transcript

Speaker 1:
[00:00] The two richest guys in the world don't just drop 400 million on a single deal. The biggest VCs, maybe, that's still like a lot. So like, how'd this guy show up with 400 million to drop? So there's $10 trillion, whether it include all the other hardware coming out, the rest of the ecosystem, et cetera. Maybe it's five to $10 trillion market cap up for grabs. And it's not an area where there's a lot of startups, which is like really shocking. I don't want it to have data centers in space. I don't want to have nuclear reactors in my backyard. And I think that, you know, it'd be a lot easier if we just made a freaking better computer. Remember how big of a mistake you made? There's always redemption. It's just you have to show the vulnerability. You have to say, I fucked up. Even if you don't say, I fucked up. You have to show a scar, a wound or bleed a little bit.

Speaker 2:
[00:41] What actually matters in AI right now? Better models or better businesses? A few years ago, the focus was intelligence. Who had the best system, benchmarks, and breakthroughs? Increasingly, the real question is economic. Who captures the value, how it's priced, and where the bottlenecks are? At the same time, we're hitting limits. Compute is getting more expensive, and new approaches from photonic computing to specialized hardware are becoming more necessary. That creates a tension. Software is easier to build, but harder to differentiate. Meanwhile, industries like finance and biotech still require deep expertise and real-world validation. In this episode, I try to understand where the real leverage is shifting across AI, hardware, and pharma. I speak with Martin Shkreli, American investor and businessman. We're talking a very exciting day. We have OpenAI about to make an announcement. I thought I'd start by just asking for you to broadly reflect on the OpenAI versus Thropic. What's happening here? How do you make sense of the ecosystem?

Speaker 1:
[01:55] Yeah. Obviously, disclosure, my wife sat OpenAI, she's one of the first people there, so I have a bit of a bias. I have a bias against Thropic as well, so I have a dual bias. But basically, just trying to be objective. If you monetized ChatGPT, both enterprise and consumer fully, you'd actually have, I think, quite a lot more revenue than today. Just case in point, my financial software company, we have about 10 licenses to Anthropic, and it's supposed to be 20 bucks a seat, and we get a bill for $1,000, and we're like $1,500, and that's good, five to seven X, kind of what we ask for. And OpenAI could do that to their customers too. Their customers will pay it, they don't really care to bargain either way. But if you try to wring every dollar out of your user, that has some adverse consequences, I think, and OpenAI I think has figured this out and wants to sort of boost their ASP or whatever dollars per seat. And so they're working their way upstream. But I think the traditional VC entrepreneur advice would be, don't rinse your customer if you can't help it. And I think that for an anthropic sort of reason, they want to price through the roof with all these overages. I mean, people complain that two prompts eats up their entire allotted amount. And so I think if OpenAI did that, they probably could get their today's enterprise revenue to something like 100 billion. I know some people would be very surprised to hear that, but I think that's the case. And then on consumer, the same thing with ads, right? So they might have, and again, this sounds preposterous to anybody listening, but if they did the same monetization effort as anthropic, their real revenue right now would be about 200 versus their 30. So I think this idea that anthropics taking the lead or something like that, I mean, it's pretty far-fetched. Now, having said that, obviously, we're a super impressive product, really great people at work there. I think Dario is, you know, he's not my favorite guy. You know, obviously, incredible entrepreneur, beautiful business he's made, but, you know, the whole, like, you know, he tried to scare the world by saying, AI is going to kill you, it's going to take over the world. Nobody really believed that, you know, nonsense. He says, well, let me shrink that down a little bit to, oh, the AI is going to hack you. And everyone's like, oh, well, you know, it could happen. And so it's kind of the greatest marketing trick, but, you know, it's, it's, it's, I don't think it's in great faith. And, you know, I do think to some extent these guys are true believers of the Doomer hypothesis. And, you know, so it's hard to tell what's real and what isn't, it's just conveniently also good marketing.

Speaker 2:
[04:31] I, you know, I was really confused by the, the anthropic DOW incident, you know, about a month ago now, why there was so much support for, for anthropic on this. To me, it felt as obvious as like a bullet maker saying, hey, we're only going to sell you these bullets, but only if we can, you know, help influence how you use them, you know, separate from US law. Now, I know I'm definitely simplifying it and we'll get, I'm sure we'll get some defenders to come on the show at some point, but, but to me, it kind of felt obvious that they, you know, we should sort of leave things to, you know, private companies shouldn't be dictating, you know, foreign policy. I get their concerns about domestic surveillance, but we have laws for that to protect us from government overreach, and it doesn't seem like the realm of private companies to get into. What am I missing there?

Speaker 1:
[05:22] No, I don't think you're ever going to say anything. I mean, I think that the idea that there would be some super governmental authority is a little scary. You know, I used to, back in the hedge fund days, when the markets were shaky around 2008, I actually asked, I would ask some hedge fund friends, would you rather buy bonds in Johnson & Johnson or bonds in US government? And it's a very interesting question because in some ways, at the time, it seemed like the US could default, the financial markets could collapse. Johnson & Johnson is not going to collapse. Anywhere in the world where there's consumer health care, pharmaceutical and medtech, they did all three. Every part of health care, every country in the world buys Johnson & Johnson. If the dollar could go worthless, J&J will survive. United States of America would not, at least financially. Now obviously, the US is very, is extremely resilient. It's the greatest country on earth, et cetera. But at the time, there was a question as to whether bond yields would be better for a high-quality corporate versus a sovereign. And certainly, if you look at some sovereigns like Brazil, J&J bonds do yield less than Brazilian bonds, they're riskier. They're more likely to default than J&J. And part of that comes from this, again, super governmental, like J&J, in a sense, is its own country. It has employees, it has a couple hundred thousand people, it's got a lot of revenue. It's bigger than the country I come from called Albania. So in a lot of ways, Anthropic, in some ways, is sort of going to be its own entity. And this is sort of a network state kind of thinking, corporate state in some ways. And with all that revenue and all that power that comes from having a powerful piece of software, I could see that sort of getting to somebody like Dario's head of thinking about, oh my gosh, with great responsibility comes great power. And it's sort of obviously way overwrought, whereas people, I think, at OpenAI and other companies, even Google, are just sort of like, yeah, it's just software, man. I have jobs to sell it. It's not that big of a deal. Meanwhile, he's like, I am Spider-Man. I must be Batman. He wants to be a superhero. And it's like, obviously, he's selling software really well, but I think the theatrics are kind of ridiculous.

Speaker 2:
[07:31] Yeah. Let's get into other parts of the ecosystem. Are you still a huge NVIDIA bull? I don't know if you saw the Jensen versus Dorkesh, but what's your latest thinking on NVIDIA?

Speaker 1:
[07:41] So I'm starting a sort of a hardware startup. So I was shorting these stocks, the quantum computing stocks, and I had to learn computing. Quantum computing is really useful for one algorithm, one Shor's algorithm, so unless you're a code breaker, and I don't know of any corporate code breakers, there aren't really too many avenues there. So I started thinking about frontier computing in general, and I stumbled upon photonic computing, and now I have a startup in this space. Interestingly, we're already a publicly traded company, but that's sort of, you know, your configuration kind of doesn't matter much, whether public, private or, you know, even a not-for-profit. So I think photonic computing is really exciting. It's not going to challenge NVIDIA anytime soon, but I think that what I've learned trying to program some AI, you know, a couple of pivots ago before we settled in on financial software, I learned that if you do the flame trace or the profetto of a really zoomed-in, like, nanosecond level instructions, basically all the GPU is doing is Matmul's so-called M, generalized Matmul on the Nvidia hardware. And light does Matmul's for free. So you have a Matmul machine in light. God kind of made light to do Matmul's whenever it diffracts. So you get the speed of light Matmul. And if we did, if we had an optical computer, I think it would have like a thousand X to a million X performance, depending on whether it's speed or speed and energy included. And so, like, you know, flops for watt or something like that, flops for joule. So I think there's like, you know, I think eventually, you know, Moore's Law, you know, Jetson's already said it's dead. I think everyone's seen that. But, you know, I don't think processor speed is going to really improve at all. And I think just shoving more energy through the system, you know, making bigger systems, these are like the only things we can do. We need to do something. I don't want it to have data centers in space. I don't want to have nuclear reactors in my backyard. And I think that, you know, it'd be a lot easier if we just made a freaking better computer. And I think that's what's going to happen. So there's Neurophos, they raised $150. There's a company called Flux Computing that renamed to Olex. They raised $250. We're out there. There's a couple of little smaller guys too. And I think that one of us, hopefully all of us in some ways, will, you know, and there's obviously unconventional, which has its, it's a little opaque exactly what it's doing, but, you know, very highly funded effort. It raised a billion dollars, I think. And, you know, I think it's time for a new computer. And ultimately, you know, we're working on the hard parts of this. So it sounds great, but there's a bunch of catches. And one of the catches is non-linearity is very hard to do in optical computing. We think we figured out a way to do it. The other catch is there's no memory in optics. So we think we figured out a way to do that. So I think that, you know, we're going to see a lot of like bankruptcies, we're going to see a lot of startups in this space, but I think it's sort of time for the end of silicon and the end of, maybe not the end of silicon, maybe the end of transistors. And so there's $10 trillion whether to include all the other hardware coming out, the rest of the ecosystem, etc. Maybe it's $5 to $10 trillion market cap up for grabs. And it's not an area where there's a lot of startups, which is like really shocking. Like the amount of agent startups is in the thousands and thousands. I can name all the, every single optical computing company on my hand. And compared to biological computing or quantum or some of the other alternative computings, those aren't really that exciting. So Photonic is, many people I've talked to at OpenAI or Meta and other places, they sort of all agree inevitably Photonic is going to be here. And I feel like whether it's five years, 10 years, or even 20 years, it's still worth making that computer because you need to hand the baton off. I don't think Nvidia is going to stay asleep, but I also don't think that this is necessarily an easy thing to sort of embark on. Most of the time these bigger companies will resist kind of uprooting themselves.

Speaker 2:
[11:31] Yeah. Well, first, I just want to say you did this epically hilarious debate with somebody on Quantum that we'll link to. And, you know, he was a bull, and you got into the technology, and he couldn't really defend himself there, although I'm sure he's a nice guy.

Speaker 1:
[11:48] Yeah, a very nice guy.

Speaker 2:
[11:48] Yeah, yeah. But it just goes to show that this is very complex stuff.

Speaker 1:
[11:53] Well, Quantum sounds futuristic, right? It's a Quantum that sounds really cool. Well, you know, it's not that cool if you understand that computers basically be measured in sort of two ways. You have the number of threads, number of operations at a time, kind of parallelization, which is why NVIDIA is so cool, 16,000 threads, versus your average computer's 16 threads. And then, you know, it's a thousand X, literally a thousand X more threads. And then you have clock speed, you know, and you basically multiply the two, and clock speed is, you know, in the gigahertz for, you know, both of those computers. So you have a thousand X speed up with NVIDIA, if you can parallelize the instructions. The quantum, the fastest quantum computer I'm aware of is a megahertz, which is, you know, your calculator is faster than that. And it's literally a thousand times slower than a normal computer, it's single threaded. And the only way the tortoise beats the hare is when you have an algorithmic advantage. The algorithmic advantage exists, you know, in shores, but it doesn't exist anywhere else. And what's interesting is you have these market caps at the bottom is about an aggregate of about 100 billion. And so, you know, obviously Naveen's company, my company, the other couple of companies I mentioned, why don't we have a market cap of aggregate 100 billion? Why not? In fact, because it's actually useful, the world, nobody wants to do integer factorization, everyone wants to do Matmul's. Why don't we have a market cap of 500 billion? I think we will, you know, and so that's kind of the pitch there. And I know, you know, your company, you know, is I think one of the major investors in this new thing, in a beans company amongst many others, you know, all-star list of investors. And, you know, I don't know what they're doing at all, but, you know, I understand that, you know, the race to make the next computer is on, and, you know, some people are sort of half stepping it, like there's some great companies doing ASICs, which is fantastic, TPUs even are sort of half steps. You know, you're still on transistors, and you're still limited by, you know, kind of the problems that come with, you know, scaling transistors, which are well known, and we'll eke out just a little bit more performance out of that, but I think to get 1000x improvement, you need to do something pretty different.

Speaker 2:
[13:57] Is there a particular milestone where you will know you're being successful or on to the right thing, or how are you, you know, don't share anything confidential, but how are you thinking about it?

Speaker 1:
[14:09] Yeah, I think you have to, you know, we talk about this internally, because taping something out is sort of a, you know, to some extent, it's a waste of time. Unlike electronics, you can actually set up these experiments in free space. So that's the nice thing about light is like a light lab, you know, an optics, photonics lab is pretty easy to set up. I'd say that unless you get 1000x or bare minimum 100x improvement, you know, it's not even worth taping out. Our rivals are using SRAM and some other, you know, analog electronics. I think you have to stay in all optics. Never, the conversion kills you if you go from electric to optical back to electric back to optical. I mean, you know, you might as well not make the chip, in my opinion. But, you know, folks want to try it anyway. They want to do their best Grok imitation. And, you know, Grok, you know, is obviously great, great success from an investor standpoint. And so I think you stay in optics. If you can make a whole LLM in optics, then you tape it out and you hopefully get the performance characteristics we're talking about. But actually, what's interesting about it is, you know, the question kind of is revealing, because what's not well known about this is that guys have been doing this in academics for many years now. So UT Austin showed a optical transformer. Guy at UCLA showed optical non-linearity very recently. So work has been done on this. That's pretty impressive. It's not like biological computing. We're still trying to get like the machine to add two numbers. I don't know how we're going to do it. The super neurons might magically start thinking. Optical computing is sort of here. It's just a matter of scaling it, some of the other challenges I mentioned. So I think if you can get some of these tricks down, we're actually closer than we think. But I still think you got to be realistic. No matter how much of an optimism you need as an entrepreneur, you have to be a little bit realistic that you're not going to kill a video overnight, and you're not even going to kill 5% of it. Look at Grok. Grok took this very small, ultra-fast segment of the market. They didn't take it all. But for people that really need very low-latency Grok's, it's a decent solution, and that's a $20 billion win, and there's nothing wrong with that. So I think you might carve a niche out that, oh, optics is really good, say, for video, because video and optics, they're naturally the same thing. So maybe you do that or something like that. And I think that you'll sort of see it. You'll know when you see it. And I think that the key, in my opinion, is investor mismatch. So the problem, if you go to your investors and you say, I'm going to have something 10x better than a video next year, well, you better deliver that. And if you don't, if it comes out at 5x or 3x, your investors are probably balking another round because ultimately video is not staying still. And your investors, your employees, everyone around you is just sort of gonna feel despondent that you're not getting anywhere. Google had the power to live through lots of crap with TPUs, right? Like they just didn't work, they didn't scale. And finally, they got the revenue break. It took 15 years. So you need to make sure investors, I've talked to some people recently where I said, I might need 20 years. Are you ready to sort of like live that through? And for most of the answers, no. And so we're focused almost exclusively in corporates instead of VCs, because I think that corporates sort of get that, especially corporates that have been through it, they've been around for 40 years, they get that like, it's worth it to stick around and wait. And look at NVIDIA, if NVIDIA had no revenue, this is a good thought experiment, zero out all the revenue for NVIDIA of gaming, and look at the annual financials. What's the net present value of NVIDIA back in 2020? It's actually the same as the net present value of it with or without gaming GPUs, because ultimately it was worth the wait for that gigantic hockey stick. So if you knew 25 years ago, oh, in 20 years, there's going to be a thing called AI, and we're going to dominate the hardware for that, you still should have invested, you still should have made it no less of your biggest investment than as if they had the revenue in the GPUs. So I think that this idea of really long-termism, I almost want to think about it like a lab, the same way OpenAI sort of thought about their business was, we're going to monetize something in AI, we don't know what it is, eventually it'll hit us and we'll know, this is a more directed effort, but I think that there's something powerful about some investors now being willing to wait 10 years and the fact that there's some entrepreneurs out there like SSI, they're willing to actually tell their investors, you're going to have to be really, really patient with this one, don't expect cursor here.

Speaker 2:
[18:46] Yeah, absolutely. To that end, you're super bullish on NVIDIA, is there any major MAG7 or major tech incumbent or emerging incumbent that you're bearish on going forward?

Speaker 1:
[19:00] Oh, good question. It's so funny because in pharma, you have this dynamic as well where it's very hard to lose your long-term edge. You have to be patient because these businesses are so valuable that they have a hard time falling apart. I thought you were going to ask me the other question, which is a lot easier to answer.

Speaker 2:
[19:23] But you can answer that too.

Speaker 1:
[19:25] Yeah, I think Facebook's met its greatest thing ever. You know, I'm just such a big fan of Zock. And this amazing thing, this distribution he has, it almost doesn't matter if he makes his AI himself. Like, all he has to do is, like, you know, he could just literally take an open-source model. And it doesn't, you know, the problem with AI, I think, is that if you're open AI or anthropic, pushing for more intelligence, like going from 150 IQ to 160 IQ, the kind of kids I went to high school with, you know, it doesn't make them more interesting. You don't want to sit down with those kids at the lunchroom anyway. You know, you mostly, Instagram users and Facebook users, most of their questions are sort of normalish stuff. You know, nobody's asking the Facebook AI, prove the remont hypothesis. You know, that's something that Dubai and all these other guys want to do, but who actually needs that? I think that the average person doesn't need questions that complicated, and it'd be cool to see it happen. But for Zuck, I mean, he's got the distribution, and he's got the best product guys in the world, you know, hence the Snapchat battle that he sort of won and stole from. And I think that he's going to continue to do that and win. So I don't know, I feel like, you know, it's so hard to like lose to the Mag-7. I do think we're going to get surprised. Like if you look at the history of capitalism and big stocks, you wouldn't recognize most of the Dow 30 from a hundred years ago. United Cigar, you know, there's some companies you've never heard of. And American Can, no joke. You know, and we laugh at that now, but like that's going to happen again. You know, some of these companies, maybe half these companies are going to disappear. And, you know, I certainly think Apple is like one of these things you raise your eyebrow at because, you know, obviously it's hard to count them out. It's hard to bet against them, but Apple's one, if I had to pick one that I wouldn't like bet on, you know, I wouldn't say I'm ready to bet against it. But Apple seems like it's now old hat. And there's nothing that special or charismatic about the company and their products. They used to be super cool and sleek and edgy. You know, that design, like in fashion, you have to stay fresh. You have to stay pushing the envelope. You have to stay dangerous. And Apple was dangerous 20 years ago. It was new and nouveau. It's kind of whatever now. And I feel like that's a monopoly that can start to fracture a little bit, depending on how other players play. Google is another one where it seems like no matter what they do with Gemini, it just kind of sucks, and people don't want to use it. They have the distribution, but they don't have... They just sort of feel like a pocket protector. They feel like Yahoo in 2000. They feel like, you know, it's not cool, it's not interesting. You know, people want to do other things. So we'll have to see if Google will fail or what will happen. Obviously, TPUs are an exciting business. Search is never going away in some ways. But it feels like there's some changing of the guard. Moments like this is when the Xeroxes become Xerox. And you know, those types of things, those transitions start to happen before your eyes and you don't necessarily feel it or notice it right away, but they do happen. And I don't know what Mag-7 is going to become, a Lag-7, but eventually you're going to get the Lag-7.

Speaker 2:
[22:29] Yeah, you could, to your point on Meta, I mean, you could tell they're a live player by them, you know, in the arena, you know, paying billions of dollars for Alexander Wang, for Daniel Gross, for Nat Friedman. Like, Apple's not doing that. Exactly. They could attract, they have the money to, they're almost sitting on a resource curse of the iPhone, but not willing to make it.

Speaker 1:
[22:49] Yeah, to buybacks. I mean, why do buybacks? Like, you know, if you can't pick the winner, and again, this may sound strange or stupid, but like, if it were me, I can't pick the winner. I don't know what company to buy. You can buy five or 10% of each of these. You could buy, you could, you know, give, you know, buy Koi or A16z, here's 100 billion dollars, you know, go invest it for us. You know, but, you know, doing buybacks is like, historically, it's the worst thing you could do. I mean, it really is kind of this admission that you don't know what, you know, you kind of don't know what to do with the capital, you don't know where the future is heading. It's a financial, it's kind of stuff I do. Like, it's a very financial engineering kind of nerdy, you know, not that innovative thing, and it kind of like, you know, it's indicative of their business, I think, being sort of like, they can't release new products, they're just refreshes the same old products. You know, all the new stuff they did has sort of been bad other than AirPods. So I feel like, you know, it's also going to be other companies. You know, Microsoft's sort of also sitting on a perch that's not great. I mean, you know, they make OS software, don't need OS software anymore. You make office tools, you can vibe coat those, you know, et cetera, et cetera. It's kind of like, you know, that's sort of a dangerous perch. I don't think anything's going to happen at Excel, but I feel like, you know, there is a fear that all of Mac 7 could be in trouble and that, you know, maybe it is hardware that rules the future.

Speaker 2:
[24:08] Yeah, we see, you know, the entropic open AI, you know, threatening startups or even very big companies on a seemingly weekly basis. And, you know, Alad Gill, the legendary investor, tweeted out the other day something like, if you're a company in AI, you know, if you're an AI company, consider getting sold the next 18 months. You're just speaking to the power that these big companies have. And I think there's also just a broader question as to whether the AI sort of platform shift will look like the Internet and that the application companies will accrue a lot of the value, or if it'll look more of like the cloud in the sense that maybe the hardware layer has the better margins, and, you know, like we're seeing with NVIDIA right now.

Speaker 1:
[24:52] Yeah, nobody knows, but one thing I think is that software's not dead yet. And I think that, you know, we're making this financial software, we've been working at it two years. We're sort of still in beta, we have some nice revenue, we're almost profitable, we haven't launched it yet, and it's this sort of, you know, many people try to make this thing, it's sort of a Bloomberg killer, you know. Killer's a strong word. You know, Bloomberg's been this like software monopoly for 40 years, and the reason it works is because it's actually very hard software to make, and your customer expects, I was just sitting down with a customer, he expects the bond prices when he looks up a bond, and he's like, yeah, you're right. And, you know, bonds are something that like 90% of these apps, like Perplexity Finance, for example, doesn't have bonds. You know, half Wall Street or even 60% or 70% of Wall Street's related to bonds, and if you don't have your bond prices in there and you vibe coded it, well, a bond API player takes six months to do a contract. Like, these guys are so sleepy, they don't want to talk to you, they just don't want a new customer. And if you want bond prices and you have to like make sure, oh, when you look up this French bond, it's correct, like, that's something that like that kind of taste and that kind of like attention to detail, it's just not something you can vibe code. And it's so hard to get right. Customers, a customer at Citadel, he's not vibe coding something. He wants to buy something. He wants to buy something that works all the time. And if it breaks, he wants to yell at somebody and take his business somewhere else. That's the kind of thing you get at a software. And the relationships matter. Sal, my wonderful Bloomberg rapper, whatever it is, you know, those kinds of things sell software. I sort of think the panic is, most panics aren't really what you expect. Something's going to change, but I think some of this panic is unrealistic. And then, again, going back to the Bloomberg example, we have like 1,800 news relationships with news companies. And if you vibe-coded it yourself, I've seen people do this, they get five. And it's like, well, the trader cares about the 1713th news source that's going to screw up his portfolio. He doesn't care about the five that everyone know about. So I feel like vibe-coding is kind of, it's a fantastic thing where you can cut your software costs a little bit or just raise the ability of your software. But it doesn't replace marketing and sales. It doesn't replace product and taste. It doesn't replace personal relationships. And it doesn't replace the customer who wants to have really well-made software. It may increase your ability to compete, but I feel like the changes where software stocks have dropped like a rock, those are probably opportunities to buy. And I feel like depending on the stock, you can get some bargains out there.

Speaker 2:
[27:38] Yeah. That's encouraging. I want to shift to biotech and pharma. If you were building a pharma company today, I'm curious what you might pursue. It's sort of one range of questions. The other thing I'm curious to get into is sort of, you know, you had this great peptide debate on TVPN, which we loved. And, you know, it seems like that's all the rage and only going to be increasingly so. And yet it seems you're bearish on it. And then also just GLP-1s and what that means. So I'll let you take it away.

Speaker 1:
[28:17] Yeah, I couldn't have more. I couldn't have higher conviction. The peptide thing is all bullshit. You know, it's, you know, self-diagnosing and self-administering medicine that you don't know anything about. I mean, this is the literally craziest thing ever, right? I mean, you know, you have to be out of your mind, you know? And I understand, like, all right, you're Patrick Carlson. You're a super billionaire. You're, you know, maybe one of my investors, Naval. You know, you kind of like you want to take health into your own hands. You have a lot of doctors around. Okay, that you're a special case. But you want to impersonate those guys. I think that, you know, that's where that's where it sort of ends. I mean, you know, the star, you know, the star asset for these guys is drug BPC 157. I spent a few days on this thing. I took a few days out of my life, wasted 1% of my year to look at the stupid drug. It's the worst drug ever. And, you know, nobody understands medicine unless you've actually made medicine successfully. My first, I got my first FDA-approved drug, you know, it was like 4 or 5 years ago. We got, you know, 3 or 4 years ago, it was like accelerator approval. We finally got final approval, which is cool. Accelerated approval is actually contingent. So, you know, I got a drug through the FDA. There are not that many. There are about 30 or 40 a year they got through FDA. So, you know, it's a pretty hollowed haul. And unless you've done that, you know, you really don't know what it looks like. Drugs are like seemingly simplistic. You know, you look at, you know, this Claritin on my desk or something, and you sort of say, oh, you know, I could do this too. And the reality is, you know, it's pretty freaking difficult. In fact, some people say, well, how hard would it be to make this drug or something like that? I said, imagine the SpaceX landing, like the chopstick, you know, or the landing on its head. Drug development makes that look easy. And, you know, because I want you to, in your mind, shrink down the molecule to the molecular and nanomolecular size, you know, where angstroms are your unit of length, you know, 100 nanometer. And imagine the atoms of, you know, some kind of organic molecule. And imagine those atoms interacting with proteins atoms. And you're telling me you understand what's going on there. It's really not easy. And I think when I see people do like 10 minutes of pharmacology and they feel like they're experts all of a sudden, no offense to the guy who's debating, Max, who's a very nice guy. You know, they, you know, sitting down with Max and like teaching in pharmacology over a few hours is like, to me, it's like you should not be advocating for this stuff. I mean, it's peptides, for example, are kind of one of the lowest desired drug classes in pharma for the longest time, because their half-lifes are generally in the minutes. And GLPs are frankenpeptides. You know, they take GLPs and they add these fatty ass, and they conjugate these different things that make the half-life better. And they sort of become biological peptides that are like, you know, they're hybrid drugs, basically. Amgen called them peptibodies. You know, so peptides in general are just this terrible asset class for pharma, because you need drugs that have long half-lifes. So BPC7 is not dangerous, because it's gone within like 60 seconds. The time you inject it, it's dead. And so it doesn't have the time to do anything useful. Any drug guy could tell you that. I think what this really is about isn't about pharmacology or science. This is about rebellion. You know, I'm a rumble. I get this. You know, it's, I don't want to pay Pfizer. I don't want to see my doctor. I don't want to pay insurance. I don't want to get a prescription for, F all that. You know, I'm going to do this myself. And I, okay, you know, I get it. I mean, I'm, you know, it's crazy that, maybe it was crazy for me to tell Congress to go F itself. And, you know, it's just as crazy to go, you know, you know, make your own medicine. But, you know, DIY medicine, this is the dumbest thing ever. So to me, as a pharma guy, I would focus on, you know, and this is where like, you know, I don't want to tear jerk anybody, but like somebody sent me a message today, or last few days, you might have seen on social media about a West Point guy, you know, such strong military network. His son just died. His two more sons, or two more kids that have the same disease, they're going to die too. I just had a baby boy. Look, you'd move heaven and earth for anything for your kids. And, you know, how many people in the world know how to develop rare disease drugs? Well, I got one approved. You know, there's maybe a hundred guys on planet Earth that can say that. And I'd happily, you know, try to develop a drug for this guy. I got banned, you know, from the farm industry for sort of political reasons. I still might want to try and do it in Europe or something or get a special exemption. But, you know, there are people dying. You know, there are people dying of hundreds of rare diseases, tons of cancers. Go where the hardest stuff is. You know, go where the people are suffering the most. Because, you know, that's where, you know, ironically, that's where the money is. And for the longest time, these kind of cosmetic diseases actually weren't money makers. And GLP's been around 20 years. And finally, a lot of farm guys are scratching their heads like, why now? We've had these for a while and they just finally got mainstream after 18 years of watching this drug class be kind of like, you know, they finally broke through because they are fantastic drugs. But for the longest time, the way you made money in pharma was you went after the most heinous, horrible, terrible illness and made a change, made a difference in somebody's life. And that's how you made money. And I think that that's still the case. It's still the case. You have to go after the toughest cancer. The Duchenne muscular dystrophy has been a passion of mine for the longest time. The first company was named after muscular dystrophy. That's, you know, it's a death sentence disease. You could get a million dollars a kid if you save their lives. I think Elon, you know, Elon's not a drug company, but he's sort of going to do the same thing with Neuralink. I mean, I think Neuralink's going to be a hundred billion dollar plus, maybe even two hundred billion dollar company because those patients, and there's many of them, they want to walk, they want to talk, they want to be productive members of society, and he's making it happen. And again, that's a million dollars per patient productivity, which insurance is happy to pay. So I feel like the rare diseases, the Alzheimer's, the terrible cancers, this is where you go. I mean, the playbook hasn't changed. I think that the science is still really hard. I think AI is going to help a lot there. Coming up with new ideas, fundamentally the drug game is an idea game. You have to sit, toss and turn and say, I wonder if you inhibited IL-1, would that work for Alzheimer's? Nah, I don't think it crosses the blood brain barrier. Inflammation is not that important, but what if we did this? And you keep thinking and thinking and you have these ideas, before you even put a drug into an animal or in a Petri dish, it's a theory game. And AI can actually do that. It can actually zip through thousands of ideas and come up with something. So I think there's something to work on there. And I feel like maybe it's Anthropic that does drug company, maybe it's OpenAI or maybe somebody else. But there could and should be a way to make this work. The drug companies themselves don't take to technology very easily or quickly. They're probably not the place to do this. But you still have to be a drug company. And I think that's what a lot of the Valley doesn't get. They're like, we want to tackle the farm industry without being in the farm industry. And it's like, no, you have to go do trials. You have to go do the stuff that nobody wants to do. And that makes drugs kind of boring and shitty. And that's why Vivek and myself are like two big interlopers in pharma. And we're both kind of not loved for lots of reasons, because pharma is this old guy's game. I mean, the average pharma CEO is like 60. There's no Patrick Collison in pharma, and there never will be. It's kind of a tough space to be an entrepreneur in. It's a pretty ivory tower. Yeah.

Speaker 2:
[35:58] Well, you guys were to some degree, so it's you guys. But I hear you on the barriers being there. So just to close the loop on the predictions, so are you thinking that in the next year, we'll sort of see that peptides don't really work? Because a lot of companies are getting funded in this space. Do you think it's just all going to collapse? And then I'm also curious on the GLP One side, Eli Lilly, this is just an Amazon announcement today. It's over a trillion dollars. Do you think it's going to be a five trillion? How big can this get?

Speaker 1:
[36:25] Drugs have a big, big problem. A friend of mine talked to Warren Buffett about this. He wanted to sell his drug company. It was privately held. He wanted to sell it to Warren Buffett. Buffett said no, and he said the problem with pharma is IP. So we have to re-create our business every 10 years. Drug goes generic. I have to go find another one. Are you effing kidding me? So imagine being open AI, or imagine 10 years after A16z. Actually, your patent on venture capital expired. You got to go find a new business. Go do PE or go do lending or something. It's like, I was just getting started. I was just hitting my stride. And in pharma, you literally have to go do something else. And so Osempic's generic in like a year or two. Osempic's pretty good. And so the problem with drugs is you make a better drug and you reach this asymptote where it's like, nobody's making a better hypertensive drug because we have solved that problem. You can lower blood pressure. You don't need another one. And so once that patent expires, it sort of permanently has lifted the bar for that disease. And eventually, GLPs, they're perfect. I mean, you really can't get a better one. So at the end of the day, Osempic going generic, Mujaro eventually it's 2040 or something, but eventually going generic. So markets price long-term asset value, long-term cash flows. So Lilly's kind of screwed. They have to find a way to perpetuate their product. And one of the ways you could do that is manufacturing. But the problem is you sort of have to design a drug that's hard to manufacture. And if you want to keep it in the long run, and J&J figured out how to do this where, for the longest time, small molecules like Claridin go generic very quickly because any Indian company can make it. But antibodies even still haven't just now are starting to be able to get disintermediated by generics. So in some ways, the more complex the molecule, the harder it is to make, the more casual you're going to get out of it. Because somebody else may not be able to make it, the patents expire, and they expire quickly in pharma because you have to develop the drug over 10 years. So your 17 to 20 year patent life gets cut in half because of your development timeline. So you have 10, 15 years of cash flow. Lily has brilliantly staged each GLP to come out one after another in this really smart way. But at some point, the consumers and the insurers say, I don't need super GLP Max Pro Plus, like good old Generico Zempik works, and it's $3 instead of super GLP Max Pro Plus, which is $250, consumers have zero brand identity with these companies. Nobody gives a crap that their drugs from Lilly versus Merck versus Weiser, and they just want the cheap version, and I think that plagues these companies. So I think that's unfortunate. I do think the peptide stuff, it's going to be weird to see this happen because drugs are, the reason we have an FDA and it's good is because snake oil works for people. People want to think, they want to believe. So if you think BPC 157 is working for you, you're not going to stop taking it and you love it, and you think, oh, well, it works for me. And you might think that, but you have to look at the trials. But I can't, nobody can convince the populace that statistics matter, you know, it's so hard. Your individual response, even though it's a non-response, it's just random noise around a distribution, you want to believe that it's relevant to you, that it changed something for you. So it's going to be hard to like defeat that. I think regulations are the way to defeat it, but you know, we're in this regulatory environment that's really laissez-faire. I feel bad for it, because there's actually drug companies, there was a Huntington's drug that was promising, maybe not perfect, and FDA rejected it. And so the Huntington's disease is deadly disease, horrible genetic disease. Companies spent a couple hundred million dollars making this drug, and the people who are suffering, dying from this disease can't get it, but I can get BPC-157, totally unregulated, stupid medicine that doesn't do anything, I can go inject that. It's sort of tough, I think that's terrible. And I feel like peptides are so safe, that's the one benefit, that maybe people will just trick themselves into doing it, it'll be this weird Veblen good where it's like, oh yeah, I take this, this and that, that's my stack. And it's not doing anything for you, but you're spending $1,000 a month on it, and you feel better about it, and it gives you some placebo effect. I mean, it's a business, it's not a business I want to be in. Business of pharma is, again, Neuralink, casting it as a pharma company for a second, I'm literally giving a million dollars of your productivity to this person. I'm cutting the costs of everyone taking care of them and feeding them and doing all this stuff for them. Now they can work. And the medical system that maybe has been spending a million a year per patient on this person and their productivity, you flip that around and it's an amazing, beautiful thing that you will get paid for. To get paid for something, you have to show the value. And if peptide, you can't show the value, yeah, you're going to get some suckers to buy it, but that's not really exciting. So I don't know why I'd want to be an entrepreneur in this space when there's, it's easy, I guess. It's like a go to market, go to consumer, direct to consumer kind of thing, where you can finagle your way into getting some demand, but this isn't going to scale to like, the reason Lily scale is product really freaking works. And that's how you went. So I feel like, good luck to it. I mean, I just don't think it's going to be a great place to play.

Speaker 2:
[41:42] Yeah, so we did some AI ecosystem analysis, and then we just did some pharma analysis. I want to end on a little bit of off the wall question, which is harkening back to your interview with Laura Shin a few years ago. Excellent interview where you analyzed sort of SBF's case and get some advice. I'm curious to do a thought experiment, which is imagine SBF gets out early. He gets lucky, there's some sort of pardon, or he just gets out on good behavior. Maybe it's unrealistic, but imagine he gets out in the next five years, something like that, a short sentence, maybe 10 years. What advice might you have for him in terms of a redemption arc or just reacclimating to society?

Speaker 1:
[42:21] Yeah, I mean, I talked to Sam. I like him a lot. I think he should be out. I think the hard part here, and having talked to even the Trump administration about this, is the hard part is Sam has got to show he's a normal guy. And I think he struggled with that. He's obviously not that normal in a lot of ways. How many of us go to MIT? How many of us start a huge company? How many of us pick Anthropic as the first? So there's something about Sam that people don't trust, and there's something about Sam that he's got to really let his human side out. And one of the problems here is maybe he doesn't have one. And I don't say that is a bad thing necessarily, but as somebody who understands his path in life, being a prodigy, being pushed on stage as this intellectual guy, and being shoved into this different room to room math camp, MIT, you know, Jane Street, you know, all this stuff, he maybe didn't have the time to like be a normal person. And I think that that's what makes it so awkward and so untrusting. People look at him, they're like, I don't know about this guy. There's something about this guy I don't trust. And he's got to fight hard to make it real, that he's a real dude with real life, with real friends, with real emotions, and then people will feel bad for him. But until then, it's like, how do we know this guy doesn't come and pull the wool over our eyes again? And I feel like, truth be told, as much as Anthropic was a successful investment, if you look back at the announcement, Anthropic was, those profits seemed like they were going to go in his pocket. And not in the depositor's pocket, not even in FTX's pocket. So if you look at the round, it was Series A by Sam Bankman. And it was like, well, at the time, that was the tell. That actually is how you would have told. Because where is this 400 million coming from? I mean, most people don't have 400 million dollars. Bezos doesn't put 400 million dollars in deals. Zuck doesn't put, the two richest guys in the world don't just drop 400 million on a single deal. The biggest VCs, maybe. That's still like a lot. So how'd this guy show up with 400 million to drop? And the answer is obvious. He took everyone's money to do it. And that's wrong. And you could almost allow it if, like, hey, he was going to go to all of his depositors somehow. He was going to give us 105 cents on the dollar. Okay. You know, whatever. I didn't ask for this gambling, but if it helps me, I'll take it. And that's kind of my case. That's what I went to jail for. And so at least I gave the profits to the people. But I think that no matter how bad, no matter how big of a mistake you made, there's always redemption. It's just you have to show the vulnerability. You have to say, I fucked up. You have to say, even if you don't say, I fucked up, you have to show a scar or wound or bleed a little bit. And I think that his thing is like, he wants to fight, but he's not connecting. And connecting is so important. And I feel like his upbringing, his sort of like MIT math camp, Jane Street life has led him to this position where he's hard to feel sorry for. He's hard to connect to him. And I feel like if he does get early, which again, I'm advocating for, but it's been met with this very stiff resistance of who really is this guy is a human. You know, that's the question I get back from the Trump camp. I think that, you know, once he does that, not only will people in politics feel bad for him, America will feel bad for him, and America will want to give him a second chance because the promise is there. The guy's obviously a genius, one of the best businessmen ever. But how do you take that and get the trust, get the, I like this guy, I want to root for this guy? That's what he's got to focus on, and he's brilliant, so I think he might be able to do it, but it's going to take some soul searching, might even take some pain. He's obviously going through pain right now. But I'd love to see him out, I'd love to see what he could do for society, and I feel like as long as he can be a human, which is, you know, some people, we joke about them being lizards or robots or whatever, and there's a human in everyone. And, you know, I think this experience may actually, it's going to sound really crazy, might actually be the best thing for him. Because, you know, it's the pain of learning who you are, the pain of going through this pressure chamber that prison is and forcing you to reflect on, like, what did I miss? What did I, what am I, what is this all about? And five years in prison, and then coming out, and, you know, I've heard a rumor that there are investors who have already lined up for Sam's Next Night. I'm not kidding. I've heard this from very serious people that, and the numbers are like, they're staggering, that they're willing to drop hundreds of millions, or even billions on Sam's Next Thing. And so obviously, you know, there's something here, and to throw a life away, you know, that like such a unique life away, you know, and I was making fun of Sam, you know, I was digging a little bit. And now, like, you know, I feel like, you know, he has to go through that journey, that hero's journey of like, he's got to hit rock bottom. And then I think, you know, perhaps, you know, the time will be where he gets out. Maybe it's this administration, maybe it's the next one. But I see a big glorious comeback if he can find this humanity, find this, like, you know, this new person.

Speaker 2:
[47:38] Yeah, that's a great note to wrap on. Martin, thanks so much for coming to my situation. Thank you, Erik.

Speaker 1:
[47:43] See you soon.

Speaker 2:
[47:46] Thanks for listening to this episode of the a16z podcast. If you like this episode, be sure to like, comment, subscribe, leave us a rating or review, and share it with your friends and family. For more episodes, go to YouTube, Apple Podcasts and Spotify. Follow us on X at A16Z and subscribe to our substack at a16z.substack.com. Thanks again for listening and I'll see you in the next episode. This information is for educational purposes only and is not a recommendation to buy, hold or sell any investment or financial product. This podcast has been produced by a third party and may include paid promotional advertisements, other company references and individuals unaffiliated with a16z. Such advertisements, companies and individuals are not endorsed by AH Capital Management LLC, a16z or any of its affiliates. Information is from sources deemed reliable on the date of publication, but a16z does not guarantee its accuracy.