AI Is Making More Millionaires Than Anything in History w/ Salim Ismail & Dave Blundin | EP #181

0:00 We may need to rewrite the whole paradigm, Peter, because what's happening now is another double exponential on top of that with AI accelerating everything so quickly, including path to revenue. 36 unicorns in half a year. I really, really hope you know the young entrepreneurs coming up appreciate what a moment in time this is. This is exponential thinking. This is entrepreneurial thinking. There is a bet to be made right now if you're an investor on where are you going to invest to ride this curve because it's not slowing down. Is it real estate? Is it data centers? Is it chips? Is it power? Cuz all of these things are going to make somebody hopefully somebody listening right now huge amounts of wealth. This is an unstoppable meta trend.

0:48 Now that's a moonshot ladies and gentlemen. Hey everybody, Peter Demandis here. Welcome to Moonshots and our WTF just happened in tech episode here with the extraordinary Salem Ismael and my partner Dave Blondon. I'm still thinking the architect of AI, the alchemist. Alchemist of AI. All right. Peter, you're the pope of hope. I'm the pope of hope. All right. Well, hey, listen. Uh, a lot going on and excited. I just want to do a shout out to our our super fans out there who are just giving us incredible love for this episode uh every week that we do and we care deeply about it right our goal is to deliver to you the news that really matters the news that's shaping our

1:33 companies our entrepreneurial journeys uh every aspect of our lives so please share this with your friends and family uh the goal is to give people a sense of optimism a compelling abundant future that's being driven by this technology. It's not all dire crisis news network out there. Anyway, welcome Dave. There was some incredible messages in last week's episode about how awesome you are and I don't think our, you know, our fans know enough about you. So, if you don't mind, I know a lot about you. I've been your roommate since, you know, our undergraduate days at MIT. But, uh, what what is Link Exponential Ventures? What do you do? just so folks can really appreciate how awesome you

2:20 truly are. Well, we we manage about a billion dollars of seedstage money, you know, here in Kendall Square, right between MIT and Harvard. Um, but you know, fundamentally what we do is when I graduated immediately tried to start a company and you know everything is working against you at that age. You know, it's it's almost impossible even to get an apartment. You know, you have no credit score. There's no funding at the time. uh there were no incubators and accelerators and so I've spent you know big chunk of my life actually now trying to identify everything that slows down a new founder especially in today's day and age almost all the deals are AI companies and we see many many companies reaching multi-billion dollar valuations in you know 2 three years so you got very young teams very fast timelines and

3:05 there's so much that we can do to help them and so you know we started with office space and funding then we added accounting we added big data. Uh a few weeks ago, we bought an apartment building. So, you know, one of our observations was Merur added from the day we invested at the seed stage to today, it added $20 million a week of value. And we're like, okay, how much does it cost to buy this apartment building? About six million bucks. Okay, buy it. Because if we can save a team like that, even one week in their growth cycle, it pays for itself times five. And so just doing everything we can at link to accelerate these young super smart AI teams. And so you know funding is is part of it. Mentoring is part of it and and so on. And the returns have been

3:51 extraordinary to to you're a unicorn incubator. Unicorn incubator. Yes. That's right. Hey, I'm curious. Uh guys, what did you do for the 4th of July? Anybody explode anything? Uh went up to Vermont. tried to get away from AI as much as possible for a day or two. It's hard to Salem, how about you? Uh, I sat in the garden with a glass of wine and a glaze look at mice. Ah, okay. Well, I was in Montana building explosives. It was so great. I had my two boys there and we were like taking apart m80s and firecrackers and separating out the gunpowder from the filler. Uh, I mean, we had an arsenal. It was extraordinary. I mean, when I remember when I was a kid, I could buy potassium nitrate, charcoal, and sulfur, make my own gunpowder, and then I remember buying potassium

4:38 perchlorate. Potassium perchlorate is great because it generates its own oxygen and if you mix it with you know charcoal and sulfur any any reducing agent and I remember throwing so this is there's a book called um poor man's James Bond and they talk about you know so you take like a old film canister you put an MAD fuse fill it with gunpowder put it in body putty and I built that threw into my friend's pool to see like this splash and What occurs next taught me something in physics. Liquids are not compressible. The explosion literally cracked my friend's pool in half. Oh no. So yeah, I'm not sure you should be talking about some of this stuff publicly at knocked

5:25 on the door from I'm not I didn't know that story after all. But you cannot buy you can't buy this stuff anymore. Uh, you can still buy what we used to do is buy the SDS model rocket engines and then just crack them open and scrape out the insides. Well, you can do that with fireworks, too. But I want to buy the pure chemicals and teach my boys how to be true, you know. Anyway, let's move on. I may need to find another podcast YouTube. All right, let's let's move on. Uh, our first area, of course, AI supremacy and speed everywhere always all at once. another and crazy week in AI. Uh Dave, uh tell us about this TechCrunch article here. Yeah, so 36 Unicorns. I I really really

6:11 hope you know the young entrepreneurs coming up appreciate what a moment in time this is. You know, I love to tell the students that, you know, last year there wasn't a single self-made billionaire under the age of 30 in America. That was the first time in 15 years that that was true. you know, prior to that, you had your Mark Zuckerbergs, then you had your your Larry Pages and and um you know, every just a whole litany of internet, you know, young billionaires, but there's this huge dead spot over about a 15-year span. And now we're in the the biggest peak of my lifetime by far, and you'll never see anything else like it again, but 36 unicorns in half a year on this kind of accelerating rate. and then you know a couple that we've invested in and then you know prior to this window you also have liquid AI and a bunch of

6:56 others from the from the prior half year this is not normal this is the opportunity of a lifetime and and you know the rate at which you need to move to keep up with it is is just you know a full sprint so the question is is this is this new normal right because uh we're seeing this crazy valuation increase and speed to not just valuation but speed to revenue like never seen before. Yeah. Look at the next slide for the for the revenue view of it. So this is also not normal, right? Like you know the uh the timeline to to profitability or to significant revenue is super super short. Uh that means you're stable. Like if there's ever a financial collapse, you can switch to profitability instantly. Uh that's

7:42 so just for those those who are listening and not watching us on on YouTube which is the preferred mechanism for you guys to see this stuff but heck uh median months to $1 million of annual revenue. Uh before 2020 it would take 16 months. Now it's taking people 5 months uh to get to 5 million. Used to take 41 months nearly 4 years. Today it's taking 13 months just one year. So this is a this is a 4x acceleration in uh in getting significant revenue as a startup. Well, Peter, remember when we graduated, you know, I went to Micro Strategy and then immediately back to Boston to start a company. I was building a company that was a true hyperrowth company of its era. Founded it in ' 91 and got it

8:27 liquid in 2000. So nine years to get to a $1 billion valuation and probably what 20 30 million of revenue run rate. Nine years that was really really fast at that time. Compare that to the numbers on this slide. Uh it's so you know here we're comparing 2020 23 to prior to 2020 which is really you know the prior decade. Uh so if you look at the decade before that it was even slower. Yeah. Selene these are all exos right? They're all exos and you know there's a step change from 2008 where you could build a company and use cloud services and therefore you could scale you could take all computing costs and off the balance sheet and make it a variable cost right and that was the birthplace of the exo

9:14 we may need to rewrite the whole paradigm Peter because the what's happening now is another double exponential on top of that with AI accelerating everything so quickly including path to revenue which is a huge huge thing as as Dave mentioned this now stabilizes companies very very early and that's a powerful point to be at. Are you proposing the third book in our EXO series? No, I hope I never write another book again. The first one almost killed me. The second one 90% killed me. It's a horrible I mean, you know, I'm more proud of it than we I thought we'd be, but but damn, it was a tough process. Dave, what you were saying? Well, if you look at the next slide too, I think one of the things that that isn't mentioned here is the headcount that gets you to these numbers is lower than ever before

9:59 by a lot. You know, for me to get to get my first company to a billion dollar valuation, I had to hire 200 people. Mhm. Uh for a young entrepreneur like the cursor team, you know, or the Merkore team, the comfort zone of these management teams, you know, 30, 40, 50 people feels a lot like college friends. And actually when you meet the people they're they are a lot like college friends. So it is so much more fun to build one of these companies because you're in that I know everybody. I know what their capabilities are. We have you know one or two kegs for the whole gang is just so much more manageable. And I I bet there's a there's a curve that looks like a peak uh of value per human being and fun value per fun per human being that basically starts falling off after

10:47 a after a company population in the 40s or 50s. Oh my god. When I have a whole presentation on this, but when you get to about 200 people and more like 100, but around 200, there are people in your company you don't really know. and in fact you can't even remember their name when you walk by them. That is a different world in terms of comfort and management from the one layer company where you know everybody directly. It's just a whole different world. You know the other direction you know we're all we're all hypothesizing the company of one person that reaches a billion dollar valuation that's coming soon. But that sounds really lonely to me too. I think the sweet spot the perfect happiness sweet spot is right where we are right this minute. Every week I study the 10 major tech meta trends that

11:32 will transform industries over the decade ahead. I cover trends ranging from humanoid robots, AGI, quantum computing, transport, energy, longevity, and more. No fluff, only the important stuff that matters, that impacts our lives and our careers. If you want me to share these with you, I write a newsletter twice a week, sending it out as a short two-minute read via email. And if you want to discover the most important meta trends 10 years before anyone else, these reports are for you. Readers include founders and CEOs from the world's most disruptive companies and entrepreneurs building the world's most disruptive companies. It's not for you if you don't want to be informed of what's coming, why it matters, and how you can benefit from it. To subscribe for free, go to dmandis.com/metatrends.

12:20 That's damandis.com/metatrends to gain access to trends 10 plus years before anyone else. I'm going to predict though that we're going to see a have a problem which is once these companies go through that super exciting phase and stabilize. They're going to go through a lot of conions as they lose the buzz that got them there, right? Um is a great problem to have. It's like we could all wish for that kind of a problem, but I think they'll start you'll see a lot of angst, a lot of founders leaving, etc., etc. No, they got to keep dreaming bigger, right? It's upleveling constantly to the next level. I mean, vibe coding startups are here to stay. That's the story here in Stripe. Uh, cursor to $500 million ARR, Lovable at$7 million, Bolt at 20 million. So, these are uh these are a

13:08 set of technologies. You know, one of the things that makes these unique is they are technologies that'll enable entrepreneurs to build other technologies and companies, right? So they're a base layer. Um any other comments on this? No, I think you're right. I think I think when you look at the list that we had a couple slides ago, uh the base layer companies are about twothirds of what you see there. Mhm. Uh you know, RHF and uh foundation models and and uh so forth. You're starting to see though uh some multi-billion dollar value vertical companies now starting to pop up and there will be many many more of those. So So to answer your question, you know, is 36 a lot? Is it going to go up or down from here? It's definitely going to go up more. Yeah, we have at least a couple years of

13:53 really good uh very rapid expansion of that number and then you know AGI and then you know who can predict after that? Yeah. Then it goes insane. I mean the interesting thing is you know the IPO market is just beginning to open but the merger and acquisitions market is wide open and and as we've seen it's gone absolutely crazy in that regard. Yep. Um, so this is another interesting slide. It says the new reality vibe valuation. So we're seeing crazy valuations that are being uh, you know, sort of opening bid $9 billion, $10 billion, $30 billion. Never seen anything like it. And so Vibe valuations are when an investor bets big on a Gen AI startup with little uh, traditional

14:40 metrics like AIR uh, you know, annual rate of return or cash burn. And this is from Vindra Matherur. Uh this is this is a problem you've solved Dave, right? Because um you know most people don't have access sufficiently early and then you're stuck at the later valuations when the company has blossomed and like do you invest at you know $120 billion into XAI or Yeah. No, we we never do. So it's it's kind of scary when you do. But uh you I was talking to General Catalyst last week actually. They they've more or less abandoned seed stage because they're managing $60 billion now and they they need to write these big checks to keep that money moving, but they're going to try and rebuild their seedstage capability. They hired a

15:26 new guy to do it. But it struck me that, you know, doing seed stage and doing 60 billion of asset management are really incompatible with each other. Yeah. And uh what's your what's your average what's your average check size at Link XPV? average uh you know our opening checks are often you know half a million a million two million something like that uh we'll allocate 7 to 10 million per company over time we've left billions of dollars of pratar rightites on the table over the years and pratarites meaning we have a right to to maintain a certain ownership stake in a company but if it gets to a 2 billion 4 billion uh you know in some cases recently 10 billion valuation we can't even come close to keeping up with our parata rights I mean one of the questions I get asked all the time. I'm I'm guessing Salem uh

16:13 you and Dave do as well is like what AI company should I invest in? It's like really you want to first of all you're not going to get an allocation in anthropic or open AI or XAI those those offerings when they do secondaries tenders or their next it's they're snapped up instantly and then you're left with the public companies right which are I mean there is growth but you're never going to see the real valuation growth so you're either in an AI venture fund like Link or or what you're going to your local university and trying to find a startup and put it in there. Well, the other option is pick a domain that you know really well that and then

16:58 invest in AI companies where you can gauge the value ad and the differentiator from others, right? That's the only other way to do it. Pick an area of existing passion that you have and go that go down that route. Well, the other thing you can do which I love just as a life strategy is is find three, four, five, six venture funds that you really really like that are seed stage. Yeah. That route their pratar rightites out to their investors and say, "Hey, look, we have capacity here. Uh, we can't keep up with it." What we do is we just email our LPs. Virtually none of them read their email and react, you know, and usually these deals like, you know, SpaceX or XAI, you have a week or two to make up your mind. You don't have a month. And so, but if you're if you're the one investor that actually reads

17:45 their email and pays attention, then there are great opportunities that come your way through those, just pop open the email. And then, you know, if it doesn't fill with those LPs, then the new investor just takes it, you know, and two weeks later it's gone. So, it's it's a pretty good strategy. All right. The big news this month has been uh the release of Super Grock. I love that. I love we went from Grock to Super Gro and and Grock 4 and uh GPT5 is coming. So this is Sam Alman on a podcast with his brother. Not much content here, but worth hearing. What is the time frame for GPD5? When are we going to see this? Probably sometime this summer. Okay, that was a big data point there. So Chad, you know, GPT5 this summer, um, one of the things that's interesting

18:31 about GPT5 is they've made a proclamation, no ad influenced answers, and they're really trying to strengthen user trust. So any predictions here gentlemen? Oh yeah. Uh I think it's funny. Poly market seems to think that it is less like you know 25% chance to Yeah. Let's go here. There's the here's the poly market bet. Yeah. Yeah. Yeah. I I was almost sure like more like 80% July because there's all these pictures of Sam, you know, giving little hints and and also the competitive pressure kind of would line up with Grock. So I was really thinking July this will be a great test of the wisdom of crowds. It is you know uh and we've written about that extensively. So here's what

19:17 it says for the you listening. Uh the poly market on GPT5 release is 26% by July 31st and 93% by December 31st. Yep. Um all right. We'll find out. I love I love this. Um this is uh uh Gro 4 scoring 35% on humanity's last exam, 45% if they're using uh their reasoning model. And you can see, you know, I'm always trying to track these models against each other. And I love it when they say the models IQ is so so and so, but they don't always do that. The last time I saw it on the uh sort of the uh

20:03 IQ scale, it was at I think it was GPT03 at a IQ of 136. Yep, that's right. Uh but so let me read this about humanity's last exam. Uh, it's a challenging AI benchmark developed by the Center for AI Safety and Scale. AI consists of 3,000 expert-crafted multimodal questions across a 100 subjects including mathematics, physics, biology, medicine, humanities, social sciences, and more. Humanity's last exam tests the limits of an AI at the frontier of human knowledge. And I found this interesting that the questions are crowdsourced from nearly a thousand experts globally. So I, you know, if you're a listener and you got an extra minute, you got to look up the actual exam and look at a couple

20:50 of the questions there. I defy anyone to answer a single question. I mean, unless you're the world's leading expert on that particular, you know, there there's questions from physics, from math, from philosophy, from history, covers every area. But I I defy you to answer one question correctly. It is so hard. And this test is supposed to last like 10 years. Yeah. So I have the I have the I have the counterpoint on all this. Please. Yeah. Um you know this this I find this mostly irrelevant. And the reason I say that is this feels to me like asking a question like oh my god that calculator can make a calculation way faster than a human

21:35 being with a piece of paper and a pencil doing long division. Well of course it can. So what? Um, I think that's true. And to Dave's point, if you put I looked up a few questions, right? One of them is, is there a god? And you ask an AI that question and they give you the standard. Here's the different ways you could look at it. Monotheism versus deism versus atheism versus whatever, whatever. It's not clear if there's a god. You could take each point. It depends on your personal perspective. Da da da da. It's a non-answer. If you looked up Wikipedia on any of these topics, you could kind of root through and find the answer and answer the question yourself. some of the details ones. This is essentially like saying um we're automating Wikipedia and so can you answer the question? Can Wikipedia answer these questions better than human being? Of course it can because it has

22:20 all the data to do it. So having it be able to answer these questions does not really make that much of it. It it doesn't feel meaningful to me at all. I think the really big question is what do we do with it? Yeah. Yeah. No, you I think you're on to something really really important and interesting and you know all the dystopians are like oh my god it's going to run away from us and you know but it actually when you're using it at this level of intelligence it feels exactly like Jarvis in an Iron Man movie where you know Tony Stark is saying here's what we need to do and Jarvis is like oh I never thought of that but I can you know I can create that for you and it it actually is behaving almost exactly the way they visualized it in that movie and so when it when it ratchets up this curve and you ask like I need to design a protein that does exactly this or I

23:07 need to design a new type of bicycle that does exactly this. It gives you great great answers back and does the work for you but it never has you know the the word reasoning is being really bastardized when we talk about what's happening in reasoning here is iterative reprompting of the same old thing which squeezes more performance out of it. But it's not really like human reasoning and brainstorming. And so it's kind of a bummer that we borrowed the word reasoning to describe what's going on here, but it is so useful as a tool for a creative person. And 100% uh in to do actions and get things done and do almost any inim imaginable thing faster and faster. Huge. For example, I've got a lot of pain in my shoulder. So I had an MRI done as to what was going on. It's taking me weeks

23:53 to get a doctor's appointment. I uploaded it into an AI and it's given me the answer in 2 seconds. That kind of speed of decision making and getting to conclusions would be unvaluable going forward. Is it an alien growing inside your body? What's I seem to have a bone spur and that's impinging on the nerve and it's causing pain and nothing. So I had I had the exact same surgery with a bone spur. A piece of recommendation for you cuz I did an actual test. I had the same surgery on both shoulders with the same surgeon 10 years apart. I did it when I was 50 on my left and when I was 60 on my right. And on my right shoulder, I injected exosomes postsurgery uh one week after and two weeks after. My recovery was like twice as fast. So,

24:40 okay, squirt some. What's the name of your surgeon? I I'll bring him. He's at Kurland Joe about here in LA. He's amazing. I So, you know, I expect to see massive. So just to come back to this the the speed of getting to outcomes and decisions and getting things done the speed of that will accelerate radically which is fantastic and now comes down to what do you want to accomplish and I think that's the bigger questions that we we're going to start pondering. So, uh, here's a big one. We keep on cursing about I keep on tweeting about how awful Siri is and, uh, thank God, uh, you know, it's lit literally I I know that you've been an Android user, Seem, for a while now. And just what we saw at Google IO was epic. And so, I am

25:25 going to buy a I've got an, you know, the latest iPhone here, but I am going to buy an Android phone as well, uh, so that I can start to play with all of Google's incredible uh, tech. And also because Siri sucks so bad they're pushing me away. Having said that, uh the here's the news article from Bloomberg. Uh Apple ways using anthropic or open AI for Siri rather than an in-house AI team. Let's take a quick listen. So Apple is now evaluating for the first time using a third-party model either Claude from Enthropic or ChatgPT from OpenAI to power Siri. Obviously, as we all know and experience in our day-to-day lives, Siri is not very good. Now it's exploring maybe using claude or chat GPT instead in order to get new features out the door more quickly and make the voice assistant more appealing.

26:12 You talk to people at Apple though the word on the street there is it's all about anthropic and so anthropic is really the focus of of Apple for this new generation of Siri. They're using anthropic to power a lot of their internal AI technology. But again, as I say in my story, Anthropic wants billions of dollars at a scale that doubles annually for Siri to be powered by claude. And so Apple is now taking a close look at Open AI as well, which has historically given Apple extraordinarily favorable terms. Sound of desperation or smart move in terms of what Apple's doing. It's a smart move in response to desperation. I love that. Yeah. So interesting, right? Because Anthropics also partnered with Amazon. Um, and if they've got Apple and Amazon

26:58 and billions of dollars of revenue, uh, they stand a good chance of of really rapid acceleration. Well, Dave, we we talked a week ago about Mirram Marotti's new startup and Ilio Suskver's new startup, both doing new foundation model companies from scratch, huge valuations. Uh, part of the logic there is the musical chairs, there aren't enough anthropics to go around, as you're pointing out. And so there's a version of anthropic where it's an independent company for 10 years, but there's another version where it gets acquired by Amazon or Apple and the other guy gets deprived. And this is, you know, just just happened recently with scale AI as well. And so, um, there's room in the market for more foundation model companies, but, you know, the price of poker is very high now and and going up quickly.

27:44 That's a really good point. I mean if you get if you get alignment of enthropic with Apple like a a you know a captured entity open AI you know has been dancing with individual partners for a while Google the same you're right there's there's room for a couple of others I have some I have some comments here on this um you know this is classic corporate innovators dilemma right you cannot do disruptive innovation in any big company to from now on there's it's not an accident that that Microsoft essentially invested in OpenAI or Amazon put money into claude or whatever. Uh you cannot build that capability. The mindset is too different between Apple being a consumer products company largely and the mindset needed and being

28:31 under 25 to even start and kind of a foundational model company. You have to partner and invest and do this. Apple I'm surprised that it's taken Apple this long because they do all their manufacturing with a partner in Foxcon. Why aren't they building a core capability like this on the edge? And part of then you go ahead, Dick. Well, I'm sorry. I didn't mean to k. So, you know, just a reminder though. So, today Nvidia is worth more than Apple like this exact day in world history. Um, Apple never in a million years could have predicted that 5 years ago. In fact, Apple treated Nvidia like like dirt. But Apple has the chip manufacturing. They have about a third of TSMC's manufacturing capacity. They were in a far better position to become Nvidia than Nvidia was, but had no recognition of the opportunity whatsoever.

29:18 So here's the problem here. Here here's the problem. And Simi and I have discussed this. It's about founder companies, right? So Jensen as the CEO and founder of Nvidia is driving it. And when he says right, everybody goes right. The same thing with Dario at Enthropic, you know, the same thing with Sam at OpenAI. But Apple has lost Steve Jobs, right? Steve would have been all over this and he would have set up the team that would have just basically driven it. And it's it's innovator's dilemma. But during this period of exponential growth, you need a dictatorial passion-driven founder who has the has the, you know, complete faith of his board and his or

30:06 her board uh and their management team. Mhm. Uh, I mean, Sim, do you agree with that? I somewhat disagree. Okay. I think I think yes, if you're in that industry, like for example, if you're Facebook and you need Facebook to do something different and you're Mark and you're Zuckerberg, you can kind of get that done and it's important to get that done. But the control systems, remember the big companies are optimized for two things, right? Efficiency and predictability. This comes from John Hegel and John Cely Brown's work. Um, and so you're trying to deliver the iPods at the same AirPods at the same quality into a million retail locations for purchase by everybody and their grandmother. Your your entire focus is on that. You're you cannot do disruptive things when you've got that machine

30:51 trying to deliver gross margins on a but you can do what Steve Jobs did, right? He took his Mac team and he he moved it off on the edge. You have to do you have to do that edge innovation in a stealth way. Google is doing it with X, right? Google X. Um, uh, this is what we try to do with Brick House or Yahoo. You have to do that type of model. And then the second level of thing is when it starts to succeed, don't bring it back in because it won't fit. If it's really disruptive, it's not going to fit in the mothership. You have to spin it off. Now, lately, companies have kind of tried. I mean, like Amazon did with with AWS in one sense. Absolutely. Or Google is this is why Google split up into Alphabet. spinning off Whimo and whatever. So, this is the only model that works. I've looked at me

31:37 in my ecosystem, we've looked at disruptive innovation across probably 200 of the Fortune 500 companies in detail. And this is the only modality that ever ever ever ever ever works. It's the only one. And so, you have to do that or you acquire that disruption and you leave it on the edge the way uh Microsoft has done with OpenAI or Zuckerberg did with WhatsApp or you will kill it inside the core organization, it won't work. And let me give you the a negative side, right? Walmart realized somewhere in the mid 2000s that it had to compete with Amazon, that this e-commerce thing was not a fad. So they set up a team inside Bentonville and they said to that team, "All right, we have the best distribution logistics in the world. Go beat Amazon." Within 18 months, the immune system had killed it. Could all the existing managers said,

32:24 "We have our own capability. We should be optimizing for our own stores. Look at that PE ratio. It's never going to succeed." da da da da. They did it a second time. Second time they put the team at the edge and they said, "All right, go to the edge, away from the core, and do the same thing, but still use our existing systems. We have the best in the world." Within 18 months, the business had figured it out and killed it. They did it a third time. Third time, they made an amazingly courageous decision. They said, "Go to the edge and build your own independent supply chain distribution, etc. Even though we have the best in the world, we have to figure this out." Started to succeed. The business got excited, pulled it back in, and killed it. And it was on iteration number four that they finally did it separately on the edge and only after it achieved critical mass did they start stitching the backends together. But in that intervening six or

33:09 eight years Amazon was gone. You know interesting story and lad and ladies and gentlemen this is why Selma Mel is incredibly brilliant and why I love him so dearly. It's just that we've seen so much of this across so many we have so many data points over 101 15 years of looking at this and looking at big companies in great detail. It just doesn't work any other way than for looking for the disruptive uh uh technologies bring buying them or investing them and leaving them on the edge. Right. Dave works. Oh, well I just happened to be in Bentville, Arkansas. You know, they were my biggest customer at the time. Uh Walmart as this was going on. So just one little interesting nugget on your story there, Salem. And when Jeff Bezos needed to recruit Rick Dzel, he needed one brilliant retail industry guy to

33:57 come out to Seattle and make Amazon like organized and structured and and retailing because everyone he had was kind of a computer science PhD type. And so he needed Rick Dazelle. But he actually took Sam Walton's book and there was a sentence in it where Sam predicted that a disruptive innovator will eventually come along and gut Walmart. And so he circled that sentence and took it to Rick's house because you know the Walmart gang is super tight. They're like a family and trying to break up a member of the family is almost impossible. And and Rick needed to talk to his wife about this like a big change moving from Bentonville out to Seattle. But the book is what put it over the top. He circled Sam's own words predicting the disruptive innovator innovator will come for us and we won't be able to defend ourselves. And can I can I flip to the other side

34:42 for a second? In in 1990 when I was starting out in my career, I was building software systems and I went to Bentonville and we were trying to sell Walmart something. We passed they had me sign all these crazy NDAs to go into the data center and I was like, why would you have me sign NDAs to go through a data center? It's just there's nothing there but racks of service. We walk past a particular room which has like flashing lights and dials and blinking things that I've never seen. And I said, "What the hell is that room?" And they said, "Oh, that's why the the NDA that's our satellite control center." Yeah. Yeah. And they had their own geostationary satellite because Benil happens to be the geographic center of the US, the continental US and they were doing everybody else was doing batch updates to their AS400s at the end of the day and and the buyers didn't know what to buy for the stores till months

35:29 afterwards. whereas these guys had real-time inventory management, credit card reconciliation, uh distribution, logistics, all that stuff figured out in and and that was delivering them 15% better margins, which in the retail industry is insane. And that's why they wipe the floor with Amazon. When you can successfully integrate disruptive technologies as a big company, you have a street uh streets ahead of the competition and they show that's what we call a moonshot, right? U exponential tech with a crazy idea and a high revenue. I would love to meet the executive at Walmart. If anybody's listening in the news, I'd love to interview whoever the executive was that got the board to do this. That's like that's hardcore. All right, we're going to we're going to move on here. So, uh let's take a look.

36:17 This is the state of the union on largecale AI models. uh we're seeing uh Gro 3 and GP and GPT 4.5 at the top of this in 2025. You know, I looked up where XAI is right now. So, the XAI cluster has uh 340,000 Nvidia GPUs today. 150,000 H100s, 50,000 H200s, 30,000 GB200s of the Blackwell architecture. uh their goal what Elon has announced is a million GPUs by December 31st. So um I mean he is scaling I think faster than anybody uh and it's pretty extraordinary. Any comments on this Dave

37:03 before I move on? Well I mean the scale on the left side is somebody out there listening should give some terminology to this so we can stay on the same page like you know calling it 1E26 or 1E27 training. Uh, so like a pedaflop is a huge amount of compute. That's one E15. So that's way off the bottom of the chart here. So then it goes Xoflop and then zetlop, which would be a 1. So you're getting near the chart. Nobody has terminology for scales of this size. But we do need to talk about it in other than an exponential. So that's my challenge for the audience like somebody name these scales. One E26 is officially a shitload of computing. It's really hard to imagine. You know, Peter, you took a tour, right, of the XAI data center. It's crazy. I'm an investor in XAI, full

37:49 disclosure. So, yeah. Yeah. I mean, it's see this this it's, you know, Raiders of the Lost Arc where there's this rows and it goes to infinity like these rows of boxes. It's like that. It just Well, one of the key points uh that we need to talk about on the back end of this is how do you power this, right? We're talking about every week that the limits are not the GPUs, the limits are our electrical available power system. So this is an article um that came out. Elon Musk purchases overseas power plant to support massive XAI data center. So XAI acquired a fully built overseas gas turbine power plant expected to house nearly a million GPUs and be powered by

38:35 that power plant. So it's an unconventional approach. to basically, you know, get power as quickly as possible. I remember um in about a year and a half ago when XAI announced its Colossus system and I was listening into an investor Zoom with with Elon, he said to, you know, this was in May before they built it and he said, "Okay, we're going to build it by the end of the summer, like within 3 months, and people thought it would take like 5 years to build what he wanted to do in a grade 100,000, you know, H100s." and he said to do this we have to corner the US market on helium and I mean just the level of thinking of doing whatever it takes and here's an example of that. Yeah. Yeah. Pretty awesome.

39:20 We had that kind of epic podcast with Eric Schmidt that should be out in about a week I hope. I hope every but we talked a lot about energy supply. He's he's very knowledgeable in this whole area of uh energy supply and here we're talking about 2 gawatts. I mean, this is a big big bold move, but remember, we need a 100 gawatts by 2029. So, there's a lot more. So, he would have to do this 50 times over just and this that just gets you the turbines, right? It doesn't get you the actual power supply. This is all going to be fossil fuels. I think the important point here is it's disruptive thinking. It's unconventional. It's no, we're not going to go through the permitting process. No, we're not going to order. We're just going to buy the and move it over here and use it. classic first principal thinking. Um, just to calibrate for the folks

40:05 listening, one gigawatt is about what a major city in the US uses. So this is like two Dallas Fort Worth, right? This is crazy. Calibrate, you know, the new Elon Musk could become Chase Lockchiller. He's the other guy thinking this way. He's doing, you know, Abalene, Texas, uh, Stargate, you know, $500 billion buildout. And he thought the same way like where am I going to go to build this? Okay, there's there's actual natural gas flare off that's not even being used and power generation that's that's so abundant that it's not even being used in this one location in the country. So, I'm just going to He was in Denver. I'm going to pick up my pick up my ass and move to Abalene, Texas. And that's where we're going to do it cuz I want So, this is this is exponential thinking. This is entrepreneurial thinking. This is when I know we're

40:50 going to get to, you know, ASI digital super intelligence. when you say to your AI, "Where can I get the power I need?" And it says, "Well, I found this power plant over in Europe. Um, it costs this much. Let's buy it and move it here." When it makes those kind of suggestions, then I'm impressed. Mhm. All right. You mentioned this, Dave. Why don't you lead us off on this one? Yeah. Uh, so, you know, we we took a bet around the office many years ago, maybe. I don't know. Um, what what company will be the first to hit a trillion dollar market cap in the history of the world? Is it going to be Google, Apple, Facebook, or Microsoft? because those are the only candidates who else could possibly get there. At the time, Nvidia wouldn't even have been like you wouldn't even vaguely consider it an outside possibility. And this is what Selma is saying all the time. It's always the outsider upandcomer

41:39 founder-ledd visionary that blindsides you and is now the most valuable company in the world. this, you know, you could argue whether it's overvalued or undervalued, but the demand for compute for AI is going to be 10,000 times higher than the supply for the foreseeable future. There's a coming wave of technological convergence as AI, robots, and other exponential tech transform every company and industry. And in its wake, no job or career will be left untouched. The people who going to win in the coming era won't be the strongest. It won't even be the smartest. It'll be the people who are fastest to spot trends and to adapt. A few weeks ago, I took everything I teach to executive teams about navigating disruption, spotting exponential trends a decade out, and put

42:26 them into a course designed for one purpose, to futureproof your life, your career, and your company against this coming surge of AI, humanoids, and exponential tech. I'm giving the first lesson out for free. You can access this first lesson and more at dmandis.com/futureproof. That's damandis.com/futureproof. The link is below. I'm going to read the article headline. Nvidia sets new milestone with a $3.92 trillion market cap, topping Apple's $3.915 trillion market cap. So I mean uh I you know I looked at Nvidia a year ago and six months ago and my answer was how much higher could it go? Well there

43:12 here's the answer. So here's the question right there there is a bet to be made right now if you're an investor on where are you going to invest to ride this curve because it's not slowing down. Uh are you going to on the chip side? Are you going to invest in the power generation side? Are you going to invest in the you know I have a friend of mine who has identified geothermal energy as the as a key source of energy and so he's going to start drilling over a geothermal bed you know bedrock hard zone and put in a large plant over there. So is it real estate? Is it data centers? Is it chips? Is it power? Because all of these things are going to

43:58 make somebody hopefully somebody listening right now huge amounts of wealth. This is this is unstoppable meta trend. Period. Just to clarify one thing on this slide too because it's even more profound than that seemed. Uh Nvidia just beat Apple's all-time high which was Christmas Eve this past uh you know 6 months ago. Um it's actually significantly higher than Apple today. Apple's down to 3.2 2 trillion while while um Nvidia is a 3.92. Uh so it's it's a significant gap that's opened up. Uh and yeah, I don't think anyone would have seen it would have seen it coming. You know, Sam Alman is always saying that you know a lot of people think about this as victims like oh if this goes up that goes down. It's

44:44 really not true. like everybody participating in this buildout is is going up and there are some companies that are getting crushed along the way but they're they're kind of rounding errors compared to the number of things that are going up. So David, where would you invest? I want to hear where would you invest right now because people are ask you know our our super fans here are asking that question where do I put my money? I would do a basket of these to get the general trend and get the general transformative trend and then I would pick uh certain verticals where AI can make a massive difference and then invest in the companies going after those. What's in your bucket? Hell no. I'm trying to put everything into Bitcoin. So, it's hard to think, you know, my bucket as well.

45:29 Yeah. I'm famous for kind of a lot of people ask me, "Oh my god, you've been tracking Tesla forever. You must must have made a fortune." And the answer is no because every quarter for a long long time it looked like Elon was going to run out of money, right? Um, I even had the chance, as you mentioned, Peter, to invest in XAI at a crazy valuation and I said no. And it's it's 10xed from that. And I'm like, duh. It's this it's a very hard game to play and you have to really set aside a lot of assumptions to do it. And and in general, I think you just buy these top hyperscalers and just sit on it. And and this is not investment advice, by the way. I should say that. Dave, where would what's in your bucket? Uh for me it's a no-brainer because we have access to some incredible investments that most people can't access. But in the software layer which is not nearly as capital

46:16 intensive, there are 10 and 100x performance improvements. So so when you're looking at, you know, Elon buying a million chips, a million and a half chips, you know, it's a 3040 billion risk. Uh if you accelerate the algorithm running on those chips by 10x, you just saved an enormous amount of money. And so it's not as capital intensive. the innovations are right in front of us. You know, Blitzy down the hall from me here. Uh is innovating, you know, writing three to 10 million lines of code a night. So you can implement these ideas in in like a week. Uh so that is where you get the the highest very quick returns. And so if you have access, that's the sweet spot. Now when you go to the conferences like when we go to Riad together or we go to you know any of these big you don't hear about that as much because it's not as capital

47:01 intensive and they're all putting much larger amounts of money into physical real estate generators power supplies chip buys spaceships you know that's where all the money goes because they're very capital intensive so it tends to dominate the agenda but the returns are far far better in the software layer. Yeah. Well, you know, just to round off a comment on this slide, I was talking to an executive in Intel who was saying, "Oh my god, look at that PE ratio. That's insane for Nvidia, etc." And my comment was, "I'd rather be them than you." Because like really, that's brutal. But that's the reality of it. I mean, it's the the moaning and winging from the incumbents is incredible. All right. Uh uh Merkor partners with six of the Magnificent 7 uh and all top

47:50 five AI labs. Super proud. Uh Dave, this is one of your key investments at Link XPV. We had Brendan Foody uh with us at the Abundance Summit last year. Tell us about about Merkore. So So uh you know Brendan is just an awesome guy and talk about getting lucky again. Uh Scale AI got acquired last week. uh what 29 billion dollars by Meta all the customers of scale are like well now I can't work with scale anymore it's part of meta all the other you know competing uh big AI labs so now Brendan's got all of them except for one actually guess which one he doesn't have he has six of the seven meta you'd think you know actually that was

48:36 his anchor customer so he still technically has it that that would I guess yeah that's that's exactly Right. It's Elon Musk doing it all himself. Elon doesn't like partnering with people. He likes to build vertically, internally. So, out of curiosity, just so folks can know how awesome you are, at what stage did you guys invest in in Brendan? How old was he then? And what and what's the valuation now? Cuz I mean, this is the story that people need to realize. You know what used to be the peak age for building a unicorn was in sort of early to mid30s and it's dropped a decade. It's now like the peak age for creating a unicorn is like 20 to 23. So tell us the story there Dave. Yeah. Yeah. So uh we were the first money in Brendan. He was actually uh he

49:22 and two of his best friends from high school started the company together. Uh two of them went to Georgetown. The third friend went to Harvard. uh that made them eligible for prod which is this really really cool studentrun studentfounded joint venture club between MIT Harvard and now Stanford. So they did prod which is where they developed the idea for Mor together. Uh and then we found them at prod because we know the founders of prod like kind of all the little clubs and things around Boston here. Uh and that's where we met them and then yeah first money in and then what was the what was the valuation at that point just to make it was like 30 million plus or minus uh you know what's their latest valuation? Uh they were they got to revenue very quickly by the way latest valuation. So they they closed a $2 billion value

50:07 round two months ago. Uh heard rumors this week that they're looking at an 8 billion or so preemptive term sheet which they may or may not choose. It's eight or 10. No one no one's quite sure. Uh, so that's a like a 5x step up in two months. But Brendan, to answer your age question, I think he was 18, 19 when we first met him, uh, when you had him on stage in LA, uh, remember he told that story, he had just raised a $300 million value funding and the VCs wanted to meet him at a bar and he said, "Well, I can't meet you at a bar." And they said, "Why now? You don't drink?" And I said, "Well, I'm only years old. I haven't haven't turned 21 yet." So, I mean this guy I mean he is such an inspiration for everyone around our ecosystem now and everybody knows his name. Uh it's

50:53 just it's just so cool to watch. I mean just again for for our fans listening here I mean the story is that we're going to see such an explosion of entrepreneurship at the younger ages. Why? I think for a few reasons. Number one um they're unconstrained thinking. They have nothing to lose. Nothing to lose. Number two, they've got the tools with vibe coding and with access to these large language models uh to iterate and rapidly build stuff and throw it against the wall. Um any other reasons why that's happening? Well, for me, those are the top two. So many of the people that we run into are worried about losing their job, losing their career, losing their whatever um rather than jumping in the

51:38 huge into the huge opportunity that opened up. Uh because you know you get invested in your career, you get invested in your job, you get invested in learning. But if you unconstrain yourself from all that baggage and just just think from a clean sheet of paper, you know, when you're when you're 18 19 years old, you have the benefit of of you don't have a job, so there's nothing to worry about. But everybody could do that, right? You're not you're not actually constrained. You just feel like you are. It's the analogy I use is like a 2-year-old skiing, right? They have no fear because they only have this far to fall. Um you got older, you got taller, you got a lot further fall, you're going to break a leg or something and you don't have that for fall and it's the same type of mentality. I remember me meeting this young kid who was the founder of Arbitum, which is one of the layer 2 blockchains, right? And he and

52:24 he was I asked him, well, why do you think you'd succeed? Why not build your own L1 rather than and why would you go with Ethereum as a layer 1 as opposed to one of the others, etc., etc. And he says to me, well, according to your exponential organizations book, a new innovation has to be 10x better than a marketplace, right? And therefore, we figured these other ones are 2x better than Ethereum, but they're not 10x better. So, and the ecosystem of developers in there. So, we decided to layer on Ethereum. And I was like, wait, is this kid quoting my own book back at me? This like how like so I'm so impressed with this younger generation of founders where they got this all. I think a third reason is they're digitally native in a way that is just not right. They live and breathe this stuff. So,

53:10 it's it's really a key part of this whole thing. Also, you know, a lot of the uh a lot of the really successful new companies, they have very specific recruiting domains where they're pulling their best friends in or their people they know. And when you happen to be that age, all of your best friends are also not doing anything. So, it's a lot harder when you're older to go to like if you think of the 10 people you would most want to work with and you get really excited about it, nine of those 10 are going to be unavailable today because of some bonus cycle or whatever. Case case in point busy on podcasts this podcast right now, right? I mean, it's like when I reached out to you Dave and See, and said, "Let's do, you know, this this weekly WTF episode together." Um, you both said yes. And it's been a blast. I mean, I look forward to this

53:56 every single week to have this conversation. Yeah. But let's note that scheduling these things is like Haley's comic coming out, right? It's like a nightmare in between our schedules for sure, but we're committed to it. Core points. Uh we'll talk about it some other podcast, but he has a really good set of core points that was in our little newsfeed here. One of them though is you have a lot of things in your life that you can cut. And if you're honest with yourself and you look at them, uh, to make space for what's going on right here, you're going to have to cut something else. Yeah. But look objectively at your life and think, "Okay, something's I cut to sleep." No, unfortunately. At least this morning I did to prep for this podcast. All right. Uh, the talent war remains on. So, uh, you know, just want to lay

54:43 this out because it's ongoing. Uh and when we think about sort of what are the constraints on this Gen AI revolution, we talk about chips as a constraint. We talk about power as a constraint. It's talent. Talent right now is is the constraint, right? So Meta offers huge $und00 million comp packages. Zuckerberg himself is going out and and finding people and bringing them in. Um, OpenAI is giving, you know, 10 to20 million equity on top of Google Anthropic hires. I love this. OpenAI spent 4.4 billion in stockbased compensation, right? It's more than their compute is costing them. Uh, they're spending on on talent. Um, I

55:30 want to quote a few things here just for people to get a sense. If you believe that access to AI is in fact the single differentiator that's going to well or to the next levels of AI, AGI, ASI, whatever, it's going to differentiate all of these hyperscalers. Um, then it's the single most important thing they can do. So Meta is at 1.35 trillion. They have 58 billion of cash on hand and they're going to spend that. Google's at 2.2 trillion. They have $101 billion of cash on hand, right? Open AAI is at $300 billion. The estimate is they've got about 20 billion of cash on hand. Microsoft's at 3.2 with 78 billion of cash on hand. Anthropic is at 61

56:18 billion with about 3 to 5 billion of cash on hand. So, I just want to make the point that there's a lot of cash on the balance sheets of these companies and this is an existential risk if they don't acquire the best talent today. So they're going to play full out. This is been a complaint on Apple because they've got like unbelievable amounts on their balance sheet of cash and they're just not using it and it's a it's a huge problem. By the way, can I point out something on this slide? Sure. Look at how much Meta has spent on AI in that third bullet point which says aggressive spending impacted margin going from 28 to 23%. I'm like boo. Oh woo woohoo. I mean I mean any company would kill to spend that much money and then your margins are impacted by 5%. You kidding me?

57:03 Yeah. Yeah. Well, point being there's a lot more room to go and uh you know as things heat up they probably will. Do you remember a few years ago uh they were all talking about doing dividends? They couldn't think of what to do with their money. There's so much of it. Hey, why don't we just do a dividend like a like a bloated old bank or car manufacturer? Wow. It's just such a different world today from just four years ago. That's But I want I want people to hear expect that there's going to be outlandish spending to get the talent and then once they've got what they can get uh uh spending on on energy and then spending on building uh you know we we mentioned before that this year is about a billion dollars a day being put into uh into the

57:51 AI arena and we expect by 2030 it will triple to a trillion dollars a year. Um anyway, it's it's insane. We've also seen this in Saudi and the Emirates and so forth. I don't know what happened with Mark Zuckerberg, but he woke up one morning and said, "You know what? We're going to win this race." And and so he has built an incredible dream team in in no time. Uh but the top offer it turned out was a billion. Remember we were talking a week ago, ohion$und00 million offers. Um, and you know, since then we discovered that no, that was actually there were several of those for sure, but then there was a $1 billion offer that got turned down. Uh, and that, you know, that's just that's a person who also has a team that would have come over, but that person, you know,

58:36 yeah, it's it's like I think the key thing there is they turned it down. Well, this is when Palmer Lucky and Palmer Lucky turned down a billion dollars. He turned down a billion dollars. the first offer that that Zuckerberg made for his his VR company. Look, there there's a really important point to be made here, right? If you're going to be building one of these companies, you have an MTP. You have a massive transformative purpose and you're really keen on that and you're driven by the passion and the emotion. Peter, you talk about the emotional engagement that that brings it to people. I remember Yankum when they when Mark when Zuckerberg and tried to buy WhatsApp, the actual valuation was about one a half billion. And so they made they started having a discussion and very quickly Apple and Google and others came to the table. So the valuation kept

59:22 going up and he kept just kept saying no because if I get acquired my MTP gets threatened which was simple communications globally and I don't want to only after like at 18 billion is the rest of his shareholders said come on and then after extracting a promise that Zuck would not touch the company or change anything for 5 years did he go okay fine right and that's I think that's what's amazing about MTPs and the founder mentality that's the same situation that Palmer Lucky talked talked about on stage at the Abundance Summit last year, right? He he said no to a billion. He finally accepted 2.2 billion, but only after Zuckerberg agreed to spend like a billion a year on the whole metaverse activity and he ended up spending like 50 billion instead.

1:00:07 I would love the framing he put on that saying if I wanted if I'm excited about the domain and I want try to get 10 billion of investment into it, it would take me forever because I'd have to go raise that money. And so therefore, by doing it this way, I got Facebook to invest or Meta to put that money in and the whole place exploded, which is amazing. Yeah. Now, quick story for you on why these numbers are so big. Uh, and this is this is hearsay, but I'm pretty sure it's accurate. You know how uh, you know, open, if you're an OpenAI user, nobody uses GPD 4.5, which was supposed to be a big deal. We're still on 40 or or 03, which is using 40 or 41 under the covers. What What happened to 45? you know, it was a it was a multiundred million dollar training run. Uh, turns out there was a bug, probably a single line of code in PyTorch,

1:00:53 and the thing was just grinding, using up compute, but and you know, they thought it was making forward progress, but it wasn't for a long time, probably. Wow. And uh, yeah, it torched the whole training run. And that's uh, that's why GPT4.5 has been a disappointment. So then you're like, okay, why would I spend a billion dollars or or hundred million on a single person? Well, like if it's the right person, the next training runs are going to be much bigger than the last ones and there's a lot of room to optimize, improve, and avoid bugs in that process. So, these are really rare, super valuable uh human beings at this point in time. Amazing. I don't get it. Why can't you just say to the AI itself, go find the bugs in your code? Well, I mean, that's that's uh certainly going to be the truth within a year, right?

1:01:38 Um yeah, the rate blitzy is moving within a year for sure. A quick aside, you've probably heard me speaking about fountain life before and you're probably wishing, "Peter, would you please stop talking about fountain life?" And the answer is no, I won't because genuinely we're living through a healthc care crisis. You may not know this, but 70% of heart attacks have no precedent, no pain, no shortness of breath. And half of those people with a heart attack never wake up. You don't feel cancer until stage three or stage 4, until it's too late. But we have all the technology

…