Tierras raras y pánico por la liquidez: por qué las conversaciones entre China y EE. UU. importan más que los recortes de tipos

[00:00] Another back and forth week as the rare earth battle continues. A lot of uh public statements made. Uh some news on dark factories in China worth keeping uh in the back of your mind as we get close to humanoid and autonomous vehicle season. Uh go through what happened for the week. We did see the credit fears turn into bank fears as some of them reported. The earnings overall have been great to start off. I'll go through that both on the financials and AI the
[00:30] podcast you guys are going to start hearing more and more about the singularity being near because uh as fast as things are happening in AI and given the fact that it's still filled with bubble uh fears uh the acceleration is coming. So, uh, you don't really have much time to start to get rid of your bubble fears if you want to make alpha. Uh, and the crypto four-year cycle fears seem to be dominating right now. And I'll kind of go through some of that. So, like I
[01:01] said, uh, a lot of back and forth. You know, when we left off last week, uh, Trump had talked about big tariffs and then everything was fine on Sunday, clearly caring about the market. uh that went on all week. You had the Wall Street Journal basically saying uh they want to deescalate, hoping to stabilize the markets. Uh nobody wants the markets to fall in the US. Clearly the administration is trying to walk a line of of being both strong and also
[01:35] recognizing that uh they don't want the market to fall. Trump has basically taken on the market side. Uh Bessant was very aggressive today, very different than uh at the bottom of the market back in April where he became kind of the the person of of reason. Um the tariffs uh they're going to go through they're going to be fasttracked if China escalates the rare earth fight. Um I wouldn't say there was escalation this week if anything aside from just kind of the back and forth. Uh it definitely
[02:06] seems to be everyone sending a message for negotiations. Yeah. JP Morgan saying they're going to plow a trillion half dollars into security industries. The Pentagon stockpiling critical minerals. You had Pakistan in bringing in some rare earth. Um and pretty much now you can finally get podcasts and everything on rare earth. It's taken the entire year uh to get to this point, but the reality is now people are starting to recognize the importance of it. Um the U we had the
[02:36] IMF meetings this week in the World Bank and basically Bessant uh and Greer both said to the world that this has to stop that they're trying to course it. You had a uh a move in from the Dutch taking control of a China owned chip maker. So there's definitely a global sense to this. Now I highlighted Hugh Xiin last week and I just want to make sure that again as you're reading things and you're trying to go through things on both sides. I think continuing to understand there's a couple things in
[03:06] this um this turning point sense of being strangled by the US feel uh feels today from rare earth export controls is exactly like the one China experienced basically everything that keeps coming out from China is that China's trying to stand up to Trump. Uh but I believe China will show full respect for US interests as long as the US is doing the same. And I just want to make sure you you recognize this has been long in the making. So when I started at Morg Morgan Stanley, China's leader was Deng
[03:37] Xiaopang. A decade later, the start of the 1990s saw numerous rare earth milestone events. So again at the start of the 1990s even before the internet uh beginning with the opening of China's national lab for rare earth materials 1991 Deng Xiaoing established rare earths as central to China's industrial policy with his ominous quote and possible implied threat of using rare earths as an economic weapon. The Middle East has oil. China has rare earths. Again, this was back in the early 1990s. If you want
[04:10] to go read something on more of a negative side of how this is going to end up and not be an agreement, China's stage two coercion of rare earth rising risks. Um, part of the thing that I think is important to notice, and this goes for two reasons. One is this is the export chart. So, China's exports were released at 8.3% year-over-year. Basically, no different than where they've been the last couple years. And even back in 1718 during Trump back here 8% good good growth number. This white
[04:41] line here is exports to the US down 27%. So there's two things about this. Number one look at this gap. These things have moved together completely over time and all of a sudden now you have this huge gap. China is not exporting anything to the US. It's also not buying soybeans. So on the one side you can understand why transport stops, PMIs, all of this stuff, China not exporting to the US is a big deal. Um it's just it it has implications in terms of all this if we get a trade
[05:12] deal. Anything that involves either keeping tariffs at the lower level or even getting rid of the tariffs if China has the ability of doing that by trading it in for investments. anything along those lines and I think you're going to get a major move in risk assets. I think this is holding up the entire picture. Uh I still believe there's going to be a deal because I don't think China can hold off the rest of the world with rare earth for more than 3 years. And even though that may be a goal, uh I don't think it is unless they're planning war,
[05:45] uh it just doesn't make any sense. So part of the thing is they were up 8.3% just showing that China has the ability to offset falling exports to the US but they're holding rare earth on everyone. So the fact that in the EU they were up 14%. They don't have a strong economy and exports are critical. They don't have a domestic economy right now to handle everything. So I still believe there's going to be a deal. And now when it happens because of what I'm about to show you on everything else that's gone on this week.
[06:16] This is holding a b a ball underwater at this point in terms of earnings, rate cuts, everything. PMI is about to go higher. This evening, Ambassador Greer and Bessant, this is on Friday night at 10:16 p.m. Michael Pettis reposted this. They had frank discussions on the trade relation talks will continue next week ahead of the scheduled. So, as of now, she and Trump are scheduled to meet. Trump spoke in an interview uh with Maria Bard Romo that'll come out uh
[06:46] tomorrow. Well, you guys will see this uh it might be out uh before this uh is released, but I I do record these on Saturday. Uh but he acknowledged that his threat and tariffs on China are not sustainable. He's blaming China, saying he has a good relationship. I think we're going to be fine. And the summit is still on track. I think that's the the main point here. Um the core of the issue according to Trump is securing a fair deal. I think they both want a fair deal. I think I think
[07:16] there's going to be one that happens and again I think there is a lot of upside because the economy right now in terms of earnings in terms of the singularity which is coming massive massive positives and I still believe that people are way too negative in general on AI and see it as a bubble. Still an 80% chance in poly market. We did head down again. and it followed the stock market. Uh so it's moving click for click. This gets into the dark factories. Western executives who visit China are coming back terrified.
[07:46] Robotics has catapulted Beijing into a dominant position. These are not jokes. These are not me any Ford seat era. There are no people. Everything is robotic. It's the most humbling thing I've ever seen. Executives describe touring facilities in China where robots do much of the work that there are few or no operators so-called dark factories that run without lights. We are in a global competition with China and it's not just
[08:17] EVs. If we lose this, we do not have a future at Ford again. Humanoids is becoming part of it. Andre Horowits, America cannot lose the robotics race. Entirely autonomous dark factories like those of smartphone automobile manufacturer Xiaomi operate in complete darkness with no humans present. China is running away with the hard power side of AI. I bring all this up. I'm still going to do a lot on
[08:47] Tesla, but Thursday, November 6th, you have Elon Musk shareholder meeting. Remember he signed this massive package and there's speculation that Optimus 3 will be shown. Remember what the release of the smartphone was in terms of Steve Jobs showing it. At some point here with Musk, we are going to see humanoids displayed in a way which is going to change the investment environment forever. Not for a day, not for a week. The iPhone changed and created the Mag
[09:17] 7. People need to start to get ready for this. Autonomous vehicles is one side. I've talked about Tesla. I've written about it and people are not paying attention. Elon Musk says the biggest ROI from AI will be human humanoid robots. China's EV exec says China roll out of Tesla's self-driving tech is a deepseek moment for autonomous vehicles. China thought they were ahead in everything relating to EVs and autonomous vehicles. They dominate where over 60% of all the cars released in
[09:47] China have autonomous capabilities and they went up for a test had 36 cars. Tesla won. Tesla escaped the test relatively unscathed. This was back in the summertime. The secret fuel powering China's self-driving cars. Is there lack of regulation? Uh I don't think it's a coincidence that the US is basically again starting to come off of some of the tariffs to help the parts of the economy that need it, which are soft, including the auto market. So for the week, S&P up 1.7% despite the back and
[10:18] forth. Q's up 2.5%. Uh IWM up 2.4. All of these are the best performance in a week since August. Uh so even though September was a good month, good week for for stocks. Semiconductors, semiconductors, semiconductors. I keep saying compute, compute, compute. Do not pay attention to the charts. Pay attention to how big the demand is and how massively massively behind we are on
[10:50] compute. So when you hear bubble talk, go by semiconductors. Semiconductors were up another 5.4%. This was, believe it or not, the biggest rally since the end of June. I wrote this paper on the fra fragile immune system. This had two purposes. One is just to highlight that I think for the next two weeks because we have the back and forth on China and the US. So don't be surprised if we get more negotiation uh ups and downs. You also have the earnings from the MAG 7 in there. uh and you've had all of the the credit risk,
[11:21] but also you have the blackouts for buybacks. You have a bunch of things going on. Um but also, this was meant to just highlight that every now and then we're going to get these VAR shocks. The VIX went from 16 to 29. Um I talked about it that I think VIX calls are always a way to play a concentrated market like this. We're going to have a lot of episodic falls because there's very few names that are running. Uh, and I think you're gonna have that where some names are going to be tripling in a year and a lot of names like Ford are
[11:51] going to be range trading for the next five years while they go out of business. Um, I tweeted this. If you haven't been following my post, I am starting to post more just with things that are happening during the week and things that are going on because more and more people are requesting more videos. I'm not going to do any more videos. This is going to be the only one. Uh, I will start recording some stuff for helping people learn AI. uh getting them some other things to hopefully get them the ability to follow markets this way. Um go from podcast to ideas. Uh know which names to
[12:21] go through. I do want to create some videos on this. I'm just trying to put it together with 22V. Uh regional banks, the regional bank ETF down huge one day this week. Uh aside from SVB and aside from liberation day was a big move lower as the fears were spreading. Jamie Diamond said there's not ever one cockroach. Um, but we also have liquidity issues that were popping up. This got X all lit up. Uh, you had Sofur verse Fed funds uh ticking higher uh in
[12:53] a way that suggests that the market is strained for liquidity, which is true. Um, the Fed's uh standing repo facility, someone tapped it for one of the largest things since CO uh you have a you have a liquidity issue. um that shouldn't be a surprise at this. It's been known. Uh I'm just going to highlight this as my move, my face. Uh the breath with inside the bond market, the corporate bond market also showed the weakness this week. Um we do have liquidity issues, but that's why the Fed and Pal said that
[13:24] QT is is probably going to be stopped sooner uh very soon at this point, maybe at the next meeting. Um the VIX did get up like I said up towards 29 which basically put it at you know the highest level since just kind of the weeks after. So you had a big move as I mentioned CTAs are modeled to sell over every scenario. They've clearly been selling at this point the buyback window as I mentioned you've also got systematic V strategies which need to sell. Um but once we get out of this point that's fine. Now, I also did
[13:54] highlight that we've seen this Goldman Sachs uh crowded long versus crowded short, which is the white line index trade lower. Now, I overlaid this because it's been moving with quality. So, this is the Morgan Stanley quality basket. This is long, high quality, short, lowquality. So, it's been the garbage things or the junky names that have been causing the pain. Now, this white line here is the crowded shorts relative to the S&P. That's the white line. So, it's had a big move. This is
[14:25] IWM over the SPY, which has bottomed since August and gone higher. This green line here is the PMI. If the PMI does go higher, as I believe it absolutely will, and head up towards 60 next year, I do believe that you will continue to see a lot of the junky names rally, a lot of stuff go. If you have the combination of this cutting two more times, I think people underestimate the importance that makes for earnings for smaller cap businesses, which obviously depend a lot on on rates. So rates moving lower and
[14:57] at the same time, you're getting blowout earnings. So far, we're about 10% into the season and we're getting strong sales growth and we're getting strong earnings growth and technology doing what it always does, but you've got financials and you had a big week for Goldman and for Morgan and for all these. And you can look and see what the surprises were for each of these names. Just to run across the board, just look how many of them were at least 2% surprises on the sales side and then where we had on the earnings side. So
[15:27] the earnings have been great. Capex for next year is going to be massive and it's going to be massive. Why fade this? Why fade the numbers coming through when it comes to the stock market? These impact the earnings. They impact everything going on at the uh at the economy level. But that's not the only thing that's happening. Now if you want to be short the mag 7, I don't have a problem with the spenders. The spenders are starting to act weak. So this white line here is an equal weight of the hyperscalers the
[15:58] four you know big boys the Amazon Google Meta uh and whatever one Microsoft I'm forgetting they're unchanged basically since mid August they're overlaid here with what they're spending on we finally have seen a divergence so the semis are all the way up here and this is on the same sc uh the the same scale so semis are up 41% but all of that outperformance came since September so you're starting finally to get the break and I think this is where the issue is and what I've said to people that I've
[16:28] met with and gone through and if anyone wants to see a presentation on this beginning at the iPhone we had these companies the mega cap 8 massively grow their price to sales have just continued to go higher this to me is basically modern-day debt to equity so if you want to find how leveraged a company is it's really in their multiple so if they're spending t tons of money they need the sales to come through this is what everyone's compl complaining about. I don't know if they're going to get the sales. I don't think all of them will. I
[16:58] think some of them are going to be in trouble in terms of the spending. It doesn't mean they're going to collapse because this isn't debt, but their multiples will contract over time. And that can happen by the S&P going higher and them going sideways regard which is what's been happening since the summer. Regardless, their earnings are coming up soon. If they can't show that they're making money, they're going to have trouble. Here's the chart again, but this time I included the one that I think is going to outperform dramatically. And since the majority of you, whether you're a hedge fund or a mutual fund, don't believe in Alon Musk,
[17:28] prepare for danger next year. I believe it's a bubble. Blah blah blah. I This just goes on. It's actually boring. I will feed everyone some bubble food, though. Um there are plenty of stocks with inside the energy space. Nuclear, um all of this, they're just Momo names with no earnings. and I don't know if they're ever going to make any big amounts of money. So, there's plenty of froth with inside this and I think retail's been having fun with it and I
[17:59] think they should trade the momentum. They seem to be very good at it, but I think that is an area that is a bubble. I just don't think the market and what's going on especially for semiconductors and the spend fading the spend is a mistake. AI is far far ahead of the supply as I'm going to continue to go through. We had these four companies TSM, ASML, Applied Digital, SKH Highex, all in different parts of business lines related to it. You got memory, you've got lithography, you obviously have the FABS, and then you've got more on the real estate and energy side. So you've
[18:29] got a really big broad thing and they all blew away numbers applied uh digital and I'll spend time on this one mainly because this is the type of thing that if you fade your when you look at Micron and it's gone from 60 to 200 and I've talked about in this podcast the entire year applied digital describe the current environment as the AI revolution hyperscalers are anticipated to invest around 350 billion this year which is basically what I just showed you going up to 450 50 plus next year if not 500
[19:02] we have sell see ourselves well positioned their revenue surged 84% year-over-year not surprising that this is what their stock has done so if you fade this you're fading literally something that was trading near lows at the beginning or right after uh the lows and then we were still sitting at $7 now at 34 as you entered the summertime time. It is a huge mistake to call AI a bubble and
[19:32] cost yourself alpha if you're waiting to try and play this on the short side, which I know very few are because they would have already been knocked out. These are just words just like stagflation and hyperinflation and shorting bonds and every other stupid word we've heard all year. The reality is there's tons of money to be made on these, but these charts are hard to buy because they're straight up. As someone told me at one of the asset managers, it takes courage to buy a chart like that. I completely agree. My only thing here is I'll show you the bottlenecks and the
[20:02] things that can slow this down, but AI is far ahead of everything that you're seeing. Samsung, SK, all of that. They're way ahead of it. We're not even close. So, the summary is the demand is deep. It's not just shallow hype. And you guys can read all of this on your own, but basically across the board, when you take five companies like that that reported this week, that's what you get. So, if people have the view that it's a a bubble, and I've I've shown Harris Co's piece uh on uh it back in August, a
[20:34] blog post that he did on the hyperscalers and the difficulties in them reaching their spending, which I agree with. What I don't agree with is the argument that the market won't reach the spending. You can make the argument that Microsoft won't get it, that they won't get it, but to argue that AI is not going to bring in trillions of dollars of revenues to me is ridiculous. And I've heard people say, and I won't get through who, that it's only brought in 20 billion this year, which is just wrong. Wrong beyond belief in terms of adding up the hyperscalers reist. We're
[21:05] well over 100 billion, way past it. And then when you add in the benefits that come on the productivity lift, and that's what Matt Matthew Seagull's going through. But when you get these bubble pieces, find someone who's on the other side or just bring it into Chat GPT as I did and say, "Hey, go through the two arguments and be the judge. Tell me who wins." Matthew Seagull's argument is more accurate. Matthew Seagull's argument is more robust because it's grounded in immediate tangible benefits of operational efficiencies. You don't have to sit there and worry about them. I get
[21:36] sent these every day. Most of these are clickbait. If they have a lot of followers and X, it's meant for clickbait. Honestly, it is. People might believe it, but they're not sitting out there shorting. And if they would, they wouldn't be going through it. So, a lot of this is just people kind of going through and and and and putting out there what they want. And for media people, it's clickbait. I'd be putting bubble stuff, too. In fact, go to my videos. How many times do I include bubble in the title? It gets more views. I hate to say it, but that's the truth. That's what people search for. People are fairly negative. So, it's an easy
[22:07] way to drive revenue to you if that's what you're doing. Um, best podcast of the week, Dave Blondon, who you've probably heard if you've listened to the Moonshot podcast, very level-headed guy, been at MIT uh, in some form since the early 1980s, working initially for Marvin Minsky. Um, worth listening to this. Some explosive things in here. He doesn't get a chance to talk about this on Moonshots as much, but he's very levelheaded and went through a bunch of things that to me are important. Two versions of AI progress happening at
[22:37] once. Uh this is the most important thing of everything I'm doing this week is for you to understand there's the public facing models. These are the ones that we use, I use, you use. Whether it's GPT4, GPT5, doesn't really matter. These are stable, heavily tested and cost optimized and they're still constrained by compute and safety filters. These are released. These are not the models that they're using behind the scenes internally. Non-public research models are further ahead.
[23:08] They're used for self-training, architecture, exploration, and automated improvement. The reason this is important is he talks about how he was there this summer. He understands is he's been in AI since 1980. He had a unicorn in the dot bubble. Recursive self-improvement, which I've highlighted before, is the step before the inflection point, which means before AGI. We train models manually with human data. Models begin training themselves.
[23:38] That is the critical transition. When we move from LLM as software, LLM as researcher when they're doing the work, Marvin Minsky, Demisaba, Sam Alman had viewed this as the takeoff point towards artificial general intelligence. Everything I'm going to show you from here for the rest of this video is about the fact that we are there. from everyone behind the scenes they're saying we are at recursive already self-improvement it's not to the point of AGI but it's the takeoff point which
[24:09] means from here the compounding that's been happening you have to ignore the models for everyone who call say the models don't do anything that's already BS but to say that you know what they're going to be able to do is just not there the weakness is compute so you have to think the models they have are incredibly compute inensive. That's why when they release V3 or Sora 2, you get to use it for no more than 15 seconds. Vo3.1 is coming out soon. It doesn't really matter.
[24:41] They're seeing real progress internally that isn't yet deployable. He's talking like I said a lot of internal compute resources were diverted towards recursive self-improvement. Everyone that I know that uses chatpt in the manner that I do and to the extent has said that it's very obvious that chatpt is restricting now the usage. They've done everything they can to slow it down. Compute is their backlog. They have research product ideas they'd like to run but they are blocked by
[25:11] insufficient computing hardware test capacity. They have products further ahead than the ones they're already making billions of dollars on. ones that will help with cancer, ones that will help with things that will blow your mind. The problem is they don't have enough compute. So go buy your semiconductors even at this. If they're not really expensive for next year's earnings, they're not expensive. Sam Alman's comments on capacity overhang. So remember, I highlighted
[25:41] this last week. Sam Alman said this on a podcast. You will not hear this from Goldman Sachs. You will not hear this from Morgan Stanley. they will not get Sam Alman and he won't be allowed to say this. We have better models than we've released. When you listen to an interview and it's just going, there are so many nuances and nuggets that that's where your information is coming from. If you're reading research from a salesside analyst who happens to be in their 30s, 40s, 50s, unless they're spending all of their time listening to podcast, they're not. They're spending their time talking about AI bubbles all day long. Trust me, because I refuse to
[26:12] do that, but that's what's going on. He described the capacity overhang of having these models broadly due to the limits in GPU and accelerator availability. High bandwidth memory, data center power, cooling, and network. That's what you invest in. That's the stuff creating the alpha. Hearing multiple independent whispers that OpenAI just had a breakthrough too big to announce casually something emergent. People inside calling it phase two, others saying level four. out of Google this week. We're doing all
[26:42] kinds of stuff with these models that the public isn't even thinking of yet. Pay attention people. He's a top researcher at Google Deep Mind. After all, attention is all you need. Yet, it seems to be severely lacking. Um, so OpenAI does all these deals and you're wondering how they're doing these deals when all they have is Chat GPT and 1620 billion of of annualized revenue. I have a feeling all of these guys, or at least the ones that they did deals with here, have seen what they're able to do.
[27:13] They're securing the compute that they say they need for the models they've already built and trained. They just don't have enough to release them to the outside world, and they don't have enough to continue to drive them. That's why they want all of this. It's not some investment that isn't happening now. They just can't release it for demand. So, use your brain. Think about what that means. And then when you hear someone say it's a bubble, they're not making any money. When something is moving exponentially, today is always yesterday. Keep that in the back of your
[27:44] mind. Especially when the models are talking about are at least 6 months ahead to a year in terms of what they can do and then go through the compounding and read what Alon Mus has said. Read what any of them said. Alman's capacity overstamp isn't a warning. It's a road map. Bottlenecks drive vertical integration. So because of the bottlenecks, he's literally trying to buy everything he can to ensure that the value he has inside the walls, he's able to actually do.
[28:15] That's the problem. He thinks he has everything. He thinks he has AGI. Dave Blondon, when asked, "When do you think AGI happens?" 20 26 to 27 for AGI strongly implied to rest on the belief that of recursive self-improvement. Again, recursive self-improvement. Go listen to what Eric Schmidt has said about it. He's basically saying it's there. So is Alex from Moonshots has been saying that repeatedly. So has Ahmad Musto. All of them have said it and they're involved. They're all AI
[28:46] people, not some hedge fund guy saying it's a bubble with absolutely no knowledge of AI and doesn't use it. So every time people send me these papers, I literally just can't deal with the stuff that comes with it. Today's forecast reflects not a gradual scaling, but an inflection point where recursive learning kicks in. You may see an acceleration that's much steeper than the prior annual gains. Again, if that's the case, you can expect to see unbelievable things happening this year. A recursive loop only works if computers abundant enough to run multiple internal training
[29:16] cycles. So it basically comes down to compute. If you want to slow AI down, it comes down to things like rare earth. It comes down to things like uh power, all of those things. MIT just released a paper this week that it can write its own code. Again, getting into the self-learning side, memory cycle. I talked about how I could not believe that experts were saying that we were going to have a memory supply over supply next year. Again, we just heard the demand is further ahead than what's actually going on. And it's all inference. So, this paper was just on
[29:48] the memory super cycle and the demand side. Memory demand of large AI models topic of concern. Uh in the past large language models training was the core driving force behind the demand for the entire AI infrastructure. But now inference workloads are rapidly becoming the dominant force. If you go into this paper again and you can find these just through X through podcast. I don't even remember by the end of the week where I got all this stuff. Um, that's how much it's just kind of a constant thing for
[30:18] me to go learn about what's going on and read about it so I can not only talk about it, so I understand it, so I can make money off it personally, so I can invest in Bitcoin, which I believe is the perfect trade on this. Um, I want to make sure that this is going and I'm not missing something. I want to make sure the China US thing is going to happen. The next semiconductor upcycle will be driven not by model trading, but by inference. Every chat GPT style inference keeps data constantly moving, creating a long duration memory cycle. It's 24/7. the explosion of inference workloads and edge deployments. And again, edge deployments when we start getting our phones, our computers,
[30:48] humanoids, autonomous vehicles, all of that is on the edge, meaning the brain is now inside the device, not coming from the cloud. We need more and more memory. This is not cyclical. This is structural. It is a super cycle that could persist through 2027. Keep that in mind. 2728 is the point where everyone is kind of thinking that we will have enough data center capacity, we'll have enough compute, blah blah blah. So let's just keep focusing that we have another year maybe year and a half of investing in this and
[31:18] we'll see where it is next year. So for all of the capacity overhang you can go look at the core issues go down here a lot of money to be made on this. Um both converge on a single thesis the AI bottleneck is physical infrastructure it's not the demand side it is literally this thing. So, all of these companies in here benefit from all of these different components that are necessary to be able to meet Sam's needs. He's been buying things within all of these
[31:48] or making deals with inside all of these in some way, shape, or fashion. The All-In podcast had Brad Gersner on, made it worthwhile completely for me to listen to the whole thing, but they also covered in a bunch of important topics that I think are worthwhile for you guys. They went through the OpenAI AM uh AMD deal since people freak out about these things all the time. Also, this round tripping concern, they talk about that, but they also go through the demand and efficiency. No dark GPUs, which again means that we don't have a we have a demand. We have a supply
[32:19] issue. We don't have a demand problem because every GPU is being used. So, when I hear people say, well, these are going to be out of date in a few years. We're using everything that's available. There's nothing that's not being used and we have demand for more. So even if it means we're using older ones, we're still using them. So there's no data centers that are going to be turned off as I heard someone say in the next 3 years just because they have chips from Nvidia. Um they went through the TAM and talked about how big it would be. So again, when you're getting into what Matt Seagull was debating, you can go
[32:51] get trillions of dollars. They talk about it. I I these guys again are experts with inside the field. If you listen to 50 people and they don't have a vested interest completely across the board, they're just speaking. Um, one thing that's important and I just want to bring this up because Nvidia, um, you know, I've tried to find ways to be negative on it, but the more I've heard everyone speak about it, include Jensen, this has become an issue where its moat is the fact that it isn't a chip, it is an AI factory, and that's because of all
[33:21] of the component pieces. They've secured the high bandwidth memory packaging. I just showed you the memory issue. They continue to make progress in terms of what Brad Gersonner says is the most important thing which is doing things in performance performance per watt. That's the way everything is being measured now is system level perform performance per watt. And that means that it's the efficiency thing that everyone was worried about with invi with DeepSeek. These are the guys winning that battle
[33:51] in terms of the efficiency. So Brad emphasized that Nvidia strength lies not merely in raw chip count but in how much useful computation you can get per watt of power. If computation is the issue and power is the issue that's what matters. Can I get performance from it? So, uh, did they bring up if competitors try another approach that they are taking a big risk on unproven commodity? And yes, if someone decides they're going to try to not use Nvidia to build a data center, they're taking risk on something
[34:23] that hasn't been done at this size before. When your customers deploy a gigawatt, that's 400 or 500,000 GPUs. Getting 500,000 GPUs to work together is a miracle. What customer would place a 50 billion purchase order on an unproven architecture? That is why they have a moat. So you can continue to watch Nvidia as as the the story for chips. they may not make as much at this point because they are big and so the dollars that they're going and how much they can
[34:53] produce maybe they can't produce as much as they need and you're better off looking at the other places that I mentioned in in here in terms of the names which are just starting to go plus the names that are going to benefit from the edge side and the NPU side but the these are the issues that show up in terms of this one of the ones that's starting to come up which and you can see rare earth here these are the things that can stop it this one is becoming a much bigger issue. The last two weeks I've debated to kind of spend a lot of time on it. I will bring it up here, but state regulation is becoming an issue
[35:25] for AI and it could slow it down. There's a lot of different things. Uh Colorado's already put something out which is they they've done um Anthropic got David Saxs going at them this week in terms of them saying doom and gloom stuff. But the reality is the midterm elections are coming up and AI is going to be a dominant part of the midterm elections. Both from the jobless growth, GDP up, hiring down. I think this is going to be a major major story. I expect the GDP to be great and I expect the hiring again to be near zero. And
[35:56] I'm going to bring this up further in episodes. I had a lot of people call up and say, "Isn't the hiring all about the lower participation due to the fact that immigration?" Just remember guys, there is a big difference between the monthly labor numbers and the sentiment that's coming from the conference board, that's coming from temp hiring, that's coming from every single uh thing that shows the hiring market is soft. So this is not a supply issue only in terms of the
[36:26] numbers. There is definitely fear of employees right now from AI and businesses are not hiring the way that they used to. The economy is showing signs of weakness if first brands and the defaults that we've seen that are fraud. There's going to be more of these because companies are not. It's a K-shaped economy. You're going to have more of these. The electricity costs are going to go up. You're going to have the fake viral videos which will be part of it. Big money to shape regulations. So again, I think the elections are going to be big on this. Uh this week's where
[36:58] are the jobs going to go and what's going to happen. Um these guys wrote an article. They're from a a a company uh mechanization mechanization. Um but just type in the future of AI is already written. It's a good report. It just goes through the fact that autonomous agents that are fully substitute for human label willy be created because they will provide immense utility that where AI tools cannot. The main thing I want to bring this up on this is that their focus is the fact that even if people wanted to
[37:29] stop AI from impacting jobs, it can't happen. And the reason is capitalism and public companies need to focus on shareholder wealth. So that has been the story of US equities for a long time is to find a way. That is why when companies announce that they are firing lots of people, generally the stock goes up because their expenses went down and so their net earnings by definition unless their revenues fall equally as much. It's generally thought of as a good thing. The problem is
[38:03] argues economics in economic incentives not moral choice drive technological direction. It states that autonomous agents that fully substitute for human labor will inevitably be created because they provide immense utility and that humanity is more like a roaring stream following the path of least resistance. This logic mirrors the US focus on shareholder value maximization. When capital is rewarded for efficiency and profit, the path of least resistance becomes labor displace displacement. This is going to be next year, guys.
[38:34] Charlie Munger, show me the incentive and I will show you the outcome. The incentive in US capitalism has long been maximizing shareholder returns. Therefore, the outcome is predictable. Firms pursue efficiency, automation, and cost reduction even when it comes at the expense of labor. This is going to be a dominant issue and that's before this. I will be doing more research for 22V over the course of the rest of this year in preparation for humanoid robots. So, I just started doing my research, deep research to come up with the reports. Just to give you an idea, again, this is
[39:06] about a third of my prompt in terms of what gets done for my deep research for humanoids. Buzz around Gemini 3 Pro. So, Gemini 3 is coming out soon. The leaprogging and the moving comes and as I've mentioned Gemini has pretty much taken over at least for me for all my deep research reports for sure. Major advancements in cancer this week. Sundar Pchai himself put this out. An exciting milestone for AI and science are
[39:36] built with Yale and based on Gemma generated a novel hypothesis about cancer cellular labor which scientists experimentally validate and live in cells. This discovery may reveal a promising new pathway for de developing therapies to fight cancer. Then there was another one this week. Another huge AI cancer breakthrough from Google. This is two days later. Two AI cancer breakthrough announced in just two days. 2025 will be remembered as the last normal year before the world went fully exponential.
[40:09] To finish up the Dave Blondon side, he went through the embodied AI, but this gets into the longevity and sequences side. So for for humanoid robots, physical labor costs fall towards energy. We're at the point of humanoids. It's going to start going. He talks about the fact that there will be hundreds of thousands by 2030. The Elon Mus thing is very important in terms of what they announced there. Figure O figure AI has been releasing videos of figure 03. They seem to be wanting to get it ahead of Musk to show what their stuff can do.
[40:39] Regardless of what actually happens, humanoids are coming. Now, in terms of longevity, something else to put into this since you're hearing how much people are uh we're not having kids, uh we're now entering the point where escape velocity math gaining two healthy years buys times for the next wave of discoveries. That adds another two on top of it. Um we're at the point of singularity, guys. I don't know how to say it. Singularity as a step function. Um there's going to be a surge of a 100 times to 10,000
[41:10] times capability in a few days once recursive ideas and software advances land without new chips. Um he says it's around 2026 27 27. The problem is the physical buildout is going to take up 10 to 20 years. So on the humanoid side we already have everything we need. We just don't have the physical side yet. That's why I keep saying the PMIs have to go higher now because the software is ready. The physical side is not go buy your PMI stuff, go buy your transports,
[41:40] especially after China US margin pressure to favor AI native orgs. And again, this gets into the fact that bigger companies are going to have problems. He shows the ones there. We're up to 90 gawatts by 2029. Uh that's higher than what was expected not that long ago. And again, the singularity, this thing is saying it's here and we start going up. 2025 will will feel pretty boring. Um, I wrote this uh in Substack this week. Um,
[42:10] this is as much for your kids as it is you. But if you want to stay on top of what's happening, do not read Dex. Do not read Mckenzie reports. Do not read any of that stuff. Just go to podcast. These guys are unconstrained. They can talk. They can say whatever they want. uh they may not say anything about the revenues of the company, but right now you can basically infer them from what they're saying about everything else. Crypto. So, we had one of the largest liquidation events in history. Uh that
[42:41] was post um the Trump stuff last week, but Bitcoin continued to trade lower uh and basically is back down to where it was at the end of 2024. It's basically unchanged since December. uh despite being up whatever uh 12% now this year, something along those lines. Um, forget the Bitcoin four-year cycle. I have been completely shocked at how many times this has been brought up to me. Um, and people telling me it's real. And I guess
[43:11] the reason they're telling me it's real is because like when you have two data points, the end of 17, look at that peak. And then we didn't take it out again until 20. And then this peak, and we didn't take it out again until 24. So basically in this period in here, I've been hearing repeatedly that you're underestimating that at the end of October, that's approximately when the having when the four-year cycle will end and will go down. So I just want to say everything to anyone in the crypto world. You're leaving one part out,
[43:41] which is the S&P went down both of those times when and then Bitcoin bottomed when the S&P did in both cases. So, I hate to say this. This is going down while the S&P is at all-time highs. And I believe the S&P is going to be powered higher by massive capex spend next year. Lower inflation than what we all anticipated. Gas at the pump, guys, is now at 305, down from 320. I don't know how you're going to get your inflation prints. You think we're about to cut rates, but most importantly, the
[44:11] benefits from AI are flowing through on a productivity basis, which means profit margins are spreading to the other companies. So, you're going to get the massive effect next year. I fully expect the S&P to have another good year. And that means that Bitcoin is going to go higher. And if I'm right, this is also the year that the singularity does the disruption. That to me is what Bitcoin is all about. I've talked about the Bitcoin miners, which I think are already getting a big thing. Now, that means next year for me is about Elon
[44:41] Musk, Bitcoin, and no AI bubble. Why people think Elon Musk Bitcoin and the AI boom. So the same people that are bearish AI right now are bearish Bitcoin and they're bearish Elon Musk and also Micro Strategy doesn't really matter. They're bearish all of them for the same reason because they think they're scams. So if you know these people and you're reading them, just know they're having fun with this. They love saying this is there. Disruption feels like deception. Hype outruns understanding. Skepticism is
[45:11] safer than belief. I just want to say that at some point if you've said Elon Musk is is a scam, which I did at some point, not thinking that a human being could possibly manage more than one business, Bitcoin. I didn't buy into it in any way, shape, or form as an investment until 2020. And AI, this one I've been on since the day I went to Silicon Valley back in 2013, and I changed my view on Amazon. The majority of people never change their views. They put stuff out in public and they really can't ever get
[45:41] off of it, so they just keep doing it, so ignore it. John Vonoman, um, if you haven't heard of John Vonoyman, he was the person that basically originally talked about the singularity. He was one of the people involved in the Manhattan Project. Um, and he wrote basically this in in in the 1950s. It centered on the ever accelerating progress of technology and changes in the mode of human life which gives the appearance of approaching some essential singularity in the history of the race beyond which human affairs as we know
[46:13] them could not continue. The key insight acceleration itself would lead to unpredictability a future incomprehensible by existing models of thought. Think leis think uh debt and deficits. Think of all the things you want. They're all being impacted by the singularity, which is this. That's it for this week, guys. Have a great week.