El desplome de la velocidad de la IA está creando la próxima oportunidad

[0:00] Um,
[0:03] stocks just continue to uh defy whatever
[0:06] goes on. Um, I'm going to go through
[0:08] that today. I'm going to remind you guys
[0:10] about bubbles, parabas, and speed
[0:13] crashes and kind of give you some
[0:15] insight into uh what we've seen the last
[0:18] four to six weeks with inside AI. And
[0:20] I'm going to say now that uh the
[0:23] midcycle slowdown is ending. And it's
[0:26] not that we're not going to see the rate
[0:28] of change, but now this is no longer a
[0:30] surprise. Uh positioning has shifted and
[0:35] I'm starting to hear people that are
[0:38] trying to be bearish start to come out
[0:39] of the woodworks and highlight a whole
[0:42] bunch of things which I'll go through.
[0:44] Um it's been a sentiment and a technical
[0:46] cleanse. Uh I'll go through why that is.
[0:49] I'm going to go through the token index.
[0:51] I I've been called more on this than
[0:52] anything over the last three weeks both
[0:54] from institutional from retail and from
[0:57] very sophisticated tech people on why
[0:59] this token index doesn't matter. Uh Meta
[1:03] Apple interesting week uh I wrote a
[1:05] paper on meta. I'm spending a lot of my
[1:08] time on consumer AI agents which I think
[1:10] is the next theme. I've talked about the
[1:12] application side but I am going to start
[1:14] to go through uh what is happening more
[1:16] and more and what I think is going to be
[1:17] a major story for the next 12 months.
[1:20] Uh, I'm going to remind people about
[1:22] exponential demand versus linear supply.
[1:24] Scott Bessant did a speech and I spent a
[1:28] lot of time on it this week with regards
[1:30] to crypto. I'm going to basically start
[1:33] to give you guys what I've talked about
[1:34] now for the last month, which is
[1:37] beginning a new um part of AI macro
[1:40] Nexus, remind you what this was all for,
[1:43] but combine it with what Scott Besson
[1:45] said, what Mark Andre said regarding
[1:47] Bitcoin, and this will be a new in the
[1:50] same format as this, but very crypto
[1:53] forward.
[1:54] And we'll do a recap on what happened
[1:56] with crypto this week. First of all, um
[1:58] just in in general, uh I've had a lot of
[2:01] success with RAAS. For those of you who
[2:04] are financial advisors who want to spend
[2:07] time uh on how the work that I'm doing
[2:10] can be expanded more than just this
[2:12] YouTube video and more than just a
[2:14] subscriber site, uh I'll show you where
[2:16] you can reach out. But uh between that
[2:18] and especially the agency side of people
[2:20] working on knowledge brains uh it's been
[2:23] a uh a really good three weeks of
[2:25] helping people not only stay on signal
[2:28] with AI but also start to uh use it more
[2:30] and more and I think that's helping. So
[2:32] I just want to remind people what the
[2:34] name was because this video this week
[2:36] gets into why it's AI macro nexus. Uh AI
[2:41] is no longer just a technology story.
[2:43] It's becoming the dominant macro force
[2:44] of the cycle. It changes productivity,
[2:46] labor, inflation, capital spending,
[2:48] energy, demand, digital assets, monetary
[2:49] policy, market structure. The nexus is
[2:51] the intersection of these forces, and
[2:53] it's the place where investors connect.
[2:54] The reason I bring this up, um,
[2:56] everything that you're dealing with when
[2:58] you're reading something on
[3:00] productivity, on labor, on inflation, on
[3:02] the Fed, on anything, AI is a part of
[3:07] every single decision that's being made.
[3:09] And that's why you have to be on top of
[3:11] this. It's changing the world in front
[3:13] of our eyes. Here's one example of it.
[3:15] So, this is uh something I've shown a
[3:17] bunch of times. I did a subscriber uh
[3:19] video update where I started to go
[3:21] through the fact that I think the
[3:22] slowdown is reached a point where we've
[3:25] done enough damage. Um, and this is
[3:27] where the damage has been. This is the
[3:29] 60-day realized V of the technology
[3:32] macro uh technology momentum factor from
[3:34] Morgan Stanley. The white line here is
[3:37] that. So, we've gone up and I've shown
[3:38] this for a while. I showed it when it
[3:40] was around 70. We're at 87 now. This is
[3:43] through today on Friday. I'm doing this
[3:45] Friday night. This is overlaid with the
[3:48] S&P 60-day realize V. So, historically,
[3:52] there's a correlation that does exist.
[3:54] Meaning, if you're seeing tech momentum
[3:56] V explode like this, you're usually
[4:00] seeing the S&P V follow it. Now, the one
[4:02] thing is tech has been going higher for
[4:03] a while. This is why what we're doing is
[4:06] this bubbles, parabas, and speed
[4:09] crashes. This is what's going on. That's
[4:12] why you have to be on top of this
[4:13] because this is a structural bull
[4:16] market. It is based on token demand.
[4:18] Everything that I built everything on is
[4:20] on token demand. This entire video by
[4:23] the end should give you complete comfort
[4:26] that we're not in a bubble that the
[4:28] compute there is no excess capacity
[4:30] because Meta is selling some off.
[4:31] Hopefully you guys get there and you
[4:33] realize that if that's true over the
[4:35] next six months you're getting an
[4:36] opportunity where this has gone higher.
[4:38] has flushed people out across the globe
[4:40] on retail, on hedge funds, everywhere
[4:43] that people were using leverage and
[4:46] probably overexposed. We've seen that
[4:48] now come out mainly because the V's been
[4:50] too high for people not to do it. S&P, I
[4:54] don't know how you can look at that
[4:55] chart and not be bullish. Um, I told
[4:58] last week I showed it as a triangle.
[5:01] You know, we get bad news out of Iran,
[5:03] it still goes higher. We whatever you
[5:05] want. We get sell off on Samsung on good
[5:07] numbers. It still goes higher. This
[5:09] triangle in the NDX is underperformed.
[5:11] Again, tech is is lagging. We're getting
[5:13] more exposure on financials, more
[5:15] movement in healthcare. Uh but we did
[5:17] finally start to see some tech bounce.
[5:19] But again, this is a bullish looking
[5:20] chart.
[5:22] Uh this is a reminder if you guys didn't
[5:24] see it, I did this video back on May
[5:26] 10th. Uh and the focus of this, which
[5:29] was right around the first leg lower, is
[5:32] about the way we should expect this to
[5:34] trade. uh when you have bubbles and
[5:36] parabas on the way up, you're going to
[5:38] have speed crashes on the way down. If
[5:40] this is all with inside a structural
[5:42] bull market, which I'm going to continue
[5:44] to say, this is a structural bull
[5:46] market. You want to go with the river of
[5:48] a macro bull market, especially when
[5:50] you're in the first inning where you're
[5:51] in the first inning of AI token demand.
[5:55] Regardless of what anyone says that it
[5:57] can't be the first inning based on the
[5:59] charts, these charts were about people
[6:01] being wrong on AI for two years saying
[6:03] it was a bubble. It hallucinates. It's
[6:05] not going to do anything. They had to
[6:07] play catch-up once the earnings came in.
[6:09] Then we hit the second derivative and
[6:11] it's not going to be as easy from here.
[6:13] So, it's not that this trade is going to
[6:14] do what it did before. We are way past
[6:16] the first inning of the speed of change.
[6:19] But do I think that you're going to get
[6:20] another five bagger in Micron and in SK
[6:23] Heinex in these things over the course
[6:25] of the next five years? It's very likely
[6:27] you will, but that's less than the 20
[6:30] bagger that occurred from March of last
[6:32] year till now. So, we're going to slow
[6:34] down because now everyone is aware of
[6:36] it. So, Samsung uh released premill
[6:41] preliminary quarterly profit up 19fold.
[6:45] Micron fell that day. Samsung was down
[6:47] that day.
[6:48] Here's the chart and this is the first
[6:51] thing I want to show. I saw three
[6:52] separate YouTubes this week on the
[6:54] market where they said this was a news
[6:56] failure. See, good news came out. It
[6:58] beat wound estimates and it sold off.
[7:01] The problem is we had already sold off.
[7:04] I don't subscribe to the fact that when
[7:06] you've already had a correction, you've
[7:08] already had weakening where we're
[7:09] unchanged over the course of this when
[7:11] you get good news that you should sit
[7:13] there and say that this is bearish. Is
[7:15] it something that showed that the
[7:17] unwinds and the deleveraging were still
[7:19] on? Yeah. But if you ask me what the
[7:21] range is going to be and where the
[7:22] support's going to be in this going
[7:24] forward, I will tell you that I believe
[7:26] this high from February, this low from
[7:28] here, and this is going to be around the
[7:30] bottom end of the range. And I think
[7:32] we're going to be significantly higher
[7:34] before the end of the year because of
[7:36] what I'm going to show you.
[7:38] This is the problem is that everyone is
[7:40] looking for a correction. So if I told
[7:42] you that over the 10 days in the two
[7:45] days after that, this was the rolling
[7:47] 10day rate of change in Samsung. Okay,
[7:50] it was the largest since the middle of
[7:54] the COVID panic when we shut down the
[7:55] world was 25 26%. This was 21. The only
[7:59] other time that it was this big on a
[8:01] 10-day basis was just after Lehman
[8:03] Brothers. So this is a speed crash. It
[8:06] didn't feel like it. You don't think it
[8:08] looks like it, but within a bull market,
[8:10] this is a big fall, particularly when
[8:12] you had these types of upsides that were
[8:14] much bigger. So, these are the rolling
[8:16] rate of 10day change. You had one at 35.
[8:19] So, this has been a correction with
[8:21] inside a bull market. Now, here's where
[8:23] we are on a PE basis for next year. So,
[8:26] everyone who wants to fade this, I
[8:28] completely agree. If you believe in the
[8:30] cyclicality of memory, you should
[8:32] absolutely sit there and not pay
[8:34] attention to this. If you believe that
[8:36] the demand is going to stay at a very
[8:38] high level and prices are going to stay
[8:39] at a high level for the next three
[8:40] years, this is way too cheap and argues
[8:44] for a much higher PE than is normally
[8:47] there. If you guys look at this and pay
[8:48] attention, my PEG ratio, when we're
[8:51] looking at the earnings growth relative
[8:54] to the PE, they've already discounted
[8:56] that this is going to slow down at these
[8:58] levels. They've already built in a
[9:00] bearish outcome in terms of multiples.
[9:03] Here are how ridiculous the numbers are.
[9:05] And this is why when people say this is
[9:07] just like every other time, Samsung's
[9:09] operating profit in 2026 could be over
[9:12] $217 billion.
[9:15] That's where analysts believe. Okay.
[9:18] The company could generate more
[9:20] operating profit this year than its
[9:22] cumulative operating profit over the
[9:24] past 40 years.
[9:27] Don't treat this as a one-off event
[9:29] where AI is just doing this and
[9:30] everyone's over buying and then we're
[9:32] going to run into a problem. It's silly.
[9:35] Um, we're not even in the first inning
[9:37] of token demand. I'm going to keep
[9:39] saying it because nobody believed in
[9:42] token demand until we got into January
[9:44] of this year. And I will highlight in
[9:46] some podcasts from Dylan Patel from Semi
[9:48] analysis how much his spend has changed
[9:50] and where he was in December of last
[9:52] year to today. That should give you guys
[9:54] a sense as to someone who's very
[9:56] sophisticated on a growing business who
[9:58] understands how to use AI and his team
[10:01] does and how he went from $100,000
[10:04] budget to $11 million over the course of
[10:08] six months in terms of what his
[10:10] annualized revenue spend is going to be.
[10:13] So Korea's Cosby valuation falls to the
[10:15] lowest since the global financial
[10:17] crisis. Those are the types of things
[10:18] you see at tops.
[10:21] Joking. Um, here's Salesforce.com.
[10:24] Everyone who is basically bearish on
[10:29] memory are the same people who wanted to
[10:32] buy Salesforce while it was coming down.
[10:36] So Salesforce is now down to a 10p to
[10:40] 2028, 11 PE to next year. So again,
[10:43] we've got a 4 PE on Samsung with their
[10:45] earnings growing this year 524%
[10:49] and for next year it's 43. They haven't
[10:51] come close to meeting the estimates yet.
[10:53] They've been above it by significant
[10:55] numbers. And yet somehow or another,
[10:57] this stock is trading at an 11p.
[11:00] This is a problem. This is why from the
[11:04] beginning of the year, I said you want
[11:05] to be short abundance. Anything built on
[11:08] code that can be replicated by a
[11:10] computer. And you want to be long
[11:11] scarcity. Anything that the spenders
[11:17] are spending on that is physically
[11:20] constrained by linear humans. That's it.
[11:24] Humans build this, computers built this.
[11:27] That theme will continue to permeate
[11:29] because that's about the exponential
[11:30] demand. So again, I put this in on June
[11:34] 8th. I started talking about the fact
[11:35] that I was getting out of Micron in the
[11:37] middle of May. Again, I got out way
[11:40] before the top, but that was not my goal
[11:41] to pick the top. My goal was to be in a
[11:43] position where I'm probably going to be
[11:46] buying some back at some point soon,
[11:48] especially after the SKH Heinix deal
[11:49] that went through today, just because
[11:50] I'm hearing people be bearish. And I do
[11:53] believe that 6 months from now, we will
[11:54] be significantly higher than we are now.
[11:57] Now that we've had some sort of
[11:58] pullback, but also since we got another
[12:00] earnings report and I'm no longer
[12:02] worried about some of the things that I
[12:04] thought could be an issue in terms of
[12:06] the bottlenecks, it looks like we're
[12:07] just going to be in a situation where
[12:09] things are going to go at a slow pace.
[12:12] Mainly because there are constraints
[12:14] that are built in from humans that are
[12:15] keeping this from being a bubble.
[12:18] Believe it or not, it's the supply side,
[12:19] the TSMC's that won't overbuild, the
[12:22] fact that the gas turbines and all this
[12:24] make it very difficult. We're finding
[12:26] solutions to every problem on the power
[12:28] side and they're finding ways to get
[12:30] this done. So I just think that that's
[12:32] going to keep prices at a higher level
[12:34] because we're not going to be able to
[12:35] get capacity on time. Uh the econ so the
[12:38] navigating the AI midsight cycle
[12:40] slowdown. I still believe with the
[12:41] second derivative change but now that is
[12:43] consensus in my opinion.
[12:46] Here's what we did and this is part of
[12:47] it. So you look not only for what I hear
[12:50] from people but you want to see how much
[12:51] damage we did. This is what happened.
[12:54] We've given back now 50 or almost 50% of
[12:58] the techo rally from Opus 4.5. And the
[13:01] reason I care about Opus 4.5 is this is
[13:04] when the change happened. This is when
[13:07] people started to acknowledge it. This
[13:08] is when people started to invest. And I
[13:10] want you to look at this because it does
[13:12] look a lot as I go through this from
[13:15] February down then up looks a lot like
[13:18] the token index as I get back to later.
[13:20] And I'm not worried about the token
[13:21] index, but I think the token index, uh,
[13:23] people have started to use it for
[13:26] something it shouldn't be used for. I
[13:27] think that should go to zero over time.
[13:29] Um, and more importantly, I don't think
[13:30] it should be rising. I never think it
[13:32] should be rising, and it technically
[13:33] isn't, which I'll go through as well.
[13:35] When these stories come out, it's
[13:37] usually a good sign. Systematic long
[13:38] short managers dropped 2.1% last week
[13:41] through Thursday after declining 3.1%
[13:43] over a 5day stretch. That was their
[13:44] worst since December 23. On most of the
[13:47] numbers that I've seen, they've given
[13:48] back. And these are again systematic
[13:50] long short managers. So these are quant
[13:52] managers. They've given back about a
[13:53] third of their year. They're still up
[13:55] for the year. This is not some major
[13:57] panic. I talked about this feeling a
[13:59] little bit uh on the webinar that I did
[14:01] this week for the subscribers about this
[14:03] feeling a little bit like 2007. Um and
[14:06] that it would end at some point around
[14:08] the same time it did there, which was in
[14:10] the first 10 days. I still think there
[14:12] might be a little bit left, but I also
[14:13] think the mutual funds have proven their
[14:15] case of being underneath, which I
[14:16] thought as well as they took down a lot
[14:18] of this SKH Heinik deal. Um, fundamental
[14:22] managers also fell and we know that the
[14:25] pods had to close down as I talked about
[14:26] because of how high the V was. Here's
[14:28] the deleveraging on the retail side.
[14:30] Even though we only had about a 23%
[14:33] correction on the uh SKH Highix and
[14:36] Samsung side down to about a 30 uh 38
[14:40] and again this is back to Opus. The
[14:41] retail side using the two times levered
[14:44] gave back 62%.
[14:46] So I think retail's been flushed. Morgan
[14:49] Stanley put this out on their AI
[14:51] beneficiaries and you can see the
[14:53] massive outflow that occurred and we've
[14:56] had three weeks of fairly large uh
[14:58] outflows in terms of the net flows. Jeff
[15:01] Degraphth, I like to combine this stuff.
[15:03] This was a great one. Nearly 87% of S&P
[15:06] 500 semiconductor issues registered and
[15:08] oversold on a 14-day basis. He uses
[15:11] these lines. You can see these have all
[15:12] been very good bottoms except for this
[15:15] one here. all the others uh except for
[15:17] these two little things here. So, could
[15:20] it go lower? Of course, it could. But at
[15:21] least you're getting also that on that
[15:23] side. Here's what happened to my side in
[15:26] terms of the technical sheet that I put
[15:27] out. This was from May 29th, the week
[15:29] ending. Look at these numbers. Percent
[15:32] 50day, 100, 91, 98. Look at the average
[15:34] RSIs. Average RSIs. Everything was above
[15:37] the 50 level. Here's where this week's
[15:40] came in. Look how low these numbers are.
[15:42] Percent above the 50-day. We've really
[15:44] sold off here. You have one that's still
[15:46] above, but the average RSIs
[15:49] are now there and we still have
[15:51] constructive none of the tech patterns
[15:53] have changed. I'll go through that as
[15:54] well. The score, they're good, meaning
[15:57] this is a good entry point. Um, I show
[16:01] this uh again, I've shown how to do this
[16:03] and I will finish this video for you
[16:04] guys in terms of how to be able to do
[16:06] this on your own using Claude Co-work by
[16:08] just updating the files. But this was
[16:10] 20-day of my 100 name thematic
[16:12] portfolio. 73 of the 100 names were
[16:15] above the 20-day on May 8th. We got down
[16:17] to 21 on Wednesday when I did the uh uh
[16:21] sub subscriber webinar to basically an
[16:24] update to kind of give people a sense as
[16:26] to where we are. Uh I like to see the
[16:29] bump at the end of the week. We went
[16:30] from 21 to 31. This definitely seems
[16:32] like an oversold level. I mentioned last
[16:34] week that when we get or I mentioned
[16:36] this week when we get below 40 I get
[16:38] very interested on the uh percent above
[16:40] the 50-day and most importantly above
[16:42] the 200 day it's back to the same level
[16:45] it was kind of ending on the 26th but
[16:47] these levels 50-day rising still at 76
[16:51] and the 200 day rising which is the
[16:53] critical one which is the bull market
[16:55] structural situation is still up at 86
[16:58] of the 100 names. We saw massive inflow
[17:01] in the SOXX
[17:04] and SKHEX today uh offering more than
[17:07] seven times overs subscribed. It also
[17:10] traded up. Here are the facts from the
[17:12] memory side over the course of the last
[17:14] three months
[17:16] uh just in terms of what's gone on.
[17:18] Memory chip crunch to persist until
[17:20] 2030. This was back during his earnings
[17:23] in March 16th. If you thought anything
[17:26] would change, this is what he said
[17:28] today. Okay, the memory chip shortages
[17:30] that are rolling the computer that are
[17:32] roing the computer car and device
[17:34] markets will probably persistent beyond
[17:36] 2030. So, we went from until 2030 to
[17:39] beyond. Customers are signing long-term
[17:41] contracts. Hadn't seen that before, but
[17:43] remember Micron Inc.'s long-term supply
[17:46] agreement is worth $und00 billion. Says
[17:47] it has no idea when the RAM crisis will
[17:49] end. This was less than a month ago.
[17:51] Actually, it was less than three weeks
[17:53] ago. and Dylan Patel
[17:56] memory capacity is only growing 20 to
[17:58] 30% a year for the next three years.
[18:00] This is the human capacity side. Demand
[18:03] is doubling. This is the exponential dam
[18:06] demand. Memory prices are going to keep
[18:08] soaring. This is from an interview which
[18:09] I will highlight which came out today.
[18:12] Remember, Elon Musk is building a
[18:16] terrafab
[18:18] because he says there will not be enough
[18:21] memory or chips for his humanoids
[18:25] and the demand that he sees by 2030. I'm
[18:28] going to get into that as we go on. I'm
[18:30] showing you these things because this is
[18:32] the reality of where we are. So, I want
[18:34] to make this clear because I had I've
[18:36] had multiple people now and I think this
[18:38] is because there's a bunch of people on
[18:40] next that have said this is just like
[18:41] fracking. this will blow up.
[18:44] The only thing that's just like fracking
[18:46] is the amount of people that are saying
[18:47] it's just like fracking. So, here's a
[18:49] normal cycle before
[18:53] AI. And I say before AI because this is
[18:56] a typical demand cycle for hardwardriven
[19:00] demand. So, hardware driven demand
[19:01] meaning physical stuff. So if the demand
[19:04] is driven by physical output and the
[19:07] memory is a physical output, it's really
[19:09] hard to get demand significantly away
[19:11] from capacity. If the demand is now
[19:14] digital and it's tokens and it increases
[19:18] purely based on the productivity of the
[19:21] models, but we need more memory. The
[19:24] capacity can't grow any faster. The only
[19:26] way to fix this problem is with an
[19:28] algorithmic efficiency side of which
[19:32] that is such a long shot that at this
[19:35] point to get it to be able to do it to
[19:37] be able to get it out there for every
[19:39] it's just a long shot. Is it a hail
[19:41] Mary? Will it eventually happen? I'm
[19:42] sure it will. I don't think this is
[19:44] going to get up to meet there anytime
[19:46] soon particularly when I go through what
[19:48] else is coming. So you have to remember
[19:50] this is the difference. This is the old
[19:52] world. This is the new world. This is
[19:55] exponential growth. This gets important
[19:57] when you start going into consumer
[19:59] agents, full self-driving and humanoids.
[20:02] These are all bigger demand for tokens
[20:04] than what we have now. Just like chat
[20:06] was here. Chat, we didn't have a
[20:08] problem. We only had a problem when we
[20:10] got into this period here when it became
[20:13] evident that we were starting to hit a
[20:15] part of the growth curve that was out of
[20:17] control. This is why when you compare it
[20:20] to the oil market, it's ridiculous. we
[20:23] had a parabolic or an exponential move
[20:26] on the supply side. The demand side
[20:28] didn't move. So again, for everyone who
[20:31] wants to compare these two, for every
[20:33] smart oil guy or smart person who keeps
[20:35] telling me this is the same as this,
[20:37] I've seen this before. You haven't seen
[20:39] this before because the demand is not
[20:41] coming from human beings. The demand is
[20:43] coming from computers. It's coming from
[20:46] AI agents. It has nothing to do with
[20:48] humans. Humans move on a linear scale.
[20:52] It's very important for people to start
[20:54] to realize this. That is where the bull
[20:55] market lies. That is why I do this
[20:57] stuff. Now, my next webinar will be on
[21:00] July 15th, next week. I will
[21:02] specifically go through names with
[21:04] inside the portfolio that I think have
[21:06] hit the levels that make sense. I will
[21:08] go through the concentrated basket, all
[21:10] 10 names, and give you guys the details
[21:13] on it. This is what the concentrated
[21:14] basket has done now. So, we've hit the
[21:17] 50-day moving average. We did this back
[21:19] in March 2 and then we got the next
[21:20] liftoff, we hit the RSI, you guys can go
[21:23] back and look and when the RSI gets down
[21:24] here, the riskreward has sh has changed.
[21:27] So when here when I thought there was a
[21:29] midcycle slowdown, we've basically gone
[21:31] sideways during this period. Now we can
[21:33] reset and go. I reduced my position in a
[21:37] bunch of these things down significantly
[21:38] with Micron being the biggest one. I
[21:40] kept some things the same. I moved some
[21:42] things into silver and Bitcoin which
[21:43] haven't worked out as well. I will now
[21:45] go back and look at some of these names
[21:47] and pick out the ones with inside the
[21:48] concentrated basket. So, this is the
[21:50] whole point of kind of helping you guys
[21:52] not just on one side but also the other.
[21:54] If you guys follow me on X, I put this
[21:56] out uh on the same day I did the
[21:59] subscriber webinar. Nvidia peaked on May
[22:02] 14th, three weeks before the AI
[22:03] infrastructure trade peaked. So again,
[22:06] if I go backwards and go over here and
[22:09] you look into the middle of May
[22:12] and basically go through, that was when
[22:13] we saw the peak in the rate of change
[22:15] and when we started to slow down. Well,
[22:17] Nvidia peaked beforehand and more than a
[22:20] month before SMH and the memory names
[22:22] peak. Today is set to close above the
[22:23] 20-day moving average for the first time
[22:25] since June 2nd, holding the 200 day and
[22:28] with a MACD buy signal at the lowest
[22:30] valuation in a decade. That is one of
[22:32] the 10 names. So in the consensus
[22:35] basket, Nvidia is one of the names. So
[22:37] again for the RAAS and for anyone who is
[22:40] new to this, if you want to reach out,
[22:42] that's where the website is where you
[22:43] can go find it. You can always reach out
[22:45] to Mark Whailing at22V. He will help you
[22:47] on anything you need, particularly on
[22:49] the RAIA side. This is where I want to
[22:51] get into the token expenditure index
[22:52] because as I showed you before, if you
[22:54] go through the es and flows on the
[22:56] earlier chart I showed with the
[22:58] everything related to this, including
[23:00] the AI themes, I highlighted back here
[23:03] in the middle of May when Nvidia peaked,
[23:05] that this would be a growing risk, not
[23:07] because it was a negative, but because
[23:10] the narrative would start to change and
[23:12] that's what we've seen. The narrative
[23:14] has shifted. So, let's go through this.
[23:17] The first rise in the token index was
[23:19] the coding agent capability shock. Then
[23:21] we had openclaw.
[23:24] So open claw led to more users
[23:27] immediately using agents that happen to
[23:29] be AI native or Chinese.
[23:32] Open claw was a big gamecher but it was
[23:34] a sophisticated level. So most of the
[23:37] token usage went to a lot of open-
[23:40] source stuff. Then the enterprises
[23:42] finally realized that they were missing
[23:44] out. So the FOMO kicked in and every
[23:47] single place whether it's Anthropics,
[23:49] ARR, everything started to see a step up
[23:52] function maxing from there. Token maxing
[23:54] became a word. Everyone was on the
[23:56] exploratory spend side. Then they saw
[23:58] the bill shock and the token index goes
[24:01] down. Not because everyone's bailing out
[24:04] partly because school ended. All of
[24:07] these different things that I talked
[24:08] about. It's when the usage starts to go
[24:10] more towards people that are using open
[24:11] source, which AI native companies don't
[24:13] stop. College students stop. People on
[24:17] Morgan Stanley, Goldman Sachs, places I
[24:19] know well, 20% of the people are out on
[24:21] vacation or working remotely or they
[24:22] don't work on Fridays. Whatever the case
[24:24] is, you're not going to have as many
[24:25] people using stuff as what was here. And
[24:27] the bill shock kicked in. The problem is
[24:30] with this whole analysis and with
[24:31] everything going through it, it doesn't
[24:33] include the amount of adoption that's
[24:35] still happening. We have massive
[24:37] adoption happening. This is not an an
[24:39] index of making or losing money as I
[24:42] will show with the Dylan Patel side. So
[24:45] again, memory shortage exponential token
[24:48] usage going higher every single day.
[24:50] I've heard people argue with me that
[24:52] nobody's going to use the most expensive
[24:54] stuff. That is a bunch of garbage.
[24:58] If you want to go hear someone who talks
[25:00] about this and I think always does a
[25:02] good job in an interview, this came out
[25:04] today. Dylan Patel or one on Thursday I
[25:07] guess late uh Dylan Patel on wisdom tree
[25:11] in Europe good interview I think the
[25:15] highlights that I have here Patel argues
[25:17] the AI ROA debate is missing how quickly
[25:20] serious users are increasing token spent
[25:22] and he highlights again that semi
[25:23] analysis spend went from less than
[25:26] 100,000 now the reason it was 100,000 in
[25:28] November which is a very similar story
[25:30] to me he was paying for everyone at the
[25:33] firm for a 200 $00 Claude model. Once
[25:37] Claude Code started to be able to work
[25:39] after Opus 4.5 and build stuff, that's
[25:42] when they went to a token spend side and
[25:44] everyone started to spend there. By
[25:45] January, it was already $4 million
[25:48] annualized.
[25:50] It stands at a roughly 11 million
[25:52] annualized today and the peak week was
[25:55] equivalent to 14 million. That's for a
[25:58] 90 person firm.
[26:00] Companies are blowing through AI
[26:01] budgets, but the right response may be
[26:03] cutting other SAS or legacy spend rather
[26:05] than cutting AI. The right model choice
[26:08] depends on the workload. Cheaper models
[26:09] for fixed workflows, smartest models for
[26:12] assistant work. This part and I h you
[26:15] have to listen to this. For assistant
[26:17] style work, the smartest model can be
[26:19] cheaper overall because it uses fewer
[26:20] tokens, fewer iterations, and less human
[26:23] time. This is why when you look at the
[26:25] labor market, we have weekly hours. If
[26:28] you can get things done in onetenth the
[26:30] time, but you're spending more money for
[26:32] it, that's the way it should work. So
[26:34] his point being, we are in the stage of
[26:36] enterprises trying to figure out what to
[26:38] do, how much they should be using for
[26:40] workflows, which will be much cheaper,
[26:42] and how much they should be using for
[26:44] assistance style work. You don't hire
[26:46] all people from the most expensive
[26:48] colleges with the best resumes. A team
[26:51] at a firm is broken up by a lot of
[26:53] different areas. You hire younger people
[26:55] which are cheaper than the domain
[26:57] experience people. That is the way AI
[26:59] will work. He goes through memory is no
[27:01] longer a normal commodity cycle. If you
[27:03] don't believe me, he goes through more
[27:04] of the technical reasons. He expects
[27:06] memory prices to keep rising until AI
[27:08] gets enough supply and consumer devices
[27:10] like smartphones and laptops are forced
[27:12] to absorb higher prices. Basically
[27:14] saying the same thing as the Samsung. We
[27:16] have a shortage where the phones, the
[27:19] cars, the PCs, we don't have enough
[27:21] memory. So the demand is coming not just
[27:24] from hyperscalers for data centers. It's
[27:26] coming from everybody. He's bullish on
[27:28] CPU. There's a CPU thing in there. I
[27:30] think it's worth listening to as well.
[27:32] Uh on the power side, he's been the best
[27:36] by far in my opinion on what's been
[27:38] happening. I've used him a lot. I've
[27:39] referenced him a lot. This was another
[27:41] great one. Basically what he said in a
[27:43] summary with me listening to it and
[27:44] going through it, we're figuring a way
[27:46] to get the gigawatts done, even if it
[27:49] means using diesel truck boat train
[27:51] engines converted for power generation,
[27:54] you have to listen to it to understand
[27:55] that if you thought we were going to run
[27:57] out of compute because people are going
[27:59] to stop the data centers, it's not the
[28:01] case. He went through and obviously I I
[28:05] I'm pretty sure he's an investor, but if
[28:07] he's not, he obviously has access.
[28:09] You're getting more and more of this
[28:10] that Anthropic is going cash flow
[28:14] positive or at least reaching operating
[28:16] profit.
[28:18] Again,
[28:20] this just means the margins are coming
[28:21] through and the RAIC should not be there
[28:24] uh in terms of a fear. But I did do this
[28:26] on the AI capex air pocket. If you guys
[28:28] didn't read it because of the vacation,
[28:30] I would go read it. I I spent some time
[28:32] I've spent a lot of time on consumer
[28:34] agents, but I really spent time on this
[28:36] Mark Zuckerberg town hall reportedly
[28:38] acknowledging the Meta Zentic AI
[28:40] progress had been slower than expected.
[28:41] And I make the argument that if they did
[28:44] end up selling any of their capex
[28:48] capacity, it would be because they can't
[28:51] get their consumer agents working the
[28:53] way they want yet. It's coming, guys.
[28:56] And if you want to go listen, this came
[28:59] out one day ago. Meta or two days ago.
[29:02] The CTO of Meta, Andrew Bosworth, was on
[29:04] with Alex Canantrowitz. It's a great
[29:07] interview and it's very simple to go
[29:09] listen. The industry is moving beyond
[29:11] one giant monolithic model. I think he
[29:13] talks about the fact that there's going
[29:15] to be open source and there's going to
[29:16] be the frontier models. Consumer AI has
[29:19] been slow because the products are still
[29:21] too hard to use. This is a really
[29:23] critical part especially for mutual
[29:25] funds and longerterm investors to spend
[29:27] time on understanding how consumer
[29:29] agents are going to change the world
[29:30] which is a dramatic shift. Whatever
[29:32] happened with coding agents which is for
[29:34] the language of agents this is massively
[29:38] bigger. This is far bigger in importance
[29:40] than opus 4.5. So think of it this way.
[29:43] All we've done is increase the ability
[29:45] to code but we haven't found the ability
[29:47] to get through all of this yet. It is
[29:51] incredibly
[29:53] important to understand what that means.
[29:54] Consumer AI has been slow because the
[29:56] product layer is still immature, not
[29:58] because the model's weak. For consumer
[29:59] AI to work, it has to disappear into
[30:02] daily life like search maps, messaging,
[30:04] or the camera. I use my uh Grock in my
[30:09] Tesla this week. I had a conversation
[30:11] with it on the drive, the entire drive.
[30:14] I cannot do that with a phone the same
[30:16] way. It's not as good until we get to
[30:19] the point that you are speaking to it
[30:20] like you were supposed to with Alexa
[30:22] where you can be sitting there cooking
[30:24] and asking it questions on what the
[30:26] recipe is. So you don't have to go to
[30:27] your phone, you don't have to go look at
[30:28] it. Until we get to that point, which is
[30:30] coming soon, it's really hard to get the
[30:32] habit formation. But once it happens,
[30:35] billions of users will start having
[30:37] agents working for them and they will
[30:39] start giving them things to do, but they
[30:41] need the context, the trust, the timing,
[30:42] the permissions. It is extremely
[30:44] complicated and the frontier models are
[30:46] not capable of doing that. The token
[30:48] curve changes because of all of these
[30:50] things. Frequency goes up, context
[30:52] length goes, multi-step reasoning
[30:54] because of all of the things you need in
[30:56] here. Every single thing will be on
[30:58] there. Every single thing that is the
[31:00] next function.
[31:02] When the consumer AI interface problem
[31:04] is solved, token usage should explode.
[31:06] AI moves from being something people
[31:08] occasionally summon to something that is
[31:09] working in the background. So it goes
[31:12] through before the iPhone mobile data
[31:14] existed but usage was constrained by
[31:15] friction. You have to go back and think
[31:17] about what happened with bandwidth and
[31:19] what's going to happen with consumer
[31:21] agents. The problem is what happened
[31:23] with bandwidth again and getting it
[31:26] there was a very slowmoving thing. Once
[31:29] we hit the phone and we're getting
[31:31] closer and I'm going to show you not
[31:32] just this is all the meta side but you
[31:35] have to go look at what happened to
[31:36] Apple stock because Apple's also a part
[31:38] of this because of the phone. So once it
[31:42] happens, these are going to start
[31:44] consuming an enormous amount of tokens.
[31:47] So coding agents were the first visible
[31:51] token explosion
[31:53] and that they would consume a thousand
[31:55] more tokens. This is why once we hit the
[31:58] coding side with Opus 4.5, everyone had
[32:01] to adjust. Consumer agents
[32:04] bigger user base lower initially.
[32:07] Goldman has reportedly estimated that AI
[32:09] agents could drive token consumption up
[32:10] roughly 24 times
[32:13] to around 120 quadrillion
[32:18] per month.
[32:21] FSD robo taxis not tokens in the API
[32:24] sense but enormous video compute. Um
[32:26] Elon Musk has talked about this to be
[32:28] able to get to the point where this
[32:31] works completely. have a lot a lot of
[32:36] work to do and this is why I went
[32:38] through a lot of things with the visual
[32:40] l the VLMs uh in terms of the instead of
[32:43] the LLMs humanoids again a combination
[32:46] of the VLMs but the step function is
[32:49] enormous in terms of what you're going
[32:50] to need compute-wise if you haven't gone
[32:53] through it yet you should
[32:55] we have a lot more demand coming for
[32:58] compute so again coding agents then we
[33:00] get consumer agents this is why we're in
[33:02] the first inning right now. This only
[33:04] started the beginning of this year. And
[33:06] you can go look at Dylan Patel's answer.
[33:10] It's only been five months. We are in
[33:12] the beginning stages of what it'll take
[33:14] five years to get to here. If you just
[33:15] do it on a time basis, we're barely in
[33:17] the first inning. It's just going to
[33:20] continue to go. The compute demand is
[33:21] exponential because each wave changes
[33:23] the unit of consumption. It grows
[33:25] exponentially because every capability
[33:27] threshold creates a new consumption
[33:29] surface. This is why humans can't keep
[33:31] up with this. We are used to seeing
[33:33] linear change. We are thinking of memory
[33:35] in a linear framework.
[33:38] They wrote a positive piece on meta. Uh
[33:40] I'm getting more interested in this
[33:43] consumer agent thing and the upside
[33:45] potential. And in particular, I just
[33:48] want to bring up this. This is Apple
[33:50] stock price as we got to all-time highs.
[33:52] We actually closed at all-time highs
[33:53] yesterday on a closing basis. That there
[33:57] is when they did their AI Siri
[33:59] announcement. It made an all-time high.
[34:01] It sold off. That was a sell the news
[34:04] event. That was something that to me
[34:07] looked bearish except for the fact that
[34:09] we're above the 200 day. So, we trade
[34:10] down, not just one day, but two days. We
[34:12] have a very weak bounce. Then we go down
[34:14] and on this day, this is the day that
[34:17] they announced they're raising prices
[34:19] 20% because of the memory prices. We
[34:21] know there's going to be less handsets
[34:23] sold. So, what happens? Goes right back
[34:25] to all-time highs. When I start seeing
[34:27] that kind of activity and I'm reading
[34:28] about consumer agents coming and this
[34:31] stock is expensive on a PEG ratio basis,
[34:35] I'm starting to think there's a lot more
[34:37] to this story than what is going on.
[34:40] Now, if you don't buy into the stuff I
[34:42] just said in terms of uh all of the
[34:44] things about the future, you don't
[34:46] believe in humanoids, you don't believe
[34:48] in in consumer agents, you think it's
[34:50] going to take years to get there, you
[34:51] don't think Meta's doing any of the
[34:53] things I said for these reasons, I'm
[34:54] just getting sick of hearing that
[34:55] there's too much capacity when
[34:57] Microsoft, this is all in the most
[34:59] recent earnings period, strong customer
[35:01] demand continues to exceed capacity.
[35:03] Google's allocating compute under a
[35:05] robust ROIC framework, meaning they're
[35:07] only allocating places where they can
[35:08] get back revenue. Uh, Trrenium largely
[35:11] sold out. Tranium 3 nearly fully
[35:13] subscribed to much of tranium for
[35:14] already reserved. Meta said it's compute
[35:16] constrained and that it had
[35:18] underestimated compute demand. Again,
[35:20] this is from the last earnings. So, this
[35:23] is not long ago. The backlog in terms of
[35:26] orders that they have, commitment
[35:29] quotes, 627 billion for Microsoft, 462
[35:34] billion for Google, 225 billion for
[35:37] Amazon, 638 billion for Oracle. The
[35:40] demand is there guys, it's already
[35:43] there. Uh if you don't believe that,
[35:45] let's just look at the prices of the uh
[35:48] the GPUs. As Dylan Patel talked about in
[35:50] the interview, he said CPUs can run to a
[35:53] price where they start to actually run
[35:54] in GPUs. We just don't have enough. We
[35:56] still have we we still are seeing the
[35:59] need for them as we're building out data
[36:01] centers and it's going to continue. So
[36:03] rental rates, we're just not seeing it
[36:06] there. Uh
[36:10] this is the way o bullish oil curves are
[36:12] supposed to to look. Um the rental curve
[36:15] again. So this is from the semi analysis
[36:18] guys just using the silicon data
[36:20] according to our data compute market
[36:22] continues to tighten.
[36:26] Um in terms of the capex buildout and
[36:28] the latest forecast semi analysis is at
[36:32] 11.1 trillion well above Goldman's 7.6
[36:36] trillion
[36:38] and pretty much everyone else has no
[36:40] figures that extend that thus far. Here
[36:43] are their numbers. So again, you get
[36:45] this big step function. It's g the
[36:47] second derivative kicks in. We're not
[36:49] going to get a doubling. We're going to
[36:50] get a slow every year is going to be
[36:52] slower growth, but the numbers are
[36:53] enormous. The spend is enormous, and
[36:55] it's just going to take all this time.
[36:57] JP Morgan put something out this week.
[37:00] Again, multiple ways to look at it. Uh
[37:04] everything is bullish. So when you go
[37:06] through stuff and you realize that all
[37:08] of this stuff is bullish and we're in a
[37:10] race around the globe, it's just not
[37:12] slowing down, guys. Um, in terms of the
[37:15] token usage, uh, I think this was a good
[37:17] article. So Aaron Levy put this out.
[37:19] Great post on how to think about open
[37:20] source AI and applied AI layer. Two
[37:22] things will always be true in AI.
[37:24] Frontier intelligence will likely remain
[37:25] at the forefront. Again, think about
[37:28] what Dylan Patel said of solving brand
[37:30] new use cases and often be used for
[37:32] orchestrating and planning in any type
[37:34] of complex workflow. I don't know how
[37:36] this is a a a shock to people. The the
[37:39] the best intelligence will always be
[37:41] used by enterprises for at least some of
[37:44] what's going on.
[37:46] And in many cases, it's going to be
[37:47] cheaper. And that's where the margin is.
[37:49] At the same time, as use cases become
[37:51] mature and predictable and become
[37:53] workflows, yeah, you'll use lower or
[37:55] close models. Yeah, I'm already doing
[37:58] that. So are most people. This is not
[38:00] something new. If you want to go read
[38:01] the story, this is from uh Jesse Zang.
[38:05] And again, we run 90% of our workloads
[38:08] on open-source models instead of open AI
[38:10] or anthropic. This is consistent with
[38:11] most of the hyperrowth app companies.
[38:14] Enterprises won't get to this number for
[38:16] a while. They're still going to have the
[38:18] best models. They're still going to be
[38:20] trying to focus on this stuff. But
[38:22] eventually, as they get more
[38:23] sophisticated to get there, enterprises
[38:25] spend as a whole is moving the opposite
[38:27] direction as AI native firms. Open
[38:29] source models just fell to 11% of
[38:31] enterprise spend down from 19%. The
[38:34] piece argues that the popular native
[38:36] open source is eating enterprise is
[38:38] directionally right over the long term,
[38:39] but wrong in the current spending data.
[38:41] Closed models dominate early stage
[38:43] discovery. Open source wins the later
[38:44] stage production.
[38:46] more in token budgeting. Same type of
[38:49] thing from uh from semi analysis, Dylan
[38:53] Patel's group. And again, you can just
[38:55] see that everyone is spending more on
[38:57] AI. It's not changing. We're not seeing
[38:59] some fall. Now, Palanteer is talking
[39:02] their book, but I think this is worth
[39:04] the read. Uh they put this out this
[39:06] week. He's been all over trying to argue
[39:08] that you want to own your own stuff
[39:10] before anthropic and open AI steal it.
[39:12] Uh obviously, he's talking his own book.
[39:15] It's still a name in my 100 name
[39:17] portfolio. I still own it myself, too.
[39:20] Um, it's been a long time. I It's I
[39:22] bought it around 130. It's still around
[39:25] 130. Uh, and I bought it back in I think
[39:28] the first quarter earnings report, which
[39:30] was Februaryish, I believe, uh, early
[39:33] February when I was on on trip. So, it's
[39:36] been consolidating and it looks like a
[39:37] consolidation, but I happen to agree
[39:39] that a lot of places are going to take
[39:41] Palunteer seriously and actually do what
[39:43] they're saying. So, worth going through.
[39:46] Um, I'm not going to read all this
[39:47] because you guys can go through it, but
[39:49] I do think there's a lot of good points
[39:50] in here that are in that are important.
[39:52] 5.6 came out this week. Uh, barely have
[39:56] used it so far. Very impressed. A lot of
[39:58] the feedback's been great. I'll spend
[40:00] time on it. Uh, this story has me
[40:03] interested. I haven't spent enough time
[40:05] on it. Uh but
[40:08] basically we're starting to get to the
[40:10] point with AI where the conscious
[40:13] side is coming up. Uh I think it's very
[40:15] interesting. If there's anything useful
[40:17] there, I will bring it, but I haven't
[40:19] spent enough time. I've listened to the
[40:21] Moonshots episode from this week, which
[40:23] there was nothing in it for me to bring
[40:24] to you guys. Uh and they talked about it
[40:27] and I have had some I've watched some
[40:28] videos and stuff, but nothing big. Uh
[40:30] I'm getting back here because
[40:33] this is the beginning of something that
[40:35] for all of you is a must. Uh very seldom
[40:38] do I say that. I know most of the people
[40:41] listening uh have views on Bitcoin. They
[40:45] have views on crypto. They're not the
[40:47] same. The traditional finance people,
[40:50] I'm sure, are more on the negative side
[40:52] and the crypto people that watch this
[40:53] video are more on the positive side.
[40:55] Well, here's the deal. You're past the
[40:57] point now of ignoring it. And the AO
[41:00] macacro nexus was two things. It was the
[41:02] disruption that would come from AI on
[41:04] the traditional world. How it would
[41:05] impact economics, government decisions,
[41:08] how it would impact earnings, how it
[41:10] impact the stocks. All of that has
[41:11] happened. Every single one of the things
[41:13] and the reason that I chose to do this
[41:15] has happened. Once we got to consumer
[41:17] agents, the next phase was that the
[41:19] financial guardrails would change and it
[41:20] would be one of the most important
[41:22] things in the history of the world. We
[41:24] have been on a banking system this way
[41:26] for a long, long time. Swift has been
[41:29] around for a long long time. This is all
[41:32] ending. It's all changing. It's all
[41:34] morphing. And there is money to be made
[41:36] on this and lost. And that is what the
[41:39] next phase and why I'm starting a new
[41:41] YouTube channel specifically starting
[41:44] from the crypto side spending at least
[41:46] 20 of the 30 minutes I expect to do this
[41:48] on what's happening with inside crypto
[41:50] so people can learn how to con speak in
[41:53] crypto how to speak in tokenization how
[41:55] to speak in stable coins how to speak in
[41:57] the new payment system. So Matt Hogan,
[42:00] who I enjoy very much, uh, and was on a
[42:03] panel with him once before,
[42:06] he posted this and it just got me going.
[42:08] And this was on Wednesday. Did and this
[42:10] is related to a Scott Bessett
[42:13] speech that he did, I believe, at the uh
[42:16] economic the New York economic club.
[42:20] Within that speech, he said, "Digital
[42:21] assets, stable coins, tokenization, new
[42:23] payment system will help to shape the
[42:25] future of money."
[42:27] He referenced that El Muhammad Alerian
[42:29] calls it a remarkably important speech.
[42:34] I posted this today. I think everyone
[42:36] whether you believe in crypto or not
[42:38] needs to read both the Scott Bess
[42:39] impeach and the Muhammad Alan New York
[42:42] Times oped about the importance of the
[42:44] speech for Wall Street. You can no
[42:46] longer have no view on crypto. You
[42:49] cannot just call it speculative and walk
[42:50] away. This will disrupt businesses,
[42:54] public stocks. we will have
[42:56] tokenization. So the point is do your
[42:59] homework now. If you don't believe me
[43:01] and you've ignored all this, then go
[43:03] read these articles. Go read the speech
[43:05] from Scott Besson. Here's the uh the
[43:07] op-ed from Muhammad Alan. Argues Scott
[43:10] Besson has articulated a new a new US
[43:13] economic doctrine. This administration
[43:16] is in place for the next two and a half
[43:18] years. His main point is that Wall
[43:20] Street and economists have not fully
[43:21] internalized how big this shift is.
[43:24] Instead, the US now sees economic
[43:26] security as inseparable from national
[43:27] security. That means trade, investment,
[43:29] technology, standards, supply chains,
[43:30] tariff, exports, and payment systems and
[43:33] the dollar's role are all becoming tools
[43:34] of strategic power. The old
[43:36] globalization framework is over.
[43:39] This is really critical. The AI part,
[43:42] they put the AI action plan. They have
[43:44] focused on getting the clarity act
[43:46] through. It's not through yet. The
[43:48] genius uh act went through last year. on
[43:51] digital asset payments. Best at Alian's
[43:52] point is crypto is being reframed
[43:55] from a speculative
[43:58] sideshow into a tool of US financial
[44:00] statecraft. The future of money will be
[44:02] shaped by digital assets, stable coins,
[44:04] tokenization, and the new payment
[44:05] systems. And the US should not let that
[44:07] infrastructure be built elsewhere. This
[44:09] is the same thing they have said about
[44:11] AI. So if you are doubting it now at a
[44:14] time when Bitcoin and the crypto market
[44:16] has no positive sentiment and as I go
[44:19] through very reminiscent of where we
[44:21] were in AI a year and three months ago
[44:24] when I was buying Micron below 100 and
[44:27] people were telling me there was nothing
[44:29] happening. I wrote a paper on inference
[44:31] in May. Crickets
[44:33] best in Stockman appears to treat
[44:34] dollar-based stable coins as a way to
[44:36] extend the dollar. Do not fight digital
[44:39] dollars. Regulate them.
[44:42] For Bitcoin, the implication is
[44:43] different but complimentary, which I'll
[44:45] get through.
[44:48] As an investor, is it important to do
[44:50] your homework on digital assets, stable
[44:51] coins, and how they fit in the future. I
[44:54] just literally went through and asked uh
[44:57] an LLM on this.
[44:59] What happens to the financial system
[45:01] when money, collateral, payment,
[45:03] settlement, and eventually AI, agent,
[45:05] commerce all become programmable? That's
[45:08] really where we are at this point. And
[45:09] the reason that I want you guys to do
[45:11] that homework is that I also want you to
[45:14] go back after you've read Muhammad Alan,
[45:16] after you've read Scott Besson, and then
[45:18] go back to Mark Andre and what he posted
[45:21] on January 21st, 2014 titled Why Bitcoin
[45:25] Matters. Eventually, mainstream
[45:27] products, companies, and industries
[45:28] emerged to commercialize it. Its effects
[45:30] become profound. And later, many people
[45:32] wonder why its powerful promise wasn't
[45:34] more obvious from the start. What
[45:35] technology am I talking about? Personal
[45:37] computers in 1975, the internet in 1993,
[45:41] and I believe Bitcoin in 2014.
[45:44] People didn't accept the internet as a
[45:46] big thing, meaning accept that it was
[45:48] real, until effectively the iPhone came
[45:51] out. That was when everything became
[45:53] real. We were in a bare market in tech
[45:55] in the from the dotcom bubble. So,
[45:57] Bitcoin, we're now 12 years ahead. This
[45:59] was 14 years ahead. He's talking about
[46:01] personal computers. You guys can go back
[46:03] and look at when they were accepted.
[46:07] To finish up further, there is no
[46:09] shortage of regulatory topics and issues
[46:10] that will have to be addressed as part
[46:12] of Bitcoin since almost no country's
[46:14] regulatory framework for banking and
[46:16] payments accepted a technology like
[46:17] Bitcoin. Well, we've made it through
[46:19] this, guys. the US government, the
[46:21] leader of the free world, not only won
[46:23] an election on it, is putting regulatory
[46:26] uh uh framework in,
[46:30] but they're doing it at the same time as
[46:31] AI. But I hope I've given you a sense of
[46:34] the enormous promise of Bitcoin. Bitcoin
[46:36] offers a sweeping vista of opportunity
[46:38] to imagine how the financial system can
[46:40] and should be worked. This is the most
[46:42] important thing. reimagine how the
[46:44] financial system can and should work in
[46:46] the internet era and a catalyst to
[46:47] reshape that system in ways that are
[46:49] more powerful for individuals and
[46:50] businesses alike.
[46:56] Andre told investors why Bitcoin
[46:58] mattered before the state understood it.
[46:59] Vesson is now telling policy makers why
[47:01] it matters. Agentic AI commerce is the
[47:03] accelerant. Once AI agents become
[47:05] transacting, consuming compute, managing
[47:07] subscriptions, settling invoices, moving
[47:09] it, the payment system becomes the
[47:10] operating layer of the agent economy. In
[47:13] 2014, Andre saw the internet's need for
[47:15] the native value transfer. In 2026,
[47:17] Betton recognized it. Guys, this is a
[47:20] no-brainer. It's a no-brainer to go
[47:22] spend the time on it. Jeremy Grantham is
[47:24] doing rants about Bitcoin, saying it is
[47:26] worthless. He has Again, I I don't know
[47:29] how to say this anymore. People do not
[47:32] pay attention. They were saying the same
[47:33] thing about AI. It was hallucinating. I
[47:35] was on trips last year hearing about how
[47:37] the hallucinations would always
[47:38] presented. It was garbage. It would
[47:40] never be adopted. it would never go
[47:41] through. So I wrote this paper in May
[47:43] 15, 2025. The inference inflection where
[47:45] real-time assets meet real world
[47:47] opportunity. We're just getting started
[47:49] with the upcoming rise of AI agents.
[47:51] This is in May of 2015. AI native phones
[47:54] and PCs and embodied AI like
[47:56] self-driving platforms and humanoid
[47:58] robots. Inference demand is poised to go
[48:01] parabolic. The implications for
[48:03] semiconductors, memory, networking,
[48:05] power, and edge computer are massive. If
[48:08] Q1 earnings, this is Q1 of last year,
[48:09] told us anything, it's that scalable
[48:11] inference infrastructure has officially
[48:13] arrived. Crickets last year when I wrote
[48:16] this piece, and I'm sure crypto right
[48:18] now gets crickets. Now, if you guys want
[48:20] to go read stuff from March of this
[48:21] year, why cons when consumers become
[48:24] agents, the open claw gateway? Well,
[48:26] that proved to be right for the agentic
[48:28] side. Tokenization, when ownerships
[48:30] become programmable. You guys can go
[48:32] look at this on your own. This is all
[48:34] what's coming, guys. have written about
[48:35] it now for the last year and a half on
[48:38] is all on Substack
[48:40] Stripe sessions and the coming agentic
[48:42] commerce economy. The reason this is on
[48:43] Substack so everyone traditional finance
[48:45] is realizing is because most people
[48:48] don't pay attention to crypto. And so
[48:50] what I do on the subscriber wall, what I
[48:52] do in the stuff that I do with 22V
[48:54] clients, I don't talk about crypto at
[48:56] all. For you, those of you on the crypto
[48:58] side, I know you buy into this. I know
[49:00] you get it. But all of these things,
[49:02] this is from June 30th. Stripe sessions
[49:04] and the coming agentic commerce. They're
[49:05] showing you what's happening. It's
[49:08] already going on from public. Well,
[49:10] these guys are not public, but it's
[49:11] there. This is why I've shown this. This
[49:14] will be part of the video. This will be
[49:15] the main framework around it. The blue
[49:17] line here is the 40 name index.
[49:21] 40 names to create the crypto index that
[49:23] I've created. These are broken up into
[49:26] eight verticals which are effectively
[49:27] sectors. If you go through them, there
[49:30] are 34 that are tokens. representing
[49:34] the private world or the crypto world
[49:36] and there are six that are now public
[49:37] companies things like circle
[49:40] equal weight it is an overlay with
[49:42] bitcoin so at the end of the day I will
[49:44] continue to say bitcoin is the S&P 500
[49:46] of the future of tokenization you can
[49:49] believe it you can doubt it you can do
[49:51] whatever but for all of you that are
[49:53] doubting it particularly the raas and
[49:54] fas that have your money inside public
[49:56] companies and inside real estate and
[49:58] inside all of these different things the
[50:00] capital structure of the world is
[50:02] changing Scott Bessant has told you that
[50:04] it is changing and it will be
[50:06] represented by this stuff. This is where
[50:09] you're going to find the growth assets.
[50:11] This is where you're going to find the
[50:12] things that are going higher. This is
[50:15] what I will be putting together for the
[50:16] next video thing on YouTube, which will
[50:18] effectively be like this, but it'll be
[50:21] starting from the place not of the AI
[50:23] side. It'll be where AI meets stable
[50:25] coins tokenization and the new financial
[50:28] guard rails. That is what I'll be going
[50:30] through. That is what I'll be showing
[50:31] you. Now to finish up, Bitcoin still a
[50:34] bare market. Technically, I showed you
[50:36] guys last week. I it it traded well
[50:39] again this week. We made the low here. I
[50:41] had divergence signals. We're still
[50:42] there. The 50-day is the next line in
[50:45] the sand as far as I'm concerned. We get
[50:47] above there and then we'll worry about
[50:48] the 200 day. Uh that's what I'm looking
[50:51] at. Doge, which I told you, and for
[50:53] those who don't care about Doge, Doge is
[50:55] retail energy. That's what I care about.
[50:57] I want to see it get above the 200 day
[51:00] as well. We're way below. We didn't even
[51:01] get above the 20-day yet. We did on
[51:03] Bitcoin. So, there's no energy yet.
[51:04] Retail is not comfortable in this.
[51:07] Michael Sailor did something that anyone
[51:08] said if he ever did, Bitcoin would be
[51:10] down big. He sold a chunk of Bitcoin.
[51:13] About 3,500 coins.
[51:16] This is when the announcement came out.
[51:18] This is what happened to Bitcoin.
[51:20] Oh, guess what? Closed higher than it
[51:22] was when that announcement came out. So,
[51:25] we had positive response to bad news
[51:27] just like I highlighted last week. So,
[51:30] I'm going to finish up the last three
[51:31] slides here. I wrote a paper on the
[51:33] academic Fed versus the inflation target
[51:34] of the future. If you guys didn't get a
[51:36] chance to do it, I highly recommend it.
[51:38] I think it is critical as we come into
[51:40] what I think is the only month that the
[51:42] Fed
[51:44] would really raise rates. That's my
[51:46] belief is that if they're going to raise
[51:47] rates and he wants to get the Fed
[51:49] credibility, why not do it uh in July?
[51:54] uh it's close to 50% but to do it I do
[51:57] think there'll be some sort of wink wink
[52:00] nod nod even though he said he's not
[52:01] going to do any forward guidance I do
[52:03] believe that something would probably
[52:05] come out because I don't think as his
[52:07] first thing he wants to disappoint the
[52:10] market especially with his uh with
[52:12] Donald Trump who just brought him in. I
[52:15] think there's a lot of things in here
[52:16] for people to read about in terms of the
[52:18] impact that AI is going to have on the
[52:19] Fed. And all I'm using is Kevin Worsh's
[52:21] words from his confirmation hearing and
[52:23] from an interview he did with the ECB
[52:25] forum last week. I also put this out on
[52:28] Substack. If you guys aren't getting my
[52:29] Substack, go subscribe to it for the
[52:31] Trady [clears throat] people. I think
[52:32] now more than ever reading what I'm
[52:34] writing about on crypto. I connected
[52:36] Bitcoin to that art of unlearning the
[52:38] Fed and why it matters. And the main
[52:40] reason why it matters is this chart
[52:42] here. So this is the rate cuts, the
[52:45] white line here. So, we were expected to
[52:47] have 150 uhund
[52:50] uh
[52:53] yeah, we were supposed to have the rate
[52:55] cuts in here and now we've got rate
[52:57] hikes down here.
[53:00] So, we went from a position of rate
[53:03] cuts, we did a cut here, we did a cut
[53:06] here, and then the rest of this is
[53:09] turning it from future cuts by December
[53:11] of this year to hikes. This is Bitcoin
[53:16] overlaid with it. So as they got more
[53:18] hawkish as we started to see an ending
[53:21] to the cut cycle and fears this
[53:23] happened. Now we also the green line
[53:26] here was growth verse value which was
[53:28] really the software side. This has
[53:30] broken and I think the Fed side is going
[53:32] to break. One or two things are going to
[53:34] happen. I don't think if the Fed does
[53:35] hike I think it's one and done and I
[53:37] think it's 25. I think you'll probably
[53:39] see long-term rates come down on the
[53:41] back of it with the Fred credibility. If
[53:43] they don't do it, which I don't think
[53:46] they will, uh I I'm going to stick with
[53:48] the fact that they're not going to do
[53:49] anything and I'm going to stick with my
[53:51] paper. Uh and I think at that point you
[53:53] can have some tailwinds for Bitcoin.
[53:59] And that's it for this week, guys. Um
[54:01] again, thanks to everyone. Uh I'll be up
[54:04] in Maine the rest of the summer and I'll
[54:07] keep you up to date. See you next week.