H1B签证对美国初创公司的影响 & NVIDIA向OpenAI投资1000亿美元

Well, I'm excited because, just like 2008 at the moment, I'm 0% cash.

You're 0% cash.

Founder-friendly has become bullshit, right? Any hot AI deal, there is no diligence provided, nor is any done. It's just done on Saturday. All you can lose is one X your money.

Having an early success is highly correlated with future success. Partly, you get the referral effect, but partly I think it's that you just have the stomach to roll the dice and you get braver.

This is an epic monopoly like we've never seen. Think how much ChatGPT already dominates our lives. It's the Standard Oil of tech.

Ready to go, guys. I am so excited for this. We have a lot to get through. I even have graphs this week. I mean, look, you see this. This is intense. I don't know how many GPUs you're buying from NVIDIA. Is that what 20VC's committed?

It's up and to the right. That's all I see from this graph. Now, what else do we start on?

NVIDIA invests $100 billion into OpenAI. I wanted to start on this. Is this an infinite money-printing machine, where NVIDIA invests in OpenAI, which commits $300 billion to Oracle, which then buys more NVIDIA chips? Is this just the way to print money these days?

First of all, it's not an infinite money machine because it will end, right? The way I thought of it on the way in is basically Sam's going to get to make the bet he wants to make, which is to apply an infinite amount of capital and see how long these scaling laws last. No one's going to call timeout until you actually hit a wall, right?

What this says is NVIDIA is going to get rewarded. Everyone's going to book gains, and if it all works, it will all be good. It's like any aggressive financing strategy: if the underlying business works, everyone looks smart, the debt gets paid back, the equity goes up in value, and everyone's a hero. If it doesn't work, it all comes back and bites them in the ass, right?

What this means is, if it turns out that those sets of OpenAI projections—that $100 billion-plus of revenue, or whatever the numbers are—are real, we're going to get to find out. No one's going to call timeout along the way, at least for another year or 2 based on this, because the capital is being made available and everyone involved is getting kind of mentally marked up.

So that's my takeaway. They're going to get to find out here.

I have to interject. You said Sam gets to see if scaling laws do continue. I thought we all agreed that they didn't continue, and that's why GPT-5 was focused on efficiency. That's why we didn't see the improvements in the way that we thought we would. I thought we were already reaching.

I think you've got to listen. The 1 thing I've learned over the course of the show is to really listen to what Sam Altman says. Elon Musk will say something and it happens. He's just a couple of years off, right? But it always happens—the self-driving cars and the rockets, right?

Sam says something and it happens kind of soon. He says it off the cuff, and he's sitting there with Jensen and Brockman this week saying, "This is just a start. We need 3 orders of magnitude more compute than this." We need not 10 times as much, but 3 orders of magnitude more, he says.

Then he writes today, calmly, that with this first slug—the $100 billion—hopefully we'll cure cancer and educate all students on the internet. There's a lot going on there. He said $100 billion, then he said Stargate, and we didn't understand it. Now, he didn't just say $100 billion. We could talk about the round-trip revenue. Is it really round-trip revenue? There are some interesting questions, right?

But he and the president said it was. They need 3 orders of magnitude more than this to achieve what they're predicting today—their goals. This is not phony baloney. This is what they have on a whiteboard and in a spreadsheet.

To be clear, I didn't say it was correct. Harry, I simply said we'll get to find out, right? I'm not sure I believe that the marginal $300 billion will earn a return on capital at all.

My point is simply that there are some aggressive business projects where you wonder, "Can we invest that?" You're just not able to make those investments, so you don't get to find out. Then there are deals where the market says, "Here, you can have the capital. Have a go." This is 1 of those deals. We will find out.

As you know, the cliché goes, that's why they play the game, right? I'm somewhat more skeptical because I think the level of heroic assumptions you have to start making to make these investments all work is high. But the truth is, we're human beings, and we tend to like it when someone is as astonishingly right as OpenAI has been over the last 6 years. It's just human nature to say, "I'm going to continue backing this bet as long as it works."

What that means, by the way, is that it goes on until it stops. In other words, you've got someone who's clearly going to keep doubling down, so the doubling down is going to keep taking place until the return on the double-down isn't there. Has that already happened? Is that going to happen 3 years from now? I don't know. I can speculate why I think no, but it doesn't matter. What I think is that the market has said, "Have a go. Here's $100 billion."

If you are Dario today, are you thinking, "Wow, we are in a significantly disadvantaged position as a result of this"? Does this $100 billion move the needle significantly more in favor of OpenAI?

The question is, what is it giving you? It's giving you capital and access to chips.

So do you feel constrained by those? You probably aren't constrained by a lack of capital. My understanding is they were beating people off with a stick on the recent Anthropic round, and if they decided they wanted another $10 billion, it would be there tomorrow morning. So they're not capital-constrained.

Maybe you'd say it gives you some preferential access to GPUs that OpenAI now has, but I don't know, right? In the "Oh my God, their capital is bigger than my capital" wars, I think you'll feel the need to respond and do something. But it would be interesting to actually pencil out what exactly, other than momentum and bigness, is the problem you're trying to solve.

Well, Anthropic isn't going to be able to build its own GPU like ChatGPT is. It's not going to be able to lease or create hundreds of gigawatts of compute without that capital. It's not going to be able to build its own GPU.

Another thing I thought through with all this news, which didn't really come up in any analysis I saw, is that Sam is very clever at toeing the line on being a monopolist. OpenAI does not want to be a monopolist, right? Nor does NVIDIA. They both want to be very careful in how they position themselves.

I think NVIDIA has an existential risk, which is that everyone is trying to take its share, right? Google has its own TPUs, Amazon is trying, and OpenAI is building its own chips to take away market share. They both need to be careful.

But OpenAI, I don't think they want 99.9% market share. I think they could be—well, they need their number 2, possibly, or their risk, right? Because they might be a monopolist in some ways today at a consumer level.

Oh, I use Claude every day, but you're weird if you use Claude, right? On the consumer level, they border on Google Chrome levels of market share. Sam's very careful not to denigrate others, other than a few jabs at Elon, because he needs a little bit of this so that he isn't ripped apart as a monopolist over time.

This is an epic monopoly like we've never seen. Think how much ChatGPT already dominates our lives. It's the Standard Oil of tech.

I feel the need to cynically say that the objection to a monopoly is the extortionate excess profits a monopoly extracts from the consumer. Never has there, on that basis, been a less successful monopoly than ChatGPT because they're subsidizing—I saw an estimate of the consumer surplus delivered by ChatGPT in the tens of billions of dollars.

To some extent, Jason, that was glib. I hear what you're saying. It's obviously not a cash-extraction monopoly at the moment, but you're right: they have commanding market share in the consumer chatbot market. Between Gemini and Perplexity, though, I think they would argue that it's not a monopoly.

I think NVIDIA, frankly, is far closer to being a monopoly in terms of market share than ChatGPT, which is why I thought you were going there. I do think it's interesting. There's a whole bunch of people talking about building processors to try to take away that revenue, and I'm sure they're hyper-aware of that.

It would be fun to know what, if any, agreement was made about OpenAI making its own chips as part of this deal, right? It's hard to imagine giving equity to someone who's literally building a competitive product when they're 1 of your 6 largest customers. So that's something it would be interesting to know. On the other hand, that's a detail that probably will not be highlighted, given the dynamics.

Well, NVIDIA is weird, too, because it has elements of a monopoly, but it has elements of a monopsony, too. It only has 2 customers. I mean, it has a long tail of customers, but it has to make sure it doesn't have only 1 customer. It has a risk of only having 1 customer.

I think you're right. I think 6 customers—I saw on the last quarterly report—accounted for something like 83% of the revenue, which, again, first of all, let's step back, is astonishing. Look at the market cap there: $4 trillion. Look at the 2 other $3–$4 trillion market-cap companies, Apple and Microsoft. Apple has 2 billion customers—everyone on the planet. Microsoft has probably a couple of 500,000-plus meaningful enterprise customers. These guys have 6.

When you say it like that, it makes you realize, frankly—and again, I don't want to be the Debbie Downer—how single-threaded the market cap of the largest company on the planet is on the spending decisions of 6 or 7 people. Now, the good news for Nvidia, and this is why I go back to the consequences here, is that none of those 6 or 7 people look like they're blinking. OpenAI ain't blinking. Google has made it clear multiple times. I saw a good quote recently from B Capital, just reminding us that Google has said over and over again, “I ain't blinking.”

Obviously, Facebook—Meta—has indicated an absolute willingness to tear up the book and do anything it takes to win. Oracle ain't blinking. So it's this really weird dynamic. You've got this company, what, $4.5 trillion, with only 6 customers. That's bad news. But the good news is all 6 of them are determined to spend themselves into oblivion to win the prize. It's a fascinating game.

What's interesting for me is that the revenue concentration expands from just these providers in this space to the data labelers as well. We just had Mercor on the show. 2 Mercor customers make up 55% of the revenue. I spoke to the other 4 main providers; it's exactly the same, and it's the same 2 customers across them all. It's fascinating. They are incredibly promiscuous with their data-labeling providers. They just use all of them, and they are the same 2 for every one of them.

Yes, because those are the only people who want to buy it at scale. I always like to distinguish: there's the AI revolution, but there's the AI capex boom. There is an AI boom at the apps level. In other words, yes, adoption is taking place, but the scale of acceleration—the real wow in the last 3 years—has been the fact that the capex boom, which the markets and these 6 deciders have been willing to let get so far ahead of revenue, has led them to say, “Let's spend, in aggregate, $600 billion of capex per year on a market that today, depending on how you add up all the revenues, is yielding $30–$40 billion in revenue.” It's amazing.

Anything attached to that capex boom has just killed it. And you're right: data labeling, which, I'll be honest, we looked at in 2016 and 2017 and thought, “Is this a really great business?” In theory, no. If you have a services business and you're only selling to 6 customers, you can make this intellectual MBA case that, “Oh my God, they'll get ground down on price.” That case is totally wrong because those 6 customers don't have time to optimize. None of those 6 customers is trying to optimize their cost basis. They're just trying to build as fast as they can.

And if you're Zuckerberg, if you're Sergey, if you're anyone in that line, you're just picking up your dollars and stuffing them in your bucket as fast as you can. It feels so nuts to me seeing $100 billion go into OpenAI, and that as a headline. Rory, when you compare it to any other time that you've been investing, does this match any other time in terms of a “holy cow” shock?

It's a good question. I think these analogies fall apart because the scale is so many orders of magnitude larger. The only thing that's similar is the sense that it's unlimited, that it's unbounded. And we actually have more skepticism today than we probably did back then. Rory's skeptical of the limits here. I totally agree. I'm not skeptical of the long-term trend.

Does it feel like 1999? It's not a complete analogy. History doesn't repeat; it rhymes. But it's more like 1999 than anything else I've seen in the last 20 years.

I remember Nortel and Lucent making big vendor-financing commitments to their big bandwidth customers to sell equipment, just as Nvidia is doing today—though, interestingly, Nvidia is doing it as equity, not debt. But I do remember that sense. Jason, you said “unlimited possibility,” like, “The future's so bright, you've got to wear shades,” and that just kind of endless belief. I remember it also collapsing very quickly in 2000.

Now, to your point, Jason, everything people said at the time was true. In the end, I actually still have the Mary Meeker book from 1996 or 1997 on my desk that called these the big trends. Content, commerce, and collaboration, I think, were the big internet trends. And yet pretty much everything happened over the intervening 20 years. But obviously, the time element turned out to be important, and there was a 5- or 7-year period where the market had gotten ahead of itself, and it was a pretty ugly retrenchment.

I mean, I think Harry's got a good point. This limitless potential—we've seen this before. It's not like other booms. This limitless potential, this feeling that it could be limitless, is what's so different. It's not the cloud boom when we were all locked up in our homes.

What's so different, though, in the Web 1.0 days is that there just wasn't enough money. Amazon almost ran out of money after its IPO. Now, bless their souls, we've got all the leaders running around guaranteeing each other stuff. CoreWeave now cannot go out of business. In 2000, CoreWeave would have imploded in months because it would have run out of money. Now NVIDIA's agreed to buy 300 years of its capacity. Everyone's guaranteeing everyone.

So, this unlimited capital—yet, of course, maybe it ends—but that's nothing like Web 1.0. No, there was just no money. Okay, Nvidia's fiscal year 2023: $3.8 billion in free cash flow. Fiscal 2023. Pretty good, right? $3.8 billion. Fiscal 2024: $27 billion. Fiscal 2025: $60 billion. $100 billion or something the next fiscal year. That's a lot of cash to reinvest from your balance sheet.

Funny you should say that, because I actually looked at the same number this morning. Yes, the free cash flow is $60 billion. I was surprised by that. On the other side, that's obviously, to state the obvious, a lot of money. Again, the bear case, that's a lot of money and a lot more than it was 2 or 3 years ago, right?

In that context, I was surprised they have a buyback program. They've been buying back stock—they bought $9 billion of stock back last quarter, which is just interesting at this period in the cycle. They've authorized a $60 billion buyback program. In other words, there's a buyback program equal to the free cash flow for the last year, which is just interesting and aggressive.

Cash on the balance sheet—I hadn't looked in forever—is $60 billion, which is a lot, but it's only 1.5–2% of the market cap, right? You kind of go and say, if this amazing market doesn't keep going just the way it's going now forever, then you could find yourself going, “Maybe I shouldn't have bought back $9 billion worth of stock at whatever it is, $180 a share.” And I'll look back on that and say that might have been a mistake.

Again, I feel like I'm the doomer here. I'm not a doomer. I think the trends are great, but it's just a very fraught time.

You're right. In all fairness, I wish I'd done the math ahead of time. What I saw when I was at Adobe—which I think Nvidia is doing the same—is that Nvidia isn't a brand-new company, right? Beside the boom, they do try to buy back an amount equal to the option dilution, the RSU dilution.

It was almost 1-to-1 when I was at Adobe. So, to maintain your EPS at Adobe, you were buying back everything equal to the dilution you gave out in RSUs, which was just, frankly, a cash equivalent until the market boomed, right? It's almost 1-to-1.

And, flagging vigorously, my opinion on that, as I've done when I've been on boards: I think the buyback—the same as your dilution—is as dumb as rocks, right? You should buy back when your stock is cheap, and you should sit on your cash when the stock is dear.

I think the idea of linking it to your equity dilution—again, reminding us, since we said we'd level up—lots of public companies obviously issue stock to their employees, and there's this ostensible rule that maybe you buy back in the market around the same number of shares as you've issued as stock to keep the share count constant. As I said—and you're right, Jason—loads of people do that. I think it's absurd, but what can you do?

It at least keeps your stock-based expense honest. A lot of folks losing money are pretending it's not an expense. If you're profitable and you buy it back, it's like, “Look, this is the same as cash.” It's just a little bit of financial engineering.

And it doesn't appear in the P&L. But the question is this: again, it's back to the zoom-out comment. It's the same in a bull market: no one focuses on balance sheets; we only talk about income statements. And when things get tough, you're like, “Oh, I wouldn't mind having some extra money around.”

Stock prices that were high can be low, and corporations historically have a terrible record. They typically buy back stock at peaks and don't buy back when it's cheap. I see no reason why humans will change this time.

Ladies and gentlemen, study Larry Ellison, the master of this.

The master who, exactly when it was cheap, bought it all back and then used the capital to totally change the game to the current game of capex and investment—and has made it work.

That’s great, but what about me? Your favorite thing about chatting to me? I’m looking at my public book, and I’ve got 28 holdings—27 are green, and I’m in the money. I’m humble enough to know I’m not that good, and I’m looking at the S&P going, “Really? It looks like it’s going to hit 7,000.”

Do you think it’s going to hit 7,000 pretty soon? Would you be selling now? How do you think about when to catch the falling knife? Genuinely.

Well, I’m excited because, just like 2008, at the moment I’m 0% cash. Nothing.

0% cash?

Nothing. I have no cash whatsoever. Just like 2008, I remember feeling—I had no cash. It was so great when the market crashed, and I literally did not have enough cash to fix the roof on my house. I had enough stock, right? But nothing was more fun than selling my stock at a 70% loss—or 60%, whatever the worst of the crash was—to fix that roof.

I don’t have a cent of cash right now.

It’s, by definition, hard to know. That’s what the data says. Look, you can throw out 2 factoids either way. The correlation between valuation and short-term return is pretty low. In other words, you can say stocks are expensive, but the correlation for predicting 1-year returns is pretty poor.

The correlation for predicting 10-year returns is pretty good. At the current valuation, your likely 10-year return on the public markets is significantly lower than average because it’s more expensive going in. That’s the long-term message, and that’s a pretty grim message.

The short-term message is, “Oh, my God, the Fed’s cutting rates,” and things are going up. Since you ask, what I do with that information is accept that the cost of sleeping at night and having some portion of cash is accepting a level of underperformance. You make an asset-allocation decision based on what you want over the medium term, not what’s going to optimize just in a bull market.

I’ll tell you what the frothiest sign is. One, we’re all invested in Nvidia, right? To Harry’s point, in the public markets, all our 401(k)s are in Nvidia. But the frothiest sign is that LinkedIn LPs are bragging about their returns on LinkedIn. When LPs start bragging about their returns, to me, that’s the 2021 moment I’ve seen—when LPs, who usually hide behind the Wizard of Oz curtain, are bragging about their returns.

Dude, I have everyone in my Twitter feed saying, “Triple-triple-double-double is dead.” That’s the top. There are many signs that people think are the top. I disagree with that.

I do want to discuss this: 75% of VC dollars in 2025 went to 19 companies. Is this just an extension of the Magnificent 7 and a concentration of capital? Is venture itself changing, where we’re all doing a Kleiner Perkins and moving late-stage to get into the surefire winner?

First of all, it’s obviously a stunning fact, but a better way, if I may say, of thinking of it is not that venture capital has changed. It’s that the 25% that’s remaining is, in fact, the same venture that’s always existed.

It’s roughly—I mean, I think in our space—1,000-something-odd Series A’s every year. A certain percentage go to Series B’s, and that business has stayed the same, plus or minus 10% to 20%, for the last 10 or 15 years, with fluctuations. What’s really happened is that, on top of that business, this totally separate business has emerged called, as you say, ultra-late-stage, private-public-style investing.

I just think of that extra $50 billion a year, or whatever it is, added on top. It didn’t change my business. It just means there’s another business that you can choose to be in or not that exists 1 layer—maybe 2 orders of magnitude—above you in the valuation world, and it’s still private. It’s still, quote-unquote, reported as VC. It’s just a different business.

It makes sense that that’s way more concentrated. Series A is more concentrated than seed. Series B is more concentrated than Series A. Series C is more concentrated than Series B, because at every step some people fall out of the game. The longer you hold private, the more concentrated it gets. In the limit, maybe we’re just left with 2 foundation models, Databricks, and Stripe raising a Series N or G, or whatever it is, right?

But it all makes, quote-unquote, sense. It’s just that late-stage is a different business from Series A, B, and C venture capital. It just gets reported in the same bucket.

You guys have taught me so much, but one thing that Jason’s always taught me is to try not to predict out several rounds. Just predict the next round and ask, “Can you see a 3x there?” I really like that framing, and we do it as a team.

The hardest thing I find is that we’re getting it wrong a lot. A lot of our businesses are growing nicely, and I cannot predict what the next round wants because it seems to be moving so much. I’m having a real problem predicting financing markets. Are you in the same boat?

First of all, I’ll go back to the rule. I actually think Jason’s rule is a very good one. We’ve evolved to the same thing. You can have a high-level, generalized theory of the case for what this company can ultimately be worth, but I think Jason’s exactly right: it’s far more useful to say to yourself, “What is this company going to do in 18 to 24 months, and do I believe it can raise a follow-on round at a step-up to our valuation commensurate with the risk?”

That’s a much more tangible discussion, and I think every VC evolves pretty quickly to realizing that’s actually the right way to think about it. You sit there and go, “If we do this round and they make that triple, despite what you say, will we be able to get a 2x to 3x step-up in the next round?”

It is a great rule because it’s deeply practical, much more so than forecasting long-term returns.

Is the market fluky now? That’s a different question. Jason, can you predict which of your companies will be hot?

I’m not challenging you. I think it’s always been easy when something is super-hot to know it’s super-hot, right? When you’re at the top now, maybe he’s right. Maybe the top 1% has changed, or the top 0.1%. Maybe we could be analytical about that.

But when you’re in it—and you can fall out of it, as we all know—there’s no doubt. What’s weird today is that level just below it, where it’s unpredictable. Someone may see this as an outlier and preempt at a very high price, while others may see the risk beneath the surface, be concerned about margins, or be concerned about churn. It’s very easy to criticize a lot of these companies on churn and margins.

If you’re just below that—whatever that S-tier is—that’s where I find my ability to predict very, very murky. Very, very murky.

The one I disagree with you on—but I agree with you from Twitter—is that I think the ones 1 layer above that, if they are triple-triple-double-double, you meet with enough people, and you’re not burning a lot, do get funded. That’s where I quite disagree with you. But it’s a lot more work.

It’s a lot more work to get funded growing 100% at $20 million or 110% at $10 million than it was 24 months ago. It’s just a lot more work because you can’t get the meeting.

I would just add 1 nuance. If you’re in a slightly weird space that’s traditionally unloved, like restaurants—as an example, no one likes selling to restaurants. It’s a hard business—then where they would have taken a bet on you before with triple-triple-double-double, now they’re not.

I would say the opposite, looking at Owner, which we know, right? They dominated. Triple-triple-double-double—it’s faster.

But my learning is that if these models are growing at outlier rates, you’re an outlier. I don’t really care if it’s tagging or at some other thing. We all want it to be AI-native, right? But if you’re hitting those numbers, people don’t even dig beneath the surface, do they? They don’t even care if there are a lot of forward-deployed agents or this or that.

If you’re the 11th undifferentiated restaurant SaaS company struggling to build a point of sale at $1 million in revenue, people don’t want to take that meeting, do they? At $50 million to $100 million, growing triple digits, they’ll take the meeting. There are only so many of those. There are only so many folks growing beyond triple-triple-double-double at $50 million to $100 million. They’ll take the meeting.

I agree. I think that whole meme is a little overdone, right? The whole, “Oh, 3-2-2, it’s not good enough.” There are examples of companies doing better than that—a small number, but a meaningful number of companies doing better than that, right?

But that’s not the only game in town. If you really have clarity on that kind of traction, especially at any kind of reasonable scale, I totally think you’re getting funded. I think you raise a separate question: Is what’s really going on a story of belief? People don’t believe that the growth will happen in the future.

But look, you just had a couple of IPOs where companies are growing 30%.

Obviously, that’s at scale. So, as long as you’re on that kind of trajectory, I don’t believe it’s as vehement, as sharp a line as some of the Twitter thread—the Harry Twitter thread—makes it, right? Yes, there’s a small number of companies growing significantly better, especially early on in the foundation models, but I don’t think it’s the only game in town.

But let’s, for sure—look, it’s sensational, right? We could break it down. I saw a version of it this week that was less dramatic, which was ICONIQ, right? ICONIQ had $2 billion in exits this week. ICONIQ Growth, okay, had Netskope, which no one talked about: an $8.5 billion exit.

And they congratulated themselves on DX, which Atlassian bought for $1 billion, although I don’t know how much they put in because they describe themselves as essentially bootstrapped, right? So ICONIQ couldn’t have owned a third of the company, right? But are those even rounding errors compared to Anthropic? They equally congratulated themselves, but with the fund sizes and expectations, do DX and Netskope meet the bar in 2025? It’s a different version of Harry’s question, isn’t it?

Well, as an absolute number, it’s a great return. Let’s start with that. Again, if you have a $10 billion fund, you don’t care. But most funds, those are excellent exits. Netskope is a superlative company and did a great—

Superlative.

Yeah, they are. No, they are.

But do they count in 2025?

Yes, of course they count because they go in your bank account, and that’s the mission, right? If you own 10% of Netskope, you have a $700 million equity position in a freely traded public stock. Maybe you paid $100 million for it; you have a 7x. It’s great. It’s a great outcome.

I don’t think of this every day, but I do think about me and Emergence 2 back in the day. Emergence 2 was an incredible fund, right? So many winners in cloud, right? The Veevas and all those. I was an asterisk at the bottom of the outcome. I was an asterisk. The returns were so gigantic. I mean, that was a 10x-plus fund. I was just a rounding error in other exits, right?

I just kind of think about who gets to be in the asterisk at the bottom of the DPI table in 2026, because I was in the asterisk back then. It’s cool, but it didn’t feel great to find out I was in the asterisk.

Interesting. I understand what you’re saying, but let’s talk about multiple, and then let’s talk about absolute amounts.

As the OpenAI cap table, quote-unquote, crystallizes, I think some of the early investors in 2019—and I’m doing this from memory based on their ownership versus the original capital—are around a 7x or 8x, right? So it’s a magnificent company. It’s the most important company of the last decade, right? But the actual multiples earned are really good.

Someone who did the Series A at Netskope also made a 7x or 8x. In the end, maybe the absolute sums might be different—you could put more money to work—but it’s just worth pointing out that, on the basis on which you invested in those, if you made those 2 bets, one of them being OpenAI from 2019 to today and one of them being Netskope from 2017 to today, your IRRs might be different, but in both cases, you made a 7x.

All 7xs are exactly the same because money is fungible. That’s why we invented it. There are very few ways to make a good return on a large amount of money, which is why the bigger the fund size, the more you have to be in only 5 or 7 deals. But there are quite a lot of ways to make meaningful equity returns on good outcomes. Even $5 billion-plus IPOs can result in a perfectly great 10x.

I’m here in defense of the idea that none of those are boring, and there are more ways than that. Scary listening to you, Rory. Sorry. A $5 billion IPO does a 10x.

Well, again, by the way, it’s not a 10x on the Series A. I’m actually just doing it because I calculated—I’m not going to name the investor, because a lot of them, very wisely in my opinion, have piled into the follow-on rounds. So your return on your first money is a 25x or 30x. Your return on the last round is a 3x. Blended across everything, you have a 7x.

But it’s a 7x on your $150 million. If you want to talk about the return on the Series A, my guess is it’s 25x or 30x at least. It’s a great return.

But it is also a company where 10% of the world’s adult population is a weekly active user.

Oh, we’re now going back to OpenAI, the other 7x. Yes.

Ten percent of the world’s population is a weekly active user.

Yes. Yes. So we’re now switching from, as it were, the 7x in the midsize, tech-centric IPO to the 7x that you get investing in OpenAI. Yes, your return might be only the same, but obviously it’s a company of far more stature and significance, just in terms of capital.

If you had a large position in OpenAI in the fund, would you be selling, Rory?

I think you’d have to think about it, wouldn’t you? I mean, if you’re not thinking about it at $500 billion, you’re probably just not thinking.

I completely agree with you. It’s another—you mentioned D.

I don’t think anyone’s going to sell. ICONIQ was saying they never got more calls than they got trying to get into the Anthropic round this last round. They never got more calls in the history of ICONIQ than from their own LPs wanting to get in.

So I’m not saying you shouldn’t sell. I’m just saying, boy, it’s hard to be sitting at the fund and sell when you’ve never felt more strongly there’s another card to play, right? I mean, the easiest thing in venture in the world is if OpenAI goes from $500 billion to $1 trillion and you don’t have to take a single meeting, right? You don’t have to show up to anything. All you have to do is open an email and your position doubles. I mean, it’s so hard to say no.

This isn’t like a—I mean, Rory’s point, look, 80% of IPOs trade down, right? So that’s a tough bar as it is, right? But when you’ve got one like that in a frothy market, how do you sell?

I don’t know, dude. On paper, you can, but in real life, good God. I mean, the 3 of us are each going to make $50 million now. But if we just wait 6 months, we can make $100 million.

$100 million. And I’ve got to pay taxes. I’d rather let it ride and defer the taxes. I mean, I can’t even. Again, there’s really only so much I can get on the Yellowstone Club for $50 million. I mean, after taxes, I’m sort of mid-hill and I’m probably under 3,000 square feet. I don’t even know about the radiant heating.

So let’s play another card. God. I think the level of—no, no, but, Jo, I think you’re right, Jason. I just think the level of money you have going in dictates your willingness to sell. Again, we’re very candid and we’re friends now, which is great, but I don’t have $50 million, and I would absolutely take it off the table because it’s really meaningful when it’s your first big hit.

Agreed.

Absolutely.

But there are also weird dynamics. Rory could educate us the most. This is what I think about: it returns the fund, right? If you have a smaller fund and you have a fund returner, it’s a weird dynamic because all the internet advice says a fund returner is what you want to do, right? But turning a fund returner into a 2x fund returner is a BFD. It’s such a big deal for carry, for performance, right?

That’s where it’s actually a nice but stressful position to have: a 1x fund returner with liquidity options. What do you do?

And I like the layout because I think you have to lay it out there. It’s a 2-step thinking process. First of all, you have to come to some kind of opinion on fair value and upside for the stock. You have to have some level of thought. It can’t just be, “It’s going to double because it’s always doubled.” You have to have some grasp of the fundamentals and say, “What do you think this is going to be worth?”

Half a trillion makes it the 15th-largest market-cap company on the planet. It can go from here, right? Whatever. Right? But then I think the interesting thing you’re saying is that you have to overlay on that the institutional or personal portfolio imperatives that come on top of the raw expected return.

Plugging a book I recently reread, actually, a book called The Missing Billionaires, which was written in part by Victor Haghani, who was the youngest partner at Long-Term Capital Management when they went spectacularly bust in 1997. He’s gone on to a career in wealth management. Truly excellent book, one of the best books I’ve read on portfolio management.

The comment at the start is, Cornelius Vanderbilt died the richest man in the world. If all his heirs had done was stick it in the S&P and live on the dividends, there would be 15 or 20 of them, each worth $1 billion today. And there are none.

Because he says people screw up bet sizing. They screw up portfolio management. It’s not about stock selection as much as it’s about the things you just mentioned, the institutional decisions: How much of your wealth should you have in one stock? What should you do with a 1x fund that’s liquid, where you might see potential to a 2x? How certain do you have to be that it could double again before you should leave all your net worth in that stock?

And it’s a super-interesting point, and you actually ended up convincing me of the marginal utility analysis. It’s what Harry effectively mentioned: the marginal utility analysis. You do have to take into account your risk aversion.

And we’re not all just trying to maximize expected return. You should have some level of risk aversion, and the question is how much? As Harry said, your risk aversion relative to the bet size changes with your net worth. If you don’t have $5 million and you have a liquid $5 million, you probably should take that $5 million. And then there’s how that changes over time.

It’s a super well-written book, quite quant at times, but it’s exactly about these kinds of things. The meta-conclusion I have is that most of us have a risk aversion of about 2, as he quantifies it. In other words, you’re not going to let the bet ride for an equal expected return, and that’s most normal human behavior.

The second conclusion I have is that some people—I lump Elon in with them—are just totally maximizing expected return. There’s literally no risk, zero risk aversion, in the system. They just want to make the bet. SBF had the same thing; he would do a 51/49 bet. Some people just have super-high risk tolerance, to the point of arguably insanity, and those are the people who make great entrepreneurs.

Most money people will take some money off the table. Do you think investors are like founders then? We say, “Hey, take secondaries, take all that stress off the table.” Do you think richer investors make for more high-upside investors because they’re like, “Fuck it, let it ride”? Sequoia aren’t here to make a half-a-fund return, whereas an emerging manager is thinking, “Fuck, I need a half-a-fund return in DPI. I need DPI.”

That may be true at the margin. It’s almost certainly true in the sense that you don’t want the emerging fund manager to behave totally irrationally or just over-optimize on that one thing. But there’s no doubt that a significant part of the advantage of a firm like Sequoia is the innate belief that something else will turn up tomorrow. You don’t have to fiddle around with this at the margin.

There are a bunch of stories about that. I believe the early offers on YouTube were significantly lower, and the SEO guys weren’t just taking that, so the offer went up. In retrospect, given that they sold for $1 billion and it’s probably now worth $100 billion, you wonder.

There’s no doubt that success begets success—not just for all the referral effects that we could talk about, like getting to see good deals because you’ve been successful, but also for this very intangible effect, Harry, that you mentioned. People who’ve been successful are more willing to take risks, and the only way you get success is by taking risks.

Right. There’s a reason that one of the strongest and most interesting correlations in venture is difficult to establish. It’s hard to correlate venture success with a whole bunch of things—education, stage, and so on—but it turns out that having an early success is highly correlated with future success.

Partly, you get the referral effect, but partly I think it’s that you just have the stomach to roll the dice and you get braver. You can over-extrapolate that and screw up, but there’s no doubt that it’s a lot harder early on to have the big cojones to roll the dice. With Sequoia, it’s probably a lot easier.

When we talk about risk-return analysis and concentration of assets, I hope he doesn’t mind me saying this, but Oren has a lot of concentration across funds in Navan, and at points that has looked very, very nerve-racking. The point being, when it went to zero in COVID, your travel company’s revenue probably went to zero in March 2020. To go from there, in a non-diversified bet, to having an S-1 on file for a perfectly doable, nice IPO—I can imagine the exhale when this puppy prices. I thought the interesting elements were $613 million in revenues, growing 32% year over year, 10,000 customers, and 110% NDR, which is good—not best-in-class, but good. How did you guys think about this S-1 announcement? Before we talk about the S-1, let's go back to Oren.

Well done, Oren. Good guy. I mean, exactly: you made a very non-diversified bet. It’s terrifying. You then had COVID in your travel company, your revenue probably went to zero in March 2020. To go from there to having an S-1 on file for a perfectly doable, nice IPO, I can imagine the exhale when this puppy prices. So, well done to him and his investors.

I mean, you guys are better experts than me. Is it really as concentrated as it sounds? If these are SPVs, opportunity funds, and other things, is this him really putting 80% of a main fund into Navan, or is this stacking a whole bunch of vehicles?

I have no idea.

I have no idea. I mean, Bonnie, I—

This may not be as concentrated as it sounds. It may be a lot of his book, but it may not be as concentrated for his early-stage fund as it sounds, right?

Well, 20% of a fund, dude, is a lot. And 20% across all funds—I think you’ve got to put it in context. If you have a small fund, how big was his core fund in that?

Do you have 20% of your fund in any company?

That’s the goal. I’m a pretty concentrated investor, so I’m going to get to almost 10% in 2 checks into any deal. I don’t think it’s that crazy, because if you have 100 bets, it makes sense. But if you have a breakout winner, you should put 20% into your winner, right?

I think 100 is risky. Rory can help me do the math, but once your fund is up 4 or 5x, 20% of the initial principal is not that much of your NAV.

Yeah.

Once you’re up, these are all trade-offs. Concentration can result in increased outperformance at significantly more risk, and the question is, again, are you getting paid for that extra risk?

I don’t know if I’d have the stomach to put 20% in one deal. I want to honor the fact that Jason clearly—and Founders Fund, in my view, one of the most successful firms—have the stomach for that. Maybe it’s just a risk-tolerance perspective. I’d find 20% hard.

Let me give you a learning I had, for what it’s worth, from Byron Deeter. I forget exactly how it works, so I’m mixing it up a little, Harry. You just had him on a show, but there were a certain number of partners, and each was essentially allocated a tenth of the fund, or a sixth of the fund, nominally. Byron put roughly 30% of his into Twilio.

That wasn’t 30% of Bessemer 9, or whatever it was; it was 30% of his carry in that fund, right? For the GP?

It is, because for the GP in question, no. But for the firm, I mean—

For sure.

Yeah.

It’s a similar risk at the GP level, isn’t it? Putting all of his chips into Twilio?

Yes, it is a similar risk for him, but it’s not the risk his investors want to undertake at the fund level. In other words, it might be fine for an individual GP to have 30% of their book.

Look, it’s the same reason—step back, guys—it’s why these macro managers, the best business on the planet, by the way, have these little pods. Each individual pod is taking wild risk, and then the macro manager is sitting back. Anyone who underperforms by more than the tolerance just gets whacked, and you put a new one in.

The person on point is taking a lot of risk, a lot of upside, and a lot of focus, but there’s always this intermediary layer saying, “I want a little bit of risk smoothing and risk management.” I do think most—not all, but most—investing vehicles have some element of risk diversification in them.

There are very few people, and there’s not a huge appetite in most markets, for undiversified, single-stock risk. Interestingly enough, there is right now in, as you say, the Anthropic and OpenAI companies of this world.

I always remember Brian Singerman teaching me that capital-concentration limits are the enemy of great venture returns, and that’s why they have 33% of their fund in Airbnb. I don’t have the balls that he does, sadly; otherwise, I’d be much richer, I’m sure.

I always remember that. On Navan, point taken, Jason, in terms of level of concentration.

Before you go on, I’m just going to make a comment back on that, citing the book again. It was really interesting because it made you run the exercise of what level of confidence you would have to have in the excess return from an individual stock to put all, half, or a quarter of your net worth in it, versus having it in the S&P as a whole. It quantifies your certainty level.

One of the exercises it did was to ask how certain you would have to be that Tesla was an outperformer to be 100% in Tesla from the 2010 IPO onward. The answer is that you have to believe it’s about a 70% outperformer on the S&P, which, interestingly, is about where it ended up, right?

You can quantify these things. Instead of just saying, “I’m going to take risk,” you say, “How certain are you that this stock is going to do 20% better than any other stock in my venture portfolio?” If you have a high degree of certainty about that, then, yes, you skew your concentration. But you have to have some rule of thumb like that, rather than just saying, “I feel brave. Let’s do this.”

Enough of that. Sorry, now back to Navan.

It’s a meaningful company. It’s been a very prominent startup for the last few years. Can I ask both of you—maybe Rory first—a question on it?

Maybe this is a little mean, and I’m a big fan of Ramp and everything they’ve done, but it feels to me—and sometimes this is a good IPO strategy—that they’re going first because, look, Brex just announced, and Brex had its slowdown, right? But it just announced it’s growing 50% at $700 million, right? It’s hard to take anything away from Ramp.

It's a little confusing, but let's assume they're growing that fast, or faster, at $1 billion. Okay, they have to be, mathematically. So if you see them as peers—and we could argue they're not, but I think the markets will—it's number 3. There is an argument that number 3 should go out now, before number 1 and number 2 are out and the public markets lose interest.

I don't know whether it's true today, but when I've been on the other side of IPOs, there's a strong desire to get—if number 3 is good, right, or number 2 is good—to get it out before number 1.

Two things. One is, I think they would say—and in my view, correctly—that while they're in adjacent spaces, Navan is very much trip- and travel-focused, with a small amount of software. Brex is very much card-focused, with a small amount of software and payments; same with Ramp. Then BILL, which I was on the board of for years, is very much accounts payable with card.

So they get lumped into the same thing, but they're actually quite different. Most of Navan's revenue comes from business travel, which the other guys aren't in. It's travel booking—business travel, almost all of it.

Right, right, travel booking. So, in other words, when I book my flight on the Navan system and I book United, that's Navan money. Now, if all I do is pay for a flight on United Airlines with my Brex card, that's Brex money.

Their adjacent space is not the same, just to put that out there. But at the same time, yes, I can see you're right when you zoom out a million miles. When you read their S-1, they're clearly trying to claim more than just travel because they're truly trying to claim payments. They're trying to claim software.

So, based on their claims, which are a little in advance of their reality, you're right: they are in the same market as the other 2. And if they are, it's pretty damn smart to get out early.

I think that expansion across the horizontal product suite is actually behind the rebrand from TripActions to Navan, going from vertical-specific to horizontal. I agree that TravelPerk is obviously a very direct comparable for them, and they are bigger than TravelPerk. I think it's perfect timing.

If I was the board, I'm like, "Mother, perfect timing. Let's go, go, go." Completely agree with you.

But this isn't the best time for Navan to IPO. They're not profitable. This isn't the best time. I think they're doing it because of—listen, I could be wrong, and I'm a fan, right?—but I think they're doing it because of the competition for IPOs.

Otherwise, why not wait another few quarters and get profitable? It's not the perfect time to IPO.

I don't know, dude. If we were sitting on the board together, I would argue with you that Ramp and Brex are not going to go out anytime in the next 6 months. But my experience is: create that gap.

You could argue that, at a high level, Netskope didn't get much attention because Rubrik is the same but better. Now, it's not the same, okay? But if you're just comparing security IPOs, Rubrik is growing faster with better economics than Netskope. Why? You know, wake me up when you have something better than Rubrik. Wake me up.

But you're still getting $7 billion of liquidity, right? I mean, I think that's the problem—

For sure. For sure. Yeah. Yeah.

So, first of all, let's consider this on a standalone basis, right? Getting liquidity, and then we'll do the game theory with the other guys, because those are different questions. On a standalone basis, yeah, you're right. Probably—I mean, I can see the pro and the con.

If you wait another year, it'll probably take more than another year to get profitable, because if you actually look in the P&L at Navan, they really held OpEx flat, in fact slightly down this year versus the prior year. That's what you do as a board when you want to throttle the damn thing and make it profitable.

So you really push, and remember, inflation's 5% or 4%. If you're running negative OpEx in real terms, you're really reducing. So this is a company that is clearly straining might and main to get profitable, right? And you're growing 30%. You could probably work out based on that: is it 1 year or 2 more years to get profitable? Basically, right? It's just too long.

Yeah. And you say to yourself, I could wait, but this is a pretty damn good market. Do I take a little haircut for not being profitable yet? But am I done now? Am I public? Am I—it's off the to-do list?

You survived the near-death experience 4 years ago. And as we've discussed, we think at the margin, public securities are just an easier place to be in terms of access to capital on an ongoing basis. So I think it's smarter than you're giving it credit for.

So you have that in the abstract, and then you have the game theory side of it, which is it's good to be first, maybe, but you definitely don't want to be last. I mean, because the other side of the thing is really hard where 2 other companies get out roughly in your space and you're the 3rd.

At that point, unless you were demonstrably better than them, it becomes troubling to get out because every public investor says, "I already have Brex and Ramp. Now you're Brex and Ramp, but not quite as good. Why would I buy you?" Right?

So you often don't want to be the last one out. You don't want to be the last player and the last one out. When people perceive it as a direct comp, even if it isn't, it's just like, "I'll just go buy Brex and Ramp if they're already public and they're better. I'll just go buy those shares today on E-Trade or whatever. Why do I need to do your IPO and listen to your road show unless there's a massive discount? Why do it?"

Yeah, no, which would argue for exactly what you're saying: it might be smart to go now. It might be smart to get the novelty value. It might be smart to take the ground.

Maybe you can go public, maybe you can acquire a second-string card-payment company and add that arrow to your quiver. Much easier to do as a public company. So, yeah, I think it's—

Sure. And first of all, yes, it is a totally funny thing because the other thing, just to put it out there, is that you tend to report as an investor your holdings. You're the named person on the thing.

So, in SEC filings, it looks as if you own not just your shares, but your partners' shares and the LP shares. So suddenly—I mean, take the Index example at Figma. You can probably Google the partner and Index and they'll say, "Net worth: Figma stock, $3 billion."

It's totally misleading and results in a whole bunch of charities calling you and saying, "Please give me some of your $3 billion." You're like, "I don't have $3 billion. I got 20% of $3 billion divided 5 ways, and it's 6 months before the lockup." So, yes, you do get that effect at times.

Yeah, look, it's just a process of time. I mean, you have a lockup most of the time. One of the attractions of a direct listing is that you don't have a lockup. The lockup is typically 6 months, and it can be waived early.

Sometimes you see these performance triggers where, if the stock trades above a certain amount, you can waive the lockup early, which is nice. And then, after that happens, the next thing is that sometimes, during the lockup period, if the stock performs well—and by performing well, it means trading well above the IPO price, which gets back to this whole thing of IPO pricing—if the stock, quote-unquote, trades well, you can probably get a secondary done.

That means you can sell more of your shares during the lockup period via a registered offering, right? And that's attractive because it's liquidity in a structured deal where you just get your capital, right?

It's one of the reasons why, as a cynic, a 10% or 15% pop in the stock isn't, quote, the worst thing in the world. Everybody who buys at the IPO is happy, and then 6 months from now, if there's going to be a secondary, which means a structured sale, you know, you can't get a secondary done if the stock's traded down.

If you go public at $14 a share and 6 months later you're trading at $12 or $10 or $9, it's extremely hard to get a secondary done. Whereas if you were public at $14 and it trades up to $18 or $19 and you make your first quarter, then you can easily get a secondary done.

That's the second way out. So, at the end of the lockup, when it ends, other than a structured secondary, your choice is to distribute the shares to your limited partners or sell.

The truth is, it takes a lot of time, especially if you're on the board. You have reporting obligations and quiet periods where you're not able to sell, and it takes a long time to get out of a position. Typically, I'd go 18 months from the IPO, plus or minus—maybe 24.

Is it not just better to sell before, in a pre-IPO secondary?

No, probably not all the time. Again, you have to have an informed opinion on the value of the stock, and then you have to figure out your risk tolerance and institutional kind of issues on top of that, right?

But, I mean, look, to the extent that the median stock pops and then trades up, you'd probably be leaving some money on the table. But there have been times when that's been the right call.

I asked him on the show, “Do you believe that you are fundamentally a better manager of public stocks than your LPs?” And he said, “No, no, but we do understand that there are some who have this rule that they have to systematically sell the minute that they’re distributed to.” In that situation, we will deliberately hold on because we’re not saying we’re arrogantly better, but they have this systematic, bluntly strange rule in certain cases, or ineffective in a lot of cases, where actually they need to hold on. I thought that was interesting.

And he’s broadly correct. Some LPs choose to automatically sell. They’re inheriting a stock they don’t know anything about. The person who knows most about it, which is the GP, has elected to distribute it. So there’s some signal in that, and it’s just not the asset they want to hold. I get the logic of the distribution. Well, I get the logic of the rule.

Holding is your best time to legally trade on inside information, though. If you’re on the board or close to it, it’s your best time to trade on that information by not trading.

Yeah, you can hold legally with inside information. It’s a privileged position. You can sit on that board, know what’s coming next quarter, tell no one, and hold. You have to be careful if you distribute or sell. You have to be a little thoughtful about your timing, but you have this special thing where you can hold on to inside information.

Yeah, right. There’s no securities law violation in holding. Equally, just to say it, the caveat is that if you have negative information, you absolutely can’t sell. So it’s not a one-way street.

But you’re right. One of the interesting examples—I remember 1 case of that—was on a public company I was on for 6 or 9 months, where we were in active M&A talks. You’d have your LPs ask you, “Why don’t you sell the stock? It seems very fully appreciated,” and you can’t say a word. You just have to say, “We’re taking everything under advisement,” while you’re sitting there knowing we’re about to get a 30% or 40% premium once this deal closes.

So yes, you do have that, because it’s a great point, Jason. Sometimes LPs ask, in my view correctly, “What’s the advantage of being on the board once you’re public?” There are disadvantages, right? But it’s not a one-way street. The disadvantage is limited trading windows, but the advantages are that you do have an inside seat on something like driving toward an M&A or an upside outcome, and you have a better sense of the company’s performance. So it’s a toughie.

I find you err on the side of getting off reasonably quickly because you do want to get on to the next business. But I don’t, on the other hand, believe in just bailing day 1. I’ve generally found you get off within 12 to 18 months of the IPO most of the time, and at that point you’re distributed, you’ve done your job, and you shouldn’t just be on for other reasons.

That said, it is worth pointing out that the 2 venture investors in the NVIDIA IPO in 1997, Mark Stevens and Tench Coxe of Sutter Hill, have stayed on that public board to this day. I believe the board package—as you know, the equity you get as a board member—has been extraordinarily worth their time. Let’s just go with that. Extraordinarily so.

And there’s 1 I’m totally unable to remember, so I’m not even going to try. But who’s never sold a share?

Yeah, I think it’s Mark Stevens who never sold a share. You’re probably looking at $1 billion plus, maybe many billions of dollars. It turns out being early in the best, largest market-cap company on the planet and never selling any shares is a remarkably good way to make money.

I think it’s impressive and all, but just like the concentration in Navan, it’s just a hint less impressive than it sounds. It’s very impressive, right? But when you’re up enough personally, it makes sense to hold all your winners in the public markets for tax and other reasons. There’s just no reason, if you’re personally up enough, to sell any winner, right? You might as well let it ride, unless you know it’s going down. Then sell it. But you want to hold on to an asset that will continue to appreciate essentially tax-free. There are a lot of advantages to it.

Portfolio theory would say something different. Emotionally, I’m with you. I like holding on to the companies I’m involved with. Portfolio theory would say something different.

I’m not breaking the rules here—I don’t think you can tell me if I am—but it’s a very significant part of our ecosystem, and it directly applies to tech. I’m not going into politics, but H-1B visas: it was announced that a $100,000 fee or payment is now needed for new H-1B visas to be granted. I’m not going to get into Trump politics very deliberately. Do you think this will have a material impact on startups, early-stage companies, and the teams they build?

It will have an impact. It will obviously, at the margin, be negative, right? Because, at the margin, immigration has obviously been extremely good for the tech ecosystem. So you can make that a definitive statement, right? You ask whether it’s material; material is a harder thing to assess. There were 440,000 applications in the last year. Those generated between $19 billion and $120 billion of GDP for the US.

Yeah, I mean, 70,000 accepted. I think there are 70,000 or 75,000 a year, one or the other, right?

I think at the margin, if you were to have any rational immigration strategy, and you were to rank-order the people you want to let into your country, STEM graduates who founded companies that employ thousands of Americans would be at the top of that list, right?

You could argue that some higher fee—or some program like the extraordinary-ability program we have, which is separate from the H-1Bs—kind of makes sense. You could argue that some increased fee might mitigate some of the arguments against the H-1B, which some folks would make: that some of the applicants are at least doing much simpler work that could be done by folks in the country, and therefore maybe they should be charged more than that.

I think the way this has been implemented, the absolute sum—all those things—aren’t great. The real truth is that rational immigration policy for something like this gets caught up in, as you say, a whole swirl of other emotions around wider immigration issues. What it means is that it seems to effectively preclude any sensible, rational policy on this kind of highly skilled immigration, when it’s pretty obvious that a rational program like this would be extremely good for the US.

Well, look, I think anyone who has been doing this for a while, who isn’t just 3 kids working 996 in San Francisco, has had H-1B folks on their team. I have. I’ve had great folks on my team. My first startup, especially, was hard. It had a material-science component. It wouldn’t have been possible without H-1B, at least on its surface.

I had H-1B folks on my team; they were transfers, right? I didn’t sponsor them. On my first team of 10, I had 2. I wouldn’t have had my first exit or my first startup, or saved hundreds more lives from my first startup, without them, for sure, right?

What we want at a meta level is everyone great coming to the US. That’s what I selfishly want: every single talented person who can help keep our NVIDIA shares high-flying coming to this country. I want it selfishly, ethically, and personally. So it sucks, right?

But at a very tactical level, we find ways around it. The O-1—if you look at the companies you’ve invested in, they’re all O-1s now. Everyone finds a way to get an O-1; at least all the founders get O-1s, and O-1s have a lot of cons, right? You’ve got to keep them going, and it’s stressful. But there are ways, and so I do think the impact will be modest. The big tech companies will just pay up, right? I think it’s terrible, but I think the impact will be modest at the moment if it doesn’t expand. We’ve all had great H-1Bs on our teams. It sucks.

Exactly—the pragmatic point. If you see the strong positive that the program brings in aggregate, you can say there’s a little bit of abuse, but it’s worth the tax. If you’re on the outside and you’re incensed by immigration in general, then you ignore the great people who’ve been enabled by H-1Bs and focus on the abuse. You say, “This is awful. Pick a company. Microsoft is using H-1Bs while laying off Americans.” It’s easy to give that speech, right?

There’s no accident that the countries that have the most skills-based point systems are the countries that have the least angst about immigration. A $100,000 fee is kind of a rough American proxy for a skills-based system.

In a rational world, what do you do? Let’s remind ourselves what Trump at one point said he would do, which is attach it to every STEM degree, right? You’re trying to find some proxy for letting in the people who are best for America. That seems to be a reasonable rule because, let’s be frank, everyone wants to get to America. God knows I did, right? So you have to have some rule that isn’t “everybody.” A logical rule is what’s good for us. You could imagine a point system and all that kind of stuff. The money thing is just a very crude proxy for that.

And I think with people being more cooperative, you could probably come up with a better scheme than picking a dollar sum and making that the deciding factor.

The final one before we do a quick-fire is Notion hitting $500 million of ARR. I thought that was really impressive. It’s accelerating.

So, okay, I’m going to say it then. Harry, they’re not triple-triple-double-double, right? Is it impressive? I think it’s impressive, but this is almost a counterpoint to the whole thing that the only thing that matters is growing faster than SaaS. This is a midsize SaaS company. We’re growing at probably, I don’t know, 30%. It’s great.

Why did I think it was impressive? I think it’s to Jason’s point: it’s hard to get a reacceleration at scale. Triple-triple-double-double is absolutely dead at early stages. Agreed. Triple-triple-double-double at hundreds of millions in revenue is phenomenally impressive to me. Agreed.

And at $500 million in revenue, you could actually squint and see my $10 billion stock, which I have since it acquired some of my companies, reasonably finally being up to the watermark of $10 billion. I’m like, 20x? It’s a bit punchy. It’s a bit punchy with that growth rate.

Pesky 2021 valuations.

Oh, Jason, $10 billion.

Klarna doesn’t even talk about it anymore.

I mean, what it points to is that reports of their death have been greatly exaggerated, as Mr. Twain would say, right? It turns out these mid-tier, significant-scale SaaS companies don’t have to just become something totally different, but they have to embrace and lean into the AI trends while still being fundamentally the thing they are. They’re not trying to be a totally different company, and you can get reacceleration.

And I agree with you. I was giving you shit, but you’re exactly right. A 30% acceleration of $10 million isn’t what the paper’s written on, but if you’re at critical scale within striking distance of an IPO and you can use an AI-enabled story to get you back over 30%, plus 40% growth, you have an IPO in your future.

They IPO today. Why did they price?

Well, we can tell, because, crudely estimating—I don’t know the growth rate, but let’s assume it’s 30% or 40%. I doubt it’s really doubling; I could be wrong. But just looking at the headcount, I mean, they have 1,200-odd people, which probably means $200 million, you know, which doesn’t get you to doubling based on the growth rate.

A 30% grower—I mean, Netskope and Navan. Well, Netskope is priced already at 7, 8, 9 times NTM, so something like that: $4 billion or $5 billion, maybe more if their forward growth rate is a little better or if I’m underestimating the revenue. But yeah, a long way from $20 billion, and it’ll be interesting to see how they digest those preference issues.

But a great outcome. I mean, $5 billion, $7 billion, $8 billion is real money.

No, I know. I got in at 10.

What? How? I don’t know if I was. That’s your problem, not theirs.

It’s not my shit problem. They bought my company. I didn’t have a choice.

Well, it’s a little bit—I mean, it does hang over the heads of most founders. The high valuations—they’ve compartmentalized it, but it’s a little bit their problem.

It is their problem. And as I said, the real question then is how does it get up? Playing that out, how does it get unwound? We talked about this in the context of a couple of the prior IPOs. I don’t think you should stay private just to earn your way back into $20 billion. I think you can go public, and then I think it will boil down to—you might go public and either the stock gets converted or, as we’ve discussed, the preference remains outstanding until it grows into it.

But yeah, I don’t think you can hold the rest of the company up just because 5% of the company paid $20 billion pre. Do you think Airtable will make it back? I got a ton of Airtable from them buying a load of my companies again. Just help me out. Just help me with my planning.

All these companies are going to be priced on the fundamentals. This does go back to maybe a more prosaic version of the Chamath comment, right? Stories at the forefront can be priced on sizzle. Stories that are 10 years old are going to be priced on fundamentals.

If they have $200 million in revenue growing at 20%, they’ll be priced at 5 or 6 times. If they have $300 million or $400 million going at 30% or 40%, they’ll get a decent 7 or 8 multiple, right? I don’t know because I haven’t seen the data. It’s hard to get to scale in these markets.

You have the dynamic of Microsoft at all times. There was a period of time when all those companies felt euphoric, right? It felt unbounded for Notion. It felt unbounded for Airtable. And then, a little bit, the tide went out of the productivity tools market. A little bit, Microsoft just started grinding away at everybody, as they’ve done in so many other markets. And then the world moved on to AI.

So now these are perfectly good companies that are just going to have to find fundamental value based on revenue multiples and, even, God forbid, free cash flow.

I think, for what it’s worth, just like there was a time when you weren’t allowed to talk about Web 1.0 anymore—it just didn’t matter—I think we can’t talk about 2021 valuations anymore. It’s time to just flush them down the toilet.

I wrote down everything, I guess, a year and a half ago—everything that had a hint of froth. I have 1 deal I’m still carrying myself because it’s over $300 million in revenue and growing. I’m holding it as 2020, but at 12/31, whatever the Lord says, I’m marking it down this year.

And it’s time to just forget about those valuations. Maybe you can’t pretend, but mark them down, even if they’re personal, and just forget about them because it’s too far in the past now. Four years—it’s time to move on from those decacorns of 2021. I mean, Klarna was what, $40 billion? Right? Time to move on. Time to move on.

Its peak was $45 billion, which SoftBank did. Sequoia brilliantly did the round at $5 billion or $6 billion and obviously 2x or 3x that, and I think overall 7x their overall investment.

I think we’re not allowed to talk about these rounds anymore. We’re getting to the end of this year. This is your last—you’ve got 90 days left to kvetch and complain about your 2021 valuations. On January 1, 2026, no one is allowed to talk about their 2021 valuations.

The only problem with not talking about your 2021 valuations, Jason, is that we seem to be determined to make exactly the same mistakes in 2025.

Yeah, we need the runway to make them again, Rory. We need to focus on making the same mistakes again and not be hampered by them in the past.

Okay. I’m seeing less diligence now than I did in 2021.

There’s no diligence, Harry. On any hot AI deal, there is no diligence provided, nor is any done, right? It’s just done on Saturday. Why would you do diligence? All you can lose is 1 X your money. Why would you do diligence?

Again, you can say what you like. As long as you don’t put this quote out there with my name, Harry, I’m happy.

I mean, look, I’m going to be the boring guy here. We’re finding that what you have to do is your due diligence prior, right? You have to come to the table with an informed opinion, which you can either pull out of your ass or try and do some pre-work, right?

1 of my companies had a term sheet pulled. This is a large company doing a lot of revenue from a Tier 1 firm. What I find more and more is that, just to get the exclusivity and lock the deal down, they’ll term-sheet it. Then, for the deep work, they’re like, “I’ll do it after, in the 30-day closing period.” And then they pull it.

This is a really bad trait of an increasingly competitive market.

I’m less critical of it now than I was in all the prior years of my career. I’m less critical of—

Shit, bad.

What? I know. But here’s my view, Harry. If you’re only giving me a Saturday afternoon to make a decision, if you’re giving me 1 hour, you won’t share any data, a lot of these contracts are paid pilots, you’re not disclosing anything, and you want me to make a decision in 5 minutes, either it better be fucking solid, okay?

And if immediately after that term sheet I see 10 things that aren’t true and you want me not to rescind my term sheet, you better let me dig in here, because the level of trust that you have to have to do a deal on a Saturday is under-discussed.

It is mocked by many, including leading accelerators. But there is a high degree of trust involved in this. So again, if you want people to make a decision in 5 minutes, it better be on the up and up.

I used to be critical of it, Harry. I used to think it was holding my place and bad, and it still happens. That’s your point: it happens. It’s bad practice. But I have a lot more empathy than I’ve had over the last decade. I have a lot more empathy in 2025.

Or just do what we do, which is say, “If you want that decision timeline, that’s not our type of deal. Sorry, we’re not going to engage.”

Yeah. But I think many firms looking at it—I mean, many of the best companies don’t provide that flexibility today. They don’t. It’s literally now the best founders wind them up, right? Give them breadcrumbs and give them data ahead of time, right? If you want a big check, you’ve got to do that.

But when they’re ready, it’s 1 day, dude. It’s 1 day. You can say no, Harry, but you’re an aggressive investor.

You're going to say yes to a couple. I'll bet you that SaaS tattoo. You're going to break your rule a couple times. But it's the right response, right—

Dude?

I'll get a SaaS tattoo if Rory gets one.

Well, then you're fine. I plan to die with my body unblemished by tattoos.

It's not just being founder-friendly. We all want anyone who's been a founder to be founder-friendly, right? I want it to be extreme, but not allowing diligence, for all intents and purposes, means you're running a risk.

Have you got less founder-friendly?

Less founder-friendly? I am more founder-friendly, but it is less appreciated by far. I am the most founder-friendly I've ever been in my career, and I catch more crap for it because everyone says, "Great job," and everyone's jostling behind the table. Everyone's manipulative, and I'm the only one—

Poor Jason.

No, no, you asked the question. Founder-friendly is as bullshitty as pulling the term sheets in 2025. Founder-friendly has become bullshit, right? But it's table stakes. It's table stakes to get into the deal.

Founder-friendly is writing the check when no one else does. That's founder-friendly. Founder-friendly is when no one else is there at the board meeting anymore, and you're there, and you still have a W on the other side of it. Founder-friendly is when you actually recruit the executive for the role, not just say, "Who are you looking for? I'll send it to my talent person," and do nothing. These are things that are founder-friendly, right? Not saying, "Great job," no matter how you do.

Funny, I find myself oddly agreeing with much of that. What I've internalized is that no matter how much you're in the game of being founder-friendly, it's not a winnable game. So what I say I'm trying to be is founder-honest. I want to be founder-honest, which is that I want to tell you exactly what I think, which I think is more useful than "Great job" if it's not doing a great job, right?

I think you're right, Jason. The only way to judge who really is founder-friendly is by how they behave in a tough deal. There's no information on how you do in a good time, right? I've just internalized that most people don't check that, and that's just the way it is right now.

You end up with this meaningless attempt to prove something in a bull market that you really only demonstrate in a bear market. But that's okay.

I am just going to say this before we do a Khosla quickfire, because Jason won't. I always remember the RevenueCat founders telling me about Jason wiring the money from his personal account, and it's just like, whoa. I remember that it was the SVB weekend. We were literally on Sunday having a partnership call, doing a whip-around to say who could cover which companies, when the rescue came in and the US government stepped up.

But, yeah, it was interesting. You only know what people are like in a tough deal. It's also true, by the way, of VCs. You can only decide which VCs you really like when you've been through a tough deal with them, because then you know.

Anyway, we've got a Khosla quickfire. We're going to do a Khosla quickfire. Bets are on: When will a final TikTok deal be reached between the US and China? Come on, Rory. You love this round.

I hate this round, as you know, but I'm going to vote for never because it's so much fun for the big guy to keep dangling it. Planning to do the TikTok deal has been the most fun anyone's ever had, because you have favors to throw to people, et cetera, et cetera. I'm sure it'll get done at some point, but it's not knowable by me.

Look, I obviously think there's a lot of complexity here, right? But I'm going to say the next 60 days, because I think this definitely somewhat toxic H-1B stuff is a lot of tariff posturing, too, to get these deals with China and India done. I think they're going to get done.

Whatever this crazy deal is, it's going to get signed in the next 60 days. I think these deals are going to get done this calendar year, and I think it's going to get signed. Hopefully, this H-1B thing diffuses, too, because now it's not for existing—everyone had to freaking fly back in 24 hours, right? Now you don't have to do that, and a lot of these things may evaporate as the tariffs get resolved. I'm betting 60 days.

Meta smart glasses: another fail for VR or a mega-success? Zero percent chance they're a success. What's the bet?

Wow. Why?

Because I own, like, 8 pairs of the existing ones.

The Luxas [?]

No, the Meta ones that already work, that are light and work. We just don't need to play Tron in our eyes. We just don't need a 7th screen. It's another solution in search of a problem, in my opinion. We just don't.

There's only so many times I need to be on a podcast with Harry looking at my notes in the size of my 11-inch-thick Gen 1 new ones. It's just not a huge problem in the real world.

Oh, no. I think it'll be a success.

These thick, huge things that you saw—

As you looked at—

I'm buying several pairs, don't get me wrong. If you look to disintermediate a phone and actually have a life where you can unify computer, vision, and living, yeah, that makes sense.

No, I think in a typical VC lifestyle—I think for the normal world—

We're not looking to—I mean, look how successful ChatGPT is. It's the same paradigm we've been using since forever.

It is. And look at Jony Ive coming out with a device. We'll see.

We'll see. It's as risky as the Meta AI glasses are. It's risky. It's risky.

But you were certain, James, that the hardware device Jony Ive was working on would work.

What's that?

You were certain that the hardware device Jony Ive was working on would work.

I think—yeah, I think there's so much thought and energy put into this one. It may create the additional paradigm. But you just asked my opinion. I don't think the AI glasses—as someone who owns at least 6 or 8 pairs of the existing ones—I just don't. I think it's rushed. I think it's clever, and I don't think it solves the problem.

I am having fun watching what Jony Ive is doing with his new $70 million Tibbran [?]. Peter Thiel just bought with his shares to add to all of Jackson Square [?]. I think he's going to come up with something disruptive, and I think it's fascinating because it's such a hard problem. We just don't need an extra screen. Only so many of us want to wear 2 smartwatches. Changing this paradigm has been tough in tech, and a lot of venture dollars will go into it, per Harry, but in the real world, we leave them on the shelf.

Final one: Will Atlassian's buying spree pay off? They've got DX, they've got smaller ones—Cycleborn [?]—and they're just going and going and going on the M&A. Will it pay off and make them an AI leader again or not? Slap in the face, in other words?

It'll help. Look, it's not going to make them an AI leader, but they're doing the smart thing that existing corporations do: buying relatively new technology to try. All they have to do is move their existing customers into an AI engineering-management world, rather than the browser-based things separately, and just try to stay vaguely relevant, defend the market cap, and grow the business 20%.

If that's what leading is, then maybe. If it's, "Do I think it's going to make them the AI-dominant coding agent?" No. But it's just trying to move the needle on a $4 billion-revenue, $40 billion-market-cap company. No, it's not enough. You've got to put chips into play, and I think it's just baby steps.

I also thought—I look for these signs, like what Sam Altman is saying, right? The other thing on the DX one: these are not easy jobs that Michael Cannon-Brookes has. These are not easy jobs. The PR pictures of him on the deal were of him just alone. They weren't even with the DX founders. It was this lonely picture of Michael, and I just thought, man, this is a tough job.

He has one of the great iconic properties of all time in somewhat developer-focused B2B software, but it ain't—the lone, lonely Mike in these pictures. I'm like, man, this stuff's hard. If this were going to change Atlassian, they'd be together, toasting, rather than the kind of wartime Mike picture.

So, I think it's a start. Dilution aside—and I don't know, I should know whether they did it with cash or stock—make the bets. I don't think this is going to change the face of the company.

Combination. Yeah. Guys, listen, thank you so much for doing this, as always.