Oracle Crashes 11%: The Deadly ROIC Gap Could Kill the AI Hyperscalers in 2026

[00:00] to do it. Go through the recap. More signs of reflation being built into the market. Factors still shifting and still lining up with a PMI rise to come. Fed dovish dovish meeting or at least more dovish than people expected, but I'm going to go through the guts of it and just highlight why I think this is important going forward when you put it all into context. More bubble talk. Bad week for Oracle. I'll go through some of the details on that. Great podcast with Gavin Baker. Lights from it, but also a
[00:33] lot of the rest of the uh the video is going to be related to that. So, the interview is worth listening to and then I'll go through the crypto at the end. Again, more signs of adoption and still consolidating at lower levels for the crypto space, but again, looks great to me going forward. S&P down 60 bips for the week. Q's underperformed on the back of Oracle and AI more AI fears down about 2%.
[01:04] Again, the outperformer small caps third week in a row being up up about 1 and 1/4%. IWM versus Q. Best week since August. IWM new all-time highs. The value side of small cap new all-time highs. Mid caps new all-time highs. The micro caps new all-time highs. Transports PMI sensitive group up to all-time highs
[01:36] continuing on the reflation theme. I'm going to just go through some proxies here that make this about global. You've got people again focused on long-term yields going higher. I think people just need to start to accept that the government spending and all the things going on are definitely leading to more of a growth trade for next year. Highly sensitive Aussie 10-year rates had been in a very very tight range all year and now since the middle of October seen a big move higher. The CRB raw industrials market a surge
[02:09] to new highs for the year up year-over-year. PMI sensitive Dr. Copper. There's copper over the PMIs. Copper breaking up to the highs last 5 years. uh Beta over quality, which is typically something that happens again when PMIs go up. Surge back up after having a fall in November. And you can see how the rise in beta versus quality lined exactly with the
[02:39] CRB raw industrials market uh going higher in in this breakout. As I mentioned, factors are a great proxy. I showed this last week. If you didn't see it, go watch the video, but when size starts to roll over, that's usually a time when PMI new orders are going higher. We had size come down sharply this week. Growth over value, same thing. I'm expecting there to be a growth versus value shift. Growth has been a dominant factor. The
[03:10] reason is because of the necessity of shifting from the software side to the physical AI side. I'll go through more of that. I'll keep talking about it as we go into embodied AI. The Gavin Baker connected back to this. So, last week's video flows well into this. This whole VLM to VLA vision language models, which is the new form of compute that we need depending on Blackwells and then moving into vision language action, which is really the connection into the machines.
[03:42] So, I expect growth versus value to continue to go through its compression time. Revisions again, another PI sent PMI sensitive. The thing I wanted to highlight this week. We not only on a global basis went higher, but Merrill Lynch highlights that or Bank of America highlights that 12 of the 16 global sectors are seeing this. So, again this is a PMI thing diffusion. It's not just a few names. It's not just a few markets. It is everything around the globe. Cantrow, I'm going to be doing his podcast next week
[04:15] basically showing the same thing. It's good that when the two of us line up, but he's just basically showing that transports reflect positive macro surprises. And again, I'll keep saying it when the transports are leading at a time when the data is not great and especially the PMIs that is a good sign. They're leaders. On the inflation side again, the Fed came out. You've still got people saying policy mistake. I literally just want to make sure rather than people sit here
[04:45] and just talk about things which are not fact. Here is gas at the pump falling sharply. We are now down 16 cents since the end of November. Very short move lower. At the same time, we still have housing market in a downtrend. This is owner's equivalent rent almost back to where it was before COVID in 2019. That continues to go lower. And the employment data. And this is basically the ECI, which came out this week, which is the white line. So, we
[05:16] continue to see employment cost come down. And then this is the quits rate, which typically leads. And the quits rate came out in the JOLTS data for this week. So, nobody's quitting their jobs because they're worried about the ability of getting a job. It shows up on everything. The The uncertainty is there. Wages are coming down. So, you have gas coming down. You have wages in the employment side on the weaker side and you have housing. So, it is possible that we'll see inflation go higher, but I also believe
[05:46] you're trying to catch a falling knife when these three things are moving lower. Small business jobs again, reminder from last week. This is how bad the situation is in the K-shaped economy. Small businesses continue to suffer. So the Fed on that side is airing in a way that I think makes complete sense. If you don't believe the data, then just use what the market is saying on expectations. Here are the inflation swaps for 2-year, 5-year, and 10-year. They continue to come down in the case of 2-year
[06:17] we are back down through the lows of this year. So, again, on all metrics Fed cuts rates with three dissents. Not really sure where the three dissents are coming from other than again just ignoring this or being political and focusing on the potential that inflation will go higher or focusing on the fact that we are above the 2% 2% target they have, but at this point they've kind of thrown that out. Now we're starting purchases. So, 40 billion going into to deal with tax season.
[06:49] But regardless Fed balance sheet is now getting active again. It's a change. Trump basically was asked is the new chair do you expect him to lower rates immediately? Yes. >> [sighs] >> So, the press conference. First point, if you guys don't do this yet in AI, you can take Powell's press conference, download the transcript, and then just go get a summary from there and that way you don't have to take the bias of economists who might have already had an opinion and they might lean a certain way. Just go into there and pick out the
[07:20] key points. Powell's view of the labor market is weaker than the headline numbers. This is one of the critical components came out at the 37-minute mark of that. The payroll data is overstated by 60,000 jobs per month, which would mean if that were the case that we're already clearly in negative category. The real number may be minus 20. Labor supply has fallen. He's talking about that with obviously the immigration situation and labor participation rates softening. Job creation is extremely low though and again, some of this gets into
[07:53] the AI side and what we just saw. AI layoffs are yet they're not yet macro significant, but they are visible. I think the important thing on this and I didn't want to include another slide in it cuz I did it a couple weeks a month ago. If you take Powell's statements and his press conference views over the course since Jackson Hole, you will see that before Jackson Hole did not even mention AI in the labor situation. This has been a growing theme. They have gotten more data. They have openly said they're starting to pay more attention to the ADP data because
[08:23] they're realizing how important it is to be real time and how real time are nonfarm payrolls when they get revised so much. So, the focus for him at least has been more on paying attention to what's happening in the labor market and specifically the way that AI is going to impact the economy. This is a critical point where again, when I talk to macro people, the majority of them know nothing or very little about AI other than their belief that based on history, this is a bubble and it will end badly and they don't need to spend time with it. The reality is now the Fed chair is openly talking about the impact it will
[08:53] have on disinflation, service inflation, everything. Service inflation is cooling. Goods inflation is tariff driven. So, he's minimizing the the inflation side. He is saying AI is one of the main drivers of stronger GDP. I would argue that at this point it is the single only driver of GDP. I can argue with people on that, but if the stock market was down, I think we'd have consumer spending slowing down significantly and we'd be much closer to a recession. Take out the AI CapEx spending and you'd have a problem. The profit margins from AI are
[09:23] contributing to the wealth effect and that's why I think this is a much bigger number, but regardless of that, he is now acknowledging that it is happening. AI likely contributing to higher productivity, but it's still early. I never thought I would see a time when we had 5 or 6 years of 2% productivity growth. This is definitely higher. So, he's acknowledging the fact that it is a real productivity booster. It's not hype. Early stage adoption already visible. A potential structural shift in output per worker. We've already seen that. Startup businesses, their output
[09:53] per worker measured by revenue per employee is through the roof. Much higher than public companies. You're going to see this trend continue, which is why the public companies are trying to focus on raising revenue per employee, hence profit margins. And the way that they have to do that is to replace either future hires as the revenue grows, not hire people, or replace with people. As AI agents come, you will see more replacements. So far, it's just been through the lack of hiring. We will get into the replacement side, particularly for businesses that are in trouble.
[10:23] Most of us have FOMC presser since 2021, according to Bloomberg, from the guys at Forward Guidance on Now, if you want to take a contrarian side, here you go. So, I don't think AI is a bubble. I do think there are pockets in there that are bubble-ish. Uh I've talked about that. I think there will be plenty of losses for investors on companies that will never get the money that the speculation is on. But, this is a sign of possibly the Mag 7 and
[10:54] the hyperscalers being maybe at a relative top. So, when I talk about growth versus value, I do think you have to pay attention to these kinds of signs, and I do think there is a legitimate worry here that the capex spenders, which have consistently not had to spend and have been able to do buybacks, and all of this cash that people are minimizing, not only the impact it'll have on their their those companies in terms of their ability to have multiple stay at the levels they are, but I think the entire ecosystem of software and things built on code will be impacted. So, at this point, I just
[11:26] want you to think balance sheets, balance sheets, balance sheets. Their balance sheets are are worsening. We know that. You've seen it with Oracle. I'm going to go through some of that. Russell Napier, who I've talked about, I love, one of my favorite historians, but I think it's hilarious when understanding the AI opportunity with Russell Napier. I had dinner with Russell recently at an event. And I think he's the first person to admit he's not an AI expert, but he's forced to talk about it. And so, on these podcasts, the way that they're talking about it again is capital. So, they go back to
[11:58] all of these times, and this is basically, again, the same time. And I think it is a huge mistake to compare what is happening in AI to all these places from the point of it ending in something badly. You can have multiple ways that a bubble can not happen the way that it has in the past. And one of the most important on this is not to have contagion, to have governments supporting it, and for it to be a military {slash} need globally to go on, and to have it just kind of grow with various places. So, as an investor, I
[12:30] think your focus should not be on the bubble side. I think it's going to cost you money every time you get focused, but I do think you need to be on the places where there won't be bottlenecks. And as I go through, one of the big bottlenecks right now is on ROIC. That is why another reason why I think we're going to have a hard time for the hyperscalers. Doesn't mean they're all going to blow up, but if one of them blew up and someone bought their assets, did the bubble burst if the S&P trades higher during that? Obviously not. The key difference in past examples, infrastructure got ahead of demand. So, this is the one of the main points that
[13:01] I continually like to focus on. There is insatiable demand for AI. With AI, demand may be infinite or self-generating. It is absolutely at this stage infinite. Until we can get through the physical constraints, which we're nowhere near close to, uh the bigger problem is the power side and the ability of getting the chips that are necessary for everything to be in those data centers. So, we're still at a situation where demand is ahead of supply. Once we get to the point where supply is ahead of demand,
[13:33] we can talk about comparing it to these places. Once we have massive job creation, which all of these did, we can talk about there being a bubble where everyone benefits. That is the way I view a bubble. I remember the dot-com bubble vividly. I remember my high school friends from a blue-collar town talking about their investments, and I remember the people at Morgan Stanley going to the finest schools and being wealthy talking about it. I don't hear anyone talking about their investments in AI other than a few retail traders who get in and out regularly. I don't see this happening otherwise. Howard
[14:04] Marks had a good article titled Is it a bubble? Uh I think it was balanced. He admits it's a world-changing thing. He talks about speculative excess. I completely agree. Uncertainty dominates. This was covered last week by me with Dario Amodei. I completely agree. We have no idea when the revenues are coming in. When you're spending tons of money, you're depending on the revenues coming in a certain amount of time. If you overspend and bottlenecks come up and you've put all this money out or made promises, and the revenues don't come in, you're
[14:34] going to be in trouble unless you can go restructure things. So, I do think that's it. Progress and losses will likely coincide. This is the part that I believe in. We will continue to have progress, and we will have losers along the way. That makes for a great long-short environment, and this is why you have to be moving your portfolio more and more. Take advantage when the panic sets in. One other thing, again, broad participation. Many companies, millions of investors benefit in bubbles, not a few hyperscalers. As once we get to the point that everyone is
[15:04] benefiting, and everyone is bragging, and most importantly, here is the all-time high in Michigan Consumer Confidence. I've shown this before, but for those of you who've joined over the course of the last few months, the peak of the dot-com bubble occurred with the highest Michigan consumer sentiment on record. This is 45 years of data. Here's where we are now, sitting near the lows. Last week was near the lows. If you take the two-month average, there's only one point that was
[15:34] less here. If we stay down here for another one, this will be the all-time worst consumer confidence. How can you have a bubble when people don't buy into it? Oracle plummets 11% on the week. This is where we get into Oracle. Their earnings beat expectations, but revenue came up short. Questions have Investors have questions on whether their investments are justified. Oracle And then on Friday, we had this story come out, which was later re- re- said it wasn't true, but who cares about
[16:05] that? Um said they were on track for everything. But for the week, regardless, remember how this was such a huge Sorry, this is the quarter. So, we had a quarterly move. These were the prior two quarters in Oracle. We just gave up so far this quarter 32%, which was the daily move we saw in Oracle back in September on the rise. So, Oracle's been hit hard. You've got the balance sheet issues. So, the CDS is up, you know, up
[16:35] at these levels. This is not some critical place yet, but uh the balance sheet issues, you've got people worried about it. I'm not going to show junk spreads, but those are still at all-time tights. So, don't worry about it from a systemic basis. Just use it as a gauge that if people want to bet on the entire AI situation, if you're long a bunch of AI names, and you want to have a hedge against this thing blowing up, Oracle CDS is not a bad way to at least have some in your hedge basket. Gavin Baker article uh or interview with Patrick O'Shaughnessy, phenomenal. As I said, I
[17:07] would spend time listening to it, and I'm going to go through all of the different ways that I've connected it now over the the better part of this video. So, first of all, I'm just going to go through the things he talked about. Um He emphasizes point, which I completely agree with. To judge progress, you have to use the top-tier paid models. So, I pay for all of the highest models on all of the LLMs. This morning, I was building a Bitcoin signal trading model using Grok and Claude. I used Gemini for
[17:39] a lot of the stuff that you'll see in the video, and I still use ChatGPT. I use Perplexity still as a main thing for finance and for the earnings commentary. But I pay for the most expensive model on all of them. You cannot have a view on what's going on with these models without paying for the best ones. Uh you can say that they're improving, you can go through it, but his thing was you have to listen to the leaders of the labs. It's essential signal. Um reasoning saved us. This was a really important point, which, again, if you've
[18:12] if you've read my work the last month on visual language models and the importance for the next phase of AI, the gateway to the embodied AI, and how important it is, he highlighted that the Blackwell delays, which continue to be there, not on bringing them there, but to get them all together. There's a couple reasons for the bottlenecks, and he goes through the difference in it. But Blackwell is such a dramatic uh efficiency gain in terms of being able
[18:42] to to use it. The problem is, for the last, you know, 18 months, 12 months to 18 months, we've been stuck getting past scaling laws and actually in- in- increasing them. He goes through the way that that was uh had to go on in terms of reinforcement learning, test time compute. All these different kind of efficiency gains to allow the models to continue to progress, but now the leap is going to come dramatically. This is really important. I'm going to go through this.
[19:13] Blackwell will reset the industry in 2026. This is really important for everyone from an investor standpoint to really drill down on it. And I thought he did a great job of going through. Hopefully, I've connected some dots in here, which will make it easier to understand. Reasoning creates the first AI true AI flywheel. He goes through the usefulness, not intelligence, that we're at the stage now where Blackwell plus lower inference costs unlock agents. Before this, again, we couldn't The agents are delayed with Blackwell. He goes through the bear case, which is
[19:44] eventually we will get to edge AI, where you're going to having have these devices on your phone, on on computer, and that's going to limit the amount of cloud uh compute demand could shift dramatically. So, he goes through that case. He talks about the data centers in space, which have you know, all of a sudden is is is gone viral. Um he talks about the fact that the compute shortages will persist because Taiwan semi refuses to over invest or get caught over investing. So, they're being very very cautious on this. Unless they
[20:14] accelerate, we're going to constantly be in this compute shortage situation. And I think because power becomes the dominant long-term constraint, you can see where Taiwan semi would be reluctant. And as I go through some of the things I built out, uh that's that's the case. SAS companies are making the same mistake as retailers did. I put out a comment on a re-rating in software, and I got a bunch of uh hedge fund people reaching out saying, "Great, you gave us something that's already happened. This has already been the case." I don't think people read the report the way that it was meant to be
[20:44] read. It was very very structural and long-term. I view the SAS companies as being in the same predicament that the energy companies were after fracking came. You are now fighting an uphill battle. It's not that your businesses go out of business, it's is that you're priced as a growth company. And the question is for SAS companies, are they going to be able to find growth from the public companies? The startup companies will not be paying SAS companies in the way that they did in the past. The startup companies will be using AI. So, the question is, where is that next customer coming from? We can see bounces
[21:15] in things like Adobe and Salesforce for dot com. I don't see bounces in SAS companies as being an investment theme comparable to things like Corning, which I showed last week in some of the other areas. And I certainly don't think for the semiconductors that are more geared towards embodied AI, it is even a close strategy. You want to focus on hardware, hardware, hardware. Gavin Baker goes through the problem with SAS companies as well. He also says that if you just look at this year, you realize how difficult it is in the AI frontier race. And this
[21:47] gets back to the Dario Amodei. And I highly recommend that you people start to really worry about the spending happening on the frontier model companies. Not um that as them going out of business, but just in the fact of it really does make it hard to know whether they're going to win or not. So, this goes through and talks about the challenges of Meta and where they thought they'd be. Meta was thought to be at, you know, a certain place. They're not even as he said in the top 100. Couldn't run a 100,000 plus GPU cluster coherently.
[22:18] Underestimated the difficulty. Remember Microsoft saying they're going to close down some uh compu- uh some uh data centers. E- Every time you go through this at this point, it's very hard to predict what's going to happen. Uh and that's really this. Predicting the winners of AI is harder than any prior tech cycle because the true bottlenecks are not capital, not head count, not cloud scale, but deep research, intuition, engineering culture, and the ability to operate massive GPU clusters at the beginning edge of physics.
[22:48] They should not have the multiple in at the high levels of certainty. When they're spending this much money, their buybacks are going to go down, and there's going to be massive IPOs, which I'll cover, coming into spaces of competition. Uh I wrote a Substack on podcast and how they are the critical way to learn in today's thing. He brought this up. I loved it. For the first time, the actual people building the most important technologies in the world are talking in public every week. They're going on podcast,
[23:18] explaining what they're working on, discussing the constraints. It's unbelievable how much information you get each week. They they speak all week. I did this whole thing on Elon Musk last week. But this is every week. You can go listen to a bunch of them. Podcast with operators and leaders is signal. Commentary about AI by people who know nothing about it is noise. There is no doubt about it. Treat podcasts as raw data. If you're sitting there reading a research piece by some hedge fund person telling you why something is not going to work, I don't remember who the person
[23:49] was that wrote something about Nvidia when it was $120 saying that was the end of it at the beginning of the year. Trust me. W- Just listen to the people in the businesses, connect the dots. Now, I will be launching the website and teaching people and continuing on my consulting business, teaching especially younger people, but hedge fund people, allocators, pension funds, whatever, in terms of organizing their world to use AI to basically do the work that I'm going to do now. It doesn't have to be a
[24:19] podcast, but these are my gems with Inside Gemini. And basically what I do is I'll take Gavin Baker's podcast, I'll upload it into here. All the instructions are in here. It's about a three-page instruction list of what to do. Act as a buy-side analyst to convert podcast transcripts into actionable investment memos, extracts macro, blah blah blah. This is starts with give me the macro parts of this that are important. So, it's meant to go in and pick out those parts. Then I take that result, I put it into this gem to do the deep research.
[24:50] And then I go in and get specific company ideas. The next one is technical analysis to go in and tell me the ones that are already trending. The next one is a commentary. So, by the time I'm done, I have a complete research report that you can use. This is the way I'm trying to help or get paid by people to train them on how to use this stuff and keep going. But this is the way you can go. This is the end result of the first one there, which basically takes what he talked about on there. And this is one of the themes, the watt-constrained era. He goes through the different
[25:20] components, the why now. And then breaks it down into all of these different places, shows you why, gets into this Blackwell complexity. The transition from Hopper to Blackwell is the most complex in history, requiring a jump from 30 kilowatts to 130 kilowatts per rack. Liquid cooling reinforced floors effectively stalling hardware progress in late 2024. Reasoning models are verified, bridge the 18-month hardware, blah blah blah blah blah, and progress would have stalled in 2024. Taking that, taking any
[25:52] of these things. He goes through the CH Robinson case study, which I brought up here, which is a transport company. Why they jumped. Can this be replicated? He goes through all of these different things. You can take any one of those and then turn it into actionable ideas. I took the Blackwell as a gateway for VLMs and VLAs. I compared it to my research paper to go through and find find ideas. The line that was the most important is Baker warns of a short-term ROI error gap. Billions in capex for training being spent with zero revenue
[26:24] coming back before inference monetization, potentially pressuring balance sheets for three to four quarters. Here's the visual. So, we've been kind of stuck in the Hopper era, focused on Blackwell. Spending tons of money, having to build out all these data centers to eventually get the gateway into all of these things here. So, we're caught now. The money's come. This is the software side. This is the LLM build out. To get the VLM build out over here, you need massive investment
[26:55] far faster. You've got Colossus 2, which will be the first pure gigawatt data center. All of this stuff, you've got Stargate being built. Meta's building a mass a bunch of massive ones. In the interim, there's a a risk here that the revenue doesn't come in. This is where the revenue is going to kind of come in. AI agents fit in here. Uh edge inference, all of this. So, this visual is meant to show two things. One this here and this here is massive amounts of hardware. Here's the software side. This is now
[27:26] going into the, as he calls it, three to four quarters of pressuring balance sheets. You add in the power and infrastructure choke points. You can build the data centers, but we still need this stuff to allow these to run to be able to get to this using the Blackwell chips. This is where this choke point is. Again, this is a hardware situation. This is the physical world, these three components. And this is what we're entering. And all we did here was build the early stages of the brain. So, this is before we get to this and
[27:56] get through it. That's why this is such a big deal in terms of seeing it. Now, if you want to use a comparison, just remember what happened once we got LTE from 3G. So, with 3G, we were kind of stuck over here with the limitations. Remember your uh Uber uh freezing or Waze not being being as up-to-date or losing. Once you got LTE, then video came. You got Instagram. You got all of the YouTube. All of these different components could be used on your phone and expanded the ability for these companies to grow. Blackwell to LTE is
[28:28] the comparison to use. He also says this. And for all of the leaders, uh I obviously have run businesses since I was 29. I've been in that seat. You have to be a leader using AI if you're going to be able to get the culture in the firm to change. He talks about how AI-native founders are insanely growth-oriented in terms of how they're using AI. They are learning quickly. This is why one of the things that I'm focused on for my business and
[28:59] through 22V is to help college kids start to use this and become AI-native. An AI-native person has a huge advantage. The way that I'm able to build things as a 58, soon-to-be 59-year-old blows me away every day, but I use it multiple hours a day. I think on these videos I've said seven. I use it every single day. It never stops. It's on the road. It's on my phone. It's here. I build things. I come back. I build them again. He talked about how different
[29:30] leaders are using AI and AI-native. You can be AI-native. You just have to use it all day long. He also talks about how investing is a search for truth. This is the power of using AI. And this is what I want to get into to to you the first example of me using it especially over the last 3 years to switch away from what I used to do. I am a systems thinker in terms of way that I like to approach problems. This is why I hated school. I do not like silos. I like to cross-reference data
[30:01] from various places. It's one of the reasons why I've built contagion models, turbulence models. I focused always on trying to predict when things were going to go by covariance matrix. Everything to me is about taking correlations of things and looking for differences. This is what AI allows you to do. So, I got a lot of reach out on HRV. It seems like there's a lot of ordering people out there. There's a lot of people who are are on the same journey I am of my iWatch and everything. So, HRV I'm going to just give you an example.
[30:31] So, I got my ordering back in 2020. I learned very quickly that the most important thing with HRV is that it is a one of the highest markers of biological age. As you get older, it declines. So, what I wanted to do was get it to rise. So, you can see that over the course of the 5 years it has been in a linear fashion. But, where it really started to accelerate was in here. This is when Chat GPT came. This is when I started seeing these blow up upside. This is me testing and working on a variety of
[31:01] different things to try and get it back up to here. You will see that the age side of it here is the age thing for ordering. This is all of their users. They have about 5 million users by age and you can see the correlation. So, the younger you are I'm currently at 80 on a 80-day moving average. That chart you just saw was a 90-day moving average. Put me at 75. So, basically based on their data, this is the midpoint of the range. This is the high end. I have taken myself and reduced it. Now, I also go through my blood work on the same
[31:31] thing. The reason I'm bringing this up is this is an example of how you can take any problem and go through this to where this is the way that you would try to do it based on reading, researching, and Google. You'd go meet with a meditation instructor. You'd go learn about your immune system, your nutrition, your exercise, your sleep. I did all those things to get it up here. But, it was with AI where you're able to be a systems thinker and talk to one thing. You don't have to go to a specialist in each one and they don't know anything about this. So, if I want to know the
[32:02] impact that nutrition has on the immune system, I can get through this. If I want to know the exercise that has an impact on my meditation, my breathing, and my sleep, I can go through it. This is the way that I learned everything on this. And so, you end up putting together HRV is a mark of biological aging. When you go through my blood work, it says the exact same thing. I'm trying to increase the probability of longevity. It's very Bryan Johnson-ish. I started it before he did in terms of at least looking at it. The approach I take is completely different, but it's somewhat similar,
[32:33] which is database. I'm just saying you can take any problem in life to solve that is very complex, do it that way. The problem is the way we were trained in school is in the silos. The gastroenterologist does not know enough about exercise and sleep. He knows something. He doesn't know enough. Schools kill curiosity. They stick you in silos. They force you to do this stuff. It's the reason why I hated school. I couldn't stand it. Went through it. Now, I'm having fun using a system that I can speak to someone all the time to solve problems. Now, we end
[33:04] up in a situation where Eric Schmidt talks gave an interview or a podcast this week. Every time he speaks, you go listen. He covered a very a few important topics. If you combine it with the artificial intelligence and you combine these two podcasts, which I did transcripts into one because they covered a lot of the same thing. It is this though that I'm most involved in right now. Recursive self-improvement. If you haven't spent time, you get to language. Then, when you get to agents in reasoning, which is where we are now,
[33:34] you will eventually get to recursive self-improvement. So, recursive self-improvement computers are learning on their own. It gets rid of the human bottleneck. So, it reduces the need for humans. This has huge implications for Bitcoin. It has huge implications for decision-making for booking.com. Anything where you're taking human biases or human nostalgia out of it and the decisions being made are based on the lowest cost, the best place to go, the fastest place to go. You can see why stablecoins will grow rapidly with AI agents. But, more
[34:04] importantly, when you get to recursive self-improvement, you start to see this thing all going on its own and problems will be solved much faster. Think of it as the speed uh the the speed limit in AI terms went from 55 in the LLM side. We're now up at 75 and soon we'll be going at 150. The compounding continues to go faster and recursive self-improvement will speed it up. They cover that there. There's a moonshots episode where they also go through it. This is a really important thing to start to understand.
[34:35] Two to four years is what this Eric Schmidt is saying. Four years. San Francisco consensus is more towards two years. This is more important than AGI because once you get to recursive self-improvement, you're getting there faster. And if you want to have one other part with it, think about Elon Musk saying once the humanoids are building the humanoids, that's when the labor problem becomes optional for people to work. So, Silicon Valley believes self-improvement is 2 years 2 to 4 years. This is including Eric Schmidt's belief.
[35:05] Uh So, I'm not even going to bother with that. It's another another recursive self-improvement uh one. Pentagon ordered to form AI steering committee on AGI. Again, next step of before AGI recursive self-improvement. Again, the military is depending on this. The reason I bring this stuff up is every time you think there's a bubble in this, just remember the government has a vested interest in in AI growing rapidly and we still have 3 years in this administration uh to make sure it happens. 5.2 was released. Not going to
[35:35] go through the details of it. But, the constant chase of this from all of them is there. The latest entry to me that has uh surpassed uh Gemini and uh and Chat GPT for now is Grok. I'm using Grok a lot more on various things. Now, Elon Musk said self-driving Tesla uh robotaxi countdown to 3 weeks. He basically said in a video conference this week at an AXI hackathon, this unsupervised full self-driving is pretty much solved at this point. There
[36:05] will be Tesla robotaxis operating in Austin with no one in them, not even anyone in the passenger seat in about 3 weeks. This is critical. This was him speaking this week before the end of the year. I've talked about this. I've talked about this as again, the opening act of embodied AI that once we got to this point with a single one, regardless of the fact that if there was an accident or anything that went on, Texas has already given them the approval to do this. So, it was up to him to wait and to go through it. He's been having software updates constantly.
[36:35] Once you get to this, you're at the race to embodied AI and I want you to go back to the Blackwell thing. This could not happen without the ability of taking all of the data he's getting and go through this computation with the physical world. So, video plus length plus text just takes a lot more compute and Blackwell allows you to get more because of the efficiency side. Colossus 2 is one of the reasons this will happen in the first quarter. This is a big deal. Gavin Baker talked about it.
[37:05] Gavin Baker is clearly biased towards Tesla. So, I'm not going to go through the um his viewpoint on the on the company and the stock. But, I do believe that Colossus 2, the first gigawatt data center, how quickly it's being built and how quickly XAI has caught up is very important to this whole discussion because of the Blackwell situation, the need for speed. How you can be bearish on this chart at this point when you're just getting higher high This thing is going to explode at some point. But, I believe embodied AI and the ability for him to scale so quickly, it
[37:37] just reminds me if you go do your work on Nvidia. The bias against Tesla is so big. But, this is critical for the benchmark argument I'm making for next year because I don't know many people that are benchmarked to the global benchmarks that even have an a position in Tesla. Most people have a bias to not own it. Either they're restricted or they just don't believe they don't know how to value it. I think it is a great risk when you're at the embodied AI beginning and it is the company that stands out the most. Now, just to use a smaller sensor-based PMI, one that you should go
[38:07] do some homework on. I just wanted to bring up lidar in this. Obviously, Elon Musk is is going anti-lidar. But, that's not the point. Most autonomous vehicles and robotaxis, which are not just Teslas. This is Waymo's. This is military vehicles. This is uh mining uh automobiles. There will be so many robotics and so many things coming out which will not be based on vision. They will be based on lidar. And what you end up with is at this point lidar demand is going to absolutely increase once the Blackwell situation goes. So, again,
[38:39] all of these physical hardware things that are necessary to power the robotics and all of these things, this is the timeline of what was needed. So, you actually needed the Blackwells to happen to be able to be the gateway to get this stuff going. So, I view lidar as part of the PMI up uh move higher. I view the up cycle or the upgrade cycle for cars, for computers, for phones to be directly related to the bottleneck that happened with this where we were finding efficiency ways to go,
[39:09] but that didn't help with the VLMs. So, just remember I showed this last week with regards to Corning. The stagnant years for Corning during this period where it's all software. So, unless you were a software company for 17 years, nothing. So, when I talk to people about a re-rating in software and they go, that already happened this year. Sorry, that's not the case. Not my view. I believe we are destroying the valuation through hypercompetition. Uh Bill Gates talked about it last week. Eric Schmidt has talked about this.
[39:39] Software companies that are big will have an uphill battle trying to compete with smaller companies that don't have the people and don't have to go through this. It is a really tough in my opinion for a software company relative to the hardware stuff where we under invested here and the demand is necessary to fuel all of these things here. So continue to focus on the fact that if you can get situations like Corning which is so far been an optical fiber but will also be a glass situation in this.
[40:09] It's really hard to pick up the demand on this. You're seeing this with gas turbines. You're seeing it with transformers. If the demand comes quickly these stocks can go up violently. I would look there. I would also spend time because of the power situation on the solar names. There's a variety of areas that got hurt that I think have been on the lower side because of PMIs. Now the IPO situation if you haven't put this into context of next year, we are going to have massive IPOs. They This is from the FT. SpaceX, OpenAI and Anthropic. All of them could
[40:41] IPO next year if not next year definitely by the end of 2027. You can include XAI in this too. The reason is these companies all need money for the hardware capex that they're spending. So they can go out and keep trying to borrow money or or do deals and get money at 1 billion at a time but they are still going to need to do this and when you think about what that will do to growth related companies the fact that these companies are either being competed with from the software side so
[41:13] code being free and they're having to spend tons of money and they're going to have IPOs coming into the marketplace. It just doesn't seem like the place. There's going to be a lot of overhead here. So unless their earnings grow rapidly, I think multiple compression is likely. If those companies multiple compress, emerging markets, foreign markets, the 497 small caps, those will be the outperformers at a time when people are benchmarked to growth. I could this is the piece I did this week that got the you know has already
[41:44] begun and they wanted to tell me that this is already happened and now's the time to jump back into software. I don't agree. I think over the next 5 years as AI agents come in this accelerates, it's only going to get worse. If that's the case, growth versus value, size factor, all of those things will start to rotate. You're already seeing that here and again, IWMs versus Qs, the black widow trade. The only time it has worked over the course at any any point over the course of the last 20 years has been this is IWMs over Q has been the times
[42:15] where PMIs go higher. So these lines are where the PMI bottom below 50 and did go higher. Not in every case did they make money. Made money here for about 2 years barely. Barely. Again, barely for a little while before it went down again. Right now I think PMIs are going up. If that's the case, worst case scenario if you went from here to here, trust me that's from here to here is is a doubling from 0.41 to 0.8. So don't minimize it. I think there's a lot of risk here. Bitcoin
[42:46] I put this Substack out this week um going through why this is the most asymmetric bet in my opinion in my career. I highlight the fact that you can believe that Bitcoin's not going to go higher. Nothing is 100%. I go through my odds on where it'll be 20 years from now in terms of the probabilities based on everything that I think will happen to fiat assets from AI. My entire viewpoint is based on what the future will look like and the fact that performance wise like Tesla despite being hated by
[43:17] everyone, it is outperformed every single thing except for Nvidia over the course of the last decade. So I will continue to go for it. On that front, we're still in a downtrend here. Bitcoin is very very technical focus cuz it needs the retail traders. Until we break above this line and stay above 92 to 93,000 for 3 days, I think we're going to stay in this downtrend which means we could have another leg down here. This looks to me like it could be a completed Elliott wave ABC down. So we could start a new wave
[43:49] but I'll leave it up and just say next year is when I would expect it. Ethereum we did break through what I highlighted last week but then we went right back down and tested the bottom end. This needs to stay above 3,000 and go through it. I think Bitcoin needs to be above 92,000 with that thing trading higher for this to go. I put this out. It kind of went viral 700,000 views. It shows how much people want to find a reason to buy Bitcoin at this point but they're not. The sentiment is still very low but this is the overlay between
[44:20] beta versus profitability. Beta versus profitability as I highlighted which is PMI sensitive have ripped back to the highs. Bitcoin is not followed yet but at least you have a technical reason that if it did start to go, you'd get a lot of people involved. Now just to finish up this week in terms of again, the news about how structural the financial guardrails are changing. Stablecoin should work better for international payments from the guy who hated it and said he saw anyone that used it, he'd fire them. Blockchain is real. He speaks out on the blockchain says that's working. JP
[44:51] Morgan arranges a galaxy bond issuance on Solana blockchain. So they're actually using it. National bank regulator says banks now can buy and hold Bitcoin for customers. So custody has come in. CFTC chair says using Bitcoin and crypto as collateral will bring trillions into the US market. So now we've got collateral. Fidelity I like Bitcoin. I own Bitcoin. It will play a role in the savings hierarchy. Michael Saylor still hated more than ever. All the major banks have contacted
[45:21] him for Bitcoin advice. You go through X, everyone says I doubt that. It's 100% true. Of course they do. They're trying to make money. This just shows that while the old guard has been selling basically this year a lot of the holders have been the corporate buying that's gone on led of course by the blue here which is strategy but regardless that is where the transfer is going and just to finish off this week I want to reiterate the importance for tokenization and the balance sheet side of the equation. So as everyone focuses
[45:53] on income the balance sheet of the equation is going to be released starting next year. Mark Rowan sees market makers coming for private credit. How does this fit with tokenization? So turning illiquid assets into tradeable ones is part of tokenization's attractiveness and what is going to happen. If all of a sudden now you can hold your commercial real estate but sell a piece of it or find eight billion potential buyers structured in some way where everyone does it the same way we
[46:24] do in the public markets where every company needs to go to where the capital is the biggest you're going to start to see tokenization have an impact on more volume in the system. Caitlin Long has talked about this that velocity will increase because the amount of money that is flowing in the system and not clogged in illiquidity will start to go. This will help endowments and pension funds and everyone along the line who wants to sell something and do it in a way where they're not having to go out to three
[46:55] potential buyers who are telling them to bid down because they need to get out of it but to actually be able to do it on tokenized world. All right guys, that's it for this week. If you learned anything and you want anything, remember reach out to 22V. The videos will be coming out soon. The help in terms of the names will be coming out and do me a favor, subscribe. It helps me. I'm building a business. I need all the help I can get. Subscribe in Substack too and reach out and I'll talk to you soon. Have a good week and I'll be back next week.