What could go wrong with the AI trade? Micron, AI extinction fears and Broadcom’s Anthropic loan

Welcome to Market Hang. >> Hey. [music] >> Hey. Hey. Glad you're all here. We have Ross coming in for the second time. Julie, two days in a row. Yeah. >> Daniel, first time here. Thank you. Same setting, but totally different. >> Yes. >> Yes. >> Exactly. >> So, it's October 1st, which I'm a huge fan of. October. We have baseball playoffs. We have football in full swing. We have pumpkin spice. We have horror movies all month long. And we have earning season. So, what's not to like? >> It's funny. I don't like I don't think I like any of those things, but I do like October. >> So, you're a seller horror movies. >> I don't like horror. >> How about Halloween? >> I do like Halloween and I love fall leaves, so I'll take October. >> Ross, what's your stance on pumpkin spice? >> Not a fan. You know, my wife will buy all this weird pumpkin spice. Like, every product has some sort of weird spice this time of year. And I'm always like, "Why are you buying peppermint and pumpkin spice crap? I'm fine with the things the way they are." >> I drink my coffee black. I I'm Me, too. I'm boring. >> Football or baseball? >> Uh I Is there a baseball? I I don't know. But I watch a lot of football. >> Come on. >> Uh I you know, I'm We got the Dodgers, so you know, I get your old job. >> I'm a Chicago. I live in Texas now, but I was born, raised, spent my first 40 years of my life. And I actually looked for the first time yesterday and saw that both the Cubs and the White Sox are in the playoffs. And I had no idea. I had no idea. I'm like I had to leave Chicago for the teams to finally start winning some games. So funny. So, uh, on earnings season last night, Micron, um, beat and raised. Uh, I think the stock is up 550% over the last 12 months. So, it seems like the only risk is not owning the stock, right? So, last night's report, does that make you more or less excited about the whole AI trade? That was the absolute least surprising result ever. I I I had been saying, I mean, I I made the the rounds and I was like, look, this is going to be a beat. The question is by how much. They're having their I called it the World Series of earnings. So I gave baseball a tip because you know Nvidia is the Super Bowl. Micron is now the World Series. Um but again I think we've hit that point. It was kind of like two years into the Nvidia earning cycles when we hit the point where it was expected for them to beat and there wasn't an amount that they could beat that is big enough that people are going to be satisfied. And the hard part is people are really misunderstanding that this is a supply issue at this point. They are maximizing revenues, profits, margins and executing extraordinarily well, but they can't bring more supply online. You know, I talked to Sanjay Moro earlier in the year actually when I was in in at the World Economic Forum and and he was like, I don't think people get that like 28 20 I mean 28 29 is when we're going to start to really see supply come online. And so we've got a couple more years that this constrained environment is going to exist. But what we didn't expect is that the AI utility and demand would actually go parabolic as quickly as it has. So we're we're hitting it from both sides constraint on the on the actual supply and go ahead. >> Well, I was going to say because memory is a unique issue different than Nvidia in that it's part of the everyday computing that we do with our phones and and computers and that demand was already like filled because it hadn't really grown. These companies hadn't invested in excess capacity because they were they were basically like a low margin business, you know, that's now at like a 90% margin business. And so when you actually think about the problems that they have and even when you get out to 2030 or 2029, you know, robotics will just be coming online in the next decade, not really this decade. And that like can you build a robot without a brain with memory? No. Like, and this is one of the issues with if every car is going to have to have like real hardware like a Tesla and every robot is going to need real hardware like a Tesla, you know, where are they going to get this memory? And you know, like like him or hate him, but Elon says all the time very transparently what he needs. And and and so he said it, you know, actually the biggest problem is there isn't even power to do all this stuff. So that's the first place you know as an investor you got to look at because you can't even do anything without power and then secondly memory is going to be one of the constrained things I think for possibly like a decade and even when they have new capacity come online prices will come down for components but demand will continue to just I think grow and so it I I don't think the market gets the story I think it's great for investors that you can buy Micron at like six times forward earnings micron will be deploying way more capital back into buying back stock which is all they need to do because it's the best return they can get on their capital is actually just buying back their own stock >> and they haven't been able to buy it back >> right >> I don't know if you know about this but the chips act I guess prevented them from buying back their own stock that changes >> until December that changes in December so now everybody's waiting for them >> to announce so keeping with >> yeah but you know you know that the supply of stock's going to go down and the demand for the stock is not going to go down. So, as an investor, like what I mean, if you're watching the show, like what a gift, right? >> Well, the the war chest grew too. What like over 70 billion this quarter? The this company could look very much structurally like Nvidia in about two to three more quarters, right? >> And they'll suddenly have all the optionality to invest in ecosystem to start playing more of a role as a VC. They will be able to of course invest heavily into their own R&D. I mean, these members >> and pay their workers and pay their workers, which is something we talked about today. Yeah. >> Yeah. I mean the the situation is just incredibly different. The scars, the battle scars that these memory companies have though are very real. They have real they their memory is very real and what they've been through is very real. And they haven't seen a situation like this, but I think what Ross said, >> structurally speaking, this cycle is different. And I know people get tired of hearing some of us say that, but it is a different cycle. But so I mean it sounds like you everyone's saying this is no no this is this is great because then I'll take the other side but um >> this is Micron in a vacuum. Are we this excited for the overall AI trade? >> Like what could go wrong? That's always what makes me like whenever anybody ever like you guys are so bullish it always makes me think what could go wrong. >> I can tell you what could go wrong. I can tell you it's so obvious. We don't have power. We I mean, you know, have you ever tried to plug in a bunch of stuff and not have power? It doesn't work. >> Well, it's not just power, right? It's power, it's labor, it's land, it's the >> Yeah, but we've got I was just I was just in Texas. All you have to do is drive an hour outside of Austin and there's land everywhere. You know what I mean? >> You need to go an hour, >> right? But but [laughter] the land being there and them actually getting the land and being able to put this thing on there is are two different things. >> Yeah. You you can build infrastructure. Listen, these guys are printing billions and billions of dollars. You can build infrastructure around a data center and so on, but building power is much harder. It's it's a more complicated thing. Like, you know, I live in an area where they're building power right now in the Palisades and they're rebuilding my city right now. And I'm watching the power people work every day and it's been for months for a year now. Like power and and distribution lines and the entire system that you're building is actually a highly complex. You need certain types of skilled labor to do all this stuff. It's super expensive. you know, you need the the parts. So, the Elon said it the other day and I just like people need to pay attention to this. China has three times the power capacity that we have here in the United States and that's a huge competitive advantage for them when it comes to AI because we can do everything else. >> But does that take the whole >> plug it in and it doesn't work? >> Does that take the whole trade down? I mean, >> well, it could growing as much. >> I would I would actually take another because I agree on energy being a problem. I think policy is our biggest potential risk right now. I mean, obviously the current administration gets it. The White House meeting this week, getting the right assortment of people there, but the popular opinion and heading into the midterms, I mean, AI has become a battleground. It's become a political divide. And in some ways, of course, you have actual surprising political unity against AI. Um, you know, President Trump wanting to rename it has some parts to do with the fact that AI has gotten such a bad name because I've said loudly, we've had the wrong spokespeople. I mean, Dario is an impressive guy and so is Sam, but I would like them to speak a lot less publicly about the benefits of AI. You know, on one side, AI is going to help accelerate curing rare diseases. We're already seeing it lower mortality rates in hospitals by supporting our doctors. But what we hear about is P Doom scores. I watched an interview this morning with a with a thoughtful Walter Isacson and he was talking about this P doom score which is probability of doom. We actually spent a week talking on the on the news about a technology that has a better the the suggestion better than 10% chance of causing the extinction of humanity. And then, you know, you go home and like I finally am having conversations at the dinner table with ordinary people asking me about AI, asking me, you know, I was I was at a uh a film premiere last night here in New York and just a random person comes up to me and >> Oh, did you see did you see the Elizabeth Holmes documentary? Is that what you saw? >> No, it's a different one. Dyn are you saying are you saying I know I'll tell you more. I do want to see that >> extinction is now a kitchen table. >> Well, I'm saying the combination of is AI safe? Is it going to ruin our resources and our communities? Is it going to take my job? I mean, if you actually listen to Daario Amade took a trillion dollars in market cap out and and turned it into private valuation by saying software would be destroyed, data services would be destroyed, security would be destroyed, jobs would be destroyed, and quietly we walked all those things back, but now they're a $2 trillion company. and he continues to say these things, but like having a 10% likelihood of extinction come out of the media's mouth in a serious tone >> for a week, maybe two weeks now without coming back and saying, "How realistic is this?" I mean, if it's true, then why do they release a model the next day? [laughter] The bottom line is we don't know. But I don't I think the people inside the company believe it. the people who say it inside the company, this whole rationalist movement, they I mean, >> say what you will about Daario and Sam and looking at it cynically and saying it's regulatory cap capture, all that stuff, >> there are pe there are certainly people who are building this who >> Well, there are certainly risks. I mean, look at these agents getting out and what they're up to. I'm just saying like if you have a technology that could legitimately end humanity, >> like you would think that they wouldn't release the next model and then file the the IPO a few days later. I mean, there was a you'd have a little cynic Yeah, but you know, listen, listen. You know, I I've been hearing this stuff now, it feels like, for months about the end of humanity. And I find it absurd when climate change is actually the biggest risk to the end of humanity. Okay, we're literally having massive climate events left and right and everybody's worried about AI. Like, AI isn't going to come wipe out your house with a fire. Okay, I dealt with that. Okay, it's it's this is the most absurd conversation that AI is wiping out anybody. Now I get if you put Daario, Elon, and Sam in a room, it's like a horror movie. These guys are like the most hated people I have ever like like known of. Like when you go into a crowd of people and you ask like who could you hate more, you know, you've got this, you know, you could even do like Trump versus these guys, they wouldn't know who to pick, right? >> Zuck starts to sound like the rational voice, Zuck. But right >> I don't know if I would argue but trust us you know but these guys are so untrustworthy and so when they go out and they start so so the only people we need to worry about are actually them because it's humans that will build something that's bad that's the Bill Gates right Bill Gates isn't worried about the AI killing us he's worried about people >> I'm not worried about AI I'm worried about the people so we'll take extinction off the table right but one of that's a pretty big risk we agree like one of the things I like to do is >> every bubble that has popped right in hindsight the risk were so obvious whether do financial collapse and so what am I missing because it seems like we all are very bullish on on the whole trade and so what could we be missing >> I I mean I there are lots of things that could go wrong I mean I don't think it's just the power thing that could go wrong first of all I think if you're talking about open AI Anthropic specifically, there is a lot that could go wrong with them. I don't know about the broader ecosystem, but with them, I mean, we're also already seeing token prices go down, right? So, that's something that they could be vulnerable to. We're seeing open- source models gain more market share and traction. It's still small as a part of the whole, but they are gaining traction. Um, >> you know, we have seen a bunch of reports, when we talked on the show about this yesterday, I talked about the Banning Company report. There are a bunch of reports coming out and saying they're spending X. they need to be g bringing in this much revenue and there is a big gap between those two numbers right and so how are they going to close that gap there needs to be innovation to close that gap what is that innovation we don't know yet right we don't know what all of the applications are going to be for AI but they're going to need to figure it out so what could go wrong the demand the end market demand might not materialize in the way that they think it's going to materialize >> yeah I think we we we we continue though to to try to kind of move the numbers around without really playing correctly the denominator. We talk about 3% of GDP and the capex spend and the ROI, but we're not talking about accelerating G GDP. We're very focused on this direct correlation of AI to revenue. And the hyperscalers largely have it covered. They largely have it covered in their backlogs. They largely have it covered in their commits and revenue that they will make that they will profit on the ROIC, return on invested capital of all the money they are spending. Where we're not hearing as much, but where I think we will see it is the utility inside of enter of of inside of enterprises and industries. >> Yes. Um it's it's when healthcare companies talk about the acceleration of drug pipelines. New drugs developed for less cost that could be distributed at scale. And it's the Jebans paradox that exists across every industry. It's being able to do things faster. Insurance companies more accurately being able to prescribe. >> And we're already doing this. >> But it's not even that. But it's also creation of new revenue. It's >> but attributing that to revenue too. No. And I'm not. >> Yeah. But it's it but it's it's not just more revenue. It's it's more margin. So, so what's happening at my firm as we employ more and more AI solutions, it's it's increasing our efficiency by like an order of magnitude. So, it's really that impactful in a financial firm. In fact, financial firms and healthcare, I think, will be most impacted in a positive way through AI. For us, it's efficiency. >> Have you have you fired anybody yet? >> No, it's the opposite. We're hiring more people than ever because every person I hire that that's AI enabled is like 10 people and we're growing like crazy. We've never grown more. >> And that's what we're doing. I mean, as a research company, I I just put no I need to hire like I can't hire enough analysts fast enough because of how much more work AI enables our team to do. >> It's actually the opposite effect. And that's what I think in every turn. We've heard like 10% of job well 10% unemployment, 90% of entry- level jobs. I'm like, I can't get enough entry-level people to actually take advantage of the acceleration. >> The issue is if you're not if you don't know how to use AI, then you have no purpose at my company. So, we we put everybody on AI the first day. >> Do those anecdotics translate to corporate America? Are we expecting blowout earnings? Well, we are getting them. Yes, I know. >> Yes, we're getting numbers I've never seen in my life. Like 50% earnings growth on top of margins. like margins in the S&P used to average 10 to 12% margins. You know, now we're at 13 and 14. And this is across industries with big companies that management is super slow, untsavvy. That's why I'm saying we are so at the beginning of this companies like mine, we move super fast. Like I just make decisions, my team makes decisions and we implement it and that's that. But our competitors, they're walking around at conferences like what's AI? What should we do with AI? And I couldn't be happier to see it, you know, and and so we're just crushing everybody. >> My comment about denominator was just that GDP as a whole will grow and subtly under the hood of why GDP is going to grow is going to be the impact of AI. And what I'm saying is the earnings across all these industries will show up and not all of them are going to get on their calls and just be endlessly talking about because they bought GPUs or because they spent more with AWS. They're going to talk about how they've built their businesses and scaled and got into new markets and and and en enabled new product development at a rate that they've never seen before. And it's already >> entrepreneurialism. >> It enables entrepreneurialism because you can take ideas now. I've never been able to program in my life. I've had to hire these people in the past, right? And now I can just go on and build stuff like >> like myself. >> Like what are you building? Well, how is it like? >> Well, the first thing we built was a financial services software because we're in my industry, we're forced on a platform called e-money, which is highly complex and super annoying and I hate it. And so, it cost me a lot of money. I have my whole team on this and we're I can tell you all the fun problems that we have with it and integration. So, we went on and we said we can build something better. And we built something better. And it took us about four weeks of rejiggering it. But but now we have the best financial planning software we've ever had. Now the problem we're having is we can't put client data on it because it's not secure and so we're working with anthropic to to get secure servers. So there's we're still at the beginning of like how do we deal with security which is I think cyber security is one of the biggest issues with AI. It's like so firms like mine I can't just like when OpenAI sends you an email and says like oh you can link your bank accounts to us now and we'll help you. That's the dumbest thing you could do you know like putting your personal information into AI. um you're gonna have problems. >> You don't want Muse reading your IME messages. >> No, no, you should not be doing similar things. Everything everything he said. I mean, we've built our APIs, MCPs, all of our research was exposed in a new layer. And we're still using enterprise software though. That's the point. Like there's a lot of this kind of idea that it's going to completely displace. I'm seeing a lot of augmentation where the experience layer, the utility and semantic layer of software was terrible. It was just awful. >> Um so what we're doing is we're building this semantic layer. That's how we would want to use software. In many cases, I you know I I I helped uh I Bill McDermott with this term, but I call it the rules and rails of business. The enterprise software companies are pretty good at the rules and rails, the governance and all those things and they're trusted inside the enterprise, but they're they were just all really bad at UI. They were all bad at building a good experience. So AI enable and by the way AI enables to what Ross was saying like you can build these things in in hours, days, weeks and have them give you insights into your business that you just never could unearth with software that used to come out of the box. It just didn't exist. >> Sort of. Well, the insights are what's interesting because you can ask AI to do analysis, you know, and and and so if you give it data, you know, and you say like, okay, so I do an a lot of analysis is my work, you know, and like and and so I was just looking at a company. I was looking at MGM Resorts cuz, you know, I've owned it on an offer 30 years and this they they went in this merger and and the thing fell apart and the stock's now like 30 bucks. And so I do my own analysis. I think the MGM is worth 50 55 bucks. But then I go into AI and I'm like, look, here's the premise. Here's where I'm at. Look at my thought process and what's wrong with it, if any, and then you do your own version of this and then come out with your own number, you know, and it breaks down the steps it's taking in its analysis, which I really like because then you can verify how if it's smart or dumb, you know, like is it doing this right? And it literally came up with this it came up with the same number yesterday that I was working on and I was like and I can justify it with all this stuff and I was like damn you know this is really helpful and great work that it does. >> I was I was talking to the IBM research director Jay Gambetta actually this morning and he actually said one of the most um profound ways that AI will change is research itself. He's talking about like the development of semiconductors. Yesterday, Synopsis launched uh it's it's GPT Synopsis in partnership with OpenAI. And remember, there's another one Sam and Dario said, we're going to displace EDA, which is the whole business of silicon design. And what ended up happening was OpenAI needed to partner with a company like Synopsis because it's the proprietary data layer coupled with, you know, the the front-end architecture now where an engineer that used to take two years to spin up, tape out a a single chip can take that process down to 6 months. So you think about how much we can accelerate what we're building when you take twoear cycles and turn them into six months that innovation cycle. And that's why I was trying to say it's like subtly this shows up in GDP, >> but we're kind of only talking about the 3% of GDP. Well, that's only if the GDP doesn't grow. It could be 2% of GDP. If GDP is meaning like you know I'm not doing all the math right here but you get the impact of if our actual economy grows then the amount of capex we're spending against that is different than the way it's being prescribed externally. >> I mean to some degree you could argue that the paradigm of what we expect out of companies should be rejiggered now and rethought. You know, I was just at an entertainment business conference yesterday called The Grill and I was with a lot of people in the entertainment industry, but it was very interesting because it was the I go to this thing every year and it was the first year that there were a lot of actually like entertainment tech people there now because the technology is now starting to like feed into the entertainment industry and it was the same discussion a lot of on the surface you hear all the actors oh we don't want AI actors and all this nobody wants this but on the behind the scenes it's about efficiency. It's about building and rendering visuals in these highly complex visual environments that are being created whether it's in video games or movies and being able to do this way more efficiently and inexpensive than in the past. And and that's why I keep saying a lot of people are looking at the wrong thing with AI. But when you look behind the entertainment industry is highly inefficient and AI will really really increase margins. Maybe it doesn't make more revenue, but now I can make more movies, maybe at a lesser cost, right? Um, and and even the analysis of all of this now can be better used for advertising dollars and so on and so forth. So, so we're at the beginning of I think of a revolution in new entrepreneurship and new companies that can be formed because AI can allow a small group of people now to build a big company. Uh, do we think that there's an issue with data and and feeding AI? Like that seems to be an issue. >> I mean, I think it's you have to be very thoughtful and protective. I know Alex Karp is a is a controversial character in in, you know, I saw, but I believe he's he said it very well when he talked about protecting your alpha. You know, just feeding your data. Ross talked about it through the lens of governance and and and what FINRA requires in his business, but every company's value sits in that thin >> data layer that sits between infrastructure and the model. And if you're just like feeding that to AI, you're basically giving them your business. >> Did you see what what Reddit did? >> Uh I think it was it came out this morning. They're they're shutting off the pipes so people can't just um get into like they're calling it old Reddit and and they realize how valuable their their da d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d d da is because you have 20 years of human interaction like that is that is so >> they were licensed they were selling it to to open and they're they're suing Anthropic >> because Anthropic just took it but the the the I'm seeing emails now and I'm I'm sure many smaller mid-size company owners are but I'm getting emails every day from companies companies that are being stood up right now that are literally trying to no longer are they trying to buy my company. They just want to buy my data. You know, services and data and they're saying, "Hey, your data might be worth anywhere from, you know, $500,000 to $50 million. You know, do you want to >> have a meeting with us?" And >> well, we just had that with with a PE firm and and they're interested in investing in my company and they're like, you know, send us the data to, you know, justify your valuation, all this stuff. And I was like, no, you know, like I I'm not letting you have this data. you like you'd have no reason to buy me if you have all this data, right? And it's like our data is like crazy now because we're fairly we serve like thousands of clients and and we manage about 20,000 separate accounts. So, you know, the fact that we've been able to organize and get all this data working, which was extremely challenging, but we started two years ago doing that was a huge value ad for the company. But now it's kind of what you were saying about having an overlay of AI and that's where Salesforce is. We're you know a lot of us use Salesforce including me and we're not getting rid of Salesforce but if AI actually worked on Salesforce that would be an incredibly powerful tool >> right although we've been giving away our personal data for what the last decade right like >> well but there's giving it away and there's giving it away. In other words, like >> yeah, it's stolen, you know, like Well, but I mean, I assume that, you know, Amazon knows everything or that, you know, the Facebook knows everything. Yes. >> But it doesn't have my bank account information, for example. >> Not yet. No, but it knows it. It knows where you're going to be on Tuesday at noon. Like, >> it probably knows where you bank and it probably has a pretty good idea of how much you have in the bank, but they don't you haven't hooked up the physical connectors for them to transact on your behalf. But by the way, that's what where where we're heading, right? That's what Muse is is is doing. Yeah, but that's a dangerous thing. I'm not doing it. I'm not I'm not setting up. And >> well, you know, we have tons of security at my firm just to prevent like actual people stealing people's money, you know? So, if you think we're going to let some AI agent transact for our clients, there's zero chance. >> But even as a consumer, on the consumer level, I think it's still pretty risky. I mean, it is the ultimate convenience when you can have a system like that to transact for you. I think we got I I the very least I think for anybody out there that's listening this and and wants an opinion is be thoughtful and kind of slow in terms of what you decide to connect to these things and and and is I've been an Apple skeptic over the years. But what I will say is that this is actually something Apple should nail. They have% [laughter] yesterday I was like John% please give me your >> give me an Apple agent. Give me those like like they have at least >> Apple has failed that they don't have this right now. Okay. I I I hate to say it, but I've been highly critical of Tim Cook. I was calling for him to resign. He actually did resign. Um and now there's this opportunity for somebody, this the new CEO of what is name to actually like do something, right? And and and the first thing I would do is build a freaking AI agent. This is actually all this could actually abdicate them from all of their Siri failures >> because really the product we needed from them was a >> I was adding to my Apple position actually recently in my fund GK and and and and in inside the firm because not that I haven't owned it for 30 years or whatever like I need more Apple but but I was like dude like it knows everything. We trust Apple. It already has a a transactional ability. You have Apple Pay, right? And if and and like the other day I was using chat to book a and it booked a a a reservation for me for a restaurant. So that's like a non-scary thing for me to connect to AI like booking reservations at restaurants. >> Will you ever trust Open AI or anthropic like like you trust Apple? >> Um I I trust their tools as as a as a semantic overlay of very very well-designed APIs. My my tech team >> that was a very tech answer. My tech team was a really tech team has built this thing. [laughter] >> My tech team is Ohio. >> Their models are great. Their models are terrific >> to expose great data. But I will absolutely not let them have read write access. I don't really even trust their zero data retention right now. And you're seeing it show up. You're seeing it show In other words, you don't believe them when they say what you say to us is is not shared with >> That's why I said I think Apple's the only one that's pretty well delivered on its promise of security and and and privacy and I you know >> I think they have I think Google in that conversation or no >> they could be but you know Google still has a bit of a reputation for for >> Google's listening to you right now. You know what I mean? >> Yeah. That's meta Google. You guys notice the weird ads? Like you're talking to your kid about like something and then next thing you know there's ads. >> Yeah, Instagram's the best at that though. I just get the most experiences on Instagram even though my mic's off. I I mean because I'm I do try to do all the privacy things. I don't >> I set it off too but it still knows. But where do you doom scroll? You don't tell me you don't doom scroll. >> Xoom [laughter] scroll. Okay. I do X is like my job so I doom scroll on Instagram. >> Exactly. Yeah. [laughter] I don't need to doom. >> I actually don't even use Instagram. I do it too much for work. I don't use Instagram either. I'm >> I'm a big fitness peptides guy like and and there's no good fitness content on Axe. You got to Instagram is where all the good working out [laughter] happens. So I'm a I'm a nonpeptide fitness guy. >> That's okay. We'll get you there. >> I'm also a non peptide [laughter] fitness guy. >> There are We'll get you there, too. No, >> I like strawberries. >> I'm all set. Don't be >> I like bananas. I don't need two shots. two words that >> might might be above my pay grade, but two words that concern me a little bit, and that's circular financing. Is that something I should be concerned with? >> Uh, I'll give a read in. >> Not if they're using cash. >> Um, well, I got a couple thoughts on this. Like, it depends who you're talking about. See, I think Nvidia has has performed what I would call some of the best corporate venture in history in how they've applied their their dollars. You look at what they're averaging in terms of return. I think it's like well over 300% return on their deployment of capital. They are investing into the ecosystem up and down. They are using a balance sheet that looks like most governments would dream of to apply to parts of the industry that they need to make sure are successful for them to be successful. meaning things like optics uh investments in coherent and lumenum the neoclouds they absolutely knew that the um Google's and Amazon's and Microsofts are going to increasingly build their own silicon while I continuously say Jevans paradox right they're still going to need a lot of Nvidia why would you arm companies that are putting meaningful efforts into building competitive products from you when you can actually build an entire industry of next generation clouds that become hyperdependent on you and then build their entire businesses On top of you, great corporate venture has always been about investing in companies that will utilize your technology to grow and then you grow as a byproduct. But on top of it, it's really circular. If a dollar comes out and a dollar comes back, but right now the way it's worked for them, a dollar has gone out and on average somewhere between $10 and $30 in between equity and cash value comes back. That to me is good investing. >> So what about but what about the newcome anthropic financing scheme? Is that you know that >> the Broadcom? Yeah. the one that was announced today where Broadcom is lending anthropic $42 billion to lease it >> to lease its chips. Now that's a dollar in dollar out it seems to me when you have that kind of a financing structure. >> Yeah, but it's convertible. It's not really a loan. It the the it's a convertible deal is the way I read it. And so I get what you're saying, you know, but in my industry that's very common, you know, like a lot of companies will support companies like mine because we make these companies lots of money and so they have every incentive to lend us money or create deals for us that are basically helping us financially because we grow their businesses for them. >> There are cases where this magnitude >> there are cases where these risks become greater. >> I wish it was 42 billion. Yes. There are cases where these risks become greater and there are cases where the choice of what companies they're investing in become less lucrative. I continue to talk about the fungeability of demand meaning that the belief is that this chip volume even if anthropic was to fail do these 42 billion in chips become useless no utility in the market a lot of people talk about this with Oracle and open AI if Oracle something had gone wrong you know with open AI what happens to everything Oracle is building but what we've seen throughout this entire AI buildout is every estimate by analysts like myself um you know we try to size markets constantly we're trying to understand it we've underestimated ated every part of this curve so far. Meaning the amount of demand and the amount of utility we've all gotten wrong. Now again, there's a reason to be cynical. There's a reason at some point to be cynical that this could all go wrong, but we've underestimated power requirements. We've un underestimated comput requirements all the way back to the internet like we've underestimated. >> I would say you can go back to the railroads. >> But the question is not underestimating the demand. The question is underestimating or overestimating the value. In other [clears throat] words, in other words, okay, so if Oracle, if OpenAI went away, if something happened, would there be demand for those chips? Perhaps, would it be at the same value at which they sold them to OpenAI? Like >> you figure less competition might actually create a higher value. >> I don't know about that. I mean, but >> Well, but look, I I agree. Less competition creates a higher value and they want less competition for sure because they know they have a monopoly and that's why they're investing so much right now is to create a moat to stop the others from competing because it's so lucrative in their mind. I find the irony in LA just recently Peter Teal bought a $130 million house that used to be owned by this guy Gary Winnick. And Gary Winnick was the guy behind a company called G Global Crossing. And if you're old like me, you'll remember Glo Global Crossing. And and this was a company that was when the internet needed infrastructure, right? It was the same thing, a massive buildout. We needed cables. And he was like, "Oh, I'm going to build these undersea cables. It'll cost billions, whatever." borrowed tons of money and the thing goes bankrupt. So it's not that the internet didn't succeed or that we didn't need that infrastructure which is actually have been used. It's whether the deals make sense today, right? >> You know, and and and like to assume that everybody's just going to be successful in AI is obviously not a reality, right? And so I like to look at the railroads as a period of time that's similar to today. This is like the industrial revolution but for electrics, electrical things, let's say. And so so many companies were in the railroad business, you know, and then JP Morgan was like, I'm going to finance all this stuff. But when stuff went bankrupt, he was like fine with it cuz he took it over, you know, and so like they built a national railroad system, which ironically is owned by Warren Buffett of all people that still makes money. Okay. like 125 years later, it's still making money for >> a lot of people lost money along the way, didn't they? >> That's right. A lot of people lost money along the way. A lot of people lost money in the gold rush. It wasn't that there wasn't gold. There was plenty of gold. >> So that's So all of these like all of these industrial revolutions that we talk about from railroads all the way up through the internet, like did any of them happen without an investment bubble at some point? >> No, there's carnage. >> Had a bubble at some point. So the question is where's the carnage this time? >> It's not that it's not >> well are we in the bubble yet? And I don't think we're in the bubble yet at all. >> Is safety part of the carnage perhaps? >> Like >> meaning >> bad things happen >> like like personal safety. >> Yeah. >> I mean bad things have already happened by the way. We're all talking about like oh what if you know and I'm not talking about the hugging face hack. I'm talking about somebody in Vancouver, BC going and and um doing a mass shooting after they talked with Chat GBT about it and Vancouver I and BC is now is now suing the company, right? Like there's a law. >> Yeah. Like if you use AI to help plan a terrorist attack and AI doesn't wasn't ISIS using it to develop weapons? I mean, wasn't that like a story that came out with the anthropic? >> Look, it's no question the Iranians were using it to help target our bases and it was very helpful to them. There's no question. And that's what I'm saying about the people we have to worry about are the people in charge that they're like letting this happen. It's not like they don't know. Okay? So, if you're if you're an AI overlord, you should put down in your notes, prevent terrorist attacks should be one of the things, >> but liability the the industry is going to police itself, Ross. And they what you know, they're taking >> I'm sure I'm sure that that doesn't work. I, you know, in my 55 years of life, policing yourself doesn't work. Um, but I I I think the bigger question is who would be the police for this and who would make the most sense to be the police? >> Interesting that you bring that up because you spoke >> and it's not the government. >> Yes. I mean, you were just talking about synopsis and the new product that came out with and I talked to CeCe Gazi, the CEO today about this and I asked him his view like self police liability, what's your approach and this is what he told me. It's always the company's responsibility to when you develop a product that you validate the product. We're in the software business. When we design a product, we make sure that we validate and invest significant amount in not only the development, the development and the validation of the product. Now with AI is happening at a pace that engineers never seen before. it the race to get to the best and next model is natural but at the same time the investment needs to happen on validating that model. So, so it feels it felt to me like Ceine was in the Jensen camp, right? The camp that like Jensen kind of said very simply, well, a company's not going to ship a product if it's not, which I think is a little disingenuous in this world to be frank because a there we've already have models that have gotten out of control of their creators. And then B, what Ross is talking about is that there are models out in the world that are being used for nefarious. We put we put planes in the sky with software that didn't work. We've put drugs into the market that have killed people for a long time before being taken off and deregulated. So I'm I'm actually agreeing with you. Like actually expecting the full self-regulation is wrong. But to say let's just pause and stop innovation complete nonsense. What we need to do is create an agency. >> We need to regulate. We need to regulate alongside. We need to build this plane as fast as we can. We need to win. And the China US thing is real. I know we some people hate to talk about it but like look west the west needs to win this it it does it's decades of economic leadership that are depending on it but having said that like we need to build like SEC to regulate yes you need an SEC you need an FDA again are they perfect no it's imperfect but it would but the idea that we're stopping was such nonsense and and that's the kind of stuff that I talk about the public calamity of of unpopularity around AI is because these are the kinds of discussions we have we don't have a real discussion about how it needs to be done. We say, "Oh, let's just pause or let's stop or let's lose or like no, like we we need to build this out and we need to put smart people alongside of it." But realistically, and I heard Ross say it's not the government, but realistically the real problem is we don't have people like our lawmakers like Bernie Sanders. I mean, are you kidding me? >> They don't know how to use their phones. So, so what I'm saying though is we need to bring like like I know David Saxs, people have opinions, smart guy, >> says a lot of really wise and intelligent things about AI. I don't know if he's the right guy, but we need people he's completely conflicted. That's what I'm saying. Like I I don't want tech people in charge of this. This should >> have people that know something from outside the tech industry. >> Where is the I love everything you said, but where do you stand on accountability? Because look, I mean, I think if a drug company kills somebody with their drug, they should be liable. I think if a car company kills someone with their their AS, they should be liable. And that's that should be that should slow them down. And so what I'm saying is I I don't know if you guys heard uh Sundar and Mark, that's what I said about Mark being the adult in the room. Actually came out and then like 2 days later, we found out Muse is hacking iMessage. But they came out and they said that like, "Hey, we haven't shipped stuff quite as fast as as others. People have been actually saying Google is falling behind and Google is doing terribly in the model race. It's because you know Google's saying like look we follow a series of protocols. We're following Apple as public companies though as public companies they will be held accountable by either their stock will be damaged or their company will be damaged or they will face litigation if they actually kill with their technology. >> That works actually that works really well because lawyers make a lot of money suing big companies and they're happy to sue Meta and all these companies as much as they can because they get rich off of this stuff. My >> economics is a factor that will work in regulation by having liability back on that is the meta example though. >> Yeah. >> Because Meta did harm for years and was sued. >> They internally they knew it was >> how much how much did that it's like the insurance industry though. The insurance industry and they have an entire actuarial industry that gets paid on the idea of figuring out whether it's better to allow the risk into the market versus what the potential liability is. These companies are doing the same things. I mean I mean Mark has paid billions of dollars of antitrust, anti-competitive and and it has the word is is sometimes they're just speeding tickets. So I obviously agree that there is a problem here. But like I said the idea that we're going to have some government overlord come in and actually help us. They will destroy the industry. And right now our only growth opportunity, the only chance we have of GDP accelerating and our economy growing is AI. You take AI out of this, we are negative growth. I guarantee it. >> Yeah, that's right. That's very very accurate what you're saying >> where oil prices are >> and and I think that things tend to work themselves out too. Like if you take the Grock uh scandal when you could like undress people with Grock and then the backlash was pretty immediate and and it was very harmful to to Grock and its popularity when it was trying to launch and become a more used consumer product and it's not now. And and so you know now it's the operating system on your Tesla whether you like it or not. And so, um, >> yeah, but it's it's actually kind of cool, you know, it's it's it actually does stuff now on your Tesla. So, having a Tesla's doing some pretty cool stuff of of using, um, uh, AI in your car, it's basically the only AI enabled vehicle at this point. But but I think that if OpenAI or Anthropic or Gemini do does something that's detrimental or like you know it is found that you know they were not protecting our national interest or people's safety. I think the backlash from the public would be quite large actually. So so I think that's a part of it too. people have to, you know, say this is what we're not okay with and do something about it if it's happening and and we have to support individual people's protection of their personal rights and freedoms because that's really where I think most of this gets trampled is the way that individuals get abused by these things. >> So, I haven't heard Musk's name throughout this whole talk. like where where does he stand and is he >> is he going to help the situation with the AI safety? >> He seemed he seemed to be he was walking a finer line cuz he kind of came out with the Daario and Sam thing and you know I think I think you kind of said like the three of them in a room and it's a horror movie. I I like that analogy. Having said that, I I think his comments were taken a little bit out of context. I don't think he was agreeing per se that there's a 10% end of humanity, but I think he was agreeing that safety needs to be prioritized, right? >> And that we can't just like run rampage. I think in some ways it was almost a subtle cut at at Sam and Daario that yeah, safety does need to be considered. And by the way, you two are the are the absolute problem children in this particular race. Those two companies. >> Well, he doesn't like those guys at all. >> I know. That's what I'm saying. Like he wasn't agreeing to help them. He was agreeing that like it was kind of like when someone says something and they agree with you to subtly like yes you're right guys and it was like a subtle undercut at like yes you're right and by the way you're the problem and that's why I said like the frontier and by the way you were comment about open source earlier Julie like open source >> looking at each other like you're the problem >> open source is dependent on the frontier okay they're doing some nice architectural and design things but they've been distilling and Jensen talks about that's free markets and capitalism you can argue but they need the next generation frontier models to build these great lowerc cost open source models. Um, >> I'll tell you what scares me though is that they announced that Elon and Palmer Lucky and some other person was getting together for a special meeting on developing military technology with AI, you know, and that's where I think things can get a little bit weird or scary is because AI enabled weapons are the future. It's happening now in Ukraine and and having what I consider, you know, they're great Americans, so I I'm not I don't question their loyalty, but like just, you know, kind of crazy people developing this stuff, you know, it's a little bit concerning. On the other side of the coin, they're making maybe some amazing weapons. So, you know, I don't know. I don't know. >> It might be this all of our benefit that Musk is busy with a lot of other things. And >> that's right. >> Right. >> That's right. He's always been a preoccupied right now. >> I mean, he's going to build data centers in a factory. >> He's going to take over the industry, you know. >> Just get him in the factory. He's great. But outside of the factory that tunnel, that tunnel between Austin and San Antonio. Give me that tunnel. >> Doesn't he have a highspeed rail coming in California? >> I don't know. But as long as he's not tweeting, it's okay. something that we were talking about on on my show this morning. I'm I'm curious in Dan to get your take in particular on this, which is like >> why do we need to keep pushing forward on models? In other words, if we all stopped at the most advanced model today and we all used it and we put it on use in the enterprise, why is because the models by all accounts that are out already are amazing. Why why couldn't you stop there? Why is there this insatiable desire to keep making them >> better? Because it keeps getting better, >> right? But if they're already amazing and they're already transformational and the risks are associated with pushing forward, >> you would be a terrible drug addict. >> I don't I don't know, by the way. I don't know, [laughter] by the way, if the better >> Yeah, that's what I was going to say. There's always more, right? There's always more. >> Is the better model necessarily reducing the safety profile? I don't know that that's the case. In some cases, the newer models actually may handle some of these safety rails better. Having said that though, >> I I think my comment about the IBM head of research talking about the future of research and and AI. I think the most compelling things we're going to do AI to develop quantum technologies, AI to to do drug discovery, AI for chemistry, for you mentioned weapons, but >> or climate change, maybe that'll help solve that. >> Science science. But my point is like >> we have solutions for that issue. People just don't want to do being as far out in the frontier is super valuable. I will also agree with you and and we built a pinnacle benchmark. We actually benchmark every model running on every infrastructure to figure out which model can perform synthetic personas, knowledge workers, data analysts, um executive tasks for the lowest cost per correct outcome. We track this and the point is is most work is n minus one n minus2 of the frontier. And so there is an interesting debate to be said about why can't some of these second level open- source lower thinking lower energy use models be deployed at higher scale to solve smaller problems. And I think I think they can. But I think again >> well that'll that'll end up happening. >> It's happening. It's actually happening. But my point >> is it's like you know my kids will go into like the highest level AI just to ask it like silly questions. You know what I mean? >> Like help me with geographers game. But like >> Yeah. [laughter] Yeah. But my but getting back to like the the question you asked is >> the promise of AI is at the frontier. That's why anthropic may be worth $2 trillion. That's why open AI might be worth $1.4 trillion. Again, if it's all about, hey, we've found everything that needs to be found and we're done here. I think the argument is there. That's where the bubble pops. Like when you've hit the edge of what this technology can do, but if you listen to the Yan Lun and other >> Yeah. I I think it's the opposite. We're just like at the surface of just scratch models are getting so much better. And by the way, they're getting better and they're getting cheaper and they're being adopted and they're being utilized more. And I saw a stat yesterday from a A16Z from reason. And AI uses AI at about 5x the human consumption rate. This is the agentic world we live in. That 5x is going to become 10x, 20x, 30x. AI using AI will be an exponentially bigger market, which by the way will consume more. You think about like Muse, how many ad dollars Muse creates by running tasks. We're keep thinking about humans using AI to create revenue. AI using AI to create revenue. >> How is this actually not terrifying? >> What's terrifying? >> What's terrifying? That that >> you know what's terrifying is is like a bear trying to eat you. I I I don't see technology as that scary. But every front, every new technology has, I think, to Julie's point, been a little scary because you don't see I mean, I got in a cyber cab. I got a new cyber cab this weekend and I I took my wife in a cyber cab and and we we went out. You talk about scary, you get in a car with no steering wheel, if you want a real application of AI technology that could kill you. It's the car actually. And so I did it this weekend and I I was the brave man to get in this cyber cab and test it and and you know I think we're there's amazing things about it's not so much this stuff [laughter] for sure autonomous driving over me. >> What I'm scared about is not being able to see the horizon, not being able to see the frontier. >> Yeah, but we would have never found America with you. you know, like [laughter] >> you know, >> but I think I think Julie's though she's making a good argument for the populace. I get where you're coming from. My point is agree any like we have no idea. >> Yeah. >> We and that's the problem when you're and when if I'm going to find America like I'm driving the boat. >> That's the fun part. >> Yeah. But I I just mean like you know like the >> he just got in a boat and he went he went arrest. But Ross, the populace right now that is running scared, that's creating a divide in our in our economy, a divide in our voter populace >> is being stirred up by the fear of the unknown and the fear of the unexpected. So why I go back to the spokes that sells magazines, >> you know, if it bleeds, it leads or whatever. But like the idea is like we do need to communicate better. We need to communicate the promise of this cuz if you look at what the internet has offered us. You go back radio, television, printing press, all the different phases of but what we never have seen as a transformation happen accelerated at this rate and humans by nature are bad at change. We are not good at change. We have not gotten better at it. And now we're seeing you know this versus the internet. And I was going to just say about the bubble. I said the reason this may not be a bubble is the adoption of a technology has never happened this fast. When you built the railroads, when you built the internet, it took years, decades in some cases years. This is taking days. See the Muse Muse usage went from 800,000 to 5 million in a week. >> Everybody uses AI. It's kind of crazy. >> But it change it doesn't mean there won't be a bubble. All I'm saying is it completely changes the math of how we would determine where and when a bubble might show up. >> We we we have >> but like let me give you an example. Let me give you an example because I do autonomous cars and I do believe that a good use of AI that could kill people is autonomous cars, right? So, it's supposed to make it safer to drive, but in theory, we're letting AI drive us around at high speeds and it could kill people. But what have we seen in reality? So, Whimo has actually not killed anybody. Okay, there's been deaths, but it wasn't because of a Whimo. And it's way safer than human drivers. So maybe it's actually the opposite thing that AI will save lives. It will make driving better. It will be make hospitals and doctors better at what they do. Maybe everybody's scared of like nothing. This is actually a technology that could actually make the world a way better place. What about that narrative? >> No, I mean that's that's a fact. Actually, there's us, Ross. We are those guys. That's [laughter] >> right. But I mean, I'm just saying we're not done. I I I drive in LA and people die every day on the roads in LA. Every day. And I've been working on autonomous software. It feels like almost a decade. And you know, we're finally at the tipping point. Austin this weekend blew my mind. There were Whimos everywhere. There were cyber cabs everywhere. There were Teslas everywhere. There were Ubers everywhere. And people are fine adopting this this >> We're a weird town. That's my That's my We're weirdos. What's insane about I mean it seems like we're all on the same page that we're just scratching, right? >> And that's not even including robotics or space or >> like that's not even like >> the other anti-bubble thesis. >> Well, we heard it called next. >> No, you know what the bubble space >> that doesn't mean the interest is not a bubble. >> The bubble is people calling it a bubble. That's right. That's that's that's I don't know if you followed if you I think Michael Bur closed all his shorts or recently did and he noted it for like tax harvesting or something like that. But you imagine >> following maybe you imagine following his Substack into all his trades and just getting absolutely nuked by it by you know >> well everybody knows not to follow Michael Bur's trades. I mean I tweet that all the time. So if you don't follow >> I mean the guy's making millions on Substack. He's making millions entertaining substack. It's entertainment. She's a great trader or the anti- Kramer. >> Well, not about she, the person who works for her. >> Whoever's trading on her behalf is great. And yes, the inverse Kramer has not failed. All right. It has not. I have you got to give Kramer a little, you know, room to the guy's got to say something every night. So, >> I just I just mean Apple for a long time. You know, it wouldn't make for a good show. >> His Bitcoin call that like literally I think he sold it like at 64. He sold it all like on the day that it literally reversed and ran back. I mean like those kinds of things happen again if you say every day. If you say stuff every day buy >> I want to pivot real quick. My show my show is you buy and you hold. >> Is that double aura ring? >> Uh no. This is this is a this is a very this is this is how I you know work out. I wear a rubber wedding ring but I love my wife very much. >> And that and then you got the >> this is the Aura. This is the Aura and the Apple Watch. I am constantly tracking my >> Are you Are you sure your vitals are are going well? Man, you seem very into this fitness thing. You got an Aura and an Apple Watch. >> I mean, look at him. >> I'm surprised he doesn't have a screen at all. >> I'm a machine. >> I'm not working. >> I have to say I have to say goodbye. But I think this might be the best market hang ever. [laughter] >> Thank you, D. >> Do it again. We'll do it again. >> Thank you, Julie. Thank you, [music] Ross. Great show. >> That was fun. >> Yeah. Thanks. >> Thanks, Ty. [music] >> [music] [music] [music] [music] [music] [music] [music] [music] >> Heat. Heat. [music] [music] [music] [music] Heat. Heat. [music] [music] [music] >> [music] [music] >> Down. [music] Downow. Down. [music] Jama. Heat. Heat. >> [music] [music] [music] [music] [music] [music] [music] [music] >> a heat. Ah. >> [music] [music] [music] [music] [music]