General partner at Founders Fund, Founders Fund
What Top VCs Actually Think About Legal AI with Keith Rabois #harvey #legora #spellbook #startup #ai
Money is pouring into legal AI, but how do top-tier investors actually evaluate the space? In this episode, Zach speaks with Keith Rabois, Managing Director at Khosla Ventures and one of the most successful investors of the past two decades, about how he thinks about legal tech in the age of AI. They discuss why elite founders are suddenly flocking to legal, the risks facing application-layer startups, and what separates real opportunities from hype. The conversation also explores why Rabois invested in Spellbook, why he’s skeptical of some of the biggest names in legal AI, and how changing i...
About Keith Rabois
In a late April 2026 podcast, Keith Rabois discussed his views on the legal AI sector. He expressed skepticism about companies like Harvey and Lora, stating that they "make no sense" because improving the productivity of large law firms that bill by the hour is "counter to their business model." He contrasted this with his investment in Spellbook, noting that the company targets in-house enterprise teams, where general counsels "love to save money" and "love productivity improvements." Rabois also warned that application-layer AI companies face serious questions about how quickly general-purpose foundation models will improve relative to vertical-specific offerings, and predicted that many investments in the space over the last three years "are not going to look so hot." Rabois also commented on geopolitical and national security topics. He stated that export controls on advanced chips are "a good thing for the United States national security," and criticized a figure he described as "self-interested" for opposing such controls, arguing that China has historically leveraged American technology to reverse-engineer and build its own industries. Additionally, he remarked on a conflict involving Israel, claiming it was "over in three" days and that "there's not been an American killed since day three," questioning the nature of a war in which "nobody dies."
Transcript (83 segments)
Z
Zach Abramowitz0:00
Money is pouring into legal AI companies, but what do the best VCs actually think of this space? Hey there, I'm Zach Abramowitz and I am the Legally Disrupted on episode 48 of Legally Disrupted. I'm talking about investing in legal startups because that's something I'm kind of back to doing. So, after initially pausing my investments in legal AI after ChatGPT launched, I've been back at it recently. So earlier this year I invested in Sandstone. That company is growing rocket speed. And I'm actively evaluating several other deals.
Now when I asked Jonathan Levy from Y Combinator last year about why YC was investing in more legal companies, I said, 'Does this have to do with the vertical?' And he said, 'No, we don't really care about verticals, right? What we care about is super top talented teams. We're like the NBA draft. We just go for the best entrepreneurs, best athlete on the board.' And he said that right now it is the case that the top entrepreneurs, the top startup founders want to build something in legal.
Now, I've certainly seen this firsthand like top talent is flocking to legal. Look no further than Harvey and Lora, right? But it's not just them. This entire space is attracting a different kind of entrepreneur. Trust me, I've been following the space for a long time, there is a significant and very noticeable upgrade. And honestly, I've got some FOMO, right? If I've been looking at this space for years and now all of a sudden it gets flooded with top teams, you think I don't want to invest? Of course not. I don't want to be the guy who's been looking at legal tech and evaluating it for years, knows all the teams, knows all the crevice, knows all the customers. Sounds like a flex. It is. But I also now have some serious FOMO.
At the same time, this is a dangerous game. Like right now, it's true like the product capability has never been better, but it's also never been easier to replicate these products. You know, before it was acquired by xAI, Cursor was already being pronounced dead in some circles and they had billions in ARR and a customer base that absolutely loved them. You can't talk to devs over the last year without them raving about Cursor. Now, what does that mean then for application layer companies like Harvey and like Lora, especially as Claude, which is the model of the month, gains serious traction and lawyer attention, right?
We've already seen litigators building incredibly powerful proprietary systems directly on top of base models like Claude. By the way, Quinn Emanuel has now been talking about their Claude product. We just saw Freshfields announce that they're rolling out Anthropic licenses to their entire team. So like on the one hand, yes, the teams have never been better, but that also means the competition is stiffer and the stakes are huge with legal AI startups raising eyewatering amounts of cash and getting valuations that we've simply never seen in this sector.
To break all this down and kind of get inside the mind of like how an elite VC is looking at this, today I'm talking with Keith Rabois, a managing director at Coastal Ventures and one of the most successful founders, angels, and VCs of the last 25 years. He's a founder of OpenDoor. He was also an early investor in YouTube, Lyft, Eventbrite, Palantir, like his angels as an angel. By the way, fun fact, Keith was also a litigator at Sullivan and Cromwell and a clerk on the Fifth Circuit Court of Appeals.
So, we discussed number one, why he invested in Spellbook, one of the fastest growing legal AI companies, why he's not an investor in Harvey and Lora, why you're seeing top VCs suddenly pouring money into this space, and because Keith doesn't shy away from politics or sharing his opinion, I asked him about his prediction that the Iran war would finish in just 4 days, where he comes out on the Dwarakish Jensen Wang podcast, and whether the Trump administration is getting good advice on AI. So without further ado, let's get disrupted.
So Keith, really excited to talk to you about this space. First of all, I noticed, you know, and doing a little bit of research on you, I actually didn't know this, that you were an attorney at a major AmLaw 100 firm. How long did that legal career last?
K
Keith Rabois4:44
You found my legal background out of my bio? But yeah, I spent a long part of my life practicing law. If you combine law school, clerking and practicing, basically spent all my 20s as a lawyer. So I was for almost four years a litigator at Sullivan Cromwell in both Washington DC and New York after clerking on the Fifth Circuit Court of Appeals.
Z
Zach Abramowitz5:12
So like you, I know you've invested in Spellbook and I want to talk with you about that, but just broadly speaking, is any of your interest, and again I don't know how interested you specifically are in legal tech, you know, other than having invested in Spellbook, does that time in the firm inform you at all?
K
Keith Rabois5:31
Well, sure. I mean, you know, having practiced law, definitely helps understand the profession, what matters to clients and customers, how lawyers actually do their work. Although, truthfully, I haven't practiced law in 25 years. I'm not so sure there's any changes in the practice of law in 25 years, but in theory, there could have been. Law moves at the most glacial pace you can imagine probably of any profession other than the heavily regulated ones like education and maybe healthcare. So I didn't even have external email when I practiced law. I had a secretary literally who would type up my handwritten notes.
That said, most of the legal tech that we see, not all but most, is interesting. It's an obvious application of LLM. We can talk about why most of the companies that are well known are focused on the commercial and transactional nature of the practice of law. I was a litigator. So I primarily was focused on disputes after commercial agreements were executed. Maybe poorly executed ones, but fundamentally there was going to be a very significant dispute either between the parties to a contract or with a regulator or the government.
When I was involved, I focused mostly on antitrust litigation, intellectual property litigation and some what would be called white collar criminal defense. And so yeah, it helps a bit. I also have a lot of connections still into the legal world. Most of the people I went to law school with are practicing law at the prime of their career unless they've sort of morphed into politics or something. And that can be helpful too in performing diligence, etc. on these kind of companies.
Z
Zach Abramowitz7:16
Yeah. And our conversation will morph into politics in the end. But I wanted to ask you, I asked this question to Jonathan Levy from Y Combinator, which was, you know, I see you guys have invested in a lot of legal tech companies even before LLMs arrived, but even more since. Is that because you're bullish about the space? And he said, 'No, not at all. We are just, it's all about investing in talent, and it happens to be that there's more top talent in legal right now.' I see you're nodding your head. Is that how you look at it?
K
Keith Rabois7:48
Yeah, I mean I have no personal interest in investing in a particular domain, but we're in the business of investing in the best founders on the planet. And historically, the best founders on the planet were rarely interested in the legal field, in reorienting law or improving the practice of law, or didn't understand it. You do see many top tier teams focused on innovation in the legal sector right now and so it does make it easier to invest, to follow the talent as Jonathan suggested. That is definitely my model. But I don't have any top-down organizing principle that says hey let's go find some more legal AI.
That said, I do think the logic of LLM being applied to law is pretty obvious. It's not surprising to me that other than coding and possibly customer support that law is at the frontier of adoption. That makes tons of sense if you think about how work and how the written language works and how law is predicated on the written language and precedent. But I'd be happy investing in a database or a consumer application. But I follow wherever the best people on the planet kind of want to take their lives. And if they want to practice law, they want to improve law, if they want to change law, that's great.
Z
Zach Abramowitz9:13
Do you think it's getting harder to find billion-dollar opportunities in this space? Meaning, it feels very much to me like yes, it's a great application of the tools, like you said, it's kind of obvious, but ultimately, I do sometimes step back and think like, how does legal compare if a VC like you is looking at a company that's in defense tech or that's coming up with new autonomous drones. That actually sounds like a bigger market opportunity to me. And I think the question is not just like legal versus other verticals but I'd like to get your sense of even software, is it obvious today that a software company is as exciting to invest in as it was maybe 10 or 15 years ago or are we kind of past that point where these are billion dollar opportunities and it feels like there's very mixed signals.
K
Keith Rabois10:20
I think the details really matter at the micro level. Any one company can be incredibly important, generate incredible value for shareholders and for investors. So for example, climate tech quote unquote was a terrible field, but Elon created Tesla, which would have been one of the best investments in the history of the globe. So I think you need to find Elon's, not find climate. And in software, I think traditional software is under a lot of scrutiny and for some substantially serious reasons.
First of all, the ease of creating new software, the ease of creating integrations, which were historically a significant moat, those have all changed in the last year or two. And so the workflows that used to protect incumbents with mediocre software I think are also under attack because either recreating workflows is pretty easy or agents quote unquote are going to perform the workflow instead of humans and or there's a bridge between a historical workflow and the new workflow so you don't have to retrain millions of users. All those things are true and so the people taking sort of a short perspective on traditional software whether in the private or public markets are probably directionally correct.
But the outlier position in one good software company could transcend the returns on all of the cynicism in all the other companies. And it's a sort of interesting place where I said to a founder recently, I'm like if I had seen your product five years ago it would have been the best legal tech product I had ever seen, I would have invested after the first call. But today the ability to replicate these tools is just incredible. And the other problem is like there are attorneys who are really starting to code. I showed a product recently to a guy who had been a paralegal on litigation teams. He came back to me nine days later. He shows me he's got the product. He's built it himself and it's working. It's working really well. And it's like that's not pie in the sky to talk about people coding their own now. Quinn Emanuel, John Quinn was just on Bloomberg talking about how they've built their entire litigation platform on Claude.
Z
Zach Abramowitz12:39
Yeah. No, I think a lot of top law firms and corporate counsel may code their own. It's a very realistic option. There are questions you have to think through about doing that and what advantages you have and what disadvantages do you have but it's conceivable where X years ago it was inconceivable.
K
Keith Rabois12:57
Yeah.
Z
Zach Abramowitz12:57
Now you're an investor in Spellbook and I want to get like your perspective on why you invested there other than the fact that Scott is obviously a super impressive founder. But I did want to flag he has been doing some very good contrarian narrative on this. So he published recently his demo analytics and it was basically showing that the week that Claude Legal came out with their announcement, single biggest number of demos. And what I've explained, I'm curious if this is your perspective or you look at it differently, to people who are asking this question like what's the moat, I'm like I don't think people rip and replace products they really like. That's generally true and certainly over a short time frame. If you expand the time horizon to years I think that's less true, but in months, in one to two years, that's probably true.
Basically, if you're addicted to a product, I could use metrics like NPS and things like that, but ultimately if you're satisfied with the product, you're not really looking for alternatives. And even if you heard about an alternative that sounded great, you're going to discount that by the pain and friction of adoption, some discount about the probability that it's actually that good. So, the more satisfied you are, the more insulated you are for a reasonable amount of time, right?
K
Keith Rabois14:12
But the difference being that the SaaS products you're locked in. But I haven't met a ton of people at law firms as an example that love their case management platform, that love their document management system.
Z
Zach Abramowitz14:24
Oh yeah. Or just generally who loves Salesforce like nobody. I mean like so you can take like an NPS perspective, take all software. Who loves Excel? Some people. Not normal people but some. But like basically you can take any software product and say like how much do people love it? If the end user is like thrilled versus quote unquote locked in. Like think about the connotation of locked in. Basically, if your users are locked in, it means they probably don't love the product. If you have to lock them in, that suggests that they're barfing. They just don't have a lot of choices. And so, yeah, if people love your product and you wake up every day and you're a satisfied customer, you're just kind of immune to like advertising and marketing and sales. And the less satisfied you are, the more open you are to, oh my god, maybe there's something better. And so I think there's a correlation there.
K
Keith Rabois15:14
Right.
Z
Zach Abramowitz15:14
Going into Scott in particular, when you saw the product, you're obviously aware of like Harvey and Lora and the other players out there. What did you like about Spellbook?
K
Keith Rabois15:23
Well, I love Spellbook because first of all, we're not really targeting the practice of law. I just think that the Loras Harveys make no sense. Trying to improve the productivity of American law 100 or 200 or 250 or whatever American law firms that scale bill by hour except for two, and the incentive of improving their productivity is counter to their business model. And even if they were agile enough to adopt to it then you're still probably looking at fee compression not market expansion. So, if you make all these lawyers more efficient that practice at law firms, clients are going to demand lower fees and the net market you're serving just isn't that exciting. You can look at stats if you want. Roughly all software sold to law firms, somewhere around maybe a few billion dollars, maybe up to 10 billion. That's it. That's not a great market, right?
At Spellbook, we've always targeted commercial contracts and we're mostly focused on in-house enterprise teams. In-house enterprise general counsels love to save money. They love productivity improvements. For example, most GCs are under budget pressure. Their CEOs, their CFOs do not want to spend more money and allocate more money to the legal team. Yet, they have more to do than ever before. They have more agreements, more counterparties, more sophisticated regulation to comply with. So what Spellbook does is it frees up time for a legal team to use the same size team or even a smaller team perhaps and get more done. That's incredibly attractive.
So we're really focused on commercial agreements and the number of people who touch commercial agreements is also in the billions. So there's not that many lawyers in the world. There's not that many lawyers in the United States. There's too many, but there's not that many in the grand scheme of things. But the people who interact with contracts, sales people, biz dev people, customer success people, huge number of people, probably a billion and a half people on the planet touch a contract every year. So we're in the business of facilitating, accelerating, improving, basically creating the infrastructure for commercial agreements, which the western world really runs on.
Z
Zach Abramowitz17:38
Meaning legal for Spellbook is more like a go to market not necessarily the end customer. You're going where the contracts are.
K
Keith Rabois17:46
The legal AI is like a category which is misleading. We're in the business of providing the fastest, safest commercial agreements in as automated a way as absolutely possible. And that does improve the workflow and it does improve the life of transactional attorneys and especially in-house corporate attorneys and should reduce costs and wasted expenditures, but it's mostly about improving outcomes like less mistakes, less exposure.
I bet you the average general counsel at a Fortune 500 company has no idea how much exposure they have in all their agreements because they have hundreds of thousands to tens of thousands of agreements. They have not scrutinized every single one of them and there are terms in each one that could be less than ideal and they don't have a dashboard or an ability to just dive in and see what the hell's going on in my company and Spellbook provides that.
Z
Zach Abramowitz18:51
So for example in Q1, that's much more of like a director of sales or a VP of sales mindset. I agree with you. The GC, even when GCs had CLMs, it wasn't like a use case they thought about. They never thought like that's something that should be my responsibility to have like a high level and in the weeds view of all my contracts in and out.
K
Keith Rabois19:12
And you know, it's extremely valuable like having this bird's eye view of every agreement my company touches. What if the law changes somewhere? What if the law changes in Georgia? How does that affect us? What if we realized that we got in a dispute with a vendor over here and you know what? We've been doing this kind of deal for like 20 years. How many different deals have we done that have similar not necessarily exact verbatim clauses because that you could probably do with a regular search maybe. Spellbook does really really well. Spellbook basically improves the quality and the visibility of your legal agreements. And the world runs on legal agreements. Like we have this kind of funny, amusing stat that we use that the average NBA game before the lights go on, when the fans show up, there's 150 contracts that go into game day. Or putting the recent NASA launch to the moon, 5,000 contracts.
Z
Zach Abramowitz20:18
Do you think anybody has the visibility into all 5,000 contracts?
K
Keith Rabois20:24
Absolutely not. But Spellbook-like products do.
Z
Zach Abramowitz20:26
Yeah. And that's super interesting thinking because I sort of think of Spellbook and what's exciting about Spellbook is the fact that it diffused to a lot of small and solo law firms who previously just did not buy these products at all, did not buy tech products, also have an incentive to sell. And what I always liked about Spellbook is it kind of proves the point that no, people will buy any technology if it's actually good and useful to them.
K
Keith Rabois20:54
Small law firms and individual practitioners have different incentives than large American law companies. Like for example, if I'm running my own little law firm, maybe I can take on new clients that I'm more productive or maybe I can just go home earlier and hang out with my kids. And or maybe I keep more money because I'm more effective. You know, I might have a fixed cost arrangement with a client. So the economics of mid-market and smaller law firms in the United States are more amenable to improvement through technology than the American law 100, which is going to be disrupted one way or the other. Clients are basically going to say, 'We're no longer paying $2,000 an hour for your associate to do effectively cutting and pasting, period.' And half the profits are going to evaporate of the American law firm.
Z
Zach Abramowitz21:40
Well, how are you thinking about that at Coastal Ventures? Meaning like I know Fred Wilson wrote this whole post where he said that VC in 2026 was going to be done without a lawyer. You know, it's felt to me like especially like early rounds a lot of times really don't need that much.
K
Keith Rabois21:57
Sure. Well, they are kind of done without a lawyer. I mean, people have been doing seed rounds on SAFE notes which basically don't require a lawyer for the most part for like 15 plus years. So seed rounds basically go on autopilot anyway. Series A's are going to be similar. The reality is most venture can be done without a lawyer. Certainly on the VC side because we just do so many deals. I think a founder might want counsel because they don't always understand the document because it may be the first time they ever raise money or second time or something like that. It's an ad hoc transaction but for a typical VC it makes no sense to have lawyers on 90 some odd percent of all investments. There's an occasional corner case, an occasional quirky investment, but they're pretty rare.
Z
Zach Abramowitz22:42
You know, we were talking before about all the capital flowing into application layer companies and some of the risks associated with that, with the shrinking cost of software development. When it comes to these kinds of companies, when it comes to the application layer companies, I've heard that a lot of VCs are getting out and there are a lot of healthy secondaries. Are you seeing the same thing when you start getting into the B and C rounds of companies like Harvey, like Lora, like Cursor even? Are VCs hedging their bets? And is there a difference maybe between like tier one VCs and perhaps more middling VCs?
K
Keith Rabois23:32
I don't know. I don't really track the secondary market much. You know, it's very rare for us to participate in the secondary market either as a buyer or seller. Maybe we should pay attention to buying more, if that's true, but it's not really something I do. I'm like a primary investor. I want to lead a seed round. I want to lead a series A typically. So I don't have a strong feel for whether that's more true or false than ever before. So I don't really know.
I do think that application layer companies do need to answer some serious questions. You know obviously the speed and the pace of improvement by the new labs or the regular foundational model labs, Anthropic, OpenAI, etc. is incredible and there are real questions about how fast do the general purpose foundation labs improve their product versus what a vertically specific application layer company can do. And I think many of those investments over the last three years are not going to look so hot because the terminal value has to be there over 3 to 30 years in the future. And I don't know too many vertical application companies where you can easily say in 10 years people are definitely going to use our product and not OpenAI, not Anthropic, not whoever.
Z
Zach Abramowitz24:58
So I asked on X, I said, 'If you could talk with a tier one VC who wasn't invested in Harvey and Lora, but is definitely paying attention to the space, what questions would you ask them?' And this was one question that I definitely wanted to ask you. I think I know where your answer is going to come out on this, but the person said, 'I want to know if LPs are chastising them for missing out on Harvey and Lora.'
K
Keith Rabois25:24
My LPs are not because I'm in the business of returning capital. I distribute billions of dollars. My fund distributes billions of dollars. Lora may be a good seed investment possibly. Harvey I'm very dubious about the future of so they can ask the question but every time I make an investment it's designed to return real money to people. And I'll do non-standard investments meaning things that are not obvious or not popular. You know, I kind of like this laugh test of trying to invest when half my friends who are VCs think I'm crazy. So, I'll find things that are under the radar, things like Ramp that people, everybody thought Brex was cool. I'm pretty convinced Harvey is going to look like Brex.
I'll find things like I just did a very cool, incredible coding replacement that is better than Cursor, certainly better than what the labs can do. It's called Factory.ai. You should read all about it. It's great. It is the future of code generation for large enterprises. It will be the future of how large enterprises run. And people thought that oh these coding things like Cursor or Codex or Claude Code, whatever, the future. I don't believe that. I think Factory AI is going to be the last.
Z
Zach Abramowitz26:55
Has it been your experience, because that question from LPs, was there a point in your career where you were having to deal with questions more from LPs and at a certain point it's like okay like he wants a contrarian take like he's looking for outliers and like he's proven so we're not going to ask those questions anymore. Or does that change? How does that work for you specifically?
K
Keith Rabois27:17
No. I do think that's common actually watching LPs. I've been a VC for 13 years now. I have a strong, pretty strong relations with a reasonable number of LPs over the years now. I do watch what questions they ask other managers and stuff and sometimes they call me for references etc. So I have a feel. For me, I had a fairly strong track record on angel investing and was able to relatively quickly parlay that into a pretty good set of investments early in my career at KSLA. And so I never went through that kind of scrutiny question of like what the hell are you doing? But it's a pretty common thing because there's a large, the rationale is there's a huge gap between when you invest and when you distribute, can be measured in five to 10 years typically, right?
And so, many times people ask questions like, well, how well are you doing? And the proof's not in the pudding. The proof's not really there yet. So, it's very common to want to bridge that with a lot of probing questions, but I was able to bridge it pretty quickly. There were some things I invested in early in 2013 and 14 that were very clearly on the right trajectory. And so, personally for me, it was not ever that fundamental, but it is very common as a refrain for individual investors or funds.
Z
Zach Abramowitz28:51
So I wanted to segue because over the last month or so when the Iran war was going on, I was watching you on X and as someone who was here on the ground in Israel, I was finding it very amusing because I felt like all right, I know what I'm experiencing like when I go outside like I've seen this before. I saw like the first couple of days, but you predicted that the war would be over in four days and then when people came to sort of be like, 'Oh, Keith, we got you. The war wasn't over in 4 days.' I believe your response was, 'Yeah, I think it was over in three.'
K
Keith Rabois29:30
Yeah, it was definitely over in three. I mean, like, look, you live in Israel. There's no detectable impact in how Israelis are living their lives. More importantly, let's take a step back. The way I was describing over, and if you read my tweets from the first day, it's like when you're in the fourth quarter of a football game and there's a minute 53 on the clock and you're winning 77 to 3, the game is over. Every analysis says, 'This game's over.' We play chess. Well, lots of founders play chess. So, it's a pretty good metaphor. You can look at a board a few moves in and know that that game is over.
Barring a catastrophic set of mistakes by one person, it is inevitable that you were going to lose that game. That has basically been the case since day two. Actually, not even. And like, empirically, let's talk empirically now. There's not been an American, thank God, killed since day three. What kind of war does nobody die? Like, literally, there hasn't been an American soldier killed by an Iranian since day one. And no American has died since day three. So, we're 6 weeks in. Like, more Americans died in training last year, literally in training, during the same period of time, than have been killed by Iran. So, it's actually been safer to be in Iran than been in training by a true population stop.
Z
Zach Abramowitz30:44
By the way, usually when Israel is not in conflict, they always say the safest place to be is in the army.
K
Keith Rabois30:49
It is probably you're free from Israeli drivers.
Z
Zach Abramowitz30:52
Well, certainly also if you're Jewish, it's definitely safer to be in Israel than many places in the United States right now. Or Canada, you know, if you've seen Harley from Shopify tweets. But another way to put perspective on wars. So when I was growing up, we invaded this third world island in the Caribbean called Grenada. You know how many American soldiers died? 31. Taking a third world island. You know the current number of American soldiers, still too high, is 13. But like literally 31 American soldiers died taking a third world micro island.
K
Keith Rabois31:28
So this has been so successful. The other piece that's been over since like day three or so is the diplomatic front. Every Middle Eastern country other than Oman is now significantly closer to business relationships or normalization relationships with Israel. That's been like the greatest thing ever. Like we've been trying to accomplish this basically since the 1970s and from 1979 to basically for as long as I've been alive. Lebanon and Israel are going to normalize relations. It's inevitable at this point. That is not going to be the only country. You know, Jared had a lot of great success in the Trump first administration with the Abraham Accords and there's going to be more members of that. There's going to be AI partnerships with Israel. This is all happening in real time. It's the greatest thing ever.
Like Latin America, so if you think about all the success of the Trump administration, Latin America is the future. Latin America is going back to capitalism which works across every single country. It's just every election boom boom boom and they're fixing the historical messes of Argentina, El Salvador, some of the worst places you could live in the world over the last 30 years are all becoming the best places and the most popular leaders. Like I saw that El Salvador president has a 94% approval and it's like legit because you fix crime etc. So that's all great places to live if you clean them up like every election it's just like boom boom boom. Brazil, while the precipice, maybe have a revolution in Brazil, this is going to be great. So I think the Middle East is next. The Middle East is going to break out, has a lot of great assets, a lot of great natural resources obviously people focus on oil but there's others and it's going to be the future of data centers, AI.
Once the Islamic Republic is completely disabled, everything's going to be great in the Middle East. And Europe's going to be left behind. It's gonna be Latin America and the Middle East are the future because they're all going to start thriving with modern economies, modern assets. And as Larry Summers said, Europe's like a museum.
Z
Zach Abramowitz33:32
Yeah. And that's the thing. This feels like something that we should all be deeply excited about. And I've tried to when people ask me about it, my usual point is I don't think this is entirely about Israel. Like I think Israel has a significant interest in this conflict for sure. But the reason that Donald Trump moved on this is not because he's doing it for Israel. I think that he saw last year, oh wait, this group can execute because I saw what they did in the 12-day war. I have now confidence that we can go in together and do this. But I don't think he was doing it for Israel. I think this probably has much more to do honestly with China than it does with Israel.
K
Keith Rabois34:15
Yeah. The idea that this had something to do with Israel like motivating Trump is like so absurd on a couple dimensions and let's go through them. One, the most consistent, there's two things that Trump in Trump's adult life the two things he's been most consistent about are tariffs are good and the Islamic Republic is bad. Literally, if you read any interview since the 1980s with President Trump, those two things are the only two things that are perfectly consistent. Tariffs are great and Iran is terrible, meaning the Islamic Republic. He literally ran an ad in either the New York Times or Wall Street Journal in 1983 suggesting we need to overthrow the government of Iran. In 1987, he did this famous interview where he talks about how you take, you know, this island that everybody talks about.
Look, Iran's been at war with the United States since 1979. They seized our embassy and took 241 Americans hostage. That alone would justify attacking the government and removing every single one of them. Then they bombed the Beirut, they bombed our marine barracks and killed 241 or something like that Americans in 1983. And the American government has been too weak and too terrified to take any action to protect Americans. There's this famous video during Obama when they literally seized an American Navy vessel and treated our soldiers like hostages. Like the fact that the president of the United States did nothing about this has just encouraged terrorism for 47 years. So Trump's been on this for 47 years before even people even talked about Israel.
Second, your second point about China is crucial. We have an existential threat posed by the CCP of China. Iran does not pose an existential threat to the United States. It's not literally an existential unless they were to develop nuclear weapons. Their ballistic missiles are not existential. Russia is not an existential threat to the United States either except for the fact that they have ICBMs loaded with nuclear weapons.
Z
Zach Abramowitz36:12
They're a nuisance.
K
Keith Rabois36:13
Yeah. Well, they're probably maybe more. Iran's now down to nuisance level. It's like a bunch of pirates running around. But Russia does have lots of nuclear weapons still. But their economy, their military, like they can't do anything to threaten the existence of United States other than launch nuclear weapons.
So, China though does pose an existential threat. They have economic might. They have billions of people. They have a first rate military for the most part. Although, if you seen the performance of a lot of their technology in either Venezuela or Iran, you do wonder whether any of their stuff works. But we're in a serious AI race with them. And if they were to outperform us in AI plus energy, we would have a fundamental problem. Like military might is predicated on economic might. That's why the United States won World War II. Ultimately, at the end of the day, we had the industrial might to combine with the will and foresight of the United Kingdom and Churchill. But without our industrial might and energy compared to the German war machine, the results would have been very different or incredibly disturbing.
And so precarious on a minimum. Two, we need to win our race with AI because whoever is the best at AI will have the most proficient modern economy going into the next century. And so everything this administration does absolutely is calibrated on and calculated about how do we ensure the United States is set up for the next American century. Which is fundamentally about AI, energy, manufacturing, and education to compete with the abilities of China. And that was important in Venezuela, it was important in the Panama Canal very explicitly and it's also important in the decisions we make vis-a-vis Iran.
This administration, supply chain, they basically been working, our minerals have been vulnerable, our supply chain has been vulnerable. We've been trying to fix that in the Trump administration and they've made a lot of progress but not everything can happen overnight. But we learned during COVID that this country couldn't manufacture drugs. Even basic penicillin antibiotics were threatened by China. We cannot have an economy that's subservient to anybody else's supply. And we became energy independent. Trump had the foresight in his first administration to ensure that. And you see that paying dividends now. And we need to be mineral, rare earth minerals and supply chain independent too. And hopefully we will be sooner rather than later.
Z
Zach Abramowitz38:45
So a couple of follow-ups. First of all, I think you talked about did the CCP even know if their machines worked? I think one of the things that's been underrated in the entire experience in both Venezuela and Iran is, you know, you can put soldiers through war games, but there ain't nothing like live fire. And I say this all the time when people talk about Israel as this great beneficiary of the generosity of the United States. It's like we love the United States and Israel and I'm an expat so I love the United States as someone who grew up there but the fact is like Israel is very often the one putting these weapons into test environments and figuring out whether they work so that when the US military goes into a conflict there's always that. But then the actual experience of working together, I said to friends here I said don't be surprised if someday there's an operation that doesn't involve Israel that really involves American interests and we get called up to participate in that because there's now joint partnership on that.
So that's on the one side on just getting things to work. But you brought up the AI point. So first of all I got to ask you have you seen this Dwarakish and Jensen Wang interview that everyone's talking about and it sounds like you're sort of leaning Dwarakish on this.
K
Keith Rabois40:08
Well I mean I think there's exchange of multiple points and multiple dimensions so I wouldn't simplify it. I do believe that export controls on advanced chips are a good thing for the United States national security. I understand that Jensen has business interests that are contrary to that. I agree with many of the other points he made on many of the other topics actually. So I wouldn't subscribe to 90%, I'd say 90 some percent of the points he made I agree with. However, I think he's wrong on export controls, but he's also very self-interested. China is a big market. He wants to sell to China. I don't think we should.
I don't think the argument that Huawei is going to be discouraged from innovating and discouraged from competing with the United States because we sell them Nvidia chips is true. Every industry that China has ever cared about, they have leveraged American technology to use that in the short term to both reverse engineer and give them time to stand up their own industry. You can look at automotive, you can look at any industry and it's like almost a silly argument actually in my mind.
Z
Zach Abramowitz41:21
I'm guessing you have a little bit of insider knowledge here. Do you feel like the Trump administration is getting the right advice on AI?
K
Keith Rabois41:32
They're getting lots of great advice both from people in the administration like David etc. but also people outside the administration have access. Jensen has talked about this publicly like he can pick up the phone and call the president and give him his feedback and advice. The Secretary of Commerce definitely cares what he thinks, etc. in treasury and all this stuff.
Z
Zach Abramowitz41:57
It feels like Sam Altman may have open lines as well and which is interesting to me because the president co-hosted an event in DC on AI last summer. So they're very amenable to advice, counsel, feedback from successful business people like period including in AI which I think is a great thing.
K
Keith Rabois42:18
That said, I think people's views on what to do about the specific topic of export controls on advanced chips, there's a lot of debate about in the administration, outside the administration, publicly, privately, in Congress, in the House, and the Senate. It's a pretty complicated issue in many ways. So I do think different people can disagree, which is a healthy thing. I think it's not a bad idea. Like, people should debate this stuff. That is a healthy way to generate policy is to have sophisticated people make arguments. I just personally don't believe like so I'll give you a proof point. My understanding is that remember DeepSeek?
Z
Zach Abramowitz42:59
Sure.
K
Keith Rabois42:59
It's hard though by the way to remember it. It feels like Open Claw and everything else sort of erased our memory of DeepSeek. Well, one of the reasons why is though it shot in very some very clever nature to DeepSeek, but it was shot at the top of the app store immediately and then it crashed and burned. My understanding from some pretty good insiders is the reason why it crashed and burned was they didn't have enough chips to support the usage. They had hacked together some second generation chips, these gaming chips and were able to propel it into the market, but then when it took off, not enough to support it. Otherwise, who knows if it'd be outranking ChatGPT or Claude or Gemini or whatever. And I think that alone is a successful application of American export controls.
Z
Zach Abramowitz43:47
Before I let you go, I want to run one DeepSeek conspiracy theory I have since you brought them up. So, this guy from Highflyer, right? He's, you know, this super quant super smart guy. He's got this massively high performing hedge fund where he's been using AI and we find out that he's read Attention Is All You Need at some point in maybe 2020 and realizes, 'Oh my gosh, these transformer models are a real thing.' Now, someone who had the math background that he has and the knowledge he has certainly had to have known that when he released this paper that it was absolutely going to tank the market on Nvidia. But he also must have known that I don't really have enough chips to make this work. So therefore, when the market drops on Nvidia, I can buy it all back and I can finance this entire thing on a short.
K
Keith Rabois44:41
I don't know. I don't maybe I don't follow the, I mean, obviously I'm familiar with their incredible success and his incredible success, but I don't follow the hedge fund world that carefully. I'm not particularly interested in it at all. I'm interested in the business of using technology to propel the future and improve society, which is very different than what hedge funds do. I don't think it's bad what they do, but it's just not intellectually interesting at all to me. So I honestly don't follow it very much. So I don't know if they followed that strategy or not.
Z
Zach Abramowitz45:14
Got it. Listen, hey, I really appreciate the time and thanks so much.
K
Keith Rabois45:18
Pleasure to be with you.