Transcription
Um, our guest today, named as hedge funded tradies, which for the non-nerds in the room means the spice must flow. And for those of you who remember spice from your freshman year in high school, it's not that kind of spice. For Gavin, spice is high bandwidth memory. His arrous is Taiwan, and his sandworm is Jensen Huang. His ex bio reads, "No investment advice, views my own." which may be the most expensive disclaimer in modern finance because he's got 288,000 followers that follow his advice.
Anyways, he ran 17 billion at Fidelity. He beat 99% of his peers, although he just told me it's actually 100% of his peers. And he now runs his hedge fund from Boston, which may be the most contrarian position of all. Ladies and gentlemen, the most knowledgeable man alive on semiconductors. Please welcome Gavin Baker.
Wow. Thank you. And and let me introduce you, Jess. Uh J is one of the most senior partners at Blackstone and he just took over a new job as heading all of Blackstone's AI strategy. Um and was formerly the head of their tactical opportunities strategy. >> We've been friends for a while, so it's fun that he gets to interview me and I get to interview him. Uh but Gavin, I do get to interview you. Um, since this is an investing conference, I wanted to start with your mental model as an investor before we get into the meat and potatoes of semiconductors and memory and the stack and the bottlenecks which I'm sure all of you uh just like me want to engage with you on.
So, you started at Fidelity in 2000 and managed it through two bubbles. What is one piece of Peter Lynch era fidelity the DNA that you've kept and the one that you had to break in order to start and run and be successful with the trades?
>> Yeah. Uh good question. I'd say the Peter Lynch DNA that is forever woven into me as I suspect um is woven into anyone who's who had the great fortune of starting their career at Fidelity, which was an amazing place for me, is this concept that if you like the store or you like the product, you're going to love the stock. And just the importance of engaging with new products, with companies as deeply as you can as a consumer and user, that for sure is embedded into me.
What I have been working on my entire professional career as an investor is there's a stock market axiom that you want to you know Peter Lynch said you want to uh you know pull up your weeds water your flowers you want to sell your losers you want to ride your winners for whatever reason that is very hard for me I'm extremely valuation sensitive I'm very contrarian I'm mo most comfortable on the 52-week low list um I've been hanging on to the memory stocks for dear life Um, and yeah, but that's that's a that that's been a lifelong journey that I try to get a little better at every year.
>> Most great investors point to a single trade, a single bad trade, uh, as the formative one in their own journey. What was yours and what specifically does it stop you from doing today that you would otherwise do?
>> Well, I would say there there are two. I had a really hard year in 2011 or 2012. And it was it was two stocks. Um one was a company called Decrete of Health and they were helping effectively small businesses do uh small hospitals do a better job of negotiating with big insurance companies. You know, you feel like it's a good cause and then you wake up and there's a front page 10,000 word article in the New York Times on Sunday about how they're denying care to uh people in need and this is just simply not true. Nonetheless, the stock went down 90 95% never recovered and just there are always unknown unknowns. There's always risks that no matter how much you try to dimensionalize an opportunity are present.
And the second one is embarrassing. I was Nexttel International. Uh so I was I was a telecom analyst in the past for a long time. You could make a lot of money in telecom. whenever a new network was built in an emerging market, it was definitionally the lowest cost, best and highest quality network because it wasn't loaded. And so Nextl International, they built a new network um in South America. It was the latest and greatest 3G network. They were migrating off something called Iden. Uh I'd made money in this kind of pattern many times. And so made it a large position. And I'll never forget I uh I wrote a letter to the board insisting they do a buyback and the company was bankrupt 18 months later. And the lesson is be very very careful of high leverage because the company was just too levered and sometimes not everything goes right. What happened was a price war broke out between um two unrelated competitors who were much larger. They got caught up in that and I will forever have the black mark of having written u my only um letter to a board asking them to board buy back stock you know 15 months before a company went bankrupt which is also a good lesson.
>> Yeah. Oh appreciate that. Thanks for the cander and the vulnerability. I appreciate that.
>> We'll mix it up a little bit.
>> Sure.
>> Usually lightning rounds come at the end. Great.
>> Put it at the beginning.
>> Overrated or underrated? pattern recognition, scuttlebutt, position sizing, sleep. You just get one word.
>> I think pattern recognition and sleep are highly underrated. I think scuttlebutt is overrated. What was the fourth?
>> Position sizing.
>> Position sizing is very important. You just have to pick your game and you have to play it. You can either be a slugging percentage player or you can be a batting average player, but you have to know what your game is and stick to it and be consistent.
>> Appropriately rated sounds like uh the production function of Gavin Baker. When you actually look at where your edge came from last, maybe it's this year, last year, what fraction is reading? What fraction is your network? Maybe not so much given the scuttle, but what fraction is just being early to one or two correct frameworks?
I think reading is overwhelmingly the most important part of it. I will admit I know I no longer I rarely meet with public companies. I I only I I meet with them if they want to meet with me and it's just they are very well trained. They never say anything that's not in a transcript or 10 Q and I can read much faster than they can speak. And so I can I read a vast amount of transcripts, primary source material and then I do think these expert transcripts are very good and a great great use great use of AI but I would say reading overwhelmingly and then pattern recognition I I do think is also important and being early to frameworks is is helpful. You know, when Nvidia reported that uh great quarter in May of 2023 that kind of kicked everything off, ironically, six months after Chat GPT had come out, I think a majority of hedge funds didn't employ a semiconductor analyst and semis are a lifelong love for me as is deep tech. But I would say now people have gotten really smart on semis.
>> Yeah. Yeah. You were so early to it. Um, and that's a good segue because that idea of a framework and your mental model today memory prices are up 60 to 70%. Micron's margins are probably what high 60s maybe the historical average was closer to 16%. And you've been talking about the compute shortage broadly and and the shortage stacking into data centers and power and extending uh into leading edge wafer capacity. What is your mental model for how this evolves? You've also said that whenever there's a shortage, there's eventually a glut. How does this evolve? Talk us through that.
>> Well, that is that's been true throughout history. Um, and I'm sure there will eventually be a a glut, but eventually is doing all the work in that sentence. Not glut. I would say based on every memory cycle we have had for the last 25 years, this is the time to be selling memory 100%. I was actually the micron analyst in the year 2000. Like I remember going to their analyst day in Sun Valley. I'm a veteran of many many memory cycles and based on history. This is the time to sell.
However, there's one cycle where you absolutely do not want to sell and that's the cycle we had in the mid-90s which is the last true capacity cycle that I I would argue we've had in memory. M
and based on that cycle, we may still be very early. I listened to my friends Alex and Leon, um, who I thought did did a great job, as did Lesie earlier, and I would just say I take the over on every number that they gave.
Every single number, as would I think, you know, they're conservative guys. Um, I I bet they would take the over, too. Nobody want nobody wants to be wrong a year later here at S. So, we may still be early. This may be the first true capacity cycle. And I and and I and I do think that these fundamental shortages are good for us as investors. The last thing anyone should want is a bubble. Bubbles are terrible. They're awful. They're terrible to invest through. The aftermath of them is even worse. We don't want a bubble. And unfortunately also the entire history of financial markets suggests whenever you have a profound new technology whether it's AI whether it's the internet whether it's the PC whether it's railroads whether it's canals you almost always get a bubble because markets are efficient investors understandably become excited about this new technology there's a Michael Moeson frames it has there's a breakdown in diversity everyone comes to believe in this you get a bubble and then that bubble funds the build out that the new technology required and that's exactly what happened with the internet.
I am optimistic that we may avoid a bubble this time. Smoother for longer is what we all want. And the reason we're going to avoid it is we have fundamental shortages of watts and wafers. We're going to address the watt shortages with orbital compute for sure in the next 5 to seven years. But the wafer shortage I think is going to persist for a long time. And the reason it's going to persist for a long time is Taiwan Semi is run by flinty old men and women in their 70s. Not to say 70's old, it's the new 50. I'm 50. It's the new 30. But they're the most important people in Taiwan. The president of Taiwan irrelevant. They are Taiwan and they view themselves as the guardians of Morris Ching's legacy. I remember going to Science Park in Taiwan more than 20 years ago, you know, asking do they did they think they could ever catch Intel and they said it's a beautiful dream, but it's probably for our grandchildren and they did it in one lifetime. So they're the custodians of this legacy. They need to preserve Taiwan. Tai Taiw a bubble and bust is a disaster for Taiwan semi and Taiwan. And so they're just simply not expanding capacity as fast as Jensen wants. Jensen goes there every 3 months and you know maybe they expand you know 5%. He wants them to double or triple and if they doubled or tripled capacity like Nvidia could probably sell one $1.52 trillion worth of chips next year. I really believe that. But the other side of that might be very painful for everyone. And so I think these flinty old men and women who are safeguarding Morris Ching's legacy are helping us all avoid a bubble by enforcing a realworld physical constraint that simply has not been present in past precedent technologies.
>> Yeah. No, it's fascinating as a monopolist provider. They're rate limiting supply at some fundamental level.
>> they just missed Sam Alman has a podcast bro after they met with him.
>> It's >> true. you and we were talking about this backstage which is uh you would take the over on open AI anthropic combined at 200 billion of revenue maybe 12 months 18 months I don't want to put a time frame on it but near-term um it it turns out that code generation was the killer app to monetize AI at least in like chapter 2 of this whole journey and buck and if we're going to go from where we are today to get to that $200 billion in the next 12 18 months where is that revenue coming from is every S&P 500 company going to miss earnings because of tokens to anthropic.
>> I don't think that is an edge case if you are not aggressively Jensen said at GTC that his goal was for his his best engineers to spend a minimum of half of their comp on tokens. And if you just look at the wage expenses of every S&P 500 company, we simply cannot tolerate that level of token spend without significant significant adjustments um you know to the to the labor force, which is the point Leon was Leon was making. But I do think a few things are probably going to lead to not having widespread misses because of token maxing. And if you're not token maxing, you should be. And if you don't know what token maxing is, best of luck.
>> but um, so point number one, all of these models are shifting to usage based pricing. It used to be that everyone in this room could get a good sense of the capabilities of Frontier AI if you spent $250 a month on the best subscription from a Frontier model provider. That is no longer true. The best capabilities are locked behind harnesses and reserved for people on enterprise plans who can pay for use. Now, this is wildly bullish for AI. It's incredibly bullish for the pricing of these frontier tokens. You know, going back to the cellular industry, uh it the reason cellular was a great industry has a has a growth investor. And the reason long distance before that was a great industry as a growth investor is that you bought a fixed amount of minutes and then if you went over you paid by the minute and people really like to like to talk to their friends and family and that was why telecom was a great growth industry for a long time. We're just moving from these all you can eat plans to usage based plans with overage where those usage tokens cost a lot more and we're finding out that there's we're nowhere near the amount of you know people's ceiling price for how much they'll spend. So I think you'll get a lot of productivity.
Um I think that this the fundamental compute shortage we have I thought what Alex said was so profound. 10 basis points of the world's population is using these models the way they should probably be used and we're in an insane shortage despite spending cumulatively trillions of dollars. What happens when 5% of the world's population >> is using these models the way the cutting edge tin basis points are? Like it's just it's unimaginable. This is why orbital compute is a necessity. But to answer your specific question, I do think um someone posted on X that coding may be the shortest path to ASI and AGI because if you can write code to do something and write it for yourself like that's a pretty fast elegant path to AGI
>> and I think it may end up being that coding app coding is not just the you know the the killer app for AI but it's the ultimate AI app and it subsumes more and more and I would just encourage you cloud codeex you will get better answers for investment questions even if you're not a coder than you will using the regular model.
>> Yeah.
>> No that's I appreciate that. You talked about silicon and what's happening on the chip side. Obviously you were very early in Nvidia. I mean you were talking about talking to Johnson in 2000 let alone 2023 when the chat GPT moment happened. Um, alternatives are coming, competition is coming. He's still going to be the dominant part of the market, but competition is coming. As you think about tranium, TPUs, NTIA, which of these is most underestimated by the market? Where does the consensus get it wrong?
>> Tranium by far. Tranium is going to be to 2026, especially in the second half of this year when tranium 3 really ramps >> as TPUs were to 2025. that if somebody is wildly bullish on TPUs today, let's go look at their 13F and let's see if they owned Luminum or Celestica, which were the best ways to invest in in TPUs.
Um, I owned one of those. Uh, um, and so I do feel like I've, you know, I have some credibility to say this, but I do think Google, for a lot of reasons, made very conservative design choices with the TPU V8.
>> Nvidia and Tranium made very aggressive design choices. And so tranium is for sure the most underestimated not only because of those design choices but because the all of these frontier models are what are called mixture of expert models. And to inference one of these not to get too technical you need something called a switched switched scaleup network. And the only two functioning switched scaleup networks in the world today are the ones that power Nvidia's GPUs and uh Amazon's traniums.
>> Interesting.
>> And this is why I mean this is why you know Google invented the ML Perf benchmark. They will not submit TPUs to their own benchmark
>> which you can just see is visibly driving Jensen crazy. But listen, the TPU is a great chip. I'm sure the TPU V9 is going to be amazing. They'll make more aggressive choices.
>> Um, I would never bet against Google. I would never bet against um, bet against Broadcom, but I do think Tranium is super underestimated right now.
>> Yeah, I appreciate it. I want to switch gears to a topic which actually is how we first got connected way back in 2022 and then 2023. Neoclouds. I called you in the summer when you were on vacation somewhere to bug you about a company called Core Weef in the summer of 2023 and get your advice and input. Ultimately that led to us investing seven and a half billion behind core to scale them up at a really pivotal moment.
Um, so first of all thank you for that. Uh fast forward to today coreweave cruso nebus lambda etc. Is the category durable today? Is it a transitional arbitrage on hyperscaler capex and token friction?
>> I absolutely think it's durable. And so first of all coreweave is a little bit of a sad subject for me. Uh I could have a trades could have invested over $50 million in the round at 1.1 billion and I was conflicted out by Crusoe and I love Crusoe and I think Crusoe is going to work and we have a large position in Crusoe but every time I think of being able to put 50 into Coreweave at a billion dollars I get a little sad.
Um, but I'm very happy you put seven and a half billion in. Yes.
>> Um, it's it's absolutely a durable category. And the way to think of running one of these clusters is like driving a Formula 1 race car. You watch Formula, you watch these Formula 1 races and it looks easy. It's like anybody could do it. You know, the same way like if you watch Tom Brady, it's like, well, you know, like why'd he miss that throw? Well, because he's in a stadium with a 100,000 people yelling. There's a bunch of people who weigh a 100 pounds more than him running 20 miles an hour at him trying to kill him before he throws the ball. And Formula 1, it's the same thing. It looks easy, but it's really hard. And that is to say that there is, you know, if I tried to get in a Formula 1 car and drive it in a race, like I would die. I would be just a danger to myself. I'd be a danger to everyone, including the people in the stands. And that is running a cluster. It's really hard to do well. And the reason that they're, you understand this, but the reason a company like Cororeweave can charge a huge premium for their GPU hours is because all GPU hours are not the same. And those coreweave GPUs are being utilized two to 3x more per hour on a on average than kind of the GPUs from maybe a like a bottom more bottom of the barrel provider. And by the way, this all goes for Crusoe and Nebus and other other highquality Neoclouds. But I think this is very underappreciated and people think oh that can't be a durable competitive advantage. Well, I was a retail analyst in the year 2005, and I observed all you had to do to create a $50 billion market cap retailer in America in any category was be able to run 1,000 stores in 50 states with different climates in different, you know, consumer preferences, that are staffed by friendly, knowledgeable employees who don't steal from you, that are stocked with the right inventory at the right time, at the right prices, and have clean stores that are well lit. That's it. Like in history like 10 companies have been able to do that and running one of these clusters is even harder. So I think it is it it is durable and the hyperscalers for a long time were stuck in a cost mentality. You know, the hyperscalers were competing with people running these Formula 1 cars and they were like, you know, doing like overnight shifts and 18-wheelers trying to stay awake,
>> deliver the lowest cost. And that's not what AI is about. Now, I think they've they're making this this mental and cultural shift and they've made it,
>> but I think some of these Neoclouds have a very durable business model.
>> Yeah. No, we certainly agree. We've talked a little bit, you've actually alluded to it already tonight, which is uh orbital compute and uh we're not going to get into the science of it because there's papers you can read online and send it through your favorite LLM, Grock, etc. to distill.
Um, when does it become commercially uh in a way that it gains meaningful market share? Orbital compute, when does that commercially gain real share? And what is your view as the most underappreciated short in the market today that is not being priced in? Is it terrestrial data center operators? If if that's the case, we're in a lot of trouble at Blackstone. But is there are there other things that today are clear shorts for you based on that?
>> I think it becomes clear that it is possible going to work and economical in the next two years. I think it starts to take meaningful share towards the end of this decade. I do think you know there is a there may come a day where we never put another like terrestrial if data centers that have been built and are in the ground are always going to be val valuable. You're going to do training and RL in terrestrial data centers but I can't imagine a day when in the next seven years when we never build another terrestrial data center.
>> Mhm. And the years leading up to that are going to be very painful for a lot of the companies in the power cooling spaces, you know, these industrial names, you know, who have massively flexed up capacity to and you know, support a buildout that could really come to a screeching halt. And in space, the power comes from the sun and the cooling comes from kind of the dark side of the satellite. If you've seen the the um there's certain names that I'm not supposed to say, but if you've seen the illustrations of a prominent potential provider of orbital compute and what their satellites are going to look like, the the radiator is three or 4 hundred feet long and it literally the it's in a sun synchronous orbit and it sits behind the satellite. So you have these big solar wings, a satellite and the satellite is a rack. It's not a data center. just a rack eightt tall, two and a half feet wide, four feet deep, and they're stitched together using lab lasers to make a virtual data center. And then you just have the radiator in the sha shadow behind the rack.
>> Yeah, Gavin, we're a ton, but thank you. This has been fantastic and appreciate you doing this.
>> Thank you. Thank you all.