumulative $10 trillion from $1 trillion. I don't know. Take 5 years to go from $1 trillion to $10 trillion. That's my bet.

I do think the inference thing—I mean, listen, there's a lot of interesting things in the math. How fast will the cost of inference continue to fall versus the consumption of inference, right? There will eventually be a Moore's law—an inverse Moore's law—where we don't see the dramatic cost decreases we're seeing. Maybe that's why we need data centers in space.

There's a lot of complexity here, but even if you think, “Why the hell does Nvidia do NemoClaw?” Now, Jensen said it's getting the most GitHub stars per unit of time, faster than Linux, faster than anything. But I think the real reason is that it uses a lot of inference.

This is why everyone in China is giving away OpenClaw, because you've got grandma and grandpa lined up on the streets outside of Tencent and Alibaba with their free OpenClaw, because it just burns tokens, right? So part of it is Jensen saying, “We want you to be burning tokens at least 72 hours a day. We want at least 3 agents running 24 hours a day.” This is probably a big part of the Groq acquisition.

We just want you to be burning tokens. So, yeah, I think it's got to be 3 orders of magnitude more inference that we run in the next 5 years.

You could have 3× more tokens, but if the price per token goes down by 6×, the revenue will decline.

Rory O’Driscoll

Just pointing out that it might be 3,000 times more tokens, not 3×. But yes, the math is the same math issue, right? Again, going back to my comment, someone unveiled a trillion-dollar number and within 10 minutes the stock market processed it to “nothing to see here,” right?

It just points out the expectation of a railway-boom-, internet-boom-level of CapEx investment continuing unabated, going at 30% for the next 4 or 5 years. It's a pretty heroic assumption. It's validated by the recent past, and there's nothing at this point to say it won't happen, but it is just worth pointing out.

If Nvidia does $200 billion, the CapEx spend is probably $400 billion or $500 billion, because they get about half of it. If Nvidia is doing $600 billion, the CapEx spend is probably $1.2 trillion, plus or minus. These are unprecedented levels of CapEx spend, and now we're forecasting them to keep going for 4 or 5 years. There is at least some probability—pick a number, 30%—that it doesn't happen this way.

Is there anything else that you think is notable from GTC before we move on?

Rory O’Driscoll

The thing that stayed with me—and I think there's a lot of interesting stuff; I hit the things I thought were interesting—but the message behind the message was that, with open-source models, everyone needs gigawatt centers. They all need massive amounts of data centers with inference, with GPUs, with everything, and that it doesn't really matter.

It doesn't really matter, because the best models are still going to win, which is where he's made his bet. He's announced another open-source alliance. Other things will win, but at the end of the day, his main message is, “We're going to win. We produce the best outputs. We produce the best inference. We produce the best everything.”

Everyone needs data centers, and there's a level of confidence that most rolls of the dice lead to Nvidia winning. You can take shots at its armor, at its moats, but I think there's a high degree of confidence that most of the roads lead back to Nvidia reaching $10 trillion in cumulative bookings over the next 5 to 7 years.

2. Meta's 20% Layoffs & Atlassian Lets Go of 1,600

Speaking of playing for this game, Jason, you said you can feel energy in rooms.

There are also tough times for certain companies where you feel other forms of energy. Two big announcements this week were layoffs at Atlassian, announcing 1,600 people, and then Meta, reportedly, a speculative 20% workforce reduction, which would be 16,000 of 79,000. We've spoken about layoffs before. Is this really just the start of the dominoes falling? Is this a sign of overhiring from 2021 and beyond, and are we just relabeling it AI? How do we think about these fairly large-scale layoffs from the biggest players?

Jason Lemkin

Well, look, I don't know that much has changed from our prior conversations, other than that everything we've said has come true. This is every conversation at scale. Here's the thing: Atlassian and Meta are both interesting, I think, because I guess they're all the same. It's not really about layoffs. Neither of these companies has to lay off anybody, okay?

Atlassian has substantial free cash flow. Meta, I don't know, their free cash flow has dipped, but it's still in the mid-40s, right? They don't have to do this. This is not a unicorn trying to figure out what to do with the last $20 million that SaaStr, Scale, and 20VC gave it, okay? This is a decision. This is a purposeful decision.

What's happening behind the scenes in board meetings and in management teams is everyone's looking at the teams they have and saying, “I just don't know what to do with half of these. I don't need half of these people. I don't need them. I do need people. I need people. I need different people.” There was a great LinkedIn post this morning from the ex-VP of engineering of Ping Identity who said, “This is going to be a sad post. I don't know if anyone's going to see it, but what I learned is over.”

“The craft, the art of creating code, creating modules, testing it, being creative, figuring out how to do something that hasn't been done before is why I got into engineering and why my teams joined me, right? That era is over. Now the AI writes the code, and all an engineer does is review the code. But now the LLMs are doing the code review. We won't need me anymore.”

Everyone's looking at their team and wondering, “Why do I need that engineer? Do I need half my sales and go-to-market team? Do I need to brute-force sales and marketing the way it was when the 3 of us met?” You brute-force sales and marketing in the enterprise. Everyone's looking, and they're saying, “I don't need to do these things.” These are leading to radically different ways of thinking about the future.

Some folks are going to be very slow on this, but everyone's talking about it. Everyone's talking about it. Whether it's 10%, 40%, or 15%, it's just an output of these conversations about what people I need in the new world. Most folks probably have teams in which half the people are not the folks they need going forward. Do you want to be kind about it like Mike? Do you want to be brutal about it like [laughter] Zuck? Does it matter? I mean, it matters to the humans, but at some point we're going to end up in the same place over the next 36 months. The pace of change is so fast.

But I do think the Block thing, you could argue, is different. They're growing 2%, like we talked about. Block is different, right? But these conversations are in every boardroom. Even if your margins are in the 40s, in every boardroom, we have the wrong people. Everyone's stressed AF right now, including OpenAI, including Anthropic.

They need to reboot their teams for the future, and they can't stick with people in the past. You just can't afford to. You're going to be obsolete in 18 months. Layoffs are just one way to re-engineer your company. As brutal as they are, they're just a small piece of re-engineering your company.

We're running out of time. Everyone knows they're running out of time. Unless you're Legora, Sierra, or Lovable, but even they know they're running out of time. In 12 months, their current products will be obsolete.

Rory O’Driscoll

That was helpful because I've actually been thinking a lot about this layoff thing. I came in with 4 different categories of what's really going on, and Jason's actually added a fifth that I'll come to last. I think if you try and be logical and use categories, it just yields more insight.

There's a whole category of layoffs that are really, “We never should have hired these people. We got fat. We're using AI as an excuse, but if you run the efficiency metrics, we just don't need these people,” right? I think there's some of that in there. Then the second category is, “We used to have a business growing at 20%. We're now growing at 2%. If we were going at 20%, we'd need all these people, but we're not. We're going at 2%. I need to give the financial market what it wants, and it wants profitability,” right?

Again, you're not making any comment on the labor. You're just conforming to capitalism, right? I think there's a lot of that going on in the SaaS world, right? I think that first category—well, Block is clearly an example of the first category and maybe Meta, but I'll come to Meta in a second. I think a lot of the SaaS world is that second category of, “What I have to give Wall Street is simple: if you give them growth, they leave you alone.”

If you don't give them growth, you better give them profitability. If you don't give them either, they're going to bust your chops. Them be the rules. If you can't give them growth, you've got to give them the second. That's the second category of layoffs.

Somewhat interestingly, the third category is starting to get into what Jason talks about. The third category is maybe you did need these people pre-AI, but now there are AI efficiencies that allow you to do the same thing with less. That's probably true in coding. I'm not sure it's true at the same scale across the board.

The fourth one isn't the one that Jason gave me, but it is the Meta example. That's something different. You spent all your money on computers. You need operating cash flow because you've got that depreciation hit coming, right? I mean, that's reallocating dollars from humans to computers, right? To compute, right?

That's not what's going on at Block, because they're not investing massively in CapEx, but that is 100% what's going on at Meta because, Jason, you're actually wrong in that the operating margins are still 40%. The free cash flow, when you honestly account for the CapEx, is almost zero, right? That depreciation's going to start hitting, and they're going to be firing people because they need to give it to Jensen. Today, compute eats jobs, and that's what you're seeing at Facebook. You literally can't afford to have Nvidia and people.

And then the last one that I didn't have—the one Jason added to me—which I think actually could be more of it than I realized is this idea that, in some cases, you actually are going to hire back. Maybe not as many people, maybe at twice the salary, but just people. Maybe there's a little bit of deck cleaning going on, in that maybe I don't need 20 engineers who all know C++, or maybe I need 8 engineers who are just really awesome at AI.

My aha is, if that's not going on, it probably should be in every company. Even if you don't have any of the first 4, even if you didn't overhire, even if business is still going strongly, even if you don't need to feed Wall Street, even if you're not spending it all on compute, you probably are doing, to Jason's point, a pretty significant talent reshuffle in real time.

I just wanted to ask Jason if the people that we want are fundamentally different. The developers that we used to hire, we don't, because AI writes the code for us. The marketers we don't want, the salespeople we don't want—who do we want, genuinely? What is the attractive profile? Your Anthropic and European teams are hiring, so what are the people that we want in the companies of the future?

Jason Lemkin

Look, I know it sounds trite, but the answer is simple. It's just the expression each year changes. We want folks who are genuinely AI-fluent. It's pretty simple.

Maybe last year we called them prompt engineers, right? That used to be a job. I don't know if you remember. That actually used to be the hottest job on planet Earth. Now no one needs a prompt engineer because it's pretty easy to prompt all these tools. That job died, okay?

Now we need go-to-market engineers. I think that job's going to die. Everyone needs so many forward-deployed engineers; you can't hire enough forward-deployed engineers. Palantir just announced at their big event that they've gotten their deployment times down by over 90% with forward-deployed engineers.

This wave of disruption for the titles and the specificity is also exhaustingly accelerating. But it's really simple. You meet anyone for any role—sales, marketing, engineering, product, QA—and they're either unable to keep all the ways they use AI to accelerate their job from spewing out of their mouth, or they're staring at you. There's nowhere in the middle.

The person who comes in and says—it's Captain Obvious, but you just had the marketing head from Lovable, who was super popular on the show, right? She's just spewing AI-native insights into Lovable, right? It's not that complicated. You hire her, Elena or whatever it is. It doesn't matter whether she's still in college, or a junior, or a senior, or a middler, or a lefter, or a righter.

Honestly, if you interview people, I would say that even among the best startups I've invested in, maybe 30% of the management teams meet the standard, at best.

30%? Maybe less. And of the interviews I do in general, it’s single-digit percentages. In that sense, it’s the same as ever: you either lower the bar in hiring, or you hire someone who’s actually great. Someone who’s actually great is so far ahead of you in how to employ the efficiencies of AI in their role that your jaw falls on the table.

The difference is that we used to need warm bodies. That’s what’s changing. We used to need warm bodies to answer the call, to do QA, to do code review, to get the blue pixel from the upper left to the lower right. You laugh, but you literally needed to brute-force this with humans.

With AI, every day that goes by, you don’t need to brute-force things with human beings on your team, and that’s another reason teams are shrinking. Why are all these new companies so efficient? They’re just not brute-forcing things with humans. They’re choosing not to.

To all the brute-forcers out there: everyone talks about how bloated teams got in 2021. I don’t agree with that. I think they got as big as they needed to be when growth was high and you needed humans to do everything. Look at these teams that doubled. If growth continued at 60%, like the rate in early 2021, for 5 years—and you can help me do the math—and every single thing a software company did required a human, you were understaffed by your 2021 head count.

You’d be sitting here in 2026. Every office in SoMa would be triple-packed, and there wouldn’t be enough humans to staff your company. It’s just that the world changed.

3. How to Test AI Fluency in Employees

Yeah. Jason, you live on the bleeding edge. I think Marie and I see that, and I think the world sees that when they hear you every week in terms of how you run SaaS, too. For all of the CEOs and executives who listen to the show, what would you advise them in terms of determining whether someone is AI-fluent when they meet them for jobs, for talent?

Jason Lemkin

Here’s my—I realized I was just asked this. I just did a review with a super-fast startup, growing just past $100 million, and I was asked this question. One of my favorite executives gave me an answer that I thought was pretty dated, because it was about 6 months old.

The answer 6 months ago was, “I look for folks on my team. I look at what tools they play with.” That was a great answer in the summer of 2025. I tried Lovable last week. I bombed. The answer in 2026 is: What commercial AI tool have you brought into your organization this month? That’s the test.

For anyone who’s on the bleeding edge that you would want to hire, there are so many great products in the market. There’s no excuse, in any role, to not have brought one tool a month into your organization. There are going to be better and better tools and better and better products as the year goes on. What’s the one you did?

You’ll see people with deer in the headlights when you ask this question. What sales tool? What marketing tool? What product tool? What engineering tool? What did you bring in? Why did you pick it? How does it work? Because if you’re remotely at the cutting edge, you’re all over this. You’re looking for the next agentic tools that will radically improve how you do your job.

Everyone thinks SaaS is at the bleeding edge, right? You know what we do? We’re just looking for the tools and trying them. We’re 1 year ahead of everybody else because we did the simplest thing in the world: we tried the tools early and we trained them. We trained them for a month.

Do you want to hear a horrible example from this week? A super-hot AI company valued at $6 billion—I’m not going to name it—told us yesterday that we had to quadruple what we spent on its product. Its agent told us. Why did this happen? At this $6 billion company, no one had trained the agent on its pricing properly. No one had tested it.

They said, “Well, we’ve been in beta.” We said, “Well, when did the beta launch?” They said, “A year ago.” These are people asleep at the wheel. You want somebody who, the instant this comes up, knows exactly what the issue is: “When I was using Lovable or Replit, we trained the agent. This is how we did it. I brought in this tool that Rory had invested in last week. It solved all these issues.” That’s what you want to hear.

If they haven’t brought in a tool in the last 30 days—or at least deeply evaluated one—I don’t really care whether they bought it, but they should have gone so far down the funnel that they can tell you: pick whatever tool—Fyxer, Regie, GCA, AIGC—I don’t care. You went through it, you looked at it, and you can tell me the 8 ways it would improve the productivity of your business and 3 ways it wouldn’t.

Just don’t hire that person, because you’re going to run your company into the ground. This is the job today. The job today is not to screw around on ChatGPT or to be a prompt engineer. The job today is to bring the best AI and agentic products into your organization and leverage all the hard work that the engineers have done building those products.

That’s your job. You don’t have to screw around. You don’t have to be a prompt engineer anymore. You have to be an agent-deployment expert, ADE. This is the new job we’re making up today: an agentic deployment expert. That’s your job from the C-level to junior. Agentic deployment expert. Don’t hire anybody else. You’re going to regret it.

They’re going to stare at the camera. He’s good, Rory. Stare at the camera—he’s on a roll. We could probably just—I could slip away, get a coffee, and come back.

No, and I sound exasperated, Rory, but the reason I am is I can just see it. I can see my best companies doing it, and I can see some companies I’ve invested in not doing it, and I want to cry. I just want to cry when they have no ADEs on their team. You’re flushing years of your life down the toilet by not approaching how you’re building this company this way.

Rory O’Driscoll

Yes, and at the risk of being positive, it’s worth pointing out 2 things he didn’t say—well, something implicit in what I said. Jason didn’t say, “Only hire people below X who get the new thing.” He didn’t commit the employment-law—I think it’s a civil penalty—of saying that, because he implicitly said anyone can do it, provided you’re willing to learn. I think that’s the big aha. That’s one of the positive statements to make here, right?

I’m always wary of coming across as, “Hey, this is the thing that you all have to do.” I think it applies to everyone, including investors. I have found that unless you’re willing to invest the time learning these tools, you actually shouldn’t be investing in them. One of my partners, Andy, had this expression: if you decide you want to stop learning new things, you probably should retire within 6 to 12 months and never earn another check again.

Jason Lemkin

For sure, but can I just add one point? It’s so important, if it helps people. And thank you, Harry. We’re going through these phases.

When AI started to blow up for real for us—call it early 2024, maybe late 2023—I wasn’t equipped. It was too technical. I wasn’t going to go in and figure out—I wasn’t smart enough to figure out how to deal with a massively hallucinating LLM API and turn that into something magical. Kudos to investors and others who got it in early 2023 or 2022.

I remember—I guess it was maybe SaaStr Annual 2023. I was with David Sacks, and I did a Q&A. I said, “How are you thinking about AI?” At Craft, he was like, “Well, we’re all in. We want 80% of investments in 2023 to be AI.” I’m like, “Great, but show me the great ones in the market.” He’s like, “They’re all prototypes. They’re all proof of concepts, but we’re all in anyway.”

That’s where you kind of had to be in 2023 if you weren’t investing at the LLM level. I wasn’t smart enough. Then we went through this weird-ass prompt-engineer era, where you could torture these products to do something good, right? But you had to torture them. You had to craft these crazy things that made no sense.

Now we’re in the era where merely ordinary smart generalists can make these tools do magical things. I literally go to these meetings and people will be like, “I don’t know how. This is so scary. I don’t know how to do this.” We show them our back ends.

Do you know how to do a workflow generator? Do you know how to do a decision tree? We’ve been building these since software in the 1990s. I can show you all of our agents. How they work is novel. They do have to be trained. You can’t be lax and have these agents work.

But honestly, the UI, the UX, and the way we interact with them—it’s just software. My point is: pick yourself off the ground. This is your time now. If you felt lost in the AI era, if you felt like you were behind, if you don’t understand what all these people are saying on X and Twitter about Claude and GPT-4.1, Mini, Nano, and o3, and it’s all like, “This is not your world,” this is your time.

This is your time for the generalist who knows how to use software tools really, really well.

And this is my last point, but it’s so important. If, at any point in your recent life—and this is why you could be all you need to be is young at heart, to Rory’s point—in the last 3 to 5 years, you have successfully deployed a piece of enterprise software of any sort, you yourself, not some agency you hired, but if you have deployed it, you can deploy any agentic tool. Any. And you can become the hero in your company, and you can become the hero in your functional area.

But I watch folks. I’m literally helping a company now. They’re adding hundreds of salespeople this year with the new pre-AI CRO. He hasn’t brought in a single tool. He’s scared of it. It’s not that hard. Did you use Salesloft? Did you use Outreach? Did you use HubSpot? You know these tools? If you can deploy these tools, you can deploy a world-changing AI agent.

And so, this is the time for the folks who were shut out of the AI revolution. Right now, the generalist folks that know how to deploy software, that don’t even know how to build software. Vibe coding, for me, was for folks who knew how to build software, but you didn’t have to be an engineer. Now, you just need to know how to deploy software to win with AI agents. That’s all you need to know.

So many people have these skills, and they’re petrified of AI. How did you do that? How did you deploy an AI BDR? Well, we bought a piece of software, we figured out how it worked for a day, we set it up in an afternoon, and then we did spend 30 minutes training it, which you didn’t do with this old software. Because in the old days, we just had to manually upload all the data, right? There was no training. The only non-intuitive part is training these things, and it’s just work.

So, that’s why when I see folks on the management team not doing this, there’s no excuse. You do not need to be technical to win with AI agents in Q2 of ’26. You do not need to be even 1% technical. Not at all. So, it’s your time, or you’re going to get laid off. Or you’re going to get laid off because you’re not mastering that.

Jason, thank you. That was an impassioned rant that I learned a lot from. And I love ADE. That’s fantastic. I think you should coin it. I would say, “Write a book,” but I don’t think writing a book is ever useful these days, given the speed of news on the internet.

Jason Lemkin

Given no one reads them.

4. Anduril Lands $20BN Army Contract

Yeah, yeah, 100%. The amount of VCs that write books, I’m like, “But seriously, what are you doing? Wasting a year doing this?” Well, you get your friends to go on the book tour with you. There’s something to be said for that. Great. Yeah, totally agree. Guys, come on. We’re actually meant to add value here. Keep going. Okay.

Anduril lands a $20 billion Army contract. The reason I said this is when you said about just mastering—I remember reading this and being like, “$20 billion contract, 10-year deal, 5-year base and a 5-year option to consolidate 120-plus separate contract actions into 1 enterprise contract.” It’s enormous. When I read this, I was like, “God, the idea just doesn’t matter.” This sweet little AI company that’s still going from $1 million to $4 million ARR. How did you guys think about and analyze this $20 billion annual contract?

Jason Lemkin

First of all, yes, it’s obviously a vast contract. But as a reminder, they have 4 or 5—they only have 4 or 5 customers. So, you better get $20 billion from each of them if you want to be a big company, right?

I mean, what it really told me is they succeeded, right? Because this isn’t so much a new program. This is basically the Army saying, “Look, we got 120 separate contracts with you. We get it. You’re now effectively a prime supplier. Why don’t we consolidate all the paperwork so that people 1 level down have less process to go through every time to buy our stuff, right?”

So, it’s more of a procurement thing. And also, there’s a systems lock-in here. The primary product I think they’re talking about is Lattice, which is a software connectivity system. You’ve got all these different physical hardware products out there, some made by Anduril, some made by other people, right?

And as it’s becoming clear in recent conflicts, a huge part of the problem is making all these systems talk to each other, dare I say it, autonomously and quickly, right? And connecting all these things in not just near-real-time, but real time. Because, as we’re learning right now, it turns out if you’re in the Strait of Hormuz, you’ve got literally seconds before you can take down an incoming drone. You don’t have time for a slow connectivity protocol. You definitely don’t have time for a human.

So, you need this integrated communication system that connects all your different physical hardware, offensive and defensive weapons. And I think it looks like the product that these guys at Anduril have is becoming at least the primary default for that—for effectively moving information between different systems.

So, it makes sense for what the Pentagon’s doing. They’re basically saying, “Look, we’ve gone from trying you out in a lot of different areas to saying, ‘Okay, damn it, you’re the dominant provider of this layer, so why don’t we just systematize the contract?’” It picks them as the clear new prime.

I’ll tell you the answer to Harry’s question, and this may be wrong. This may be a flaw in me. I completely concede this, but part of Harry’s question was, “Hey, am I investing in things that don’t matter when Anduril has a $20 billion contract?” My version of it—I feel that, and my version of this is I have given up on an investment thesis I had for 10 years because I was a B2B founder, which is that a smallish TAM is okay with a great founder. Start small, but you know what? We can all point to small things.

For me, when I started EchoSign, the TAM was $2 million. It was $2 million.

Obviously, if you just look at Rory’s investment, DocuSign, it is doing more than $2 million today. I haven’t checked the latest quarter, but even with some challenges, the TAM certainly grew, right?

Jason Lemkin

So, as soon as I realized that, I’m like, I’m investing in areas that are going through phase transitions with great founders, and they will grow the TAM, right? And for sure, we can show—if I hate to do trite things—that certainly, the legal tech space is one that has shown an explosion in TAM, right? Because of AI. So, there are many examples.

But in my heart and soul, I can’t do any of those investments anymore. I can’t invest in anything that is midsize or smaller. I just can’t. This is the Anduril problem, and this is also why I think a lot of funds are going to have terrible returns.

A lot of early-stage funds are going to swing so hard for the fences that they’re going to invest in the number 3 or the number 4 and get just zeros after zeros, because there isn’t a chance to stair-step your investment. There isn’t a chance to go from the $50 million TAM to the $150 million to the $500 million.

So, I think there’s going to be a lot of zeros, but I can’t help myself. Literally, I can’t even—I don’t like that I can’t even bring myself to take a meeting with a startup where I don’t believe the TAM will be utterly massive. I just can’t take the meeting anymore.

Jason, there’s a lot in there. I want to unpack it and try to be more precise, right? I think you’re saying 2 things, just to be clear. When you see what a big TAM feels like—and let’s agree that being the comms system for the Army probably is one of those big-ass TAMs, right?—you’re saying you just can’t get excited by super-small TAMs. That’s 1 statement.

And the 2nd statement you made is, if everyone’s thinking like that, they’re all going to swing more aggressively at the big TAMs, but invest in the 3rd or 4th player in those TAMs and probably lose. Is that what you’re saying? I didn’t understand the venture.

Jason Lemkin

And even worse, they’re going to pay up $100 million pre for these seed investments, because, as the best accelerators tell us, it doesn’t matter who has these big outcomes. It doesn’t matter whether you pay $60 million post or $100 million post at a top accelerator, because when it’s a $100 billion outcome, that’s a better return than a unicorn, right? Mathematically, it’s true, but I think it’s going to lead to a lot of zeros.

But Jason, do the best markets not start small? Do you agree with that, when you say, “To a point, do the best markets not start small?”

Jason Lemkin

I don’t believe it anymore. Rory, correct me if I’m wrong, I still think defense is the largest segment of our spend in the country. Maybe healthcare is 1st and defense is 2nd, right?

Rory O’Driscoll

No, defense isn’t—shockingly, defense isn’t that large. We just spend a little over 3% of GDP on it. And it turns out most of that—not most of that, a good slug of that, almost half—is people.

I’ll make a comment: this is not to knock that. I think Anduril is going to do amazingly well. The Pentagon budget for new entrants is fairly finite, and I think there will be 5 or 6 big winners in defense. I’m not sure there will be 100. But isn’t that the game of venture? Is it 4 or 5, not 100?

Jason Lemkin

What? But that’s the game of venture: to find the 4 or 5, not the 100. That’s my point. The thing about power laws is they get in your head—it’s getting in your head—and you can overproject from “nothing matters except a $100 billion outcome.”

Well, if you want 10% of a $1 billion outcome and you are a $100 million fund, that’s a 1X fund. So, I hear you. I understand what you’re saying. The bigger your fund size, the more you have to be a power-law junkie.

At some level, you want to be a power-law junkie because, in the end, even if you have a $100 million fund, wouldn’t you prefer to be in the trillion-dollar outcome than the billion-dollar outcome, right? The question is, when there’s that focus on the power law, could it become overly myopic and lead you to swing—you know where gamblers can go on tilt, where they’re so desperate to earn their money back that they start swinging at anything, right?

And you know, if you’re like, “Oh my God, I just got to get—I wish I’d done OpenAI or Anthropic. Therefore, I’m going to fund 8 next-gen foundation models because maybe 1 of them will be like that,” right? That’s another failure mode that, as you said, could be—

We might be seeing that happen right now, right?

Jason Lemkin

Yeah, that’s my point.

So, I'm not sure I fully hear you on the small TAM. No one wants to be in a small TAM, but almost as important as TAM size, there are 2 other things: TAM velocity, for lack of a better word, and your ability to dominate that TAM. There are mid-small and mid-tier markets that are wildly profitable software markets for the winner, and you can make really good coin in a $4–$5 billion market with a great outcome.

Jason Lemkin

No, that's the thing: now that Y Combinator is productized to a $60 million post-money valuation, right? Nothing wrong with it. I'm not criticizing it. More power to them, right? But as a seed investor—I mean, Harry made this point—all these classic seed investments can't make money, right? Just agree. You can't go into a game paying power-law prices for mid-tier market outcomes, for mid-tier markets. Because you're exactly right.

That's the game today.

Jason Lemkin

That's a fair comment. As a combined comment, that's why I was trying to unpick what you're saying. As a combined comment, that's fair. If you price every deal like, quote-unquote, it might have a $2 billion or a $20 billion contract, most of them won't.

Harry wouldn't do this, but if a classic founder came to me with a structure that made sense and I believed in them, I might still take the bet. But it's a $60 million post for a pre-seed investment, right? You can't risk that it's not Anduril or better. You can take the risk that it doesn't happen, but you have to believe the opportunity is so large. How else are you going to get your 100× with dilution, right? That's 250× in today's world, post-dilution and everything. What's 250 × $60 million post? Like, lots—$13 or $14 billion.

Yeah, so let's round up to $20 billion. How many public tech companies have market caps north of $20 billion? Not as many—fewer than when we started this podcast. The math is grim. I have that count somewhere: sub-50, yes. No, I mean, I think that if you look—Jason, if you take this mindset, though, genuinely, when a deal comes through the door, what is big enough?

Because you've talked about your Qualifieds, your Artisans, and your Monacos. Are they big enough? SDR AIs, you know, I'm seeing constant call-center replacement VAs, healthcare assistants for auto manufacturers. What is big enough?

Jason Lemkin

Well, listen, there are arguments that they are, right? But what I think is why the growth fund is the winning strategy at the moment: they wait for the proof. I'm getting bored of talking about the same companies, but to have a thread through our conversation, I certainly wouldn't have believed that Decagon and Sierra would be doing what they're doing. Now, granted, the revenue multiples are very high, right? I'm not even sure I would have believed Rory that Intercom would have reaccelerated as it has, or others. I wouldn't have believed it, but the beauty of doing the deal when Rory did it is he gets the proof points.

But hoping that a slide and a vibe-coded website proves it is tough. But, Harry, going to your point, here's the counterargument where I'm wrong, right? I mean, I barely know Decagon, but they just talked about it this week, and Owen talked about it last week. All these spaces are converging. So Owen said there's not going to be any difference between support and sales in a lot of what we do. All these agents are converting to a meta-agent that does more, replaces a lot of humans, and is worth a lot, okay?

If that bet comes true—and it's already happening—then your TAM explodes, right? Your TAM explodes, right? The question is, will you invest in a space where there's no evidence that the TAM is exploding due to AI? That's where it's tough, right? Will you invest in it? It's kind of the question that you're asking, Harry. I don't understand what you're actually articulating.

I mean, yes, it turns out seed investing is harder than A and B on a deal-by-deal basis, which is harder on a deal-by-deal basis than C, D, and growth investing. I'm just saying you can't take the smaller TAM—the stair-step risk. The classic VC seed investment stair-steps the TAM risk: take something that starts small, that has a nuclear core that's strong, and then build—add TAM layers over time, right?

But, Jason, if I go to Replit or Lovable at the seed, vibe coding at the time was a very new and nascent category. I don't think you could say the TAM was particularly massive.

Jason Lemkin

It was a small market. Well, look, first of all, I'm not sure this is necessarily a good thing; I'm criticizing myself. I have changed my perspective. I used to stair-step everything, right? And even the investments I made that weren't stair-stepped, I would invest in something that had terrible comps because I believed they were being remade by the space and it would be much better. I just don't feel that way anymore, right?

I actually think the Replit round blows my mind when I think Craft and some other folks did it at a $1 billion valuation in Replit in, like, '21, pre-AI. They did Replit at $10 million, with $8 million in revenue, pre-AI. I would not have been that omniscient, okay? It's not that I wouldn't have bet on Amjad—I mean, he's a force of nature—but pre-AI, this weird web IDE that doesn't do much and doesn't even finish any software? I ain't that visionary.

But I think if you ask Paul Graham or David Sacks, they would say, "Harry, listen, revolutionizing how we do software development on the web is massive. I know you don't see it today, but I actually think you could argue the TAM is very large, right?" If we really believe Amjad is going to take this kid—the wonder kid out of Facebook—who brought the co-founder, the guy who created React, with him to co-found the company, these 2 guys might just change how we build software. That's a big TAM, right?

Now, where I struggle is when I meet founders doing little niche things in vibe coding. That's where I'm struggling today. "I've got a little thing that does a hint of security on these platforms," or makes the outputs prettier, okay? Design is terrible in Claude Code. It's unacceptable in OpenAI. There's a lot of folks trying to tease at design, right? But is it big enough, and will the models just take it over? You better show me something so hyper-disruptive that my jaw falls on the ground, or I just don't believe, right?

Replit I might have believed, but this little nit fixing the fact that the icons all look like cloud artifacts, I just don't believe. But you could say it's bigger than Figma.

But what am I meant to do with all this? I'm just trying to understand how I do my job differently tomorrow because of this information. You raise a growth fund, Jason, like that?

Jason Lemkin

Yeah. You raise a growth fund like all of our friends. You raise a $10 billion growth fund and just wait until you have extreme product-market fit, right?

But again, I think you should assume that—look, the great thing about American capitalism is money fills any void. And it's the Barton Biggs quote I've made very often: "There's no investment opportunity so good that excess capital won't destroy it." There's a period of time—look, I think you'll look back and go, the growth investments in '17 and '18 were awesome because they sold in '21. The growth investments in '21 were priced in '21, and then the revenue slowed down and the exit market died.

The growth investments from '23 on that were LLM-centric were amazing and, from the growth investor perspective, needed more capital than was easily available. There was a period of 2 or 3 years where the capital needs of Anthropic and OpenAI were unprecedented, as were those of some of the other companies. And when capital needs are unprecedented, the people selling the capital can actually do better at the margin. Take that and have some insights as well about the future, and you can make money.

Now the wall of money's come back in, and I'm sure growth will go through the same thing again, which is everyone will pay 100 times one-year revenues for late-stage stuff they think is going quickly. Just as in '21, people thought they were going quickly, and fast-forward, some will work and make people look really smart, and some won't, right? Whenever you're investing, you have to have some marginal insight—more than the other guy—about why this thing, at whatever stage it's at, can outperform and be bigger over the next X years.

Once it becomes consensus and the capital arrives, then it's just very hard to have excess returns. And look, one of those ways of being non-consensus is finding these small markets that can expand, because I don't agree that—one of the things we often think about, stepping back, is that because the stage we invest in is early product-market fit, our highest-level rule of thumb has been: the bigger, the better. We want big-picture trends and near-term traction. We don't want to compromise between the 2.

[Speaker?]

We want something that's working right now, which could be something small that's working right now. You want something that's actually a thing right now with early product-market fit. But you're right, Jason. Where you are correct is that you don't want something that's a cute little thing now but could converge and there's just no white space.

You're taking that and saying, using the Anduril example, that if you had made a post-product-market-fit investment there in 2018—which, unfortunately, we didn't—it would have been their near-term product. Everyone, and half the country can get mad at this, but what right now was watchtowers for the US border, right? That was the near-term traction. They had really good traction on that, and that's the near-term product.

If you apply the lens of, "How big is the market for watchtowers on the US border?" the correct answer is bigger than you think, but it goes up and down every 4 years depending on random exogenous events, right? The US Department of Defense has purchased things on cost-plus for 40 years and is wildly inefficient. Anduril is going to let them purchase things on a Silicon Valley model: We build a product up front, and then we sell it to you on a per-unit basis, right? That's the big-picture trend.

From the angle, you have to have both some near-term traction—something to hang your hat on—but then, to avoid the risk you're talking about, you have to be able to articulate an expansion story. Now, the trick becomes that anyone can use words. I could take your Replit example. You're right: If you squint one way, in 2021, it's a tiny little tool for a niche case.

If you squint another way and use big, highfalutin words, it's the future of software and the democratization of software. It turns out the trick in investing is to figure out which of those words is true and which is not.

5. Travis Kalanick Returns With Atoms

I do want to talk about Travis Kalanick, who came back in force with Atoms. To be clear, it was 8 years in stealth, 1,000 employees, and a rebrand of City Storage Systems—better known as CloudKitchens—into this new company, Atoms, building gainfully employed robots for food, mining, and transport.

He came out with a pretty—I don't know if you'd say scathing, but an opinionated—piece, "I Bled, but I Did Not Perish." He wants to be more aggressive than Waymo. How did we read this, guys?

Jason Lemkin

I thought, on the merits of what he said about robotics, that he was correct, and I was pleased because it's something we've believed. Again, to remind the viewers, Travis Kalanick, the wildly successful founding CEO of Uber, was famously terminated by the board and is in a feud with Bill Gurley, which can cycle back to the Anthropic-Pentagon discussion because Travis's number 2 is now driving at the Pentagon and also showing an ability to maintain a grudge, which is just one of those things you have to admire in people.

But anyway, he went away, founded CloudKitchens, surfaced last week, and basically said it's not about CloudKitchens anymore; it's about robots. Atoms is building robots for a variety of industrial use cases. And he has an investment in Pronto, an autonomous-driving company founded by Anthony Levandowski, who was with him at Uber, right?

So, he's basically coming back and saying, "I was doing CloudKitchens. I'm now using that information to build robots, and I'm also thinking about autonomy," which obviously cycles back to Uber. So, that's kind of the background.

I thought he was spot-on on the robotics call. Let's be clear what he said: He didn't fundamentally think humanoids are the answer. He thinks robots on wheels are the incremental next step. The big aha that Travis had was, "It's not clear, if you're building an industrial machine for a lot of use cases, that you add legs."

The humanoid robots that we all see have these legs, they consume a lot of battery life, and they're pretty unstable. Most of the time, in factory work or logistics warehouse work, you don't need legs; you just need wheels, right? They're a lot more efficient, and I think it's a big-picture insight from Travis to say, "This is the direction things are going."

What he's pointing out is that he's effectively making a call against a whole bunch of the humanoid companies—not saying it's not going to happen ever, but saying it might take a lot longer than you think. The path to humanoid robotics might be through specific-purpose, non-humanoid-type machines, maybe expanding over time. So, big-ass call on robotics that, for what it's worth, I agree with.

And it's interesting—just one more comment—because Sunday, which just raised recently, is another one of these robotics companies focused on the home. If you actually look at the form factor, they've gone with wheels, too. It's very cute. It's a really cute robot. It's kind of nice, happy plastic, but you look at the ground and it's actually running on wheels, because they, too, have recognized that spending the money to give feet to many robots is just a waste of money here, right? You don't need them.

6. If Travis Kalanick Ran Uber Today, Would it be $1TRN Company?

I think, actually, on trend, he's correct. The main thing when I saw the TV interview, the main thing I thought was that if you were running Uber today, it would be a trillion-dollar company. Without question—to me, without question—in today's world.

Uber's a $160 billion company with massive free cash flow, and it is epic. I think his hyper-aggressiveness—which, in a different era, led to his downfall, which I'll come back to—but it worked, right? It destroyed Lyft. His view that if you're not at a 7 on the 1-to-10 scale of hyper-aggressiveness, you're not going to win this—this is the era we're in today, right? He was just too early.

There was toxicity to him, and there were elements of treating women and other things that were probably terrible and not okay. But putting that aside—it's hard to put that aside—he was just early for his time. Look at Uber today. Despite that, it's wildly successful, but mostly it's been engineered since then, right? Acquire, get into food delivery, which is huge for it, but a lot of that's through acquisition and managing its existing fleet.

When Travis was CEO, all he wanted to do was get into autonomy. He said from the beginning, "Our business is dead at its terminal state. No one is going to be driving cars around for Uber." And now they're years and years and years behind where they could have been.

When I look at today, trillion-dollar companies are becoming commonplace. I think Travis's Uber would be a trillion-dollar company today because it would be 5 years ahead of where it is today. And that's all I thought.

Do I actually think this CloudKitchens, 8 years out in the wilderness, is going to make it? Probably not. But that guy would be running a trillion-dollar company today if Gurley and his buddies hadn't forced him out. That's what I thought.

[Speaker?]

7. When is it Right to Replace Founders

I disagree. Picking apart the argument, I'll come back to that at the end of this conversation. I would answer that question, but first of all, it's interesting you made a comment here: CloudKitchens, I don't know if I believe in it, but Uber would be a trillion-dollar company. That's what you said, right?

Implicit in that statement is a repudiation of the logic of it, which is that you're saying the great founder is everything, but what you're saying is the great founder can't make this company worth a trillion dollars. So, implicitly, what you're saying is it's a great founder plus a great opportunity.

Rory O’Driscoll

You're not starting from scratch now. Yeah, yeah, but yes, that might be my point: 80% market share in ride-hailing is a pretty good platform to start from if you know everything about the industry—every inch of it—you know it cold.

Dara had to drive Ubers to learn it, and bless his soul, he's great, but he had to drive cabs. Travis didn't have to drive any cabs to learn how Uber worked. He already knew how it worked.

Let's ask a question here. I want to leave aside the personal-behavior stuff, where I frankly don't have visibility and I'm genuinely not going to try and make a call on that.

The hypothesis that Uber, instead of being a $160 billion company, would be a trillion-dollar company implicitly has to be something. The only thing that could bridge that gap would be autonomous driving, correct?

Rory O’Driscoll

Well, no. You would be 5 years ahead of autonomous driving, which is already taking off, and I think you would dominate food delivery more than it does. You already had such a head start ahead of your competition. You wouldn't dribble-dabble into it, and you wouldn't then go buy this or do that. You would just dominate it.

You would use capital, weaponize it, and use incumbency and some dark arts in your mobile app so that the competition would get blocked and all this crap. But you would dominate it. You would just go for 90% market share in food delivery, because it's a better market than autonomy in the short term. Maybe not in the long term, but it's a better market in the short term than ride-hailing.

I agree, for what it's worth. I agree with that. So, 2 separate arguments, but do they have to get public? They have to get them cash-flow positive.

Hard-nosed coming: Another spin on autonomy is that you could say, "Yes, we need to do this thing, but we can't afford to spend at the level they were spending in 2015 on something that we might have 10 years later, which is still only now finally doing a meaningful number of rides sub-economically in San Francisco," right? It's the trend.

It is the future. It was 10 years away. They had to get public, and they had to have a plan to cash flow even to get public, to dominate Lyft. It wasn't an option then, as perhaps it might be today, to stay private longer.

Do you think they had to cut back on their autonomous spend? Maybe not to zero, which I think probably was a mistake, but dramatically focus on getting cash flow positive, giving Wall Street what it wants to get public, and having the acquisition currency to do the food things. Growth at all costs had reached its limits there, and maybe a different manager was the right person for the next stage of that journey.

Rory O’Driscoll

Listen, I think in this universe, outside management, just like in many companies, performs as well or better through early 2022. Remember, it was a decade where no products changed. Even Uber didn't change much, right? Uber seemed feature-complete about 2 years after it launched. Great, they added UberX. Now, for 5 bucks, I can get to work and nothing changed.

This app was frozen in time for a decade, right? That's why outside management can run it. I believe what it would probably look like is that, as late as 2 years ago, both might have led to similar outcomes, or maybe even a better outcome with outside management, right? But today, it'd be a trillion-dollar company.

My point is, now would be his second time, because he wouldn't have quit. He wouldn't have stopped building. One of his best friends is Elon Musk. Maybe he would have owned autonomy with Tesla in a way that—you can't know. Who knows? But, good God, it would be a trillion-dollar company today.

I actually think now I would agree with you. I think there was a period of time where they needed to conform to that reality in the market at that time, get cash flow positive, and run it like financial engineering, and he clearly was unwilling to do it.

It turns out, in retrospect, they should have done the Steve Jobs thing. They should have swapped him out, got it public, got it cash flow positive like crazy, and sometime in ’22 should have got him back and said, “Now is the time to do autonomy.”

I do agree with you in the last 2 or 3 years. It's still unproven, by the way, whether or not Uber needs to own the technology to still make autonomy work, but I agree. I like your framing that there was a period of 5 or 6 years where the best manager for a lot of companies was a professional executive with a financial bent.

I can see why they made that change. If there was no more private capital to be raised, if your CEO just was unwilling to focus on convergence at the expense of long-term projects, and you had a risk of going bust, I can see why you made the change. But I agree with you: in the last 2 or 3 years, the financial-management game is out and the product-innovation game is back in. It's accelerating, and it's accelerating.

You said you would answer the question which Jason posed. Would you have made that decision and switched him out?

Rory O’Driscoll

You should be very wary of ever swapping out a founder, right? It's like I tell people: it's like open-heart surgery, and 50% of people die, right? It's a shitty business. Occasionally, I've done it for a bunch of reasons, and it's hell on Earth.

Forget morality. Forget whether I'm a good guy or a bad guy. It's just the most exhausting thing you do. It is easier to just lose money, right? So I hate doing it. There are only 2 reasons why you do it. If the business decisions they're making are literally going to bankrupt the company, right? If you feel that they're investing in a way whereby there simply just isn't going to be any more private capital, and we could run out of money, and they were unwilling to change course, then at some point you have to consider that.

The second option is the thing we said we wouldn't discuss, which is if any terminal behavior arose to the level of a really systemic problem, with a high bar, right? If one or both of those things is present, then, wildly reluctantly, you have to do it, right? And you have to take the heat.

Would you have done it? If one of those 2 things were the case, then reluctantly, you would have. You'd move heaven and earth not to, but if it's the former, at some point you've got to say, “We're going to run out of private capital. You're not doing what it takes to get profitable. We need to focus on profitability. Give Wall Street what it wants.” Then, yeah, you might have to. You'd hate doing it.

Again, I hate this positioning. I'm not the guy—9 times out of 10, I'd be like, “Sell the company. Get a president. Get therapy. Be better.” All the other things. But my logic is this: knowing what it's like in those board meetings, and knowing if you're a founder-friendly firm, like I would say Benchmark would like themselves to be, you don't do that lightly. So you've got to believe that 1 or 2 of those issues was on the table for them to have to do that, and they felt that this was necessary.

I think you obviously see that with Adam and WeWork being the first one there, in terms of just the fiduciary responsibility, in terms of how they spent money and the financial profile. So, totally get you there.

Are we bullish on Atoms now? Do we look at this and think this is exciting, this is going to work? Awesome. Jason, you're the fanboy. What do you think?

Rory O’Driscoll

Look, I know. I get the big bet, and certainly making this bet seems to make a lot more sense than betting on the WeWork founder, right? He just is not as deeply product- and software-focused—building co-working spaces and then having everyone figure out the finance.

Listen, if I were a huge fund and he wanted my money, I would give it to him. Don't get me wrong. But my smell test from watching the interview is that some of the things he said felt like I was back in 2017, when this happened. The world is different now, so I've just got to use my gut.

Then the question is, do I think he's past it? At some point, you do lose it. You lose the ability to create, and you're better off amalgamating, and this is in the middle, right? This is a combination of amalgamation and creation. So you just asked my opinion: would I invest? Based on the interview, no, I wouldn't.

It's so funny, the way you answer that question, because I had the same question. I didn't process it into any kind of internal analysis of Travis Kalanick's soul. I literally found myself thinking, “Do I—I kind of come from the market side—do I buy the market?”

I think there's 2 businesses going on here. One is the Atoms robot business. As I said, I like the approach of more bounded industrial robots versus general-purpose humanoids, so I think he's on the right broad track. But I think all those markets tend to be very different, and trying to do 1 robot for all of them is hard. So I think that's a hard road and will be less amenable to any kind of, quote-unquote, blitzscaling.

The autonomy thing is interesting. I think we are at the stage now where the other company that he's invested in is Pronto—or, I'm sorry, I should remember the name.

Yeah, Pronto. That's super interesting because we're now at the stage where autonomous driving on freeways is hovering on the edge of being a thing with Waymo, but autonomous driving for mining and for industrial equipment is a category now, and there are some players in that space.

So I buy that that market is there and doable. I'd have to do the next level down: Why is our technology different and better? Credible market and focused for autonomy. Believable market, but lots of subsegments in industrial robotics to really play there. So that's my comment.

Invest at 20 billion post? It's probably what he's looking for. That's what Claude thinks he's looking for in the round. He's fundraising. That's why I did the interview. He's fundraising, right?

Rory O’Driscoll

God, no. No.

Yeah, because the last round was at 15 billion. Claude thinks he's looking for a flat-or-up round, so up to 20 billion. The personality? You do that every day of the week? You'll put in your own 20VC money at 20 billion?

Rory O’Driscoll

I'm absolutely not. If I'm General Catalyst raising 10 billion now, or I am Coatue with the multibillion-dollar fund, do I want to chuck a couple of hundred million into Travis, one of the greatest founders of our time? Yes. I have to move hundreds of millions of dollars a month. Rory, I completely agree with you from our fund size.

Yeah, yeah, yeah. This is so—everyone's a victim of their fund. If you're a seed-stage fund, you assess the guy. If you're a growth-stage fund, you assess the opportunity to put quantities of money to work. And if you're in the middle, like us, you're looking at the market, trying to be intelligent and maybe overthinking it.

It's funny. If I have 9 billion dollars in my growth fund, putting 250 million with Travis here, 100% I'll do it.

Okay, good to know. Would you not agree with that rationale, Rory? I always struggle with the “oh, just have a go at it” argument.

Rory O’Driscoll

You've got 4.5 billion dollars a year. That's 400 million dollars a month you've got to move. I understand, but whenever you find it, I'll give you an honest answer. If you find your logic being reduced to, “I've got to get rid of this much money this month and this entrepreneur is amazing,” so I'm not sure about the opportunity and I don't like the price, but have a go.

If that's your logic, let me give you some advice: halve your fund size.

Rory O’Driscoll

And that's why you're not playing the AUM game.

That's right. [laughter]

Jason Lemkin

Now, look, for what it's worth, I'll say 1 thing. I wrote this on Quora years ago, and I had to write so many disclaimers about it because I was going to get hazed. The question was, “Who's the best entrepreneur you've ever met?” This was a few years ago, before I met more people, and I wrote, “Travis Kalanick—the best one I ever met.”

I met him when he was at Red Swoosh, his startup that mostly failed, right? For the record, he made money for his investors. I remember the deal. He made money for his investors. I went by his office in San Mateo. He was down to 2 employees, I think, and we had lunch. This was before YouTube even launched.

I sat down, and I had never met a founder who explained the entire future of video on the internet to me with the clarity and insight that he did. This was the first time I'd met a founder who could see the future and all the elements, how it all came together, and explain it in a way that blew my mind. If I look at my investing mistakes, it's when I've invested below that line—when a founder has not come in and utterly blown my mind about the future of voice or the future of sales in the age of AI.

I took some heat, but that was the first time in my life I met a founder where I walked out of the meeting with my jaw on the ground because he explained the whole future to me. That's why I go to Harry's point: if I was General Catalyst, I'd do the bet, unless I thought he was washed up.

I've got to give it to you, Jason, because you're basically in the category of, “I, Jason, wouldn't invest, but I'd be delighted to let you.” I think General Catalyst should stick $400 million in, because earlier on you were like, “I don't know if it's dated or not.”

Jason Lemkin

I think for me to invest at $20 billion with Travis, I have to believe he can still do it. I have to have 100% conviction that he can still do it to justify it. I think General Catalyst does not have to have 100% conviction, or Harry's math destroys me, or I end up with a 1x fund if I invest with anything less than 100%. I mean conviction at all levels, okay?

Not just believing in it. I have to know Travis is 1,000% in for doing this for 20 years at 51, or whatever he is—48. I have to believe—

We're back again to Kelly betting, inadvertently. What you're basically saying is the percentage of your bankroll that you play—it's edge over odds or whatever it is. I'd have to think about it again. What you're saying is that, for your fund, fund size is strategy, right? Someone said it is: “Fund size is strategy.” And in every dimension—

Jason Lemkin

Yeah, but it's just so true. What you're basically saying is you wouldn't because you don't have the edge, and the return and the odds aren't great. But if you have a different kind of fund, where instead of being 20% of your fund, it's 1% of your fund, maybe you do.

I hate that.

Jason Lemkin

No, no, no. I'm saying if I smelled some risk that Travis isn't in—

I think I know what you're saying. I'm translating it into—

Jason Lemkin

He might be a little bit in the past as well as the future. That's a flag for me at the rate that I have to win.

8. Adobe CEO Exit Shock

Okay, we can choose 1 final topic. We can go for Adobe. They beat earnings, their stock tanked, and as a result, the CEO is leaving. I bet Shantanu steps down in 2026. That's my bet: Shantanu.

I think you've got the interesting comment, and I want to hear Jason's comment because he's worked there. Be precise in your sequencing: they beat earnings at the same time the CEO announced his resignation without a successor, and the stock tanked. It wasn't the stock tanking and then them whacking him. The 2 were announced together, which is odd, and in response to 1 or both of those events, the stock went down. Correct?

Rory O’Driscoll

Correct. Yes, correct.

Listen, over to you, because I think your take is right here.

Jason Lemkin

Well, look, we don't really know exactly why he stepped down without a successor, other than that it's odd, okay? It is odd because a lot of folks on the internet are saying David Wadhwani, who runs Creative Cloud, is going to be a successor. But this is David: he quit Adobe and went to AppDynamics as CEO when he was passed over for CEO the first time. Then he came back to be CEO, and then Shantanu steps down and he isn't made CEO.

So that suggests to me the odds that he becomes CEO are less than 100%, right? Because certainly the most elegant thing is to hand it off to your president, right, of your largest business unit.

[gasps]

Jason Lemkin

Then to say the board's going to do a search while I step down—there's just no way that's confidence-inspiring, right? I don't know what happened. There could be multiple things, but to me it suggests he left the keys on the table after 18 years. Not 5 years, not being fired quietly by the board. I don't think that, for a million reasons. I don't think he was fired. He's very competent, right? Very insightful.

I think he calmly and respectfully left the keys, maybe after a quarter of discussion, and now they're going to recruit somebody. It's not a good sign, and I think we're going to see a lot of these. Dustin Moskovitz quit his own company, Asana, in a huff and left the keys—threw the keys on the ground, right? No successor, no anything. These are not fun times to run most public companies.

There's a handful—I mean, Alex Karp seems to be having the time of his life. He moved to a $50 million mansion in Miami. He's crushing it. Yeah, yeah, because he's winning. But who else is having fun that's public? Not too many are having fun.

The correlation between winning and having fun is pretty high. Pretty damn high, right? You're not having fun if you're the CEO of Adobe, and you're not having fun if you're any CEO. Going back to you, I think you're spot-on. Going back to where we started this conversation, the only way to make my— I have an obligation to my shareholders to make the stock go up, and the only way to make the stock go up, given my growth rate's gone down, is to sack 20% of the people I spent the last 10 years hiring.

You might not know it's the right thing to do. You might even be willing to do it, but it's not going to be the best week in the office. So I agree with you. It's tough. I don't know if 10 quit this year, but I can totally see more than 10 saying, “Should I quit this year?” going home to their spouses and saying, “Now, tell me exactly why you're doing this.”

And to be clear, hang on. Actually, 1 thing: we're assuming something. Is he continuing in the role until you get a new CEO? I just didn't check that. I think he is. Is that— I believe, I think so. In which case, it's not keys on the table. It's more making an announcement before you have the answer. It's a little less herky-jerky, you know.

Jason Lemkin

Right, just to be clear. It's leaving the keys on the table, but still grabbing a drink at the bar in the kitchen.

Yeah, no. It's—I'll do the right thing, but I'm—

Jason Lemkin

—leaving the keys, but I'm hanging out in the house until Dad gets home, or whatever the extended version is.

Yeah, metaphors never work. You should just say the thing instead.

Where is Adobe in 5 years' time? It has a massive installed base. You have a set of core—here's the meta issue.

Jason Lemkin

I've got someone on Twitter—the guy you just had. Who'd you just have? The super-smart guy? What's his name?

The last one. What's his name? Sorry: Sriram Rajaram.

[Speaker?]

There is a lot of truth to that, as we've all seen. I think it's true in all software, but it doesn't mean you grow just because it's hard to churn. Here's the mistake so many people are making right now in the age of AI.

Just because your GRR stays high, just because your nominal churn is low, does not mean you're going to grow. I see no evidence that Adobe will grow. I see no products that show they'll grow. Nothing.

I want to come on here with 3 things, and then we'll wrap. One is, I just checked: he's actually staying on as CEO until the new successor is named. So it's more of an announcement without an answer. In which case, it's changing my opinion here: it may well be that the board also felt they wanted to send a statement.

They felt the results were disappointing. They didn't want to attract a whole bunch of, “Oh, you need to make a change.” Activists could show up. Maybe they thought that, given these results, they actually needed to state now that the change was happening. Maybe they felt they needed to pull it forward, right?

Yeah, in other words, please don't call us, Elliott, and do your Elliott thing. We know we need to make a change. Shantanu wants to make a change. Here it is. That would actually fit the facts pretty well.

Jason Lemkin

You are right. If you've been the number two who's been passed over a second time, because if you need to make a change and you have a good number two, and you feel the activists are circling, I tell you what you do: you hire the guy.

But be that as it may, going to your second comment on Adobe and Intuit, I really want to key in on that. In my less-than-amazing opinion, I think Intuit has pretty good durability over the next 5 years, because I think the thing they automate, which is accounting and tax, has some AI impact, but it's not infinitely large, and I think they can adapt to it. We can argue that another day.

I think the challenge for Adobe is this: what kills you as a software business is if the work that you automated gets done in a totally different way, right? And Adobe—you know, what we've seen is one of the areas where AI has had traction. I saw a really good post on this. I can't remember who wrote it.

They said most of the AI traction so far has been individual users, creators, individuals even within the enterprise. It's not yet an amazingly great workflow tool. Not all of it, but most of it, right? Adobe is the classic creator tool. There's a whole new way to create. By definition, they're playing catch-up.

So, more than most companies, they are under the gun to figure out how to meet their creators in a totally different way. I think the disruption risk on those guys versus Intuit is a lot higher. 5 years from now, there will be some disruption to it, but we'll still be moving money around, we'll still be producing quality accounts, and I tell you, we'll still be paying taxes to the US government.

I don't know if we'll be doing pixel-by-pixel removing in Photoshop. So, I think the Adobe AI risk over the next 5 years is pretty large. And that actually probably should be figuring into their search.

Oof. We didn't cover much, did we? I know, I know. We've had a bit of a materialist show, haven't we? Guys, thank you so much. That was fantastic.