Opendoor CEO, Kaz Nejatian: OpenAI and Oracle, How Can Either Afford to Do This

I said, “Hey, sweetheart, I’ll be back, I think, Wednesday or Thursday.” Then she said, “Do not come back until there’s a plan to break even.”

I am so excited for this. You all know it’s my favorite time of the week. My mother even loves these shows now, which I think is incredibly heartwarming because she listens to everyone. We’re going to start with news item number 1: Kaz, from COO of Shopify to now CEO at Opendoor.

I never thought I would leave Shopify. I thought it was going to be my job forever. If we can make buying, selling, and owning a home easier, less friction-filled, and less terrible, the world will be a better place.

We will get to a point where you will buy a home from Opendoor, and if you don’t like it, you’ll be able to return it. Businesses should not exist to make money. Businesses should make money to deliver on a mission.

Corporate executives should basically only get paid in options. It’s a very weird world where we create an incentive for corporate executives to be bad at their jobs just to get paid. Opendoor is priced in the public market for its potential, and it’s an incredibly fair price for that potential.

The bull case for Opendoor is obscene. It’s hard to exaggerate how big this company can be. I don’t optimize my life for money. I just don’t. When my wife and I got married, we agreed we would optimize our life for making a dent in the world.

Ready to go, guys. I am so excited for this. You all know it’s my favorite time of the week. My mother even loves these shows now, which I think is incredibly heartwarming because she listens to everyone. We’re going to start with news item number 1: Kaz, from COO of Shopify to now CEO at Opendoor. Kaz, first, thank you so much for joining us for this first little stint.

Thanks for having me. I appreciate it, man.

I wanted to start with this question: You left a great and huge business in Shopify. You must have the ultimate conviction that iBuying is a good business. Can I be so blunt as to ask what led to that conviction, when it’s been a challenging conviction to have over the last few years?

I never thought I would leave Shopify. I thought it was going to be my job forever. I genuinely loved the company, loved Tobi, and loved what it stands for.

When I joined Shopify, everyone was like, “What the hell are you doing? This is a tiny company. It’s going to go nowhere.” It was among the most shorted stocks in the tech market. It was very much, “This thing will fail.” In fact, Citron wrote a short report the day I joined Shopify about how terrible a business it was.

I think Shopify is a wonderful company. But the fundamental problem that Opendoor will solve is an incredibly important problem, not just as a business but for the world. If we can make buying, selling, and owning a home easier, less friction-filled, and less terrible, the world will be a better place.

We will figure out how to make money along the way. This will be a great business. We will make money. This will be a great business. But the problem space is a valuable one to explore, and I think you need to do it unapologetically.

The goal here isn’t to minimize risk. We have a mission to go after. We’ll go after it incredibly aggressively, and we’re going to make it easier for both people who are buying and selling homes to engage in that transaction because it matters to the future of our society.

Can I be so bold and just ask: Does that not feel like a bit of a boom-time mindset? “It’s a valuable problem, and we’ll figure out how to make money along the way.”

No, definitely not. I’m not saying we’re going to figure out how to make money along the way. We will make money. We have plans to make money. We have a couple of good bets already, and we’ll launch more of them.

I’m not saying I have a hope and a dream of profitability. This is a for-profit company. It’s very much a for-profit company. We’ll figure it out.

What I’m saying is that businesses should not exist to make money. Businesses should make money to deliver on a mission. We are a mission-driven company, and we’ll make a profit such that we can deliver on our mission.

The shareholders will be happy, buyers will be happy, and sellers will be happy. But that’s the goal. I’m not saying, “I don’t know how to make money.” I know how to make money here. What I’m saying is the mission is more important than the money.

As a wildly successful software executive, how do you deal with this meme-stock element to the company? I’m not a meme-stock guy. I don’t trade; I’m long on everything. This isn’t GameStop, but this meme-ishness premium is crazy, isn’t it?

Look, I’m not a trader. I owned one—literally one—ticker. When I was at Shopify, it was Shopify. I expect I will own 2 tickers for the next little while: Shopify and Opendoor. I’m not a diversified trader, guys. That’s not what I do. My job is to build great products and get people to pay for them.

Let’s back up. I fundamentally reject the premise. Opendoor is priced in the public market for its potential, and it’s an incredibly fair price for that potential. I think we will earn the right to realize that potential. We’re going to execute against it.

That’s not that different from how startups are valued by VCs. You don’t value the current cash flow of a company; you value the potential and say, “What are the odds it will get there?” That’s what you do. It’s a discounted cash flow: What are the odds this thing will be big?

I think this is the single largest market in the world. Tesla was not above 10% in any market in which it was selling cars until last year. It is above 10% in some markets this year. Opendoor was above 10% in many markets in which it was selling homes a couple of years ago. We will get back there.

It’s a significantly bigger market, a significantly higher attach opportunity, and a significantly longer ability to have a relationship with a buyer and a seller. If you just do a rational analysis—by the way, I’m a math nerd. I was a Mathlete growing up. I’m not a wordcel—if you do the math on the potential of the company and discount it back, I think the stock is reasonably priced.

I would have bought at this price. In fact, I did buy at a higher price than the current price.

One is, I think the meme-stock thing is actually 2 parts. There’s the price relative to value, and then there’s obviously the—let’s call it a combination of meme stuff and activism.

I think the interesting thing here is that a set of investors from outside have been able to agitate and drive change. In this case, I think it was very interesting and good change. But it’s kind of like this is what the evil activists would be like if they were VCs.

It was just kind of fun, from a distance, to watch it: rattling the cages of the board and saying, “You need to do something different.” Eventually, the board said, “Hell, you’re right. Let’s do something different. Let’s call Kaz and put these other guys back on the board.”

That’s an interesting game. I think you’re absolutely right. The bull case for Opendoor is obscene. Just obscene. It’s hard to exaggerate how big this company can be.

But companies require good stewardship, operational excellence, and aggressive execution. There are companies that can be run by professional managers. If you’re a widget factory, professional managers are great for you. If you’re a software factory, professional managers are death, right? I’m no one’s idea of a professional manager. I’m just not that dude.

So we’re going to create alpha.

But are you a software factory, though? You have a lot of real assets on your books. You have a lot of illiquid assets on your books. This is a very real-world, financial-mechanics business in a lot of ways. It’s not a software factory.

I disagree. I fundamentally disagree. Opendoor is a software company that happens to have some assets.

I think there’s an asset-light model here that could work incredibly well. There’s an asset-heavy model that can work incredibly well, and we’ll have literally all of them.

I fundamentally believe you must judge companies based on where leverage comes from. You look at a company and say, “That’s where leverage comes from.” The leverage from Opendoor will come from software.

We will build excellent software products for buyers, sellers, and owners of homes, and we will attach services to those products.

And one level down, do you think excellence is the key? There are 2 things you do, maybe 3. You have to be able to predict the price of an asset. You have to have a top of funnel to drive in buyers and sellers. You have to have a really big AI brain to figure out what these assets are worth.

Then you have to have transactional efficiency to make all of this happen: the processing, the selling, the repairs, all of that. How do you think about those 3 things? Maybe those are the software factory you’re talking about. Which do you think is the hardest part? Which do you think is the most important part?

I think the last part is very software-enabled. Generally, I’m 24 hours and 12 minutes into the job, and I’m generally impressed by the last part of the business already. It could be significantly more software-enabled and significantly more AI-enabled, but I’m actually generally impressed.

The long-term leverage for this company comes from offering a fair price for a home, not trying to make all your money on buying homes at a discount.

No one wants to sell a home for a fair price and add value-added services on top of it, right? Add things that homeowners and home sellers want as part of that transaction, and you’ve earned the trust because the price is fair, the house is good, it’s valuable, and you’re not selling lemons. You’re offering, “Hey, we will get to a point where you will buy a home from Opendoor.” I’m not saying tomorrow, but you will be able to buy a home from Opendoor and, if you don’t like it, you’ll be able to return it. We’ll get to that point.

We’ll get to a point where you’ll be able to buy a home from Opendoor, and we’ll stand behind it for the life of that house. You’ll be able to treat it as though it’s a guaranteed asset that we will take care of—and take care of you as a buyer of that house. As a seller of that home, we’ll be able to get to a point where we will find you a new home. You don’t have one? Great. We’ll find you a new one. We will find a way to make the liquidity work in the long run, and we’ll focus on both buyers and sellers.

For what it’s worth, this isn’t uncharted territory in the public markets. Carvana has a very good model of this, where they offer a fair price to the buyer, a fair price to the seller, and make margin on added services.

Well, first of all, I’ll make a big-picture comment. I think this is one of the hardest business models out there. I totally see the upside, but when we look not at Opendoor but at some of the others, this is one of the most challenging problems I can imagine, right?

You made the Carvana analogy, and it’s fair. As people know, Carvana was another high flyer in 2021 that plummeted in 2022, selling cars and doing the same thing for cars. Investors, including Thrive, held and bought some more stock well north of 10x. Everyone looks really smart who held and bought, and obviously, if I was you, that is the mental model I’d be pushing all day, every day, right?

But the core—I always have one of my many nagging worries on this—is that it’s easier to price a car than a house, right? The nuance you have to get right. I would question whether you can get within 7%, 8%, or 9% pretty accurately on comps, but it’s the little shit that, when you walk around the house, you think, “Oh, they have an X. That garden’s nice.” It seems to me the last 8% of the price, which is where you make all your margin, has a lot of variety. I could be wrong, but that seems to me the hard part of pricing. I think you are wrong.

Cool. I think you were right 3 years ago. I think 3 years ago this was an incredibly difficult problem that required human beings to visit homes, look around, and look at the shape of the garden. That’s the real thing. If you’re on one side of the street that has a slope and the other side doesn’t have the slope, those 2 things are priced differently. But there’s a reason why God invented AI.

This is a solvable problem today. We don’t need to limit ourselves to what human beings can see.

Yeah.

This is a real thing. Human beings are variance-creating machines. We can build software systems to solve these problems. Do you know what the funny thing is? Shopify is notoriously a bad business. It’s a high-churn business selling to small businesses at a massive discount.

Look, I made a decision: you could buy Shopify for $1. It’s a SaaS product you can buy for $1, right? There are no seats. There’s no seat expansion. Shopify is the best deal in SaaS. When you buy software, it’s the best deal in SaaS. Shopify makes money when you succeed, from services in Shopify. Opendoor will be the best deal in buying and selling homes, and we will make our money by adding value to those homes in ways that other people cannot because they don’t understand a home. They don’t understand buying. They can’t underwrite the risk. They can’t provide additional services.

Agreed.

The problem with having to make all your money in one transaction—

Correct.

—is that, by necessity, you have to be shady. This is why used-car dealers—people who sell used cars—aren’t typically awesome people, because they have to make all their money in that one half-hour. Whereas, if you make your money in the long run from a long-lasting relationship with the counterparty, you have an incentive to do right by each other.

This is actually a key differentiator. We will create a network for buyers and sellers—homeowners, home buyers, and home sellers—where they will have a long-term relationship with us over a series of products we will launch. Some of them will be free just because they’re good. Some of them will not be, and we’ll make money on them.

I totally get it. I mean, title, mortgage—there’s a whole bunch of things that happen right there that are wildly profitable on top. Title insurance is the world’s best business because no one ever pays out a claim, right?

How do you think about the role of the estate agent in this? That role is changing also, and obviously it’s 6% in a world where you’re making 8%. How do you think about how estate agents fit in?

I think there’s a structural issue that we need to think through, which is this: transactions and relationships that have many intermediaries are typically not great ones. They’re just not awesome transactions, usually. You usually want to look the person you’re dealing with in the eye and deal with them, right? That’s the real thing.

Now, do I think there’s a place for experts to help either side? Yes, in some cases, but not in all cases. In some cases, you will. I’m not dogmatic about this.

Okay?

But I do insist on an excellent product that a buyer and seller can use and be proud of. If some set of those buyers and sellers want to have someone else help them, great. I’m fine. I helped my mom sell her Amazon Prime account. It was fine. She shouldn’t get a worse experience from Amazon because of that.

But we will provide an excellent service to buyers and sellers. If, in some cases, we have other people involved, that’s fine.

Cool. I didn’t really fully get the Shopify analogy you just made, but Shopify is only 25% of its revenue from software, and that’s going down, right? It is the cheapest. It’s one of the best deals out there, right? Maybe ChatGPT is a better deal, but they’ve got to earn the—you had to earn the other 75% through merchant services and others. You had to earn it.

In 3 years’ time, will you have more of an asset-light model, or will you be more of an asset-heavy model? Do you have—

I mean, the company doesn’t have an asset-light model right now. So, yes, by definition, yes.

Well, no, but you can choose to stay asset-heavy, or you can choose to transition to asset-light.

We will not choose to stay in a solely asset-heavy world. Look, guys, I don’t have a magic plan that I’m executing against right now. This is 24 hours and 19 minutes now, but we are going—

More minutes, at least. The last 19 have been wasted. He’s mentally saying to himself, “But keep going.”

I think what you need to do is look at the problem space and solve all of it for the user. You need to look at the problem space and solve all of it. It’s incredibly hard to build good products. Therefore, it is incredibly hard to build good businesses if you’re solving the tiniest problem that you happen to think is profitable.

By the way, if you do that, you end up with adverse selection. The market eventually clears properly. If you think for a very long time that you’re going to have a 20% margin buying homes cheaper than someone else, you’re just straight-up dumb. The market eventually clears appropriately.

What you need to do is be incredibly efficient on the first transaction and incredibly valuable on every transaction thereafter. That’s our job here, and we’re going to get there. It’s going to take us a second. We’re going to screw a bunch of things up. We’re not going to be perfect. We’re going to launch things that won’t work, but we’re going to start launching things.

Jesse, how did the comp discussion go? It’s quite notable that you aren’t taking a salary, pretty much, or $1, or whatever it is.

I would take less. I’m not allowed to take less than $1. I would gladly. When you hit $30, you get remunerated a lot.

Yeah. Look, if I could make it the following way, I would have made it the following way. I think corporate executives should basically only get paid in options. I think it’s a very weird world where we create an incentive for corporate executives to be bad at their jobs just to get paid.

It’s just so weird when the thing you’ve created is being inoffensive enough not to get fired. That’s very weird. I would have gladly just—

I own no RSUs. Zero. Actually, zero. Literally, the entirety of my performance money is based on the stock price going up and the different cliffs. Some of it is a phantom option where, if the stock price goes down, it’s worth zero. Some of it vests based on price.

This is complicated because there are Delaware and SEC rules, but I would have gladly taken just options. In fact, that was my preferred state, but we have tried to construct a thing that looks like just options, with some upside and some vesting over time based on the stock price.

But look, I don’t have Yahoo Finance on my laptop. I will not look at the stock price every day. What I will look at is delivering value for users and shareholders of this company over a long period of time. And that’s what we’re going to do.

I think this company is incredibly valuable. I thought when I joined Shopify that the market misunderstood the opportunity. I think the market massively misunderstands the opportunity for Opendoor, just by an order of magnitude.

Can I ask how important you feel it is that you have Keith and Eric coming back into the fray as well?

Wouldn’t have taken a job. Would not have taken the job without them. Straight up, I wouldn’t have done it. In fact, I said I wouldn’t do it without them.

I love that.

We’re going to do things that look odd, but what they look like isn’t important. What they are is important. How things look is less important than what things are. We need people who will be with us while we take those risks.

I think that’s super interesting, and it actually gets to the whole private-versus-public thing, right? There was a great article in The New York Times recently on basically why being public is a pain in the butt, right? I think you’re exactly right, Kaz. This is a refounding of the company, and if you didn’t have that air cover from those guys coming on, it’s just really hard for a standard corporate board of a public company to take the kind of pain you’re going to see in the next 12 months to turn this thing around, right?

I would agree with you. If I were advising someone as a CEO to take this on, your first question should be, “Where is your air cover from entrepreneurial people?” Not the kind of people you find doing a public board for $200,000 who will allow you to do this. Because if you don’t have that, invariably you’re like the poor guy at Intel: You have this big, ambitious plan, you tell everyone on the board you have this big, ambitious plan, they all agree, and then one year in it’s like, “Oh my God, it’s terrifying. We’re out of here.” I think you’re exactly right, and it’s a rare combination to be able to fix the—to be able to, I won’t say fix, to orient the board around the task at hand.

Yeah, dude. We have great board members, but I view them as colleagues and coaches. There’s a board member coming into the office in a couple of hours because we’re going to go through literally every house we own, line by line, together. They’re going to sit next to me, and we’re going to go through everything.

One of the first things I did was get a board member and go through literally every invoice the company had paid for the last 12 months. Actually, every single one. I’ve reviewed every line item. Our board members are going to join us along this mission, and they’re going to work as hard as we do. I’m incredibly excited.

Do you want Tobi to come back into the fray?

Dude, the world can’t have enough Canadians, man. It’s a real thing.

The man wins. Kaz wins. That was a killer answer.

Final one from me, and then we’ll let you go. When you made the decision, you had a huge amount on the table at Shopify, and you left that for this.

Yeah.

How big is it?

A few hundred million. Exactly.

Do you believe that you will make more here, and what is that upside here when you made that evaluation?

I mean, look, I don’t optimize my life for money. I just don’t. When my wife and I got married, we agreed we would optimize our lives for leaving a dent on the world. We will leave a dent on the world, and that’s what we optimize our lives for.

Having said that, I’m incredibly bullish on the bet. I would not have taken it if I didn’t think it was going to pay off. I’m very bullish on the stock. I’m very bullish on the company.

I think we will build a generational company here, and this is my ask of everyone who has bought the stock and is cheering us on: We must be held to account for doing that. We must build a company that makes the world a better place, and we must deliver shareholder value. Both of those things are incredibly important. I’m genuinely asking people who have bought this stock in the hope that we will do the right thing to hold us to account and call us out when we don’t do the right thing.

Look, this man has put his money where his mouth is, and that’s pretty damn impressive. You’ve walked away from an exciting, challenging, extraordinarily well-paid, very manageable, safe gig to double down on, as I say, a truly challenging problem. It will be fun to watch you figure it out.

I optimize my life. At the end of every week, I write down a note to myself: Was the week hard? Was it valuable? Was it fun? At the end of every week, I judge my week based on hard, valuable, fun. That’s a goal every week: to have all 3. I find that you can’t have fun unless you have the first 2.

I had my first all-hands yesterday in the company. I told the team that we would value this company’s next year over hard, valuable, fun. I don’t know where the offices will be. We’re going to figure that out this week. We’ll announce them to the company on Monday and to the world on Monday.

Opendoor will be the most aggressive, innovative public tech company. If you are a builder, if you want to build a future that is better for homeowners, that tilts the world toward owners rather than renters, find us. We’re going to build an exceptional team and ship exceptionally fast. My DMs are open. Send me a DM. I don’t know where the offices will be. I will know on Monday, but we’ll figure it out.

Kaz, you are a hero, man. Thank you so much for doing this on your second day. You are a hero.

Thanks, guys. Have a great day at work.

Good job.

Back to work.

Bye, dude. Have fun. I want to start with Oracle and OpenAI. Oracle touching $1 trillion. Rory, you’re always quite good at setting the scene, and I got some good feedback the other day, which is that we need to set the scene for the stories. People love the analysis, but they like to know what actually happened.

Can you help set the scene for Oracle and OpenAI? What happened? I think it is good feedback. We often forget; we dive right in. So, step back.

What happened is Oracle announced—I want to say it was its Q2 results. They have an off-year; they have an off-cycle year, so it might have been Q1. For what it’s worth, they were actually a little light on the quarter’s results, but they announced future RPO, or revenue performance obligation, of more than $300 billion.

In other words, they said that, in the future, they have orders as of now that they have to deliver for well north of $300 billion. Most of that—and they didn’t say this, but you figured it out—is a big OpenAI order for around $300 billion of future cloud compute for its AI platform.

The stock exploded. The stock went up 36% or 38%, I think, briefly making Larry Ellison the richest man in the world. Yay, Larry. Oracle touched $1 trillion. It’s an unparalleled move for a top-10 company to jump by 38% in one quarter—a huge jump. Since then, there has been some skeptical commentary, but that’s what happened.

How did we analyze it subsequently? How did you feel when you read it? As you said, there were skeptics. How did you feel?

I suppose I was a little skeptical. I mean, positive spin: If you believe the revenue—$300 billion, let’s say over 5 years—and you think OpenAI has $300 billion to spend, plans to spend it with Oracle, and does, in fact, spend it over 5 years, that’s $60 billion of revenue a year—$300 billion divided by 5. Even at a 5- or 6-times revenue multiple, that gets you to a $300 billion delta in market cap, which is exactly what happened.

If it’s 100% money-good, and you can say it’s some kind of recurring revenue thereafter, then it roughly corresponds to the increase in value. It’s just that then you say to yourself: The customer promising to give you $300 billion is doing $12 billion in revenue, has raised—I can’t remember—$40 billion total to date, is still losing significant money, and therefore is going to have to raise, as Sam Altman has said, a couple hundred billion dollars. He said $115 billion, but who’s counting? That’s what it will take to be able to pay you that money.

You look at it and go, “This is a very leveraged bet on everything at OpenAI working.” To some extent, it’s like a proxy for OpenAI stock. I can’t buy OpenAI in the public markets, so if OpenAI is successful, it will have $300 billion. If it has $300 billion, it will give it to Mr. Ellison. If you give it to Mr. Ellison, Oracle’s stock will go up.

It feels plausible, but not risk-adjusted. You’re applying 100% certainty to 2 or 3 things, each of which has a fair amount of uncertainty associated with it.

My real question was: Have any of us given up caring whether any of this revenue is profitable? I think the bet that OpenAI can come up with the money is a reasonable one. They may not; there’s some risk there. But so far Sam’s found a way, right? I mean, Jesus Christ, everybody these days says they have insatiable demand. So that’s okay.

But no one cares that this adds nothing to Oracle’s bottom line and may never. No one cares that Oracle is basically a fungible set of server services for folks who don’t want to bother to bring it in-house.

I mean, for the foreseeable future, this is a zero-net-margin business for them, isn't it? No one cares. It's good for VCs: we don't have to look beyond the top line. It makes our lives much easier. We don't have to worry about silly things like inference costs and gross margins, because the public markets don't care anymore.

Well, a couple of things. You are right: from the perspective of, say, 5 or 10 years from now, the people selling cloud computing to the people who own the models probably will not make as much money as the people who own the models. That's a pretty obvious statement, right? Even though OpenAI is losing a lot of money now, being a commodity provider of services to them seems, over the medium term, like an even more commoditized, less profitable business.

So, at some point, you agree that I'd prefer to own the model company. I'd prefer to own OpenAI than CoreWeave, and to the extent that this is just CoreWeave, too, you're totally right, Jesse. Now, the fun thing is, we are dealing with the man who most successfully on the planet extracts gross operating margin from software companies, which is Larry Ellison. That's why he owns 41% of this damn thing, up from 27% a decade and a half ago, right? Just getting that free cash from recycling it back.

But you're right: this does feel like you're getting $300 billion in revenue and, to your point, Jesse, in return for which you have to spend a whole buttload up front on capex, in return for a business that, however profitable it will be—and maybe it is profitable—it won't be as profitable as the 41% operating margins that they currently get from selling databases.

Nothing today, is it? I mean, literally, this business is consuming cash.

I'm sure. Well, yes.

Yeah.

Yes. And I'm trying to recollect from—and you probably have some kind of positive accounting gross margins—but really, it's all about the assumptions you're making on your capex depreciation. If you know over how many years you should depreciate the latest NVIDIA chip, then that would be the key question. But you're right, it's on a cash basis.

This isn't like a bunch of memers getting excited about CoreWeave. This is Oracle. This is a company founded in the '70s, where the public markets are like, "We don't care that your new GPU-hosting product, which has massive top-line growth, is contributing nothing to the bottom line. We don't care. We don't care even an iota," is what the public market said, didn't they?

Agreed. No, you're right. It is, Jesse. You're exactly right. It's just like—I think it's equivalent, but with more success than the Facebook/Meta story. You have an existing business. It's freaking awesome. It's got nothing to do with AI. It kicks off 41% operating margins for Oracle, high 30s or 40s for Facebook.

The market is saying to you, "If you've got all that free cash flow, have at it, big guy. Throw it back in. See how it goes," right? In the case of Meta, they're not penalizing him. In the case of Oracle, they're actually rewarding him for doubling down on unprofitable growth. So, entirely rationally, he's doing it right. And it's got to—

It's not that great. One thing, one last thing: it's got to feel freaking great when you've been cranking like that guy has for 50-plus years. Whatever it means long term for the stock, the fact that you sprinted ahead and, for a brief shining moment, were probably the richest man in the world, that alone is worth whatever future damage you've done to your operating margins. It's great.

Is this not an ultimate sign, though, of where public-market irrationality and exuberance are? Sam Altman makes this announcement, and with the margins associated with this order, it jumps 38% at this scale. Is that not the height of irrationality?

You mean, unlike us, sober, careful, and sober-minded private-market VCs? I'm shocked to discover gambling going on in the stock market, Harry.

Yeah, it's pretty frothy, and I think you're dealing with—look, there's been froth in the private market for AI for 2 or 3 years, and probably the public market is saying, "I'd like to get in on the game," and this is one of the few ways to play it. So, I agree, it's frothy. Is it more frothy than any of the things you're seeing or I'm seeing? I don't know. Who am I to judge?

The one thing I do think is important—we mentioned the margin element—I have never before seen such a lack of investor diligence on anything except top-line revenue growth. Ever. Astonishing, where it's not included in updates: hundreds of millions of dollars wired without a discussion on margin, number 1.

And then, number 2, it's just like growth is amazing until it doesn't grow anymore. And when you look at this, for OpenAI to pay this, if they 2X between now and July 2026, and then 2X again between July 2026 and 2027, they'll be at $48 billion and still $12 billion short per year to do this.

At some point, the growth does taper.

Yes, agreed. Do I think they're going to collect $300 billion in orders from OpenAI? Absolutely not. I don't think that turns into all of that. I think they've got a business. I think they'll get more revenue from OpenAI.

Do I think the full last $300 billion will be wired in 5 or 6 years? No. I very much doubt it. But—

Is this market not just all about shouting as loudly as possible? Do you remember Stargate? "We're going to have $500 billion." What happened to that $500 billion? It seems like there's a lot of cynical questioning. Isn't this one of those situations where everybody just announces the same thing?

It used to be SoftBank, but wasn't Oracle part of that? And maybe now this is—

Larry was right. Larry was—it did happen. This is a big part of it.

Stargate. This is Stargate.

Honestly, to me, it's a reminder. We've talked about this: really scrutinize Sam to see the future, because he goes to Trump, right? And he's got Son from SoftBank and Larry Ellison in that awkward photo. Musk wasn't there when he was still running things, right? This was a couple of months ago, announcing the Stargate thing.

I'm looking at this uncomfortable Larry Ellison, who hasn't put on—this doesn't look like he's put on a suit in 3, 7, or 8 years—sitting there squirming in the White House. This was—we should have bought Oracle stock that day. Why didn't we put the whole fund into Oracle? Why did Larry—and why did Sam make him come, right? I mean, you've got to do this, right?

That was a telling moment to me, but I didn't get it at the time. This just seemed like the oddest photo op, and it seemed like dissing Elon Musk, but it wasn't that at all, right? This was the future. You're right, Harry.

I mean, I think one of the things that's obvious from this is that everyone is getting what they want from this press release, right? Larry Ellison is becoming the richest man in the world because his stock's going up 37%. Sam Altman is getting, frankly, leverage in his negotiations with Microsoft by making it clear he has another capex provider willing to spend vast amounts of money.

So, to some extent, this is happening because everyone involved wants it to. And I'm willing to bet they want it enough that they're not sitting there going, "Are we 100% sure this is going to happen?" or "Do we just have a credible case that says this might happen, and it will be good for us in the short term and build momentum?"

Look, to be fair, especially when you're trying to do something as ambitious as OpenAI, to some extent momentum is your friend. Momentum is a necessity. And things like this just keep building on the momentum and keep building on the sense of inevitability.

And you're right, Harry, it's the role of investors to be a little more cynical and scrutinize and say, "Do I really believe that that's going to turn into $60 billion a year of revenue? Should I really mark Oracle up by that amount?" Clearly, they forgot to do that this week.

This may be a broader point, but I just feel like the whole venture landscape has moved from a game of investing to trading, in a way that we're all just hoping someone pays a more irrational price than we paid and suspends disbelief in the meantime.

Well, we've got a couple—I think we still have a couple of good years until something happens, so we might as well play the game.

Are you going to be the Chuck Prince of our generation? Remember, while the band keeps playing, you've got to keep dancing—the CEO of Citigroup in 2007. Harry, you might regret that quote, but you—

No, I think you've got to play. I mean—

I've kept dancing with Hopin, BeReal, and Clubhouse. And you know what? The party stops. I'd rather—

Yeah, but no. What happens along the way is that there are exits. So, I think the big question for venture—and I think where we're going to make a lot of mistakes in venture—is not taking billion-dollar exits over the next 24 to 36 months.

We're going to triple down. We're going to have board members saying, "Rory, I know we've got a $4 billion offer for our latest AI tagging and categorization software at 3% gross margins. Let's not. I want 8 or 12; that would be better. You know what's better than 12? 24."

And we're going to say no to those fund-returners, and we're going to wake up and they're going to be worthless. I think that's what's going to happen. If you're in the game now, you've got to have a couple of big exits and IPOs in the next couple of years or quit the game. You've got to have a couple.

First of all, you are right that that will invariably happen. But we're just at that stage in the cycle where you've got a large amount of euphoria, fairly untethered. The positive spin, which I think is true, is that you don't make these kinds of technological leaps and technological investments without a fair amount of accompanying euphoria.

This is a means to the end of moving the needle forward on AI. Right? We have to try a lot of stuff. A small amount of it will work, a large amount will fail, and a lot of money will be lost in the end. But the good stuff will ultimately outweigh the bad stuff. This is the way it happens.

But to Harry—your point in particular—I’ve been around long enough. I remember a bunch of those companies in 1999 and 2000, right? You got the offer for $2 billion for some fiber-optics company or some communications-equipment business, and the board said, “Damn it, we want $8 billion. We’re turning it down.” Then you’d meet the team 2 years later, and the companies were worth zero. Most of those folks were shell-shocked. It was like winning the lottery and then losing your ticket. It was brutal, right?

So, I think being canny and shrewd about what chips you take off the table in the next couple of years—I think you’re exactly right, Harry—will be a key part of the game. I also think people mistake valuation for liquidity. Just because it goes up doesn’t mean you can get out, and if you can get out, it’s often at a discount to that price round. If you can, it’s often in a strip where they’ll give you 10% or 20%, but you can’t get out as much as you could by selling it all now for $4 billion. I don’t think people think about that enough.

True. And I think one of the things we’re going to discover is that when all this action was happening in the public markets, you had this weird, somewhat positive phenomenon, which is everyone was up with constant liquidity, right? You can buy in at $80 a share, it goes down to $70, and you can say, “I was wrong. I’m out.” Someone else thinks, “I think it’s going to be okay at $70,” and they go in, and they’re wrong, and it goes to $60.

You can kind of parse out the pain among various investors who come in and out of the stock. The interesting thing on the private side is that you’ve bought in high, and the whole point of private is there is no meaningful liquidity. If you’re wrong, you’re going to own it all the way down.

It’s going to be a lot more fun on the upside and a lot more pain on the downside. When you don’t have liquidity to fulfill its role, part of the role of liquidity is to allow you to alleviate risk. That’s just the game we’re in, and that’s just the nature of being private for longer, with bigger bets.

Every single founder—

—that I met this year who has a strong M&A offer, I’ve told them to take it 100% of the time. Now, I don’t necessarily want them to take it, but I don’t want to be the guy saying, “I want to double down and quadruple down.” I don’t want to be the guy who, a couple of years later, finds out it didn’t work out. So, I’m telling them to take it.

If they come back and say, “No, I’m confident,” here’s my new heuristic. It’s so simple: “No, Rory, I’m confident we’ll be worth 10 times as much. No, I’m confident we will IPO. I’m confident.” Kaz just left hundreds of millions of dollars behind at Shopify. Okay, right or wrong, that dude’s confident. We just heard it, right? He knows.

It’s not all about the money, okay, but he’s confident this was the right decision. But I want to tell them the opposite. I’m telling every founder, “Take it. Take it. Take any massive offer.” And I want them to come back and say, “F*** no way. F*** no way. This is going to be bigger.” I don’t want to have any of those regrets.

I think it’s the right thing, and I also think telling them to do that makes sure that I don’t screw it up as a GP, right? It’s just so easy when times are good to say, “Oh, Oracle’s going to get another $500 billion, aren’t they?”

Totally. And you know, I don’t love your sentence, but I actually think it’s right. Your advice should, to some extent, be tempered by the time, right? The Bayesian prior at a time like this, when valuations are at an all-time high, should be some version of what Jason said.

Now, it’s still possible that you’re the 1 in 10 for whom an all-time high is just a step on the journey, and maybe you should play the game out because you’re so confident it’s going to be amazing. But you’re right most of the time.

Look, it’s very noticeable that 3 or 4 times in the last year, you’ve seen significant M&A driven by founders at the same price at which VCs had either just invested or were about to invest. In other words, that’s a founder saying, because the VC is investing, thinking, “Oh, you’re worth $2 billion. I’m going to give you money at $2 billion. I think you’ll be worth $6 billion.” Then 3 weeks later, the founder is saying, “You mean I can actually get the $2 billion? I’m out of here.”

To some extent, that happened at Scale AI. I think there was a pending offer on Windsurf, and those founders may be listening to you and being shrewd, saying, “I could be in a local maximum here. This is a good time to take the chips off the table.” There’s a little bit of information in that about how much, dare I say it, more shrewd they are about the value of the asset than the investor who was about to write a check at the same price.

I completely agree. I also think it goes back to incentives and what we said last week with Jeff. Investors have many options, and we want to ride them as much as possible, whereas founders have one. It’s their whole net worth in it, and that drives a lot of decision-making, I’m sure, tied to that.

I do want to be cognizant in terms of how we discuss the topics we have on the agenda. We mentioned Oracle and OpenAI. Microsoft and OpenAI’s relationship is slightly changing, it would seem. On the announcement, Jason, why don’t you give a snippet on what this is in terms of the news between their relationship and how it’s changing?

We have more to learn. It’s going to be interesting. Microsoft, I think, today said they’re moving not just parts of Office to Anthropic, but that Anthropic is the default choice for several of their products. Microsoft also said that several months ago they told their teams to start using Claude Code. So, they’ve been breaking up at some level for a while.

I guess it’s fine. I guess it sounds like—we don’t know the details of the deals—OpenAI is going to get some of what they want. They’re going to get this revenue share reduced, and they’re going to get the freedom to partner with whoever they want. I’m not quite sure what the price in blood is going to be the other way.

But Microsoft’s already moving on, right? They already have the IP. They’re going to keep the IP, I guess, right? They’re going to keep whatever IP they’re allowed to keep before AGI. So, they’ve got all the code if they want to do anything with it.

They’ve already moved on to Anthropic, which is good because I think ChatGPT is probably going to end up being almost as good for coding as Anthropic. So, the whole thing—the shifting sands of AI—it’s a lot to process.

Yeah, it is. And it’s worth pointing out that it’s some kind of interim move. It’s not a final, done deal. It’s both sides saying, “We’re making progress here toward where we’re ending up.” But you’re right, Jason: this is moving on. This is consciously uncoupling here.

If you look at what the endgame is, it’s some version of Microsoft saying, “OpenAI had this weird structure for reasons we all understand. Microsoft ended up with this weird investment that, in my view, gave them a fair amount of blocking rights in a lot of different ways. It gave them access to 49% of all the profits to a certain amount. It had revenue share. It had a lot of things that were effectively a bit of a poison chalice for OpenAI in terms of making it a real, proper standalone company.”

What’s clearly happening now is that it was a marriage of convenience for a while. Both parties are moving apart, right? I’m not sure where it ends up, but the rational endgame for Microsoft is: “We got something along the way, but we don’t need the model. We don’t just need the model anymore. Anthropic looks to be more useful for this. We still have access to their model, and we like that, but we can buy that on a third-party basis. Just give them money. Be a customer of the model.”

We got some lift from AI in the short term. We have some business with them as a hosting provider for Azure. All those are good things, but in the end, my guess is they convert that $13 billion investment from a blocking kind of thing to a 20% to 35% ownership stake in OpenAI. If OpenAI is worth half a trillion dollars, Microsoft will have put in $12 billion and it will probably be worth $100 billion to $150 billion.

It’s a 10x venture return on $12 billion. In one sense, it’s a great return, but as you guys know, we’ve talked about this: it doesn’t move the needle when your market cap is $3 trillion. You don’t get paid as a large-market-cap company for making, oddly enough, $100 billion, because it’s like you’ve got a $3 trillion market cap. It doesn’t move the needle.

When you zoom out 3 or 4 years, I think the conclusion will be, “Wow, that was an interesting investment. We got a bit of a lift from Microsoft’s perspective. We made a lot of money, but it doesn’t move the needle. We got some buzz on AI in the short term, but we didn’t really get what we wanted, and we don’t have what we need in terms of AI. We’ve probably got to keep cranking on that 3 years from now.”

How do you think the relationship will look between OpenAI and Microsoft, and then Amazon and Anthropic—the pairings that we’ve seen so far?

I think it’s clear that Microsoft will be a large shareholder and a commercial partner. Hopefully, OpenAI for them will be both a customer and a vendor, because they will be selling Azure—they will be selling cloud capacity to OpenAI—but not on an exclusive basis.

They will be buying access to the models from OpenAI, but probably not on an exclusive basis, because they'll also be moving to Anthropic. Whether there's a perfectly normal relationship between the largest software company on the planet and this entity that they helped form, fund, and start, where it's now grown up and left the house, they're no longer dependent. It's just a perfectly fine, arm's-length relationship with a massive equity ownership that's probably going to make them $100 billion to $200 billion, which is a lot for anyone else.

I mean, the lift—the real value to Microsoft—has been the lift in its perceived market cap and its actual market cap from the perceived AI buzz in these 2 years, where frankly they had nothing. I think the real question will be—and to put the heat back on the Microsoft team—when OpenAI finally pulls away and you can't rely on your complex, proprietary agreement with them to get access to the AI you want, have you built your own AI that matters? Have you done something? If you have, that's great. If not, then you missed that market. That's where that one will be.

I don't have as good a sense of Anthropic and Amazon. I'm not as close to it or as informed on it—maybe I'm not close to either—but my guess is much the same. Often, when a big company partners with a small company, the small company gets smothered. In both these cases, the small company won: they got the money, the critical mass, and the credibility, and now they've pulled away.

Neither Anthropic nor OpenAI needs Microsoft or Amazon at this point. They don't need them for money, because the industry will give them infinite money. They don't need them for compute, because Larry Ellison will give them infinite compute in return for money. They may not even need them for distribution. In fact, both of these companies, ironically, just like Microsoft 30 years ago used IBM and then left it an empty husk, have used their large corporate relationships and gotten the value out of them.

It's still a bit sticky because the agreement is weird, but fundamentally they've made it. They're independent, standalone companies. “Thanks for your help, guys. Here's your equity position. Call me for the IPO.”

And Rory, thanks for your description of an empty husk with Scale AI. I had them in my inbox this week asking to come on the show and tell us why they're not an empty husk. Thank you for that. That'll be coming to 20VC soon, which should be an interesting one.

We all did think for a long time that it was a weird deal. Microsoft buying 49%, in essence, of OpenAI was the first of those deals—the style of deals that Scale AI, Windsurf, and others have done. It didn't turn out that way. They didn't leave a husk. There were some superficial similarities, but in the end, we all thought—in fact, the first time we did the show, we probably would have still said that Microsoft basically acquired OpenAI, that it was an acquisition in disguise. As it's turned out, that's not the case—not remotely the case.

Kudos to Sam Altman for dancing his way out of one of the greatest bear hugs of all time. He basically had to sell his company to Microsoft to get it off the ground, and now he's going to get out of it. He's going to get out of it—wow. No, you're exactly right, Jason. In a way, that's because when the founder goes with the acquirer, it's a bear hug. It's an empty hug, an empty husk. When the founder stays independent, like Sam, it's a stunning win.

I think you're exactly right, Jason. You people—I mean, it is Paul Graham. You've got to give him credit. You could put him on an island with a bunch of cannibals and he'd be king. You put him on a plane to Seattle, and he came back with a bunch of money, and now he's king.

This is what I find funny. We go back as the Oracle and OpenAI skeptics, and I'm like, “Fuck it. I'm not betting against him.”

Yeah. No, he might find some of that money. I mean, do I think Oracle will book a fair amount of cloud compute from OpenAI? Absolutely. Maybe not $300 billion, but if he's willing to provide it at a lower price than Microsoft, my guess is those fine people at OpenAI will take it.

Another thing, when we had Cass here, with him leaving $200 million behind, it's difficult to fully predict the outcomes of these things when you have post-economic people, right? Not only did Sam not leave the husk, but his lack of equity in OpenAI meant Microsoft couldn't pay him enough to move over and dehuskify it, because he had no equity to husk, right?

I mean, Cass may make a billion. Harry was kind. His job is to make a billion dollars from OpenAI, right? But even risk- and time-adjusted, leaving $200 million behind at Shopify and another $70-something million in what he might still get—you can only do that if you're post-economic. I hate it, especially when some guy who was a senior marketer at some company and spun out with a couple million bucks tells everyone he's post-economic. But Sam and Cass are taking post-economic actions.

It's very interesting in today's world, where a couple billion isn't very much.

In market cap—in valuation, not in personal income. Yeah, a couple-billion-dollar seed round is barely going to make Harry's show.

We're going to dive into the application layer, and there are 2 that Jason is passionate about. I'm really excited for this, actually. One is Higgsfield, and the other is Replit. Jason, which one do you want to start with?

Either one, man. It depends what you want to talk about. I think there is one thread—the Higgsfield-Gamma thread, right? I just think it's somewhat interesting that both are slightly under the radar. The CEO of Higgsfield was not quite complaining, but sort of shouting this week that he had gotten to revenue, and certainly to users, even faster than Lovable and friends by doing quick video.

I've been a small investor, and I've been a user since it launched. I love Higgsfield. Gamma is at $60 million this year—Gamma going from $0 to $60 million this year. That's pretty good for slides.

Just to be clear, Higgsfield is an AI video-creation company that raised $50 million and also announced $50 million in ARR in a faster time frame than Lovable and Replit.

Yes. Maybe Gamma is close too, right? We could talk about Higgsfield being under the radar if you're not a creator using the app. People on this podcast may have used Gamma more often.

My meta point is, my God, if you don't see some of these numbers, it's not just Cursor and friends, Replit, and the others, which we could talk about. Who would have thought that? Going back to Cass's case, homes are a big market, restaurants are a big market—e-commerce, like Shopify. Sometimes we don't even realize that short video and slides are massive, just massive.

The fact that Higgsfield can do this in the shadow of Google, even though they're using Google models, and that it can do it in the shadow of so much competition, is just incredible. This is why it's so hard to do a triple-triple-double-double. It's not just Lovable. There are so many Lovables. There might be 20 or 30 Lovables.

What you're seeing here, stepping back, is that with AI, there are a series of things that “normal people,” ordinary people, couldn't do before, whether it was coding or video creation. These tools are making it accessible to everyone. You have this step-function—10x, maybe 100x—increase in accessibility to creativity or coding, and that's the positive momentum.

The negative momentum on all these deals—Replit, Higgsfield—is, “Oh my God, you don't control the underlying model. If you don't get enough compute, will you have customers?” There's a lot of risk in those deals. But sometimes it pays to zoom out to the big picture. The big picture is that anyone with internet access today can create cute videos, edit videos, and be creative in a way that 5 years ago you couldn't do unless you were a trained special-effects editor. That's huge.

Maybe 7 months ago. It's your exact point. You couldn't do these things, and now you can do them for pennies. You can do them for pennies. It's beyond disruptive, right? Anytime you're dealing with something that everyone can do, you have the potential for these exploding growth rates.

You guys are living it with Replit and Lovable in coding. I think Higgsfield is an example of that in creativity, and I think there'll be more of them. Sometimes I'm just shocked by venture. Sometimes I look at some of these things and think, “How the hell could that get to $60 million this year?” Do we really make enough slides?

But to Rory's point, it's a way to create content you couldn't create before AI. I get it, but so many of these markets are bigger in the age of AI than they were pre-AI.

I know it’s Captain Obvious, but Higgsfield for short videos and Gamma for slides. I even like Opus Clip—it was the first little AI investment I did. I got it, but I didn’t think it could add up to so much revenue to make these clips, right? But you’ve got to get the spreadsheets right. You can’t use the TAM spreadsheets from 2021.

I find this time a little bit like COVID, though, in terms of market forecasting, which is a real difficulty in understanding what is a sustainable market trend that will be meaningful and enduring versus what is an experimental market that is cool to create but ultimately whimsical and doesn’t last a cycle. I’m finding that uncertainty very challenging as an investor, to be quite honest.

But the thing is, we agree, but that’s a conversation we’ve been having since the beginning of AI, and a lot of thin-wrapper apps died for that reason, right? What we are seeing—it’s a valid concern for venture—but we are seeing, at least, is that nominal NR is pretty high in these apps.

So it may still crash and burn, don’t get me wrong, right? But if you’re using Higgsfield and Gamma and Lovable and Replit and you’re seeing triple-digit NR, even if it’s not the NRR we used to talk about for B2B, it’s hard to say no as a VC, isn’t it? It’s hard to say no when you see 140% to 180% revenue. You can say no and just sit at home and knit.

I don’t know, dude. When the margins are where the margins are, you have to project out and go—

But Higgsfield is profitable, is cash-flow positive. It varies based on the application. It varies, basically.

Not all of them—not all Replit and Lovable—have negative margins. They’re not, but it’s a fair concern. It’s just, I guess my learning is it’s non-obvious what the margins are. They’re all over the place, right? They’re all over the place.

Agreed. Right, and you’re right. I mean, Harry, again—yeah, it’s hard. You’ve got to figure out which exploding-growth company is going to be sustainable and which exploding-growth company is not. But at least you’re dealing with the problem of exploding-growth companies, right?

And, you know, yes, yes, yes, but I’m an investor in a business, Airwallex, a very similar business to Stripe. This is an unwaveringly enduring, growing, strong market—fantastic. Comparatively, these others are incredibly experimental, potentially groundbreaking, and potentially whimsical. Very different.

Well, you’ve got to be able to tolerate a loss ratio, if nothing else.

Yeah, they are different.

You’ve got to be able to tolerate losing 30% to 40% of your money on it.

They are very different, and I’m not naturally good at these creative deals. I get all your points, but I tend to be in the steady compounders.

But I think different deals have different attributes. And, again, think about it from a portfolio: you’ve got to look at these and go, some of these will be flashes in the pan. The question is, what’s the distinguishing characteristic of the companies that explode and then sustain, right?

My gut—I’m kind of riffing here, and we can talk about it—is it’s probably 2 things. It’s probably having expansive white space that you can grow into, where your customers need more things you can do, versus getting cut off shortly. And then, secondly, having a founder who’s maniacally focused on doubling down and adding all the rest of the stuff.

Something you guys said about Lovable and Replit a while back stuck with me, which is you can envisage a whole bunch of ancillary products around that. So as people build their websites, all the other things it takes to make that website work, you can envisage building an economic model around the combined thing, right? I think it’s a combination of the opportunity.

I think some of these things will be flashes in the pan. To use Hopin and COVID, it will be a temporary phenomenon that goes away. But finding the ones where you have that explosive growth and then can parlay it into something enduring is going to be pretty damn interesting. We’ve never seen competition like today. Never. Never.

You used to have 6 months, maybe 12 months. You’d launch something, your competitor would look at it, decide if it was worth their time to build it, commit to trying it, then 6 months later get it. But you had a full year. Now, now you have like 2 weeks.

Everyone complains about competition, and everyone’s prepared to do anything to solve the problem except the one thing that will solve the problem, which is step away from the table yourself, Harry. So we’ve got too much competition, but you’re not quitting. I’m not quitting.

Yeah, I mean, tough luck. It turns out, yeah, it’s a very competitive time up and down the stack. It would be a lot easier if there were half the number of people, but the opportunities are compelling, and people are going to show up and try to play.

Let’s take an extreme example. I don’t think this is going to happen, but it’s not impossible. Anthropic could lose half its revenue in the next 12 months because GPT-5 Codex might be just as good. They could lose half of it. Literally, all you do is turn it on in Cursor or Lovable or Replit or a million other apps, or whoever—Higgsfield or Gamma—moves over to GPT-5 Codex instead. If GPT-5 Codex is as good as Claude Code, which just launched, I could imagine it is 95% as good. You could imagine in today’s crazy world Anthropic could lose 30% or 40% of its Claude Code revenue in one year. It’s very imaginable. I’m not saying it’s going to happen, but the switch could happen. Even that may not be stable. Going to Harry’s point, forget about whether Lovable and Replit and Base44 are stable. I’m not even sure Claude Code is stable. And to state the obvious, that would be a very different level of stability than we saw in the SaaS era, where these things lasted forever. They churned 5%.

If you’re right, Harry, if it’s even 30% to 40% probable that something like that could happen—and I don’t have a developed opinion yet on it—that’s obviously a very different world you live in in terms of risk.

Yeah, I think it could be high risk. The only interesting take I’ll make is that I interview, honestly, 3 to 4 decacorn CEOs every week, and on this topic, there are 2 interesting elements.

Number 1: they’re completely price-insensitive as to how much they spend on Anthropic and Claude Code in particular. They’re like, “I would spend 10x. Don’t even look at the line item.” But at the same time, you—

These are the hyper-growers. What about Cliff at Canva, who just said on this show he’s super sensitive to it? Just let me roll with this.

But then, second, there’s this duality of super-low switching costs and a complete awareness that they would very happily move tomorrow to someone else if it were a comparable service. This goes to explain the urgency around capex for these companies, because their belief—and so far it’s been correct—is that the only way to outperform, if you’re competing with Claude Code, is more reinforcement learning, more pretraining, more whatever, which means more capex.

Hence the insatiable demand around capex, because it’s the only way to win. All the players are being rational in the game, but you can see it adding up to something pretty scary.

Kind of bringing this all together, the Higgsfields and the Replits, the thing that could also kill them is actually the fact that Wix has Base44, which is doing incredibly well. Adobe or Canva could do what Higgsfield does and, with the existing distribution they have, do it.

Going to the distribution and going to the incumbent versus startup, how are we thinking about incumbent versus startup and the core crux there?

The market would say that the Wix acquisition has worked very well, and it’s probably going to be $50 million in ARR by the end of the year. So that’s a win.

The market would also say, just based on the fact that Adobe and, let’s be honest, Salesforce aren’t seeing that kind of growth, they’ve tried to announce AI products and have gotten some traction but aren’t able to access that explosive growth. I think the market reaction—the Adobe stock price—kind of reinforces that.

So it’s not a simple binary answer. Some folks have pulled it off, and probably—in fact, as I think of it in real time—it’s easier for $100 million to make a difference at Wix than it is to make a difference at Adobe or Salesforce: $23 billion in the case of Adobe and $40 billion in the case of Salesforce. It’s hard for those incumbents to move the needle significantly, which is why the stock prices have been down, because you’re not getting the AI explosion.

Yeah. But I think the Base44-Wix thing is interesting because we’ve been asking for a while: can the incumbents benefit from AI the same way the startups have?

We’re looking at ServiceNow kind of faking it, and we’re looking—we had Marc Benioff, who we all love—but we’re not seeing it in the numbers yet, right? And we see Palantir, but Palantir is a completely different company that is AI-first, so we’re not seeing the same thing.

We’re seeing in our own portfolios a lot of our 2021 high flyers not exactly crushing it in the age of AI. There are Dialpad, Talkdesk, and others that have, but a lot haven’t.

But what’s interesting is that you see Wix come in and buy a little Lovable/Replit clone that is just a cheap clone built by 1 guy. I mean, kudos to him, right? Solo founder, right? Bought for $80 million, using the same underlying technology, Claude Code, right?

But they bolt on what Wix is good at, which is safety and identity.

Okay. Then they bolt on the friggin’ funnel. They push it out to their base. And if that’s gone from nothing to $50 million in a single-digit number of months, imagine it’s $200 million or $250 million. More importantly, to Rory’s point, if it’s 10% market share, that’s a lot for a big company to get through distribution in a couple of months.

So, if Adobe could do that, if Figma could do it with its thing, could it be the revenge of the incumbent? I don’t see much evidence, but Base44 is one example of the revenge of the incumbent. I don’t want all of our friends to go down into irrelevance. I want to see Zoom back. I want to see everybody come back and be roaring in the AI age, not just the new guys.

Speaking of innovating as an incumbent, I’ve been sent by a lot of people—Workday acquiring Sana Labs for $1.1 billion. Wow. They were at, like, $50 million in ARR. That’s a pretty great outcome for everyone involved: $1.1 billion to Workday. I think Sana’s great, Joel’s great, but this was a second to Glean, very much so.

But, you know, in these times, being number two can be a great place for M&A.

Yeah, absolutely.

When number one is unacquirable, you get so many offers. I wish I’d realized that as a founder. I mean, knowing Rory, even if EchoSign had just done okay and we’d gotten up to $100 million, the folks that DocuSign turned down would have come and bought us for a billion bucks. I wouldn’t have had to do anything. Just being number two, I would have had to open the email and sold for a billion at a moment in time.

Today, it wouldn’t have worked. In the age of AI, you can’t buy Glean—it’s unacquirable. Lovable’s unacquirable. Replit’s unacquirable. So, being number two is great for venture. The other thing is, in frothy times, they’ll pay up, too. They’ll pay 2 or 3 times what they would have paid otherwise.

Agreed.

Number two is great when number one’s not available. Just don’t raise too much if you’re number two. Don’t make yourself unacquirable.

Can I push your thinking? Is number one really unacquirable when Scale gets bought for $14.9 billion? Are we not seeing the limits pushed for what is acquirable?

It’s got to be a hyperscaler or someone like Workday. Workday can only pay so much, Harry. Workday cannot pay $26 million to buy Glean. What’s Workday’s market cap today? It just doesn’t have the capital to make all the VCs what they want.

Agreed.

I’m not saying you’re wrong. I think everybody’s right. For sure, the hyperscalers can pay up, but a lot of folks won’t sell at any price. Being number two, just as founders, is a cheat code. Just don’t raise too much. Be acquirable if you’re number two. Be kind, and you will be shocked in frothy times by the offers you’ll get. You’ll be shocked by the offers you get being number two. You’ll be shocked.

Have you ever regretted selling, both of you?

I’m pausing and thinking: yes, there are times. A friend of mine said to me this week, “Harry, I’ve never regretted selling and making millions of dollars.”

As a founder or a VC, though? Those are different questions—as a founder versus a VC. He’s a VC.

Listen, Rory might have more to say, but in my experience as a founder, I would say more than 51% of founders regret it. I regret it. But as a VC, here’s the hubris in this: you’re going to tell a founder they can’t sell and expect them to work twice as hard after you tell them to fuck off? At the end of the day, unless you’re a total douche, if the founder wants to sell, you sell. It’s not your decision. It’s not your decision. That’s the height of hubris, in my experience.

Agreed. I don’t think you even do tell them. You ask me, “Do I regret it?” There have been times when I look back and go, “I think if we’d held, we would have compounded and been bigger.” You have objective facts that make you believe that, which still isn’t the same as saying you regret selling.

In one of the cases—I’m choosing my words in case my dear friend who is relevant to this is listening—I talked to the CEO 3 or 4 years later and said, “Our competitor is now worth 4 times what we’re worth. Do you regret it?” He wisely said, “No. I took money off the table. I bought a house, I got married, I’ve got kids, I’m wildly happy, and I’m doing another deal here. Do you want to invest? Here are the terms. My life is great.”

As an objective matter of fact, you go, “Compounding would have been good in that case,” but it wasn’t to be, and I don’t regret it vehemently. I’m just like, most of the time—most of the time. I’m trying to think: there are only a few where you look back and go, “There was a lot more compounding in it.” A lot of the time, you go, “Yeah, that was a good call,” right?

Again, I think the more salient fact is this: if the founder calls, my MO is actually very different. My perspective is not to have an agenda one way or the other. The first thing I always say to founders is, “If the liquidity window opens as a private company, you should pause and take it seriously, because most of the time it’s not open.” The mere fact that it’s open means you have to pause and think.

You have to change your game from 90% heads-down, work hard, and now someone’s made an offer. It’s now time to get real.

And figure out—

Or do you not have to apply that same mindset as an investor?

Yeah, you do.

With the secondary market.

You do. But I’m saying for the founder, it’s more significant because it’s all or nothing. You can get it all off the table, and then you have the real comfort. How do you feel about things? As I always say to them, “Now would be a good time to voice any of those concerns that you’ve been suppressing deep in your sternum here, guys.”

And, Rory, are you having active partnership discussions about much more liquid secondary markets and a trader mindset—maybe selling in a way that you wouldn’t have done years before because secondaries are available?

They are, but the truth is that, except for a small number of companies, they’re not as easily available. They’re available for a small number of companies at a discount. There’s a much larger number of companies where you have your winners and look at them and go, “I don’t know if the market fully appreciates what this is worth yet. I think I’m probably a holder here.”

I have no vestigial attachment to selling. I have no reluctance to sell. The opportunities are, again, a little like the founder window opening. Everyone likes to talk about this. LPs like to ask about this, but the truth is, the number of times that, as an individual investor, you’re in a company that’s so freaking amazing that there’s a free, liquid secondary market, where you’re in early enough and have a big enough hit to move the difference, is low.

If you have 1 of those per fund, you’re doing great. You probably have 3 or 4 winners per fund, and only 1 of those is that super-marquee, amazing one. The people who bought in at $50 billion at Stripe aren’t looking to sell at $90 billion just because it’s liquid. Conversely, the people who bought in under a billion probably are, and should.

So my point is this: yes, you have these conversations, but it’s not like you spend most of the day out there trying to keep your $10 million ARR company going, helping it get funded, and hoping it doubles. That’s a new world. Getting those companies funded is harder than ever when they’re going from $10 million to $20 million. But that’s an entirely new discussion.

Speaking of moments of liquidity, we had a $3.5 billion IPO for Via. We had Gemini go out at $4.4 billion, with a 32% bump on the first day. Then we had Figure Technology Solutions raising close to $800 million in its IPO. Gosh, how exciting—and nice to have IPOs again. Which do you think is most interesting to discuss out of those?

Can I add 1 thing? Rory will know this: the busiest IPO week since 2021 isn’t to be ignored or taken lightly. It isn’t the busiest year yet. It could end up being one; it’s already September. Next year might be as big as 2021. But getting back to where you were is a big moment in time. If you’ve ever been a founder or worked at a company that’s decelerated, getting back to where you were is a big moment.

You should celebrate when you get back. Last week, we were back to 2021 for a week. Maybe we’ll get there for a month, and then for a year. It’s a big micro-milestone to have 1 week of 2021.

I think they’re all interesting for different reasons, but I think Figure is the most interesting, just to start with that. Looking at those 3 IPOs and setting the scene, 2 of them are vaguely crypto-related: Figure and Gemini. One of them is a different kind of SaaS company, selling to governments around transportation. It has a lot of complexity, and it’s not just SaaS; that understates the complexity of the Via business.

Of the 2 crypto companies, Gemini is the Winklevoss twins, of social-network repute, and Figure is Mike Cagney, who was the founder of SoFi. The first thing you note here is that, of these 3 IPOs, 2 of them are second-time founders, depending on how you adjudicate the Winklevoss case.

F. Scott Fitzgerald’s line that there are no second acts in American lives is wrong. 2 out of 3 are second acts. Cagney is interesting because SoFi was an interesting company—very differentiated—and has gone on to be Chamath’s only successful SPAC and a perfectly great, successful public company. What Figure is doing is using the blockchain as a settlement mechanism for home equity loans and other nonconforming loans.

At heart, it’s a financial services company. It’s a fintech play. It’s lending money, which is one of the core things fintechs have done for 2,000 years. But the interesting twist is that it’s using the blockchain to process the back office more efficiently. In the end, it will rise and fall based on credit. If you make bad loans, you lose money in the lending business.

There is an interesting twist around using the blockchain to instantly settle these loans and to be able to securitize them. There are some securities law issues, but it’s an interesting company—a clever, good twist on blockchain. Fine.

I mean, first of all, yay, everyone. Finally, a use for the freaking blockchain that’s standalone and independent of its being a trading asset. That’s what Mike Cagney has done here, and all credit to him. It’s a good company. It popped nicely, right?

Everything about that one, to me, is the most interesting. Gemini popped high and then dropped down fast. It was fun. All 3 stocks behaved very differently. Even at the trivial level of day trading, Figure performed perfectly. Gemini popped way up and then came way back down intraday, right?

Revenue is declining on that one. It’s another crypto exchange. I never can tell them apart. It’s just not Coinbase and not Binance, so I don’t know why it matters. Revenue is declining. But God bless those guys—they’ve hung in there a long time.

Interestingly, from a stock price performance perspective, Via actually opened low—below its offer price—and then bounced up during the course of the day, proving that it’s not always free money to buy at the IPO stock price. It was a fun week for stocks, but definitely, to me, Figure was the one where I thought, “Oh, that’s interesting. I’d like to learn more.”

I guess the interesting question in the age of AI with Via Transportation is this: It was founded in 2012, so that’s 13 years. It’s classic B2B—not just SaaS, but B2B—with $493 million invested and a $4.2 billion valuation, which will fluctuate. So, 13 years: $500 million in, $4 billion out. The investors, in aggregate, 10x the total capital invested. That was an A-plus but not an S-tier investment until 18 months ago. Is it good enough today? Thirteen years, $500 million in, $4 billion out—that was great until recently.

I would take it in a heartbeat.

Harry’s not sure. He’s not sure whether he likes the fact that it has a defensible platform and big enterprise customers, or whether he wants to put more into Lovable. He’s just not sure which one he wants to do. They both have pros and cons. He’s doing both, actually.

I think at the end of the day, I’m thrilled to see IPO markets open. I’m thrilled to see exuberance. I’m aware that it will take 12 months for people to truly get cash back, but I am an incredibly selfish, self-centered individual who wants LPs to have more money to put back into venture. That will align perfectly with my fundraising cycles. Inshallah, bring home some IPOs from the previous generation and fill up the LPs’ funding, because we’re coming for it.

Good to know, Harry. Good to know. And, yeah, it was a good week.

That’s how I felt. I was very excited for one, though, which was Bending Spoons buying Vimeo. Bending—I mean, sorry, you Americans love to always claim dominance over European technology. The Italians buying the public American company for $1.38 billion.

First of all, Bending Spoons is a quirky name but a wildly successful buyout shop that specializes in these old assets that have some kind of brand recognition but no obvious model. The most obvious one is Evernote, which they bought, streamlined, raised prices on, and, I believe, gotten to a more profitable business model. So, it appears to be a successful formula.

I think the interesting thing for me is that they tend to buy assets very cheaply, centralize them, and just eke out the profits. I don’t know how cheap Vimeo is at $1.4 billion. StreamYard they bought pretty cheaply. Evernote they bought pretty cheaply, and eking out the profits was much easier. At a $1.4 billion entry price, you’re moving into a different scale. That’s a different game of roll-up.

I just have an emotional attachment to the brand. As a creator, I’ve been using Vimeo since the beginning—since when it competed with YouTube. That’s a smaller portion of the business, but it’s a $420 million business. It’s basically flat. It’s an annuity, right? And they’re buying it for, what, 3 times revenue? Less than 2.5 times revenue.

So either they have to make it more profitable, or they have to apply some folks who care more about software to inject a modest amount of growth at a time when Higgsfield and Gamma will do a third of their revenue in 12 months. Inject a little of that Higgsfield and Gamma love, and maybe they can do it.

Although I don’t know that they’ve made people use Evernote more—the Bending Spoons guys, the Spoons brothers.

Right, team. Are we ready for a quick-fire?

Fire away.

This will be great. Rory, you’re going to love this. We had the main man Kaz on the show. Opendoor today is sitting at about—I’m just going to check for accuracy—$9.30 a share today. Where is it going to be at the end of the year, December 31? It’s $9.30 today. Where’s it going to be?

Kaz seemed like a very nice man. I’ve got $24.

You got $24?

That’s my bet. My bet’s $24.

He just asked AI. Rory, do you not know this?

No. I just did it the way I do venture investments. This is the beauty of being a late-seed investor. With early seed, you’ve got to squint. With late seed, I just draw a line. It’s pretty accurate. Kaz is pretty damn good. The memers seem to like him, as near as I can tell from Twitter. I’m going $20. I’m going $24. I don’t know what the bet is exactly, but I’ve got $24.

I think that’s great. I mean, somewhere between $9 and $24. I think it keeps going up. I think it’s extraordinarily hard, and I think this guy is so smart and Keith’s talented. They’ll make noise, move momentum, and I think in the near term the stock will appreciate because they’ll just be—because you can make the story feel big.

I think it’s a brutally hard business. Over the next 3 to 5 years, you want to believe in that vision of being able to help people with the most important financial decision of their lives. I just think it’s an extraordinarily hard business because there’s a huge amount of arcane detail on every house, and everything’s a special snowflake. I hope I’m wrong, because he seems like a great guy. I just think it’s a hard thing to build massive enterprise value in.

Adobe is getting crushed by the markets, but revenue is strong and continuing to increase. Will the share price be lower or higher in 12 months for Adobe?

Adobe’s been growing 10–12%, right? And the multiple went way up a year to a year and a half ago. I think they convinced investors that AI would be good for them, or whatever. There was some amazing post-COVID story. They were obviously really high in 2021—I want to say a 45-ish P/E, with a high price-to-sales multiple. It dipped in COVID—sorry, post-COVID, in 2022—and then it came back strong.

All that’s happened now is that it’s reverted back to what it should be: 5 or 6 times revenue. It’s a wildly profitable, slow-growth company. They’ve done some things in AI—enough to not feel stupid, but not enough to actually move the needle significantly—so the value just feels right. It hasn’t crashed. It hasn’t fallen. The euphoria has just faded away, and now it’s, I want to say, a 5- or 6-times-sales, 15-times-forward-P/E company.

I think it’s still vulnerable to medium-term disruption from AI, but I don’t have an “Oh my God, it’s crashing from here” view. It’s just reverted to what it’s probably worth.

Rory, higher or lower?

Logically, my answer has to be no more than 10% higher, because if it grows 10% in sales, it should be 10% higher if it’s fairly valued today. But I feel it’s a low-confidence comment. My point is the salient one: your narrative. I’ve seen the narrative on Twitter: “Oh my God, Adobe’s crashing.” The real narrative is Adobe’s return to Earth. Jason hates that when I’m boring you guys.

I don’t think you believe that the stock price doesn’t assume a certain amount of forward growth in its current stock price. I don’t believe you believe that.

Yeah. I mean, I think it’s been growing.

You don’t like that? You don’t like the call? Fair. Fair enough. You don’t like that I’m betting down. I’m betting down at least 10%, and I’ll tell you why, if you want my—

Yeah, yeah.

Two reasons. One—

Scott Belsky leaving is a bad sign for AI at Adobe. The guy is not 70. He’s still got the fire. He goes out to be more creative. Give him a couple hundred million. Getting Scott to leave was a big blow.

And listen, maybe behind the scenes I'm exaggerating his influence, but I've worked with a lot of the folks who are still executives today. It's a big loss for Adobe losing him in AI, and maybe he came up short—I'm confident he came up short—of what he wanted to accomplish, right? It's a big company, with a lot of ships to move, but you're just at a loss without him.

Two, this is my biggest tell for being worried about public companies today. This is my least favorite metric, even worse than Google's margin—I mean, Oracle's margins on its cloud services. Adobe announces $5 billion of AI-influenced ARR. You don't have to say this, but if you have AI ARR, AI-influenced ARR, to me, any public company quoting billions of AI-influenced ARR does not believe it will have billions of real ARR. I just think both are bad signs at the margin.

But this is a business that can't be killed, right? It is an enduring business. Listening to you, Jason, you could be right: there could be a little more deflation. It's not going away, but it's 10% growth and it's slowing down. In retrospect, the real question is: How the hell did people think, less than 2 years ago, that it was worth 18 times revenue when it's worth 6 times today? I mean, what were people smoking, right?

Yeah. A lot of Adobe's history is financial engineering and the move to the cloud, and they got a lot of run out of that, right? But in the end, Scott wanted to buy Figma. It didn't happen, and he left.

They did at least try. They weren't left.

They tried. They tried. But yes, just think: if they bought it, you'd be getting those little pop-ups from Figma, too, telling you, “You've used up your license for Figma Adobe Cloud. Please pay $79.99 for the next 12 months.”

I find the Adobe products to be the most annoying and least understandable set of licensed products out there, right? Which harasses Jason. It's because Jason's not there, right?

It's because Jason—

Well, it is because, if you want to know what happens, you sit in a conference room with spreadsheets, right? At 11% growth, Salesforce is probably getting the majority of its growth from price increases today. It's hard to parse the numbers, but my rough view is that Salesforce, with 6% to 7% annual price increases on average, is growing 9%. You can't trace it perfectly. Maybe Rory can, but I'm assuming the majority of Salesforce's growth is from price increases.

You're sitting at Adobe at 11%, and you can figure out how to get people to accidentally prepay for 3 years or confuse monthly and annual. That could give you hundreds of millions of dollars of revenue, right? They don't care about the cash. They care about GAAP-recognizable revenue, so they want to lock you in for life at the highest price. Those meetings do happen, and they last days. They last days on those things. It is not just one product manager going to market.

I actually think this is an interesting discussion for the last minute here because you're right, Jason. It's $23 billion in revenue. I was roughly right. And, you know, if you compare the Wix thing, right, what does it take? If you don't think AI fundamentally threatens your business here, you're delusional, because it's what we discussed earlier: you're going from a small number of professional creators to an infinitely large number of amateur creators.

You've already allowed Canva and Figma to get under your skin. And now you have a whole next generation coming up, right? What do you do that's aggressive enough to move the needle here? Because for Wix, you know, I don't know, doing $1 billion or $2 billion in revenue, a $100 million revenue acquisition is somewhat meaningful. What do you do if you're Adobe, and what do you buy in AI now that you can't buy Figma in pre-AI? How do you think about the threat here? I mean, they've announced some products, but it just doesn't appear to have landed.

One of the areas we've been looking at a lot is next-generation creative tools. For a while there, I was like, maybe Adobe will land some punches and be relevant. They were talking the right game, but they don't appear to have done so.

Yeah. But Jeff Lawson made the great point last week, which is obvious, but I think Adobe's in this trap: you can't cannibalize your seats.

Yes. Adobe wants to add Firefly and AI to your expensive suite. Microsoft's half giving up on getting people to buy AI Copilot for Office, but what we really want is not even to buy those seats. I don't have to buy a seat with Higgsfield or with Rive or with Gamma. I don't have to buy a seat at all, or it's just 1 seat, right?

Going to Kaz's point with Shopify, I only need 1 seat to run my whole freaking store, right? And so there's a triple level of comp. We love the seat, man. But it's tough in the age of AI because you don't want to cannibalize your seat. Jeff's right. It's a tough one for Adobe, right?

Guys, this has been fantastic. The joy for me is you can see the progression of every show, actually. I think this has been amazing. Thank you so much, as always. I've loved this.

Rock and roll.