Transcription
Uh, this may be the last video of the year, depending on, uh, whether there's something important that happens next week that's impactful for the following year. This will be the last one.
Uh, was a busy week. At the very beginning, I want to say thanks again to, uh, everyone who's been watching, sharing it, uh, subscribing. If you haven't subscribed yet, please do. It helps me out, especially as I'm kind of moving into the next phase, uh, of helping people more this year with specific things in AI and websites and everything along those lines. So, um, let's keep, uh, trying to make money and keep trying to navigate the AI movement.
So, the weekly recap. S&P barely changed for the week. Not a big week in terms of movements. Q's were up about 60 bips and Russell underperformed after a few weeks of outperformance, down one and a quarter percent. Bondvall or rates V continues to decline. We are now back to, uh, where we were before the rate hike cycle began in 2022. Uh, and junk spreads remain at all-time highs.
Uh, the dollar is going to be important for next year. Um, especially as part of kind of the overall theme that I think is happening with reflation. Uh, you have kind of a MACD potential sell signal here. We got a little bounce in the dollar last week. But as I go through this, uh, I really do think AI is going to have an impact on things in the US in particular, more so than globally, on the labor front. Also on the inflation side, I think the adoption here is going to be much, much faster. Uh, but we'll see what happens. Regardless, we've got, uh, MACD's pointed down in the dollar. Uh, and because we have MSCI World X the US, because we have raw industrial commodities going higher, plus we have the dollar weakness, uh, this is my risk grow index, which, uh, has a lot of those factors. Uh, and it is continuing to rise. This is it overlaid with the, uh, US PMIs, the ISM PMI. Uh, so again, I expect PMIs to not only go higher next year, but to see a strong move higher, uh, driven on the back of this move into the physical AI world.
Uh, I wanted to bring this up. Uh, for those of you who don't get 22V research, this was something that Jeff Jacobson published on June 5th. The reason I want to bring it up is because I want to highlight again why, as we go into next year, now that I've been able to do this, uh, this year, there are two advantages to the way that I'm approaching this stuff. One is using it every day. And number two, listening to an enormous amount of podcasts with the people thinking about it. I very seldom meet anyone on the, in the hedge fund world that listens to as many podcasts on this front as I do. Um, and that's just based in the conversations I'm, I'm having. Uh, and again, they don't have the time the way that I do. And the second thing is nobody I know uses AI as much as I do. And the reason that's an advantage is because we are going to continually be in an AI bubble world.
So, I put these two ideas out, or I gave these ideas to Jeff, um, which was really kind of what I'll be doing more of this year on the website. So, this was Cisco and Micron. Cisco and Micron are two names, blah, blah, blah, continue to believe will be big winners from the AI revolution. So, this is on June 5th. Um, as AI shifts from cloud-based model training to real-time inference across physical systems, robots, autonomous devices, and edge computing, the infrastructure layer is emerging as the most investable component of the AI value chain. Cisco and Micron are two critical beneficiaries of the transition.
So, I just want to bring this up as I go through the rest of this, um, video. This kind of stuff only can happen because of the exponential nature of what's going on. The amount of people that disagreed, especially on Micron, but can't see still the connection of Cisco. We are entering a different stage. So, as I've mentioned, these two names benefit from VLMs, which is what we started this year. The reason that Micron has basically almost tripled since this note came out, uh, or at least doubled since it did, but more than tripled it, uh, from where it was at the beginning of the year during the, uh, during the, uh, the fall in April. The reason this is happening is because the dollars that are flowing in are massive, and we're in the VLM stage. So, with VLMs, you have a different scenario. The agentic world is different. The upgrade cycle is different. The robo taxis are different, and the humanoids are different. This is all different than the LLM. So, the world's not ready for this in terms of the investment side. They have not spent the time because they still doubt these things because of the AI bubble fear. So, it is a huge arbitrage opportunity.
What we did in this one is we just highlighted at that point, high bandwidth memory capacity sold out through 2025. I'm going to highlight what Micron said this week about being sold out not only through 2025 now, but into 2026. And I'm telling you, there were funds telling me that experts were coming in their offices saying there would be excess supply next year. So, experts have been wrong because they don't understand the demand side of AI. And that is the critical thing for all of you. If you don't understand the demand side that's coming, the compute needs, which I go through every single week, it will never stop. We will never catch up to demand ever. There will always be a need for, for the computers to get smarter. Just like for us, it's like saying we've learned everything. It's impossible to learn everything. So, compute and the demand will go. We'll have efficiency gains. We'll solve problems. We'll get to the point where you can't make money investing on these things. But for the next three years, and I would say until we get to the point that you're seeing humanoids walk down the street, you're probably gonna have an arbitrage opportunity to fade the AI bubble talk and to look for areas where it's not built in yet. I still think Cisco is one of them. I'm not going to go through all the details because I'm going to save some of this stuff, but to highlight the Micron theme. This is when that article came out. So, the white line here is the price, but this is the earnings. Look at what happened to the earnings forecast for a year out or two years out. So, you had this situation where they were nowhere near this. Now they just came out with earnings and they blew them away, and all the estimates went higher. So, this year $13, next year $38. Look where the PEs are now. This is where they started before this move. Even with that price move, because the earnings are growing in an unprecedented manner. And this is very similar to Nvidia. This is the Cisco chart, basically closing in on the all-time highs. I showed you Corning two weeks ago. Corning fits in in the same way of Cisco. I'm not going to go through all the details now because for people that are more interested, they can reach out to 22V and they can get the details, but Cisco fits in in a major way along with a, a bunch of other names like it as we get into the next stage of AI. That's where the video is.
So, just as a reminder, we will launch this sometime in January. This will be a way for people to get more of the research that I show on here, but get it more specific into names. The way I'm approaching this is again, the thematic discovery, staying on top of the podcast, going through things, coming up with information, taking a technical confirmation, using Grock to look for names that at this point are not just ideas that may come like humanoid, Actuate, all that stuff. I want the technical charts to be moving in the same direction. I want the companies in the earnings commentary be be confirming that this AI theme is already, uh, playing out, and I want estimate revision momentum to have start on a relative basis. So, as long as that's the case, again, this is, uh, Google Trends for AI bubble using the YouTube search, which is what I've been using all year, just to make sure that the bubble talk is high. We're going to be in this for a long time. You will constantly be climbing a wall of worry in the bubbles.
A reminder, last week we got the most dovish FOMC presser. And the reason it was dovish is because basically Pal said, "I'm concerned about the labor market. I'm noticing the impact that AI is going to have on the labor market, and I'm not as worried about inflation." So, we got jobs data and inflation data this week. The jobs data continues to be on the lower side. Now, this is the rolling four-month average, and you can see there's no way to read this any other way. The jobs market is around zero. We're still fine from a GDP basis. This is the aggregate payrolls, which are still back in the range of where they were. And again, this takes into account not just the amount of people getting hired, but the wage component and the hours worked. It's still a stable market. This is the rolling average of that chart for the six, uh, months. Uh, we're right at where we were, kind of the midpoint back in the decade before. So, the jobs market is fine, but you do have the problem showing up. The youth unemployment rate, no doubt about the trend that's going on here. So, you have this K-shaped economy, you have people being disrupted.
We got the CPI numbers, very large surprise on, on the downside. Rather than get into where it is, it lines up perfectly with the 12-month sticky X shelter Atlantic Fed. I've showed this chart many, many times. We're under 3%. Anyone who's worried about inflation, to me, is just, is just going to be again, climbing or sitting there creating issues for themselves. There's no inflation coming. This is not some hyperinflation world. This is not some, we're above the 2% target. This is a policy mistake. I've reiterated this over and over again. You are fighting housing affordability. Housing prices are coming down. You're fighting the jobs market. Wages are coming down. And you're fighting this chart. Gas at the pump at the lowest level since 2021, early in 2021. We're almost back to where we were here. Until gas at the pump goes higher, until oil goes higher, you're fighting something that just isn't there. And if it did go higher and it's just a temporary thing, doesn't matter.
We got more wage data. This is from, uh, Gerard Mc, uh, McDonald, uh, at 22V. And just highlighting again that one of the, the Indeed numbers we got out this week continue to emphasize this. And so, let's see what the market thinks. Well, the market literally has changed dramatically in their expectations of inflation. So, everyone was expecting into here, into September, that we'd be seeing the tariff inflation. That hasn't happened. So, you're not only fighting the data, you're actually fighting the market now if you're still talking about inflation. And just the fact that this floating around that we've got, not all the data in there, that'll keep people on the conspiracy side and leaving it alone.
The back and forth continues. We now have a three-horse race. Hasset was a given. Then Walsh took over during the week, right here. He's come down. Now Waller's got a. We'll see what happens. Regardless, Trump is going to make sure he picks someone that's listening to him. Uh, we did get diffusion data, which was positive with inside the payroll numbers. So, again, the diffusion number inside, uh, the jobs was always something I've used to see whether we're deteriorating or not as a leading indicator. This takes all of the different industries within there and says whether you're increasing jobs or decreasing jobs. The white line here is the, uh, three-month average, which is now back up to 55%. It had been trading lower. This is it overlaid with another diffusion index. This is the ISM PMI breath. Uh, this week when we had a bad day where the S&P was down more than 1%, more than 1%, we had 2% of stocks making 52-week highs, none hitting 52-week lows. I only bring that up because we always hear the other side on it. But you can see the numbers, if you go back to there, very positive. Mike Wilson been positive talking about next year.
I bring this up because he's really talking a lot about, uh, the impact for small caps. Uh, I've talked about small caps in here. I think small caps are going to benefit from PMIs going higher. Mike just had some other things in here, which I think are also critical. Uh, just on the earnings side, and again, I don't think small caps are in some structural thing where all of them are going to go. I think these indices are dealing with a lot of companies that are going to go out of business. But it doesn't take many companies or much on the front of a short squeeze to make the index go higher. And I think that will occur next year.
If you ask me my one big risk for next year, it's this rotation trade. I don't think the fact that over the course of the last really 12 years where we've had this massive move into software, where Nvidia right now is close to $5 trillion while the Russell 2000 is close to $3 trillion in market cap, one name versus 2,000 names. I worry about any rotation because I, this is the most crowded trade. I wrote a paper for 22V on this this week on the current market structure, how momentum is dominated, what's in there. I think we're going to see multiple compression, but I really think there's risk here that if the hyperscalers don't see their ROIC come in this year and there's a rotation, this could lead to some problems in liquidity on these small cap names, and you actually could see some issues for larger hedge funds. That will be my risk for this year, is that the ability to risk manage this situation, which is more structural and not a temporary shock where small caps outperform for a short amount of time because of a squeeze, but actually one because it's driven by fundamentals, I think is a big, big issue. Uh, I believe that's going to happen. It either happens slowly or it happens quickly. I'm going to guess it happens quickly at some point this year. Uh, and this is one of the reasons why you can see over the course of the last three years, while the PMI has been low, it made sense to be in large caps because small caps and midcaps could not have earnings. Well, that's changed. Forecast for next year. You've already started to see the small caps in terms of, uh, the estimates go up. So, we just had this inflection point. If we have any risk in the large caps, and this is why the hyperscaler spending is so important to this, it will be an issue. Uh, another way small midcap stocks are as cheap relative to large caps. So, their starting point is as cheap as they've been. Um, I think people just need to be re, uh, paying attention. Catalyst for rerating all of these types of things. In my opinion, I agree with the number one, though, is that I think three years out, because of AI, it will be very, very challenging to value things as this acceleration goes. And in particular, the benefits from AI are going to start to show up in other companies, in particular the Fortune 500, the non-hyperscalers, and I think that rotation will match in with the rotation that's happening in small midcaps.
Uh, I did something this week on the back, back on the way from Boston, and again, I'll be, uh, releasing videos that people can purchase, uh, on how they can use AI. And in particular, I've been working with a lot of college kids and a lot of parents' kids who I've known in the industry on this particular thing about how you can start incorporating it. I'll get more into this as I get into Bill Gurley's conversation that I heard this week on Tim Ferriss. This is just an image I put together of what I wanted to do. Those of you who have known me for a long time, I'm very big on covariance matrix and also on looking at regime shifts with inside the market by taking a bunch of stocks, looking for clusters of movements that signal a regime shift. That usually shows up in factors, but sometimes it just shows up in names. So, I did this on train with Grock. I basically went in and said, "Construct a portfolio of 20 US stocks from the S&P 500." I want all of them to be basically ones that are at close to or already have just gone to year-over-year price positive from negative. So, Tesla is a name that a year ago today, it was about the same level, it just went above. So, that's a name that just goes positive. John Rog does this type of stuff in his technical analysis. I used to do this when I managed a macro portfolio a long time. I also want things to be above their 20 and 50-day moving average. So, I'm looking for things that have worked recently, very, very well, and have now turned basically positive. So, it gave me a bunch of these names ranked by the last 30-day performance. If you go in then and say, based on the names that you've chose that you've given me and the sectors they're in, highlight 10 potential macro regimes this may be associated with. And it goes through and you can see reflationary environment, uh, value and cyclical rotation, small caps, all of these different components in there. Again, I did them a bunch of ways, potential entry into a new one, uh, late cycle expansion, transportation, logistics, revival. It picks the names. It looks for the common thread between all of them, and then it gives you an idea. All of them have to do basically with some kind of reflation, economic expansion phase. It goes back then, as I ask it, and find the places that are most correlated. So, 10 correlated trends in other assets supporting this regime. So, outside of just those stocks, if I was a macro trader and I wanted to go trade something, what would I do? I would play long small caps. I'd play value over growth rotation. I'd be long the industrial metals. I'm just highlighting the ones right now that fit the theme at this point, and they all line up the same way. Weaker US dollar. I'm not even going through like emerging market gains, steepening yield curve, the the twos 10s made the highest level in the last two years, and narrowing credit spreads. We've gone through this. That's what AI can do quickly. That was without me doing super extensive prompts. That was just me sitting on a train and doing that on my phone. All of that work was done on my phone.
Uh, the big negative going into the year is we do have indications finally after this entire year. Ryan Dietrich pointed out from the fund manager survey from Merrill Lynch that you have low cash levels. You have sentiment on the higher side. You're really not looking for people that think the economy is going to slow down next year. The biggest risk that is mentioned investors have never before been in such agreement about the biggest market risk for the year ahead than they are now. AI tech bubble risk towers over everything else. If you ask me what's the most crowded view, the AI bubble trade is the most crowded view. Go make money on it.
So, Oracle scares everyone this week. Michigan data center in limbo. Blue funding talk stall. The amount of emails and text messages I got. Is this it? What do we do? Oracle CD. Guys, I don't know what to tell you except go listen to Jim Chanos. He did an interview on Monetary Matters. I think Jim Chanos is a well-thought-out person who, when he's not posting on X and trying to make things hyperbolic, he's very rational. He's very thoughtful, and he thinks in bets, and he talks this way. He did this, this entire conversation. So, if you go through and you listen to this, this is the bear case specifically on Oracle. I think it's a good place to go. And he says, of all of the companies, and just so I get through this, he's positive pretty much on all the hyperscalers. This is the one he's negative on, and for legitimate reasons. This is the one that does, already has free cash flow negative, they don't have revenues coming in yet. So, the one thing that you can get angry about Meta, you can do all these things, but they do have lots of revenue coming in that was coming in before AI. So, his argument, which I agree with completely, he says, "If we don't get significant revenues in by 2028, Oracle will have a significant problem." I completely agree. And that's where the risk is. If there's no revenues coming in the door, if the buildout can't happen for some reason, there is significant risk unless they're going to get the money that they need to offset the debt that they've taken. So, there you go. He makes a logical, rational argument. He also says, "We just don't know." If you listen to it, in my opinion, he's saying that we've never been in this situation. We've never seen companies try to do this. This is an enormous bet. If they don't get it, it's going to be a big problem. I completely agree. He does say he's not worried about these companies. What he's worried about is the data center side. He talks about how these businesses are not great. If you want to be long AI, own the hyperscalers or the model companies, not the data center operators. You read X, you get the impression that he thinks this is an AI bubble and the whole thing's going to collapse. It's not the case. He literally ranks out the ones that he likes. Major concern, only real bearish call. He even says Nvidia is a good business and not where he's focused for short ideas. They're much better places for skeptics than that. I completely agree. I think Nvidia is very, very cheap relative to the next three years of growth that's likely to happen. Micron came out and basically, I don't, you know, his, his interview was done before Micron, but these numbers are insane in terms of what's come through. So, the demand is there, whether or not they can actually get the revenues, which I'll go through as well, in terms of where you are going to see them next year and where people need to start focusing their attention in terms of making money. If you're thinking about the part that Jim Chanos is talking about, you're literally just doing an anxiety fest over what may happen. Everything that we do in our, in our job, investing, everything is forecasting the future, which means it's all probabilistic. The question is, what do you know about the demand side? Jim Chanos knows very little, if anything, about the demand side. This is the demand side from the companies that want to build these out. So, this is part of the data center trade, but this is the VLM side. So, we need to actually see the revenues from the model companies. There's no way to refute what's going on. If you're going to fade these companies, remember the United States Department of Energy announced a bunch of companies that are part of Genesis Mission. This is a government thing. So, even though these companies may not quote unquote be backed up, you really want to fade something that's important to the government in terms of competition with China, where they specifically said that national mission to accelerate science through artificial intelligence. These are the companies on the list. These are all companies that people worry about right now as part of the OpenAI trade. So, he brings up the.com bubble. And for, I don't know if I've done this on the videos, but I want to make sure there is almost no comparison for me in terms of going through the.com bubble when you get into the revenue side of the equation. So, I just want to go through, you can read all of these on your own, but I think the two most critical parts and differences. Who took the debt? The telecom companies for bandwidth, not the people that were actually going to be the companies using it. Who created the demand for the bandwidth? The VCF funded startups, the think of them as the future Amazons, the future Apples for the, uh, apps business. Everything that's an app, everything that's, who had demand. It took a while to get this. We didn't get mobile until 2007. That's when we used up all the debt that they built. In this case, cash flow rich companies that are massive in size. They're the ones doing the buildout for the intelligence, which they're using. They also get revenues from the people using the intelligence as they go out and sell it to people. This is a completely different situation. And right now, demand exceeds supply at every level, as I go through this. So, there's nothing comparable.
Now, the key utilization catalysts. So, let's use a timeline. It wasn't until mobile apps that we used all of that, we were able to offset the debt with revenues. AI agents and digital employees are the revenue point. That's what happens next year. That is the most important component, and that's what I'm going to be spending my time with next year, making individual investors and hedge funds and mutual funds and pension funds and anyone who wants to make money on it, trying to give them high probabilistic places where they can benefit from the transition from the cloud to the agents in the enterprise, from the cloud into the upgrade cycle on edge devices, from the cloud into robo taxis, humanoids, and everything else. There are so many places for this transition. You do not have to focus anymore on overvalued stocks with inside the data centers where everyone is crowded into the trade, and they know it. There are places that are not crowded yet.
Sam Altman on the first interview I've listened to since the Brad Gersner debacle. Uh, this one was with Alex Canitz, who does a nice job. And again, the most important points on this. We are at the enterprise inflection, and it has arrived. So, remember, he just declared a code red. He goes through all of that stuff. Talks about how the frontier models will not commoditize. I don't care about any of this stuff. What I care the most about is companies now request a single AI platform. Again, I can't say this. Unless you listen to all of these podcasts, and you're listening to Jim Chanos, he doesn't know what this is. Most of the people who are bearish. Michael Bur is not spending time going through and understanding what this is. And if he is, he's not talking about it. Organizations still behave like GPT4 error users. Could not agree more. All the things that I'm doing in it, these are not allowed for the most part in the workplace, and people are not using them this way. Humans are not doing it. So, you need to have digital employees. GPT, uh, GPT 5.2, which is so much better than what 5.11 was, which came over the summertime. This just came out last week. I'm already back to using ChatGPT as much as I'm using the other one. 70 to 74% of knowledge work tasks at expert level. We're getting to the point next year where all of the benchmarking will say that 95% of knowledge worker jobs can be replaced. We're getting to that point. Next interface shift, agents. Remember, we couldn't get the agents at full scale because we're on a black whale delay. The new AI first device is coming. So, again, we're going to be getting the upgrades to the phones, to the computers, to the cars, every car, not just the Tesla robo taxis, every single car. There will be a massive difference between having AI in your car, the ability of working using AI as you're driving, and also having assisted driving. If you don't have those in your car, the technology is going on all of them. Regardless of what it is, it will give you the chance to be doing work at the same time and accelerating every single year. His biggest excitement is in AI-driven scientific and medical breakthroughs. AI drug discovery, pharmaceuticals going to be a massive part of this. You're going to see that revenue tracks compute. We need compute. OpenAI has never had excess compute capacity. You're going to hear this repeatedly as I go through many podcasts from this week.
So, again, just a reminder, apps versus AI agents. If you haven't spent your time on AI agents, you're missing the app store. This all starts next year. Early adoption began this year. The workflows start shifting. This is where everything changes. This is where you get into the point. Apps were built for a world where computers couldn't think. Agents are built for a world where they can. As enterprises adopt an operating system, the entire software and cost stack rewires, and revenue flows away from SAS towards AI platforms, compute, and agent orchestration. I could not agree more. This is not a question. This is a given. And I'll go through the reasons why as I go through all of these other guys. But the agent age has arrived. Prepare to make money off AI agents. The seven essential skills that you need to keep your job in this world. In the new world, this competition is no longer human versus human. It's human with agents versus human without agents. This is why I'm trying to help people, and especially college kids, become AI native. If you don't use it every day for at least a few hours a day, you are falling behind at the speed that everything is going. If you doubt it, don't use it. Don't use it at home. For every single decision you make, I use it all day long. I was telling someone as I'm launching an HRV Substack and read some of the stuff, and they said, "Where did you get most of this information?" I said, "The early days before 2023, it was all through podcasts. Now everything is something I read a brief thing or I hear something on a podcast about AI, about where we are with drug discovery, where we are with anti-aging, where we are with mitochondria, and then I immediately take that information. I go into ChatGPT and I get all my information there. I learn it. It is more efficient. It is time. You have to be doing this every day."
Maker, remember, is training the, uh, they're training to make specialized language models or at least to train, use, uh, give the labs training data, which allows them to then sell these from OpenAI, Google, Microsoft, Meta, the clients of Maker, directly to a Morgan Stanley, a Goldman Sachs on that side, to a drug company. They are basically having the lawyers, the people in the medical professional, the investment bankers, the traders, research people, they are paying them lots of money, hundreds of thousands of dollars, to make new data being an expert. So, they're taking the domain experience. So, if you're someone who's a young kid, the domain experience is actually now going to be commoditized. What they're going to need is AI native individuals who can do this and think and ask the right questions. Being great at prompt engineering means learning how to say why, learning to ask new questions, learning to be good at it. That's what I'm going to try to get people to do. What I just want an, uh, an AI platform means for Morgan Stanley. So, I just wanted to give you guys an idea about what's necessary. So, Morgan Stanley, since I worked there, are shifting away from scattered AI pilots to adopting a single AI operating system that unifies models, data, workflows, agents, and governance. To do this, it must invest heavily in cloud compute, vector databases, API gateways. It needs new network hardware. They're going to spend lots of money paying the model companies. This is what the next three years of the AI infrastructure spend intensity as we get more into AI agents. As we get into stable coins, AI agents transacting in stable coins. Right now, GPU compute is the highest spend intensity. That's why Micron, that's why Nvidia. These are a little bit less. So, where do you want to invest next year? A lot more in terms of these areas, and then eventually all in these on-premise. I'll be expanding and giving names on these. So, these are the names that worked the last five years. There isn't a single name on here that everyone hasn't talked to me about this year, uh, in some way, shape, or fashion as a potential name or one there. Bloom Energy was a name no one cared about. This is basically just taking, uh, the Stargate and all the partners that are being used on it. I want to remind people, PMIs. So, we're down here with the amount of build-up that's necessary to go from the cloud, which is what this period was in here, was all the cloud, and you still got these episodic, uh, inventory rebuilds. We are now going into hardware. We are going to have AI inside every company. The models themselves will be sold by the big companies. So, they're developing the intelligence. You're going to get them in your company, but you're going to want security. It's not going to be over the cloud. It's going to be within your walls. You're going to need servers. You're going to go back to a lot of the old school things that were necessary that were part of the overbuild going into the.com bubble. But we got up to 60 on the PMI before the dot bubble, and the new orders component got in well into the 60s. You're going to see PMIs go higher with what's happening.
Bill Gurley, one of my favorites, because he went to the Santa Fe Institute. I talk a lot about approaching things from a system, systems thinker perspective. It has served me extremely well in this day and age as a macro person who tries to look and break things down into the verticals. Bill Gurley had a great interview on Tim Ferriss. I just want to give you the critical points. He was asked if it's a bubble. Asking if it's a bubble or not is the wrong question. Every, and I agree, technological revolution has real innovation and speculative excess at the same time. Speculation is not a sign something is fake. It's a sign something is transformational. It is not faking binary, guys, and that's the whole point. It's a probabilistic thing. Annie Duke, thinking in probabilities, and then putting on there. Avoid the areas. Jim Chanos does a good job of picking on the data centers that they're not going to have the same kind of growth as Micron. That's proven to be true. So, I think you can go focus on it. He doesn't want to be short Nvidia, but he's not saying Nvidia is going to double anymore. I think it's going to be really hard, which Bill Gurley also brings up. We are no longer at the early stage where you can get 10 times your money. But at the same time, he says institutional investors have zero interest in non-AI deals. Zero. That's how you know there's too many dollars flowing into AI companies, especially startups. I do not agree with that. I think that's going to become an issue this year as the agents roll out. Who must use AI? I don't care what field you're in, you should be playing with this stuff. The best way to protect your career from AI is to be the most AI-enabled version of yourself. He's agreeing with things that I've said repeatedly, and why I have a website coming out to help the people who want to get help. Remember, if you get on there and you learn how to use it and you use the simple things that I do to do these videos and to create the things that I create, which I'll be showing you, you will be ahead of 99% of the people not coding. So, the people who at this point have been using AI to the degree I am, they're for they're mainly technology people. There are people in the industry that use it, but if you have a full-time job, you don't have that, uh, ability. For kids in college, they should be using it every single day. Unsafe careers. This gets back into the Daniel Pink thing I've talked about. Many of the pragmatic jobs that well-intentioned parents push their kids towards, think lawyers, think doctors, think investment bankers, are at risk. The new advantage isn't financial security, it's learning velocity. I have learned so much in the last year doing these videos, doing all of my podcasts with Anthony Pompliano, doing my substack, adding another substack on the HRV, all of this stuff. I can learn topics very quickly. Things I didn't know, I learn all the time. AI lets kids learn anything faster. And parents should focus on teaching them how to use these tools, not steering them into safe professions. There is no such thing anymore. You need to be an entrepreneur and you need to have an entrepreneurial mindset. I've talked about how Dennis Sabis is one of my favorite people. He is a gaming kid who is thinks like a scientist. So, when he approaches AI, he is not hyperbolic. He is not missing timelines. If anything, he's very conservative, but he is a joy to listen to because he tells you what's going on. And what he talked about is multimodality is one of the core frontiers of AR progress. It's the biggest thing coming over the next 12 months. This convergence of modalities. This is VLM. This is what I've been writing about. If all of a sudden, if you haven't spent the time, this is going to get into everything. This is robo taxis. This is humanoids. This is AI edge devices. This is everything, and we're going to progress a lot. So, the compounding is going to go faster. The multimodalities involve vision language, VLM, and then audio, video, all of this stuff together. Video understanding is the next major leap. I completely agree. This is foundational for autonomous agents, world models, robotics, understanding environments, everything at the nano level. This is where the great, great changes in the world come. He specifically says something which I'll reiterate, uh, later with Mustafa Suleyman, but I want to get through this. Multimodel architectures expand compute needs. So, the same way that Sam Altman said, we've never been at a point where we've had enough compute. This is reiterated by Demisabis. It takes tons of models under the hood. The compute needs are basically endless. Custom chips are mandatory for front, uh, recursive self-improvement is the compute bottleneck, and again, recursive self-improvement where models are learning on their own. We don't have enough data for that. The pipeline requires massive and continuous compute throughput. Closing that loop will speed up development for sure. We're not there yet. The industrialization of compute is a structural shift, not a cyclical one. AI for science will require new physical compute. So, how does it directly go into VLMs? The convergence of modalities, the text and vision, Genie3, I'm not going to go through this, but you can start to understand the importance of this. All of this stuff is made for what is going to happen with humanoids and with robo taxis in particular. Compute is never enough. Even at Google, scaling is mandatory path to AGI. If you don't understand this, we need more compute to get to AGI. It's not optional. This is not a bubble. We are not overdoing it. We are not not going to use it. And we are going to get revenues on it. And it's going to accelerate in a way. And remember, Anthropic has gone 10 times and 10 times to currently get to $10 billion. If they do another 10 times, which Dar Mod is not forecasting, that would be a H100 hundred billion. Next year is the year that people realize the revenue is coming in. Even if AGI requires new ideas, compute scaling remains the backbone. Multimodality, exponential compute requirements. I'm not going to keep going on this. Uh, humans must prepare as a species for AGI. So, the one thing that moonshots, Demisabis, Mustafa Sali, they all talk about this, that we are not ready. The species. This word gets used repeatedly. Us as a species. This is the potential for an existential event, not from the risk, but from the societal breakdown of what's happening in the job market. We're seeing the voting happen. This is all directly related to AI. And you can sit there and call it a bubble, but that is a world of denial. And that is a world of sitting there. You can hurt yourself, but don't put the same drug in your kids. Give them a chance to at least be embracing this and not say it's a bubble, not say you can't use this because you won't. Schools are not allowing it. Schools and government are preventing people from actually learning at rapid pace. You should be do using this for everything. In the substack that I'll do for HRV, I will show you how you can do this by giving you prompts at the end of every one of the posts. Think of this as the way books are going to be in the future. I've often said books are a waste of time. The reason they're a waste of time is because they take a long time to get the information. And a lot of times, particularly in books by Malcolm Gladwell, it's the same 10,000 hours thing for 300 pages. When you do a Substack and you do one every week, you give them a prompt. You can go into AI and have a conversation about it. So, you learn exactly the same thing that I learned, but in much greater detail.
Mustafa Suleyman from Microsoft also spoke with Moonshots. Compute requirements will be unbounded for true recursive self-improvement. Again, recursive self-improvement is beginning right now. AI agents are part of that. The frontier race is now a capital and compute race. This is from Microsoft. Over the next decade will require hundreds of billions of dollars by them. This is by Microsoft. Massive compute footprints, dedicated chip programs. It's going to take hundreds of billions to keep up on the frontier. They don't have a choice. The cost of inference continues to collapse, and that's where the ROIC is going to come. ROIC is going to come because it's allowing them to bring this stuff out faster. He believes Frontier AI next, oh, this is just emphasizing the point in one line. He talks about the fact that the first step to come is the deflation side. Labor deflation is coming, and it's coming fast. Why labor deflation hits before the disinflation comes. So, even though the deflation is coming first, you're going to get downward, uh, downward pressure on wages. This is the inverse of the industrial revolution. Automation hit physical labor first. Here, it's the middle class professional labor first. So, the educated, meaning college educated kids, and this is what you're seeing. They're the ones getting hit hard. The knowledge workers will be hit hard in this. They need to be entrepreneurs. They need to make sure that they're not sitting in a seat where they're not learning. It's not just the government. It's not just the schools. It's not just the parents. It's also the organizations that they work for that are not allowing them to use the tool. You will learn, lose the best talented entrepreneurs. You have to get them to do this. And you can do it after work. You can do it before work. You just have to be using it all day long. The world is unprepared for the speed and scale of the transition. Short-term human response will be unstable. I completely agree. Societal impact is going to be massive. The governments are not ready. The public institutions are not ready. And I asked, is this like the asteroid that wiped out the dinosaurs? Yes, it's exactly the analogy. His worldview supports, but we're both the dinosaurs and the asteroid, meaning we're the ones creating this. And this is why we have the ability of using it and going through it. But to ignore it and pretend like it's not going to happen, it's just not right, uh, for anyone. And that's the point of kind of going through this. If you say it's a bubble, it means you don't have to worry about it because it's not going to happen for a long time. Why this mismatch matters for AI agents? This is cognitive labor deflation. You now have AI PMs, AI coders, AI strategists, AI sales analysts, AI founders, AI assistants, AI researchers. This is for every job, guys. Um, and again, the next two to seven years will be politically and socially volatile. And that two to seven years is before the, the abundance comes, until the deflation comes to make people be benefiting in some point to get to the Elon Musk part of option we'll be working on. So, McKinsey plots thousands of job cuts and slowdown for consulting industry. They're seeing the impact. An engineer, the best engineers, all models so far were okayish at best, but now with Opus 4.5, really is something else. People who haven't tried it. I don't know what's coming for us in the next three years. This is from a software engineer. Basically, the coding people now realize there a lot of people have been fighting it, and now it's not fighting anymore.
All right, let's get into the final part. Tesla made new all-time highs this week. I'm going to keep pressing the button on Tesla. It is the number one theme for me as a MAG7 stock for next year, but more importantly, for the next five years, I have no idea how high this stock can go. And when you combine it with SpaceX, I have no idea. This is a thread you should pay attention to.
I'm not going to read you all of this. Whimo never really had a chance against Tesla. This will be obvious. Whimo has reported over 100 million fully autonomous miles driven while Tesla has 6.8 billion. Whimo has a fleet of 2500 cars. Tesla has 5.1 million cars. This means Tesla has an enormous advantage. I'm only going through these so that you guys can go do your work.
Whimo relies on LAR. Elon Musk knows LAR well. He uses it at SpaceX. He rejected them at Tesla for good reasons. Whenever I hear people say this isn't going to work, I am I am just shocked at how many people not only doubt Alain, but it already works. Oh, people get mad that I don't call him Elon. So, I'll call him Elon for now as opposed to Elon. Uh, it's from my buddy Alain who I worked with who worked for me in Brazil. Uh, Tesla's approach is vision only. End neural nets. We've talked about that, but let's talk about the cost per mile wins on cost. It's the amount of drives driver uh uh miles per minute. Then the cost of the car. Then I want to get into the SpaceX component because since he chose not to do lidar, let's go get why he didn't use lidar in cars, but why he's using it in uh in SpaceX.
Expanding on LAR, it's not good for road vehicles. Calling a crutch or a fool's erum that's doomed for anyone relying on it. It's been design the roads have been designed for biological age. More importantly, it struggles in adverse weather like rain, snow, blah blah blah blah blah. In contrast, for SpaceX, it's used SpaceX is a vacuum with no atmosphere. So, no weather, no related scatter or interference. It excels here for precise. Again, Alon Musk is the person you want to believe on anything engineering. I don't understand why people have it. I Google has a lot of smart people, but if they were really that smart, they'd be doing what Alon Elon Musk is doing.
Um, there was a a u a patent that basically got published today from Tesla which I think is really interesting and it's about developing aluminum alloy compositions and manufacturing message that allow for recycled scrap metal that has an impact on basically one of the most expensive and energyintensive individuals. This patent enables te Tesla to these are not incremental savings. It compounds across millions of units and larger casting. Why this matters for why this matters for embodied AI. The only reason I wanted to go through this rather than read it to you when you have to compare or compete with Elon Elon Musk on the scaling issue and how much he's thought about this, how many cars he can pump out. He has thought about everything thinking forward as a systems thinker taking all of the components so that when it gets to this point he's already there on the space data centers.
I heard five or six different hedge funds over the the two weeks since this came out saying this is why you can't trust Elon Musk for this even though Google has said the same thing. And this was Elon Elon Musk's reply. SpaceX has over 9,000 satellites orbiting Earth right now, which is twice as many as the rest of the world combined. We may know a thing or two about the subject. Other people have told me it's not going to work, and they gave me reasons, and I just honestly kind of laughed.
Uh, all right. Bitcoin just been brutal since October. It was actually brutal in here, too, because the market was going higher. So, this pain, this is kind of the the point where people got negative. I still have people going. We're still making lower lows. I've said before until we close, which I believe we will at some point soon, whether it's the beginning of next year, even the second quarter. Once we close above 92,000, three days in a row, I think we'll start heading higher again.
Um, this thing has been held up. I believe this has been held up by a lot of the growth versus value rotation that's happened in the in the US. This has happened for a lot of different reasons. Palanteer, Oracle, the speculative names, the fact that we've had tremendous OG selling, but next year is tokenization year. Next new year is stable coin years. Bitcoin continues to have all the news stories that I highlight compounding and compounding. I wrote a substack again this week just highlighting that one of the benefits it has is it's not part of this. I would draw a trend line down here that comes in. We're getting closer and closer, but we're not there yet. And as Jeff Degraphth said, artistic license and a PhD in denial needed to call this an uptrend. I completely agree. I don't like buying things when the trend is negative uh in any kind of size. In this case, even if it were to go down here, I'm just going to keep buying more because I think the eventuality is that it's going higher. Think Micron earlier in the year, think Nvidia earlier in the year. I have the exact same view. This is an AI trade. I view it as being the number one winner when we can't find any company that has a moat on the technology side. I believe Bitcoin has one. At least there's a positive sign from Jeff Degraphth in terms of the excessive outflow has historically been a buy signal that has carried an 85% historical rate.
Um, thanks again guys. Subscribe. Uh, stick with me. Have a great holiday season. I'll be coming up and support the products that I'll be launching if you're interested in learning more. Um, I'm really trying to help people make money and I like all the suggestions and the help and I'll try to get better technology and better microphones. But, uh, everyone have a great holiday season. Spend it with your family. Uh, remember some people you haven't thought about in a long time, particularly people that have passed away. Uh, it's been great to be with you all year and I'll see you in the new