Transcription
Um, stocks just continue to uh defy whatever goes on. Um, I'm going to go through that today. I'm going to remind you guys about bubbles, parabas, and speed crashes and kind of give you some insight into uh what we've seen the last four to six weeks with inside AI. And I'm going to say now that uh the midcycle slowdown is ending. And it's not that we're not going to see the rate of change, but now this is no longer a surprise. Uh positioning has shifted and I'm starting to hear people that are trying to be bearish start to come out of the woodworks and highlight a whole bunch of things which I'll go through. Um it's been a sentiment and a technical cleanse. Uh I'll go through why that is. I'm going to go through the token index. I I've been called more on this than anything over the last three weeks both from institutional from retail and from very sophisticated tech people on why this token index doesn't matter. Uh Meta Apple interesting week uh I wrote a paper on meta. I'm spending a lot of my time on consumer AI agents which I think is the next theme. I've talked about the application side but I am going to start to go through uh what is happening more and more and what I think is going to be a major story for the next 12 months.
Uh, I'm going to remind people about exponential demand versus linear supply. Scott Bessant did a speech and I spent a lot of time on it this week with regards to crypto. I'm going to basically start to give you guys what I've talked about now for the last month, which is beginning a new um part of AI macro Nexus, remind you what this was all for, but combine it with what Scott Besson said, what Mark Andre said regarding Bitcoin, and this will be a new in the same format as this, but very crypto forward. And we'll do a recap on what happened with crypto this week. First of all, um just in in general, uh I've had a lot of success with RAAS. For those of you who are financial advisors who want to spend time uh on how the work that I'm doing can be expanded more than just this YouTube video and more than just a subscriber site, uh I'll show you where you can reach out. But uh between that and especially the agency side of people working on knowledge brains uh it's been a uh a really good three weeks of helping people not only stay on signal with AI but also start to uh use it more and more and I think that's helping. So I just want to remind people what the name was because this video this week gets into why it's AI macro nexus. Uh AI is no longer just a technology story. It's becoming the dominant macro force of the cycle. It changes productivity, labor, inflation, capital spending, energy, demand, digital assets, monetary policy, market structure. The nexus is the intersection of these forces, and it's the place where investors connect. The reason I bring this up, um, everything that you're dealing with when you're reading something on productivity, on labor, on inflation, on the Fed, on anything, AI is a part of every single decision that's being made. And that's why you have to be on top of this. It's changing the world in front of our eyes. Here's one example of it.
So, this is uh something I've shown a bunch of times. I did a subscriber uh video update where I started to go through the fact that I think the slowdown is reached a point where we've done enough damage. Um, and this is where the damage has been. This is the 60-day realized V of the technology macro uh technology momentum factor from Morgan Stanley. The white line here is that. So, we've gone up and I've shown this for a while. I showed it when it was around 70. We're at 87 now. This is through today on Friday. I'm doing this Friday night. This is overlaid with the S&P 60-day realize V. So, historically, there's a correlation that does exist. Meaning, if you're seeing tech momentum V explode like this, you're usually seeing the S&P V follow it. Now, the one thing is tech has been going higher for a while. This is why what we're doing is this bubbles, parabas, and speed crashes. This is what's going on. That's why you have to be on top of this because this is a structural bull market. It is based on token demand. Everything that I built everything on is on token demand. This entire video by the end should give you complete comfort that we're not in a bubble that the compute there is no excess capacity because Meta is selling some off. Hopefully you guys get there and you realize that if that's true over the next six months you're getting an opportunity where this has gone higher. has flushed people out across the globe on retail, on hedge funds, everywhere that people were using leverage and probably overexposed. We've seen that now come out mainly because the V's been too high for people not to do it. S&P, I don't know how you can look at that chart and not be bullish. Um, I told last week I showed it as a triangle. You know, we get bad news out of Iran, it still goes higher. We whatever you want. We get sell off on Samsung on good numbers. It still goes higher. This triangle in the NDX is underperformed. Again, tech is is lagging. We're getting more exposure on financials, more movement in healthcare. Uh, but we did finally start to see some tech bounce. But again, this is a bullish looking chart. Uh, this is a reminder if you guys didn't see it, I did this video back on May 10th. Uh, and the focus of this, which was right around the first leg lower, is about the way we should expect this to trade. uh when you have bubbles and parabas on the way up, you're going to have speed crashes on the way down. If this is all with inside a structural bull market, which I'm going to continue to say, this is a structural bull market. You want to go with the river of a macro bull market, especially when you're in the first inning where you're in the first inning of AI token demand. Regardless of what anyone says that it can't be the first inning based on the charts, these charts were about people being wrong on AI for two years saying it was a bubble. It hallucinates. It's not going to do anything. They had to play catch-up once the earnings came in. Then we hit the second derivative and it's not going to be as easy from here. So, it's not that this trade is going to do what it did before. We are way past the first inning of the speed of change. But do I think that you're going to get another five bagger in Micron and in SK Heinex in these things over the course of the next five years? It's very likely you will, but that's less than the 20 bagger that occurred from March of last year till now. So, we're going to slow down because now everyone is aware of it.
So, Samsung uh released premilliminary quarterly profit up 19fold. Micron fell that day. Samsung was down that day. Here's the chart and this is the first thing I want to show. I saw three separate YouTubes this week on the market where they said this was a news failure. See, good news came out. It beat wound estimates and it sold off. The problem is we had already sold off. I don't subscribe to the fact that when you've already had a correction, you've already had weakening where we're unchanged over the course of this when you get good news that you should sit there and say that this is bearish. Is it something that showed that the unwinds and the deleveraging were still on? Yeah. But if you ask me what the range is going to be and where the support's going to be in this going forward, I will tell you that I believe this high from February, this low from here, and this is going to be around the bottom end of the range. And I think we're going to be significantly higher before the end of the year because of what I'm going to show you. This is the problem is that everyone is looking for a correction. So if I told you that over the 10 days in the two days after that, this was the rolling 10day rate of change in Samsung. Okay, it was the largest since the middle of the COVID panic when we shut down the world was 25 26%. This was 21. The only other time that it was this big on a 10-day basis was just after Lehman Brothers. So this is a speed crash. It didn't feel like it. You don't think it looks like it, but within a bull market, this is a big fall, particularly when you had these types of upsides that were much bigger. So, these are the rolling rate of 10day change. You had one at 35. So, this has been a correction with inside a bull market.
Now, here's where we are on a PE basis for next year. So, everyone who wants to fade this, I completely agree. If you believe in the cyclicality of memory, you should absolutely sit there and not pay attention to this. If you believe that the demand is going to stay at a very high level and prices are going to stay at a high level for the next three years, this is way too cheap and argues for a much higher PE than is normally there. If you guys look at this and pay attention, my PEG ratio, when we're looking at the earnings growth relative to the PE, they've already discounted that this is going to slow down at these levels. They've already built in a bearish outcome in terms of multiples. Here are how ridiculous the numbers are. And this is why when people say this is just like every other time, Samsung's operating profit in 2026 could be over $217 billion. That's where analysts believe. Okay. The company could generate more operating profit this year than its cumulative operating profit over the past 40 years. Don't treat this as a one-off event where AI is just doing this and everyone's over buying and then we're going to run into a problem. It's silly. Um, we're not even in the first inning of token demand. I'm going to keep saying it because nobody believed in token demand until we got into January of this year. And I will highlight in some podcasts from Dylan Patel from Semi analysis how much his spend has changed and where he was in December of last year to today. That should give you guys a sense as to someone who's very sophisticated on a growing business who understands how to use AI and his team does and how he went from $100,000 budget to $11 million over the course of six months in terms of what his annualized revenue spend is going to be. So Korea's Cosby valuation falls to the lowest since the global financial crisis. Those are the types of things you see at tops. Joking. Um, here's Salesforce.com. Everyone who is basically bearish on memory are the same people who wanted to buy Salesforce while it was coming down. So Salesforce is now down to a 10p to 2028, 11 PE to next year. So again, we've got a 4 PE on Samsung with their earnings growing this year 524% and for next year it's 43. They haven't come close to meeting the estimates yet. They've been above it by significant numbers. And yet somehow or another, this stock is trading at an 11p. This is a problem. This is why from the beginning of the year, I said you want to be short abundance. Anything built on code that can be replicated by a computer. And you want to be long scarcity. Anything that the spenders are spending on that is physically constrained by linear humans. That's it. Humans build this, computers built this. That theme will continue to permeate because that's about the exponential demand.
So again, I put this in on June 8th. I started talking about the fact that I was getting out of Micron in the middle of May. Again, I got out way before the top, but that was not my goal to pick the top. My goal was to be in a position where I'm probably going to be buying some back at some point soon, especially after the SKH Heinix deal that went through today, just because I'm hearing people be bearish. And I do believe that 6 months from now, we will be significantly higher than we are now. Now that we've had some sort of pullback, but also since we got another earnings report and I'm no longer worried about some of the things that I thought could be an issue in terms of the bottlenecks, it looks like we're just going to be in a situation where things are going to go at a slow pace. Mainly because there are constraints that are built in from humans that are keeping this from being a bubble. Believe it or not, it's the supply side, the TSMC's that won't overbuild, the fact that the gas turbines and all this make it very difficult. We're finding solutions to every problem on the power side and they're finding ways to get this done. So I just think that that's going to keep prices at a higher level because we're not going to be able to get capacity on time. Uh the econ so the navigating the AI midsight cycle slowdown. I still believe with the second derivative change but now that is consensus in my opinion. Here's what we did and this is part of it. So you look not only for what I hear from people but you want to see how much damage we did. This is what happened. We've given back now 50 or almost 50% of the techo rally from Opus 4.5. And the reason I care about Opus 4.5 is this is when the change happened. This is when people started to acknowledge it. This is when people started to invest. And I want you to look at this because it does look a lot as I go through this from February down then up looks a lot like the token index as I get back to later. And I'm not worried about the token index, but I think the token index, uh, people have started to use it for something it shouldn't be used for. I think that should go to zero over time. Um, and more importantly, I don't think it should be rising. I never think it should be rising, and it technically isn't, which I'll go through as well. When these stories come out, it's usually a good sign. Systematic long short managers dropped 2.1% last week through Thursday after declining 3.1% over a 5day stretch. That was their worst since December 23. On most of the numbers that I've seen, they've given back. And these are again systematic long short managers. So these are quant managers. They've given back about a third of their year. They're still up for the year. This is not some major panic. I talked about this feeling a little bit uh on the webinar that I did this week for the subscribers about this feeling a little bit like 2007. Um, and that it would end at some point around the same time it did there, which was in the first 10 days. I still think there might be a little bit left, but I also think the mutual funds have proven their case of being underneath, which I thought as well as they took down a lot of this SKH Heinik deal. Um, fundamental managers also fell and we know that the pods had to close down as I talked about because of how high the V was. Here's the deleveraging on the retail side. Even though we only had about a 23% correction on the uh SKH Highix and Samsung side down to about a 30 uh 38 and again this is back to Opus. The retail side using the two times levered gave back 62%. So I think retail's been flushed. Morgan Stanley put this out on their AI beneficiaries and you can see the massive outflow that occurred and we've had three weeks of fairly large uh outflows in terms of the net flows. Jeff Degraphth, I like to combine this stuff. This was a great one. Nearly 87% of S&P 500 semiconductor issues registered and oversold on a 14-day basis. He uses these lines. You can see these have all been very good bottoms except for this one here. all the others uh except for these two little things here. So, could it go lower? Of course, it could. But at least you're getting also that on that side.
Here's what happened to my side in terms of the technical sheet that I put out. This was from May 29th, the week ending. Look at these numbers. Percent 50day, 100, 91, 98. Look at the average RSIs. Average RSIs. Everything was above the 50 level. Here's where this week's came in. Look how low these numbers are. Percent above the 50-day. We've really sold off here. You have one that's still above, but the average RSIs are now there and we still have constructive none of the tech patterns have changed. I'll go through that as well. The score, they're good, meaning this is a good entry point. Um, I show this uh again, I've shown how to do this and I will finish this video for you guys in terms of how to be able to do this on your own using Claude Co-work by just updating the files. But this was 20-day of my 100 name thematic portfolio. 73 of the 100 names were above the 20-day on May 8th. We got down to 21 on Wednesday when I did the uh uh sub subscriber webinar to basically an update to kind of give people a sense as to where we are. Uh I like to see the bump at the end of the week. We went from 21 to 31. This definitely seems like an oversold level. I mentioned last week that when we get or I mentioned this week when we get below 40 I get very interested on the uh percent above the 50-day and most importantly above the 200 day it's back to the same level it was kind of ending on the 26th but these levels 50-day rising still at 76 and the 200 day rising which is the critical one which is the bull market structural situation is still up at 86 of the 100 names. We saw massive inflow in the SOXX and SKHEX today uh offering more than seven times overs subscribed. It also traded up.
Here are the facts from the memory side over the course of the last three months uh just in terms of what's gone on. Memory chip crunch to persist until 2030. This was back during his earnings in March 16th. If you thought anything would change, this is what he said today. Okay, the memory chip shortages that are rolling the computer that are roing the computer car and device markets will probably persistent beyond 2030. So, we went from until 2030 to beyond. Customers are signing long-term contracts. Hadn't seen that before, but remember Micron Inc.'s long-term supply agreement is worth $und00 billion. Says it has no idea when the RAM crisis will end. This was less than a month ago. Actually, it was less than three weeks ago. and Dylan Patel memory capacity is only growing 20 to 30% a year for the next three years. This is the human capacity side. Demand is doubling. This is the exponential dam demand. Memory prices are going to keep soaring. This is from an interview which I will highlight which came out today. Remember, Elon Musk is building a terrafab because he says there will not be enough memory or chips for his humanoids and the demand that he sees by 2030. I'm going to get into that as we go on. I'm showing you these things because this is the reality of where we are.
So, I want to make this clear because I had I've had multiple people now and I think this is because there's a bunch of people on next that have said this is just like fracking. this will blow up. The only thing that's just like fracking is the amount of people that are saying it's just like fracking. So, here's a normal cycle before AI. And I say before AI because this is a typical demand cycle for hardware driven demand. So, hardware driven demand meaning physical stuff. So if the demand is driven by physical output and the memory is a physical output, it's really hard to get demand significantly away from capacity. If the demand is now digital and it's tokens and it increases purely based on the productivity of the models, but we need more memory. The capacity can't grow any faster. The only way to fix this problem is with an algorithmic efficiency side of which that is such a long shot that at this point to get it to be able to do it to be able to get it out there for every it's just a long shot. Is it a hail Mary? Will it eventually happen? I'm sure it will. I don't think this is going to get up to meet there anytime soon particularly when I go through what else is coming. So you have to remember this is the difference. This is the old world. This is the new world. This is exponential growth. This gets important when you start going into consumer agents, full self-driving and humanoids. These are all bigger demand for tokens than what we have now. Just like chat was here. Chat, we didn't have a problem. We only had a problem when we got into this period here when it became evident that we were starting to hit a part of the growth curve that was out of control. This is why when you compare it to the oil market, it's ridiculous. we had a parabolic or an exponential move on the supply side. The demand side didn't move. So again, for everyone who wants to compare these two, for every smart oil guy or smart person who keeps telling me this is the same as this, I've seen this before. You haven't seen this before because the demand is not coming from human beings. The demand is coming from computers. It's coming from AI agents. It has nothing to do with humans. Humans move on a linear scale. It's very important for people to start to realize this. That is where the bull market lies. That is why I do this stuff.
Now, my next webinar will be on July 15th, next week. I will specifically go through names with inside the portfolio that I think have hit the levels that make sense. I will go through the concentrated basket, all 10 names, and give you guys the details on it. This is what the concentrated basket has done now. So, we've hit the 50-day moving average. We did this back in March 2 and then we got the next liftoff, we hit the RSI, you guys can go back and look and when the RSI gets down here, the riskreward has sh has changed. So when here when I thought there was a midcycle slowdown, we've basically gone sideways during this period. Now we can reset and go. I reduced my position in a bunch of these things down significantly with Micron being the biggest one. I kept some things the same. I moved some things into silver and Bitcoin which haven't worked out as well. I will now go back and look at some of these names and pick out the ones with inside the concentrated basket. So, this is the whole point of kind of helping you guys not just on one side but also the other. If you guys follow me on X, I put this out uh on the same day I did the subscriber webinar. Nvidia peaked on May 14th, three weeks before the AI infrastructure trade peaked. So again, if I go backwards and go over here and you look into the middle of May and basically go through, that was when we saw the peak in the rate of change and when we started to slow down. Well, Nvidia peaked beforehand and more than a month before SMH and the memory names peak. Today is set to close above the 20-day moving average for the first time since June 2nd, holding the 200 day and with a MACD buy signal at the lowest valuation in a decade. That is one of the 10 names. So in the consensus basket, Nvidia is one of the names. So again for the RAAS and for anyone who is new to this, if you want to reach out, that's where the website is where you can go find it. You can always reach out to Mark Whailing at22V. He will help you on anything you need, particularly on the RAIA side.
This is where I want to get into the token expenditure index because as I showed you before, if you go through the es and flows on the earlier chart I showed with the everything related to this, including the AI themes, I highlighted back here in the middle of May when Nvidia peaked, that this would be a growing risk, not because it was a negative, but because the narrative would start to change and that's what we've seen. The narrative has shifted. So, let's go through this. The first rise in the token index was the coding agent capability shock. Then we had openclaw. So openclaw led to more users immediately using agents that happen to be AI native or Chinese. Openclaw was a big gamecher but it was a sophisticated level. So most of the token usage went to a lot of open- source stuff. Then the enterprises finally realized that they were missing out. So the FOMO kicked in and every single place whether it's Anthropics, ARR, everything started to see a step up function maxing from there. Token maxing became a word. Everyone was on the exploratory spend side. Then they saw the bill shock and the token index goes down. Not because everyone's bailing out partly because school ended. All of these different things that I talked about. It's when the usage starts to go more towards people that are using open source, which AI native companies don't stop. College students stop. People on Morgan Stanley, Goldman Sachs, places I know well, 20% of the people are out on vacation or working remotely or they don't work on Fridays. Whatever the case is, you're not going to have as many people using stuff as what was here. And the bill shock kicked in. The problem is with this whole analysis and with everything going through it, it doesn't include the amount of adoption that's still happening. We have massive adoption happening. This is not an an index of making or losing money as I will show with the Dylan Patel side. So again, memory shortage exponential token usage going higher every single day. I've heard people argue with me that nobody's going to use the most expensive stuff. That is a bunch of garbage. If you want to go hear someone who talks about this and I think always does a good job in an interview, this came out today. Dylan Patel or one on Thursday I guess late uh Dylan Patel on wisdom tree in Europe good interview I think the highlights that I have here Patel argues the AI ROA debate is missing how quickly serious users are increasing token spent and he highlights again that semi analysis spend went from less than 100,000 now the reason it was 100,000 in November which is a very similar story to me he was paying for everyone at the firm for a 200 $00 Claude model. Once Claude Code started to be able to work after Opus 4.5 and build stuff, that's when they went to a token spend side and everyone started to spend there. By January, it was already $4 million annualized. It stands at a roughly 11 million annualized today and the peak week was equivalent to 14 million. That's for a 90 person firm. Companies are blowing through AI budgets, but the right response may be cutting other SAS or legacy spend rather than cutting AI. The right model choice depends on the workload. Cheaper models for fixed workflows, smartest models for assistant work. This part and I h you have to listen to this. For assistant style work, the smartest model can be cheaper overall because it uses fewer tokens, fewer iterations, and less human time. This is why when you look at the labor market, we have weekly hours. If you can get things done in onetenth the time, but you're spending more money for it, that's the way it should work. So his point being, we are in the stage of enterprises trying to figure out what to do, how much they should be using for workflows, which will be much cheaper, and how much they should be using for assistance style work. You don't hire all people from the most expensive colleges with the best resumes. A team at a firm is broken up by a lot of different areas. You hire younger people which are cheaper than the domain experience people. That is the way AI will work. He goes through memory is no longer a normal commodity cycle. If you don't believe me, he goes through more of the technical reasons. He expects memory prices to keep rising until AI gets enough supply and consumer devices like smartphones and laptops are forced to absorb higher prices. Basically saying the same thing as the Samsung. We have a shortage where the phones, the cars, the PCs, we don't have enough memory. So the demand is coming not just from hyperscalers for data centers. It's coming from everybody. He's bullish on CPU. There's a CPU thing in there. I think it's worth listening to as well. Uh on the power side, he's been the best by far in my opinion on what's been happening. I've used him a lot. I've referenced him a lot. This was another great one. Basically what he said in a summary with me listening to it and going through it, we're figuring a way to get the gigawatts done, even if it means using diesel truck boat train engines converted for power generation, you have to listen to it to understand that if you thought we were going to run out of compute because people are going to stop the data centers, it's not the case. He went through and obviously I I I'm pretty sure he's an investor, but if he's not, he obviously has access. You're getting more and more of this that Anthropic is going cash flow positive or at least reaching operating profit. Again, this just means the margins are coming through and the RAIC should not be there uh in terms of a fear. But I did do this on the AI capex air pocket. If you guys didn't read it because of the vacation, I would go read it. I I spent some time I've spent a lot of time on consumer agents, but I really spent time on this Mark Zuckerberg town hall reportedly acknowledging the Meta Zentic AI progress had been slower than expected. And I make the argument that if they did end up selling any of their capex capacity, it would be because they can't get their consumer agents working the way they want yet. It's coming, guys. And if you want to go listen, this came out one day ago. Meta or two days ago. The CTO of Meta, Andrew Bosworth, was on with Alex Canantrowitz. It's a great interview and it's very simple to go listen. The industry is moving beyond one giant monolithic model. I think he talks about the fact that there's going to be open source and there's going to be the frontier models. Consumer AI has been slow because the products are still too hard to use. This is a really critical part especially for mutual funds and longerterm investors to spend time on understanding how consumer agents are going to change the world which is a dramatic shift. Whatever happened with coding agents which is for the language of agents this is massively bigger. This is far bigger in importance than opus 4.5. So think of it this way. All we've done is increase the ability to code but we haven't found the ability to get through all of this yet. It is incredibly important to understand what that means. Consumer AI has been slow because the product layer is still immature, not because the model's weak. For consumer AI to work, it has to disappear into daily life like search maps, messaging, or the camera. I use my uh Grock in my Tesla this week. I had a conversation with it on the drive, the entire drive. I cannot do that with a phone the same way. It's not as good until we get to the point that you are speaking to it like you were supposed to with Alexa where you can be sitting there cooking and asking it questions on what the recipe is. So you don't have to go to your phone, you don't have to go look at it. Until we get to that point, which is coming soon, it's really hard to get the habit formation. But once it happens, billions of users will start having agents working for them and they will start giving them things to do, but they need the context, the trust, the timing, the permissions. It is extremely complicated and the frontier models are not capable of doing that. The token curve changes because of all of these things. Frequency goes up, context length goes, multi-step reasoning because of all of the things you need in here. Every single thing will be on there. Every single thing that is the next function. When the consumer AI interface problem is solved, token usage should explode. AI moves from being something people occasionally summon to something that is working in the background. So it goes through before the iPhone mobile data existed but usage was constrained by friction. You have to go back and think about what happened with bandwidth and what's going to happen with consumer agents. The problem is what happened with bandwidth again and getting it there was a very slowmoving thing. Once we hit the phone and we're getting closer and I'm going to show you not just this is all the meta side but you have to go look at what happened to Apple stock because Apple's also a part of this because of the phone. So once it happens, these are going to start consuming an enormous amount of tokens. So coding agents were the first visible token explosion and that they would consume a thousand more tokens. This is why once we hit the coding side with Opus 4.5, everyone had to adjust. Consumer agents bigger user base lower initially. Goldman has reportedly estimated that AI agents could drive token consumption up roughly 24 times to around 120 quadrillion per month. FSD robo taxis not tokens in the API sense but enormous video compute. Uh Elon Musk has talked about this to be able to get to the point where this works completely. have a lot a lot of work to do and this is why I went through a lot of things with the visual l the VLMs uh in terms of the instead of the LLMs humanoids again a combination of the VLMs but the step function is enormous in terms of what you're going to need compute-wise if you haven't gone through it yet you should we have a lot more demand coming for compute so again coding agents then we get consumer agents this is why we're in the first inning right now. This only started the beginning of this year. And you can go look at Dylan Patel's answer. It's only been five months. We are in the beginning stages of what it'll take five years to get to here. If you just do it on a time basis, we're barely in the first inning. It's just going to continue to go. The compute demand is exponential because each wave changes the unit of consumption. It grows exponentially because every capability threshold creates a new consumption surface. This is why humans can't keep up with this. We are used to seeing linear change. We are thinking of memory in a linear framework. They wrote a positive piece on meta. Uh I'm getting more interested in this consumer agent thing and the upside potential. And in particular, I just want to bring up this. This is Apple stock price as we got to all-time highs. We actually closed at all-time highs yesterday on a closing basis. That there is when they did their AI Siri announcement. It made an all-time high. It sold off. That was a sell the news event. That was something that to me looked bearish except for the fact that we're above the 200 day. So, we trade down, not just one day, but two days. We have a very weak bounce. Then we go down and on this day, this is the day that they announced they're raising prices 20% because of the memory prices. We know there's going to be less handsets sold. So, what happens? Goes right back to all-time highs. When I start seeing that kind of activity and I'm reading about consumer agents coming and this stock is expensive on a PEG ratio basis, I'm starting to think there's a lot more to this story than what is going on. Now, if you don't buy into the stuff I just said in terms of uh all of the things about the future, you don't believe in humanoids, you don't believe in in consumer agents, you think it's going to take years to get there, you don't think Meta's doing any of the things I said for these reasons, I'm just getting sick of hearing that there's too much capacity when Microsoft, this is all in the most recent earnings period, strong customer demand continues to exceed capacity. Google's allocating compute under a robust ROIC framework, meaning they're only allocating places where they can get back revenue. Uh, Trrenium largely sold out. Tranium 3 nearly fully subscribed to much of tranium for already reserved. Meta said it's compute constrained and that it had underestimated compute demand. Again, this is from the last earnings. So, this is not long ago. The backlog in terms of orders that they have, commitment quotes, 627 billion for Microsoft, 462 billion for Google, 225 billion for Amazon, 638 billion for Oracle. The demand is there guys, it's already there. Uh if you don't believe that, let's just look at the prices of the uh the GPUs. As Dylan Patel talked about in the interview, he said CPUs can run to a price where they start to actually run in GPUs. We just don't have enough. We still have we we still are seeing the need for them as we're building out data centers and it's going to continue. So rental rates, we're just not seeing it there. Uh this is the way o bullish oil curves are supposed to to look. Um the rental curve again. So this is from the semi analysis guys just using the silicon data according to our data compute market continues to tighten. Um in terms of the capex buildout and the latest forecast semi analysis is at 11.1 trillion well above Goldman's 7.6 trillion and pretty much everyone else has no figures that extend that thus far. Here are their numbers. So again, you get this big step function. It's g the second derivative kicks in. We're not going to get a doubling. We're going to get a slow every year is going to be slower growth, but the numbers are enormous. The spend is enormous, and it's just going to take all this time. JP Morgan put something out this week. Again, multiple ways to look at it. Uh everything is bullish. So when you go through stuff and you realize that all of this stuff is bullish and we're in a race around the globe, it's just not slowing down, guys.
Um, in terms of the token usage, uh, I think this was a good article. So Aaron Levy put this out. Great post on how to think about open source AI and applied AI layer. Two things will always be true in AI. Frontier intelligence will likely remain at the forefront. Again, think about what Dylan Patel said of solving brand new use cases and often be used for orchestrating and planning in any type of complex workflow. I don't know how this is a a a shock to people. The the the best intelligence will always be used by enterprises for at least some of what's going on. And in many cases, it's going to be cheaper. And that's where the margin is. At the same time, as use cases become mature and predictable and become workflows, yeah, you'll use lower or close models. Yeah, I'm already doing that. So are most people. This is not something new. If you want to go read the story, this is from uh Jesse Zang. And again, we run 90% of our workloads on open-source models instead of open AI or anthropic. This is consistent with most of the hyperrowth app companies. Enterprises won't get to this number for a while. They're still going to have the best models. They're still going to be trying to focus on this stuff. But eventually, as they get more sophisticated to get there, enterprises spend as a whole is moving the opposite direction as AI native firms. Open source models just fell to 11% of enterprise spend down from 19%. The piece argues that the popular native open source is eating enterprise is directionally right over the long term, but wrong in the current spending data. Closed models dominate early stage discovery. Open source wins the later stage production. more in token budgeting. Same type of thing from uh from semi analysis, Dylan Patel's group. And again, you can just see that everyone is spending more on AI. It's not changing. We're not seeing some fall. Now, Palanteer is talking their book, but I think this is worth the read. Uh they put this out this week. He's been all over trying to argue that you want to own your own stuff before anthropic and open AI steal it. Uh obviously, he's talking his own book. It's still a name in my 100 name portfolio. I still own it myself, too. Um, it's been a long time. I It's I bought it around 130. It's still around 130. Uh, and I bought it back in I think the first quarter earnings report, which was Februaryish, I believe, uh, early February when I was on on trip. So, it's been consolidating and it looks like a consolidation, but I happen to agree that a lot of places are going to take Palunteer seriously and actually do what they're saying. So, worth going through. Um, I'm not going to read all this because you guys can go through it, but I do think there's a lot of good points in here that are in that are important. 5.6 came out this week. Uh, barely have used it so far. Very impressed. A lot of the feedback's been great. I'll spend time on it. Uh, this story has me interested. I haven't spent enough time on it. Uh but basically we're starting to get to the point with AI where the conscious side is coming up. Uh I think it's very interesting. If there's anything useful there, I will bring it, but I haven't spent enough time. I've listened to the Moonshots episode from this week, which there was nothing in it for me to bring to you guys. Uh and they talked about it and I have had some I've watched some videos and stuff, but nothing big.
Uh, I'm getting back here because this is the beginning of something that for all of you is a must. Very seldom do I say that. I know most of the people listening uh have views on Bitcoin. They have views on crypto. They're not the same. The traditional finance people, I'm sure, are more on the negative side and the crypto people that watch this video are more on the positive side. Well, here's the deal. You're past the point now of ignoring it. And the AO macacro nexus was two things. It was the disruption that would come from AI on the traditional world. How it would impact economics, government decisions, how it would impact earnings, how it impact the stocks. All of that has happened. Every single one of the things and the reason that I chose to do this has happened. Once we got to consumer agents, the next phase was that the financial guardrails would change and it would be one of the most important things in the history of the world. We have been on a banking system this way for a long, long time. Swift has been around for a long long time. This is all ending. It's all changing. It's all morphing. And there is money to be made on this and lost. And that is what the next phase and why I'm starting a new YouTube channel specifically starting from the crypto side spending at least 20 of the 30 minutes I expect to do this on what's happening with inside crypto so people can learn how to con speak in crypto how to speak in tokenization how to speak in stable coins how to speak in the new payment system. So Matt Hogan, who I enjoy very much, uh, and was on a panel with him once before, he posted this and it just got me going. And this was on Wednesday. Did and this is related to a Scott Bessett speech that he did, I believe, at the uh economic the New York economic club. Within that speech, he said, "Digital assets, stable coins, tokenization, new payment system will help to shape the future of money." He referenced that El Muhammad Alerian calls it a remarkably important speech. I posted this today. I think everyone whether you believe in crypto or not needs to read both the Scott Bess impeach and the Muhammad Alan New York Times oped about the importance of the speech for Wall Street. You can no longer have no view on crypto. You cannot just call it speculative and walk away. This will disrupt businesses, public stocks. we will have tokenization. So the point is do your homework now. If you don't believe me and
You've ignored all this, then go read these articles. Go read the speech from Scott Besson. Here's the uh, the op-ed from Muhammad Alan. Argues Scott Besson has articulated a new, a new US economic doctrine. This administration is in place for the next two and a half years. His main point is that Wall Street and economists have not fully internalized how big this shift is. Instead, the US now sees economic security as inseparable from national security. That means trade, investment, technology, standards, supply chains, tariffs, exports, and payment systems and the dollar's role are all becoming tools of strategic power. The old globalization framework is over. This is really critical.
The AI part, they put the AI action plan. They have focused on getting the clarity act through. It's not through yet. The genius uh act went through last year on digital asset payments. Best at Alian's point is crypto is being reframed from a speculative sideshow into a tool of US financial statecraft. The future of money will be shaped by digital assets, stable coins, tokenization, and the new payment systems. And the US should not let that infrastructure be built elsewhere. This is the same thing they have said about AI. So if you are doubting it now at a time when Bitcoin and the crypto market has no positive sentiment and as I go through, very reminiscent of where we were in AI a year and three months ago when I was buying Micron below 100 and people were telling me there was nothing happening. I wrote a paper on inference in May. Crickets. Best in Stockman appears to treat dollar-based stable coins as a way to extend the dollar. Do not fight digital dollars. Regulate them. For Bitcoin, the implication is different but complimentary, which I'll get through. As an investor, is it important to do your homework on digital assets, stable coins, and how they fit in the future? I just literally went through and asked uh an LLM on this. What happens to the financial system when money, collateral, payment, settlement, and eventually AI, agent, commerce all become programmable? That's really where we are at this point.
And the reason that I want you guys to do that homework is that I also want you to go back after you've read Muhammad Alan, after you've read Scott Besson, and then go back to Mark Andreessen and what he posted on January 21st, 2014 titled "Why Bitcoin Matters." Eventually, mainstream products, companies, and industries emerged to commercialize it. Its effects become profound. And later, many people wonder why its powerful promise wasn't more obvious from the start. What technology am I talking about? Personal computers in 1975, the internet in 1993, and I believe Bitcoin in 2014. People didn't accept the internet as a big thing, meaning accept that it was real, until effectively the iPhone came out. That was when everything became real. We were in a bare market in tech from the dotcom bubble. So, Bitcoin, we're now 12 years ahead. This was 14 years ahead. He's talking about personal computers. You guys can go back and look at when they were accepted. To finish up further, there is no shortage of regulatory topics and issues that will have to be addressed as part of Bitcoin since almost no country's regulatory framework for banking and payments accepted a technology like Bitcoin. Well, we've made it through this, guys. The US government, the leader of the free world, not only won an election on it, is putting regulatory uh uh framework in, but they're doing it at the same time as AI. But I hope I've given you a sense of the enormous promise of Bitcoin. Bitcoin offers a sweeping vista of opportunity to imagine how the financial system can and should be worked. This is the most important thing. Reimagine how the financial system can and should work in the internet era and a catalyst to reshape that system in ways that are more powerful for individuals and businesses alike. Andreessen told investors why Bitcoin mattered before the state understood it. Besson is now telling policymakers why it matters. Agentic AI commerce is the accelerant. Once AI agents become transacting, consuming compute, managing subscriptions, settling invoices, moving it, the payment system becomes the operating layer of the agent economy. In 2014, Andreessen saw the internet's need for native value transfer. In 2026, Besson recognized it. Guys, this is a no-brainer. It's a no-brainer to go spend the time on it.
Jeremy Grantham is doing rants about Bitcoin, saying it is worthless. He has, again, I, I don't know how to say this anymore. People do not pay attention. They were saying the same thing about AI. It was hallucinating. I was on trips last year hearing about how the hallucinations would always be presented. It was garbage. It would never be adopted. It would never go through. So I wrote this paper in May 15, 2025, "The Inference Inflection: Where Real-Time Assets Meet Real-World Opportunity." We're just getting started with the upcoming rise of AI agents. This is in May of 2015. AI-native phones and PCs and embodied AI like self-driving platforms and humanoid robots. Inference demand is poised to go parabolic. The implications for semiconductors, memory, networking, power, and edge computing are massive. If Q1 earnings, this is Q1 of last year, told us anything, it's that scalable inference infrastructure has officially arrived. Crickets last year when I wrote this piece, and I'm sure crypto right now gets crickets. Now, if you guys want to go read stuff from March of this year, "Why Cons When Consumers Become Agents: The Open Claw Gateway?" Well, that proved to be right for the agentic side. Tokenization, when ownership becomes programmable. You guys can go look at this on your own. This is all what's coming, guys. I have written about it now for the last year and a half on is all on Substack, Stripe sessions, and the coming agentic commerce economy. The reason this is on Substack so everyone, traditional finance is realizing is because most people don't pay attention to crypto. And so what I do on the subscriber wall, what I do in the stuff that I do with 22V clients, I don't talk about crypto at all. For you, those of you on the crypto side, I know you buy into this. I know you get it. But all of these things, this is from June 30th. Stripe sessions and the coming agentic commerce. They're showing you what's happening. It's already going on from public. Well, these guys are not public, but it's there. This is why I've shown this. This will be part of the video. This will be the main framework around it. The blue line here is the 40-name index. 40 names to create the crypto index that I've created. These are broken up into eight verticals, which are effectively sectors. If you go through them, there are 34 that are tokens, representing the private world or the crypto world, and there are six that are now public companies, things like Circle, equal weight. It is an overlay with Bitcoin. So at the end of the day, I will continue to say Bitcoin is the S&P 500 of the future of tokenization. You can believe it, you can doubt it, you can do whatever. But for all of you that are doubting it, particularly the RIAs and FAs that have your money inside public companies and inside real estate and inside all of these different things, the capital structure of the world is changing. Scott Besson has told you that it is changing, and it will be represented by this stuff. This is where you're going to find the growth assets. This is where you're going to find the things that are going higher. This is what I will be putting together for the next video thing on YouTube, which will effectively be like this, but it'll be starting from the place not of the AI side. It'll be where AI meets stable coins, tokenization, and the new financial guardrails. That is what I'll be going through. That is what I'll be showing you.
Now, to finish up, Bitcoin still a bare market. Technically, I showed you guys last week. I, it, it traded well again this week. We made the low here. I had divergence signals. We're still there. The 50-day is the next line in the sand, as far as I'm concerned. We get above there, and then we'll worry about the 200-day. Uh, that's what I'm looking at. Doge, which I told you, and for those who don't care about Doge, Doge is retail energy. That's what I care about. I want to see it get above the 200-day as well. We're way below. We didn't even get above the 20-day yet. We did on Bitcoin. So, there's no energy yet. Retail is not comfortable in this. Michael Saylor did something that anyone said if he ever did, Bitcoin would be down big. He sold a chunk of Bitcoin, about 3,500 coins. This is when the announcement came out. This is what happened to Bitcoin. Oh, guess what? Closed higher than it was when that announcement came out. So, we had positive response to bad news, just like I highlighted last week.
So, I'm going to finish up the last three slides here. I wrote a paper on "The Academic Fed Versus the Inflation Target of the Future." If you guys didn't get a chance to do it, I highly recommend it. I think it is critical as we come into what I think is the only month that the Fed would really raise rates. That's my belief. Is that if they're going to raise rates and he wants to get the Fed credibility, why not do it uh in July? Uh, it's close to 50%, but to do it, I do think there'll be some sort of wink wink, nod nod, even though he said he's not going to do any forward guidance. I do believe that something would probably come out because I don't think, as his first thing, he wants to disappoint the market, especially with his uh with Donald Trump, who just brought him in. I think there's a lot of things in here for people to read about in terms of the impact that AI is going to have on the Fed. And all I'm using is Kevin Warsh's words from his confirmation hearing and from an interview he did with the ECB forum last week. I also put this out on Substack. If you guys aren't getting my Substack, go subscribe to it for the Trady [clears throat] people. I think now more than ever, reading what I'm writing about on crypto. I connected Bitcoin to that art of unlearning the Fed and why it matters. And the main reason why it matters is this chart here. So this is the rate cuts, the white line here. So, we were expected to have 150 uhund uh, yeah, we were supposed to have the rate cuts in here, and now we've got rate hikes down here. So, we went from a position of rate cuts, we did a cut here, we did a cut here, and then the rest of this is turning it from future cuts by December of this year to hikes. This is Bitcoin overlaid with it. So as they got more hawkish, as we started to see an ending to the cut cycle and fears, this happened. Now, we also, the green line here was growth versus value, which was really the software side. This has broken, and I think the Fed side is going to break. One or two things are going to happen. I don't think if the Fed does hike, I think it's one and done, and I think it's 25. I think you'll probably see long-term rates come down on the back of it with the Fed credibility. If they don't do it, which I don't think they will, uh, I, I'm going to stick with the fact that they're not going to do anything, and I'm going to stick with my paper. Uh, and I think at that point, you can have some tailwinds for Bitcoin.
And that's it for this week, guys. Um, again, thanks to everyone. Uh, I'll be up in Maine the rest of the summer, and I'll keep you up to date. See you next week.