The Exchanges

Q Let's get into real estate. Let's, let's get into Opendoor. Um, a lot. We recently had Kaz on as well. Why did Opendoor lose its way? What were the mistakes that were made? And let's talk about how we're gonna figure it out.

A Well, there was two. One was predictable. Vinod actually warned Eric and me about this in 2015, that real estate has a cyclical, cyclical nature to it. And you need to make sure that your cost structure will be Acceptable when you hit the lows of that cycle. So he's basically get as many variable costs built into your company culture and as low fixed costs as possible. And then you don't care. Like, so real estate, residential real estate, people's intuition on this is very, very off. At the low market, like when people think real estate's dead, four million homes transact a year. High in the market, six. So you have four to six million transactions a year. You need to build your cost structure that you can be breakeven or not, you know, incredibly unprofitable at four million transactions a year. And Eric and me collectively did not do that from 2000 to 15, 2019. So when the Fed started raising interest rates and it actually did raise interest rates six times in a very compressed period of time, which is the fastest rate of interest rate hikes. The company then went from five and a half million transactions a year as a market as a TAM to four. And immediately started burning money. An extreme example of this that clarifies my point, I think is what happened to Airbnb COVID the first month. Airbnb is a wonderful business. One of the best network effect businesses, maybe the bes…

AI assessment note: “Well, there was two. One was predictable. Vinod actually warned Eric and me”

Q Let's get into real estate. Let's, let's get into Opendoor. Um, a lot. We recently had Kaz on as well. Why did Opendoor lose its way? What were the mistakes that were made? And let's talk about how we're gonna figure it out.

A Well, there was two. One was predictable. Vinod actually warned Eric and me about this in 2015, that real estate has a cyclical, cyclical nature to it. And you need to make sure that your cost structure will be Acceptable when you hit the lows of that cycle. So he's basically get as many variable costs built into your company culture and as low fixed costs as possible. And then you don't care. Like, so real estate, residential real estate, people's intuition on this is very, very off. At the low market, like when people think real estate's dead, four million homes transact a year. High in the market, six. So you have four to six million transactions a year. You need to build your cost structure that you can be breakeven or not, you know, incredibly unprofitable at four million transactions a year. And Eric and me collectively did not do that from 2000 to 15, 2019. So when the Fed started raising interest rates and it actually did raise interest rates six times in a very compressed period of time, which is the fastest rate of interest rate hikes. The company then went from five and a half million transactions a year as a market as a TAM to four. And immediately started burning money. An extreme example of this that clarifies my point, I think is what happened to Airbnb COVID the first month. Airbnb is a wonderful business. One of the best network effect businesses, maybe the bes…

AI assessment note: “Eric and me collectively did not do that from 2000 to 15, 2019.”

Q Yeah. Are you, are you bullish on sovereign AI? Sort of that these, are you sort of, you know, think that you might invest in that type of-

A We have a KV. We invested in a sovereign AI company in Japan called Sakano. The thesis of AI is too important to the future of nations. To allow an American company to dominate in certain regions, I think is going to prove to be true. I think you still need to marshal the hard part. That's easy to say at a conceptual level. To implement that though, You still need a martial, a critical density of talent of AI researchers, and there's not that many world-class AI researchers across the globe. I saw, you know, a public presentation by Jensen where he said there's a 150 people on the planet who could actually build a foundation of Waddle. He's probably more right than wrong, and so you still, if you're gonna build one for Germany or Europe or Israel or wherever, you're still gonna need a critical density of those 150 people to be able to compete At the cutting edge frontier models. But I think most countries are gonna try.

AI assessment note: “We invested in a sovereign AI company in Japan called Sakano.”

Q Exactly. Um, what is your mental model of open AI as a company? What is it? Doing right now. What is, yeah, how do you sort of see it going forward?

A Well, I think there's several things they're doing. Obviously, it's a phenomenal company. It's the most important company of the last decade. I think ChatGBT becomes a monopoly. It's the fastest growing consumer product of all time. It is transforming how everybody, every normal person basically does their work, runs their life. And so that product alone is a several trillion dollar company. Then they do all these other things. And some of them may pay commercial dividends. Some may pay societal dividends. But ChatGBT alone, there is no competition for. It's not like Anthropic or all these other companies have any competition for that. There are vertical applications where some foundational models or some geos where some foundational models may thrive. And AI may be important enough that even having a vertical application like coding or having a vertical geo application may be worth tens of billions of dollars.

AI assessment note: “I think ChatGBT becomes a monopoly. It's the fastest growing consumer product of all time.”

Q I want, I want to segue to fintech. You guys are among two of the best fintech investors and, you know, founders, operators, uh, in, in the world. What is the, you know, what is the state of fintech in terms of our, where are we actually looking for, for, for investments, um, and how we think about it, Keith, when you start?

A Well, I think there's lots of problems in fintech companies. It's always been difficult. Like the reality is I personally have a view that to be successful as a fintech innovation, you want an underwriting advantage and a distribution advantage. If you have both Then you can build a really epic company. If you have one, you might be able to build a pretty solid company. Affirm, which we were both involved in, um, had both. Underwriting, misprice, certain kinds of people were mispriced. We're gonna figure out a different way to underwrite based upon other things. And a distribution, a very clever distribution hack that other people have now copied, but basically was very innovative at the time. And that's why Affirm's, you know, a twenty-five billion dollar company was still upside. Um, but I think those are rare to find. Um, if you find them, The world's still open for innovation, and it's not like the world's incredibly efficient around financial services. Uh, we have some very successful ones that have been built over the last five years that now need to infuse more AI, but they're not, they're not predicated on AI. And I actually kind of like that because you could argue there's an AI over investment sort of cycle going on, and arguably one way to kind of create Uh, a portfolio that's not as sensitive to AI valuations is financial services innovation, which still requires di…

AI assessment note: “to be successful as a fintech innovation, you want an underwriting advantage and a distribution advantage”

Q Exactly. Um, what is your mental model of open AI as a company? What is it? Doing right now. What is, yeah, how do you sort of see it going forward?

A Well, I think there's several things they're doing. Obviously, it's a phenomenal company. It's the most important company of the last decade. I think ChatGBT becomes a monopoly. It's the fastest growing consumer product of all time. It is transforming how everybody, every normal person basically does their work, runs their life. And so that product alone is a several trillion dollar company. Then they do all these other things. And some of them may pay commercial dividends. Some may pay societal dividends. But ChatGBT alone, there is no competition for. It's not like Anthropic or all these other companies have any competition for that. There are vertical applications where some foundational models or some geos where some foundational models may thrive. And AI may be important enough that even having a vertical application like coding or having a vertical geo application may be worth tens of billions of dollars.

AI assessment note: “I think ChatGBT becomes a monopoly. It's the fastest growing consumer product of all time.”

Q So what's your, what's your wish, you know, uh, in our last episode with Kaz, Alex fleshed out kind of his Amazon for Homes thesis, why he got so excited about, you know, leading our investment in the company. What's your, um, where, where are you excited?

A Well, here's my, I'll give you the top down thesis, which I've shared publicly before, is in 2017, I had dinner with Yuri and Max Levton. And Yuri asked this question, he's like, what's the largest market cap company on the planet in residential real estate? And I assume there must be something outside the United States, and that's why I was asking the question. Turns out the time the answer was Zillow at eighteen billion. And it's insane. The largest asset class, period, that of 30, we've basically been innovating in technology, consumer technology for call it 30 years, like 1995, I'd say, to now, 30 years, that in the largest asset class in the world for consumers, that the largest market cap company that would take advantage of any technology would be eighteen billion dollars makes Absolutely no sense. Like, that's like insane. Um, so our goal is to reinvent the process of buying and selling on a home, and if you do that, you're gonna be worth hundreds of billions of dollars. Now, it's difficult to do that successfully, completely reinventing the process of buying and selling a large asset, but it's not impossible. Um, things that, you know, eBay did back in 1995 to 2003 apply to homes. Carvana has mastered this applied to automotive, which for a normal American, there's actually, sometimes people are like, oh, there's nothing like in common between autos and homes, but that…

AI assessment note: “our goal is to reinvent the process of buying and selling on a home”

Q I want, I want to segue to fintech. You guys are among two of the best fintech investors and, you know, founders, operators, uh, in, in the world. What is the, you know, what is the state of fintech in terms of our, where are we actually looking for, for, for investments, um, and how we think about it, Keith, when you start?

A Well, I think there's lots of problems in fintech companies. It's always been difficult. Like the reality is I personally have a view that to be successful as a fintech innovation, you want an underwriting advantage and a distribution advantage. If you have both Then you can build a really epic company. If you have one, you might be able to build a pretty solid company. Affirm, which we were both involved in, um, had both. Underwriting, misprice, certain kinds of people were mispriced. We're gonna figure out a different way to underwrite based upon other things. And a distribution, a very clever distribution hack that other people have now copied, but basically was very innovative at the time. And that's why Affirm's, you know, a twenty-five billion dollar company was still upside. Um, but I think those are rare to find. Um, if you find them, The world's still open for innovation, and it's not like the world's incredibly efficient around financial services. Uh, we have some very successful ones that have been built over the last five years that now need to infuse more AI, but they're not, they're not predicated on AI. And I actually kind of like that because you could argue there's an AI over investment sort of cycle going on, and arguably one way to kind of create Uh, a portfolio that's not as sensitive to AI valuations is financial services innovation, which still requires di…

AI assessment note: “to be successful as a fintech innovation, you want an underwriting advantage and a distribution advantage”

Q Yeah. But before we get to tech, just to close the, the politics front, do you think that this will end up defining Trump's legacy? Or what do you think his second term will be remembered for?

A Well, you know, I was kind of thinking last night, you know, he's gonna solve all the problems he inherited in the first two years. Like, if you think about immigration, basically solved. Uh, if you think about the Middle East, on its way to being solved, he'll get Russia and Ukraine solved, I'm pretty sure. Tariffs are bringing down the debt. The growth rate is gonna accelerate. We're gonna have four percent or more growth all of next year. That will make the debt irrelevant. You can grow your way out of deficits. Um, interest rates still need to come down. Trying to fix the federal, the Federal Reserve has been obstinate. That'll eventually get fixed. He will replace Powell at some point. Dustin's going to continue to be outstanding. So I think after the next year, it's going to be all upside.

AI assessment note: “he's gonna solve all the problems he inherited in the first two years.”

Q Do you, do you know what that device might be?

A I personally am a fan of putting something in your ear. Um, I think, first of all, I believe in the science fiction movie adoption of, like, like, Mission Impossible. It's always in his ear. The glasses have done reasonably well, but I, you know, I had Lasix. Um, I think most people who wear glasses try to get Lasix at some point. And so I think putting another device here all the time, the bar on the value proposition goes up versus put something in your ear and might actually go down. Um, there are people with like pendants and necklaces, which have some value and you may have multiple, instead of having a phone, maybe you have something in your ear and something here, something here, depends what the use case is, what the ideal form factor is. There's issues around battery innovation too. Like some of the stuff's really hard because you need power, constant power. Depending on where something's located, it may be very difficult to power it or not. So I, I don't know how to do it, but if I had a vote, if I was running a company that's doing hardware, I tried the year first.

AI assessment note: “I personally am a fan of putting something in your ear.”

Q Do you, do you know what that device might be?

A I personally am a fan of putting something in your ear. Um, I think, first of all, I believe in the science fiction movie adoption of, like, like, Mission Impossible. It's always in his ear. The glasses have done reasonably well, but I, you know, I had Lasix. Um, I think most people who wear glasses try to get Lasix at some point. And so I think putting another device here all the time, the bar on the value proposition goes up versus put something in your ear and might actually go down. Um, there are people with like pendants and necklaces, which have some value and you may have multiple, instead of having a phone, maybe you have something in your ear and something here, something here, depends what the use case is, what the ideal form factor is. There's issues around battery innovation too. Like some of the stuff's really hard because you need power, constant power. Depending on where something's located, it may be very difficult to power it or not. So I, I don't know how to do it, but if I had a vote, if I was running a company that's doing hardware, I tried the year first.

AI assessment note: “I personally am a fan of putting something in your ear.”

Q Yeah. But before we get to tech, just to close the, the politics front, do you think that this will end up defining Trump's legacy? Or what do you think his second term will be remembered for?

A Well, you know, I was kind of thinking last night, you know, he's gonna solve all the problems he inherited in the first two years. Like, if you think about immigration, basically solved. Uh, if you think about the Middle East, on its way to being solved, he'll get Russia and Ukraine solved, I'm pretty sure. Tariffs are bringing down the debt. The growth rate is gonna accelerate. We're gonna have four percent or more growth all of next year. That will make the debt irrelevant. You can grow your way out of deficits. Um, interest rates still need to come down. Trying to fix the federal, the Federal Reserve has been obstinate. That'll eventually get fixed. He will replace Powell at some point. Dustin's going to continue to be outstanding. So I think after the next year, it's going to be all upside.

AI assessment note: “he's gonna solve all the problems he inherited in the first two years.”

Q I guess I asked that to ask, Did Microsoft sort of drop the ball on their relationship with it? Like, could they have capitalized in a much bigger way, or was it just?

A I don't know. You know, the talent, there's a complexity around talent. Like, you know, individual people, if there's only a 150 people that can build research models successfully, talent's gonna have a lot of power. And if the talent doesn't want to work for Microsoft, there, there may not be economics there. I mean, Meta's trying this of like, can you use economic leverage to force people to work there? Um, you know, and we'll see, we'll see what the results are. It's not a bad idea. Like, I actually think Mark's clever. He means more money than most. Take the money and use it. It's like, imagine you're competing in sports. Mark doesn't really have a salary cap.

AI assessment note: “if the talent doesn't want to work for Microsoft, there, there may not be economics”

Q But if Elon couldn't do it to some degree, does it speak to the intractability of actually?

A Well, you need political will, but I think one of the things the last two weeks is showing, Trump's gonna be more popular than ever. He's literally posting the highest popularity and approval rate of his career right now, and the government shut down. One of the reasons why there hasn't been political will is when you have sort of a diffuse benefit When, and, and concentrated aggravated people, politicians, elected officials are afraid. So when you cut something, the people who like whatever that program is, they may only be one percent of the population, but they're very loud and annoying. And the 99% of the people who benefit are quiet, typically. But if you can convince people that, hey, we don't need all this stuff, then the 99% people are like, hey, why do we need all this stuff? Why am I paying all these taxes? Then the one percent you can safely ignore. And because the one percent have historically been channeling through the legacy media, which is no longer relevant, it does, you're going to need to find a new megaphone. Like, whether, you know, New York Times or The Wall Street Journal writes some attack piece, nobody on the Republican side cares anymore. We've learned that that stuff is just garbage, it's fake, and irrelevant. And so if you have the, if you're on the right side of history with the right evidence and the right ideas, you can just ignore the legacy medi…

AI assessment note: “I think it's going to be a lot easier to cut things.”

Q Yeah. Let's get deeper, just on the effects of AI and the economy, if, if scaling laws hold to some degree, do, you know, will it show up in the productivity statistics directly? Um, you know, what does it do to, uh, wages?

A Yes, but you do have the, I do believe in the, you know, sort of the bottom of the cost thing, where, like, the limiting step is the most, the most difficult, most human things actually put a constraint On like, uh, AI growth or jobs going away. I think that is actually true. Like someone still needs to construct all these power plants or data centers. And as long as there's manual electricians, as long as there's man, maybe you don't need manual architects, but probably need manual plumbing for a while. Manual electricians. I don't think robots are building the next data center, at least not right away. And so that is going to be the constraint. And so these jobs are not really going to go away. So some, if you read the Bommel, you know, Curse of the Bommel stuff, I think we're not gonna see radical disruption. We're just gonna see the golden age that Trump talked about in his inaugural address.

AI assessment note: “these jobs are not really going to go away... not gonna see radical disruption”

Q And, and why doesn't Powell want to reduce rates? What is a steel man for?

A Well, his argument, well, I don't know if he has a good argument. I don't, I think he has Trump derangement syndrome. He just hates Trump. That's why he artificially lowered rates to try to get Kamala elected. It is indefensible cutting 50 basis points off in, you know, last fall. But he does see employment is running pretty hot, but the difference is with AI and productivity gains through technology, hot employment does not necessarily yield inflation. It's not a, it's not just wage growth that's fueling GDP growth. And so I think you're seeing some of the substitution where productivity gains can be made. GDP can run four or five percent without triggering inflation. Because it's not just, you don't grow GDP just by paying humans more, which is the historical precedent.

AI assessment note: “he does see employment is running pretty hot, but the difference is with AI”

Q But if Elon couldn't do it to some degree, does it speak to the intractability of actually?

A Well, you need political will, but I think one of the things the last two weeks is showing, Trump's gonna be more popular than ever. He's literally posting the highest popularity and approval rate of his career right now, and the government shut down. One of the reasons why there hasn't been political will is when you have sort of a diffuse benefit When, and, and concentrated aggravated people, politicians, elected officials are afraid. So when you cut something, the people who like whatever that program is, they may only be one percent of the population, but they're very loud and annoying. And the 99% of the people who benefit are quiet, typically. But if you can convince people that, hey, we don't need all this stuff, then the 99% people are like, hey, why do we need all this stuff? Why am I paying all these taxes? Then the one percent you can safely ignore. And because the one percent have historically been channeling through the legacy media, which is no longer relevant, it does, you're going to need to find a new megaphone. Like, whether, you know, New York Times or The Wall Street Journal writes some attack piece, nobody on the Republican side cares anymore. We've learned that that stuff is just garbage, it's fake, and irrelevant. And so if you have the, if you're on the right side of history with the right evidence and the right ideas, you can just ignore the legacy medi…

AI assessment note: “Well, you need political will... So I think it's going to be a lot easier to cut things.”

Q Yeah. Let's get deeper, just on the effects of AI and the economy, if, if scaling laws hold to some degree, do, you know, will it show up in the productivity statistics directly? Um, you know, what does it do to, uh, wages?

A Yes, but you do have the, I do believe in the, you know, sort of the bottom of the cost thing, where, like, the limiting step is the most, the most difficult, most human things actually put a constraint On like, uh, AI growth or jobs going away. I think that is actually true. Like someone still needs to construct all these power plants or data centers. And as long as there's manual electricians, as long as there's man, maybe you don't need manual architects, but probably need manual plumbing for a while. Manual electricians. I don't think robots are building the next data center, at least not right away. And so that is going to be the constraint. And so these jobs are not really going to go away. So some, if you read the Bommel, you know, Curse of the Bommel stuff, I think we're not gonna see radical disruption. We're just gonna see the golden age that Trump talked about in his inaugural address.

AI assessment note: “if you read the Bommel, you know, Curse of the Bommel stuff”

Q I guess I asked that to ask, Did Microsoft sort of drop the ball on their relationship with it? Like, could they have capitalized in a much bigger way, or was it just?

A I don't know. You know, the talent, there's a complexity around talent. Like, you know, individual people, if there's only a 150 people that can build research models successfully, talent's gonna have a lot of power. And if the talent doesn't want to work for Microsoft, there, there may not be economics there. I mean, Meta's trying this of like, can you use economic leverage to force people to work there? Um, you know, and we'll see, we'll see what the results are. It's not a bad idea. Like, I actually think Mark's clever. He means more money than most. Take the money and use it. It's like, imagine you're competing in sports. Mark doesn't really have a salary cap.

AI assessment note: “if the talent doesn't want to work for Microsoft, there, there may not be economics”

Q And, and why doesn't Powell want to reduce rates? What is a steel man for?

A Well, his argument, well, I don't know if he has a good argument. I don't, I think he has Trump derangement syndrome. He just hates Trump. That's why he artificially lowered rates to try to get Kamala elected. It is indefensible cutting 50 basis points off in, you know, last fall. But he does see employment is running pretty hot, but the difference is with AI and productivity gains through technology, hot employment does not necessarily yield inflation. It's not a, it's not just wage growth that's fueling GDP growth. And so I think you're seeing some of the substitution where productivity gains can be made. GDP can run four or five percent without triggering inflation. Because it's not just, you don't grow GDP just by paying humans more, which is the historical precedent.

AI assessment note: “he does see employment is running pretty hot”