AI’s Speed Crash Is Creating the Next Opportunity

[0:00] Um, stocks just
[0:03] continue to, uh,
[0:04] defy what's
[0:05] going on. Um, I'm
[0:07] going to look into
[0:08] that today. I'm
[0:09] going to remind
[0:10] you guys about
[0:11] bubbles, parabasas,
[0:13] and velocity collapses,
[0:14] and kind of give you some
[0:15] idea of ​​what we've
[0:16] seen the last
[0:17] four to six weeks with
[0:19] internal AI. And I'm
[0:20] going to say right
[0:22] now that, uh, the
[0:23] mid-cycle slowdown
[0:24] is ending. And it's not that
[0:26] we won't see the
[0:28] pace of change. But
[0:29] now it's not surprising anymore.
[0:32] Uh, the positioning
[0:34] has changed and I'm
[0:34] starting to hear people
[0:36] trying to be
[0:37] bearish start to
[0:39] get out of
[0:40] hand and highlight a
[0:41] whole bunch of things that I'll
[0:42] address. Uh, it was a
[0:45] mood and
[0:45] technical cleanup.
[0:47] Uh, I'll tell you why that is
[0:48] . I'm going to
[0:49] look at the
[0:50] token index. I've
[0:51] been called more about this
[0:52] than
[0:53] anything else
[0:54] over the past
[0:54] three weeks, from both
[0:55] institutional
[0:56] retail firms
[0:57] and very
[0:58] experienced
[0:58] technical professionals,
[0:59] asking why
[1:00] this token index doesn't
[1:01] matter. Uh,
[1:02] Apple's Meta, interesting
[1:03] week, uh, I
[1:04] wrote an article about the
[1:05] meta. I spend a lot of time on
[1:07] consumer
[1:08] AI agents
[1:09] , which I
[1:10] think is the next
[1:11] topic. I've already talked
[1:12] about the applied
[1:13] side, but now I'll
[1:14] start looking at something
[1:15] that's happening
[1:16] more and more, and what
[1:17] I think
[1:18] will be the main theme of the
[1:19] next 12 months.
[1:20] I will remind people about
[1:22] exponential
[1:23] demand versus linear
[1:24] supply. Scott
[1:25] Bessant gave a
[1:26] talk, and I
[1:27] spent a lot of time on it
[1:29] this week,
[1:30] regarding cryptocurrencies
[1:31] . I'm going to start by telling
[1:34] you what I've been talking about
[1:36] for the last
[1:37] month, which is the beginning of a
[1:38] new part of AI Macro
[1:40] Nexus, remind you
[1:41] what all this was for, but
[1:42] combine that with what
[1:44] Scott Besson said, what
[1:45] Marc Andre said
[1:47] about Bitcoin, and it's going to be
[1:48] new, in the same
[1:49] format as this, but
[1:50] very crypto-
[1:51] focused. And we'll
[1:54] summarize what
[1:55] happened with
[1:56] cryptocurrency this
[1:57] week. First of all,
[1:59] overall, I have had
[2:00] great success with RAAS.
[2:04] For those of you who are
[2:05] financial
[2:06] advisors and
[2:07] want to take the time to see
[2:08] how my work
[2:09] can be expanded
[2:11] beyond just this
[2:12] YouTube video and
[2:13] subscription site, I'll show
[2:14] you where you can
[2:15] go. But, uh,
[2:17] in between that, and especially on the
[2:18] part of the agency that's
[2:19] working on the
[2:21] knowledge brain, it's been, uh, a
[2:22] really good three
[2:23] weeks where I've been
[2:23] helping people not
[2:24] only stay up to date
[2:25] with
[2:26] AI but also
[2:27] start, uh,
[2:28] using it
[2:29] more and more, and
[2:31] I think that's
[2:31] helping. So I
[2:32] just want to remind
[2:34] people what it
[2:35] was called, because
[2:36] this week's video
[2:37] explains why it's an
[2:38] AI macro. Uh,
[2:40] AI is no longer just a
[2:41] technology story.
[2:43] It becomes the dominant
[2:44] macro force of the cycle. It
[2:45] changes
[2:45] productivity,
[2:46] labor, inflation
[2:47] , capital spending,
[2:48] energy, demand,
[2:48] digital assets,
[2:49] monetary policy,
[2:50] market structure.
[2:51] Connectivity is the intersection of
[2:52] these forces, and it's where
[2:53] investors
[2:54] connect. The reason
[2:55] I say this is
[2:56] because, um,
[2:58] in everything you
[2:59] deal with, when you
[3:00] read anything about
[3:01] productivity, about
[3:03] labor, about inflation,
[3:04] about the Fed, about
[3:05] whatever, AI is
[3:06] part of every
[3:07] decision that's
[3:08] made. And so
[3:09] you should be
[3:10] aware of this. This is
[3:11] changing the world before our
[3:13] eyes. Here is one
[3:14] example. So, this is something
[3:15] I've, um, shown
[3:17] many times. I
[3:18] did
[3:18] a video update for
[3:19] subscribers, um, where I
[3:20] started to address the
[3:21] fact that I think the
[3:23] slowdown has reached a
[3:24] point where we've
[3:25] done enough
[3:26] damage. Um, and that's where the
[3:27] damage was. This is the 60-day
[3:29] realized V macro-
[3:32] factor
[3:33] technology
[3:34] momentum from Morgan Stanley. The
[3:35] white line here is
[3:36] it. So, we grew up,
[3:38] and I showed it for a
[3:39] while. I
[3:39] showed this when it
[3:41] was around 70. Now
[3:42] we're at 87. This
[3:43] continues until today,
[3:44] Friday. I do this on
[3:46] Friday evening. This
[3:48] overlaps with the 60-
[3:49] day S&P
[3:51] V. So historically
[3:52] there is a correlation.
[3:54] That is, if you see the
[3:55] tech
[3:56] momentum V explode
[3:57] like this, you
[3:58] usually see the S&P
[4:00] V follow suit. The one
[4:01] thing we have is that
[4:02] tech
[4:03] companies have been
[4:04] growing for some time now. That's
[4:05] why we make these
[4:06] bubbles, paralysis,
[4:07] and speed crashes.
[4:10] Here's what's happening.
[4:12] That's why you need to
[4:13] be on top of this
[4:14] , because this is a
[4:15] structural bull
[4:16] market. It is based
[4:17] on the demand for tokens.
[4:18] Everything I
[4:19] built everything on depends
[4:20] on the demand for tokens.
[4:22] This entire video
[4:23] should give you complete
[4:24] confidence that we
[4:25] are not in
[4:26] a bubble where there is no
[4:27] excess
[4:28] computing
[4:28] power because
[4:30] Meta is selling off some.
[4:31] I hope you
[4:32] guys understand
[4:33] that if this is true, then
[4:34] within the next
[4:35] six months you will have an
[4:36] opportunity when
[4:37] this rises higher.
[4:38] This pushed people all over the
[4:39] world into
[4:40] retail,
[4:42] into hedge funds, everywhere
[4:43] where people
[4:43] were using
[4:44] leverage and
[4:45] probably taking on excessive
[4:46] risk. We've seen
[4:47] this come out now
[4:49] , mainly because
[4:50] V was too
[4:51] high for people not to
[4:52] do it. S&P, I don't
[4:53] know how you can
[4:54] look at this
[4:55] chart and not be
[4:56] optimistic. Um, I
[4:57] said last week
[4:58] that I showed it as a
[4:59] triangle. You know, we
[5:01] get bad
[5:02] news from Iran, and
[5:02] stocks still
[5:03] go up. We do whatever you
[5:04] want. We
[5:05] sell
[5:06] Samsung shares on good
[5:07] performance. Stocks are
[5:08] still rising. This
[5:09] triangle in NDX
[5:10] shows low
[5:11] results. Again
[5:12] , technology is
[5:12] lagging behind. We
[5:13] get more
[5:14] information about
[5:14] financial indicators,
[5:15] more movement in
[5:16] healthcare. But we are
[5:17] finally starting to
[5:17] see some
[5:18] technological
[5:18] rebound. But again
[5:20] , this is a graph that
[5:21] looks
[5:21] optimistic. Uh, just a
[5:22] reminder, if you
[5:24] guys haven't
[5:25] seen this, I made this
[5:26] video back on May 10th. Uh,
[5:27] and the main focus here,
[5:28] which was about the
[5:30] first end down,
[5:31] is on how
[5:32] we should expect
[5:33] this trade to go. Uh
[5:35] , when you have
[5:35] bubbles and
[5:36] parabasas on the way
[5:37] up, you're going to have
[5:38] velocity collapses on the
[5:39] way down. If all of
[5:41] this is happening
[5:41] within a
[5:42] structural
[5:43] bull market,
[5:44] which I will continue to
[5:45] talk about, then it is a
[5:45] structural bull
[5:47] market. You want to
[5:47] ride the wave of a
[5:49] macro bull market,
[5:50] especially when you
[5:51] are in
[5:52] the first innings of
[5:52] demand for
[5:53] AI tokens.
[5:55] Even though
[5:56] someone says it can't
[5:56] be the first
[5:57] inning based on the
[5:58] graphs, those graphs were
[5:59] evidence that
[6:00] people
[6:01] were wrong about
[6:02] AI for two years,
[6:03] saying it was a
[6:03] bubble. These are
[6:04] hallucinations. This will
[6:05] not do anything. They
[6:06] had to catch up
[6:08] as soon as
[6:09] profits appeared. Then we
[6:10] reached the second
[6:11] march, and it won't be so
[6:12] easy from now on. So, it's
[6:13] not that
[6:14] this deal will do what it
[6:15] did before. We are well
[6:16] past the
[6:17] first inning of the
[6:18] speed of change. But do
[6:19] I think you'll
[6:20] get five more
[6:22] shares in Micron and SK Heinex in
[6:23] these things over the
[6:24] next five years
[6:26] ? It is likely that this will be the case
[6:28] , but it is
[6:29] less than the
[6:30] 20% drop that has occurred since
[6:31] March of last
[6:32] year. So we'll
[6:34] slow down, because
[6:35] now everyone knows about it
[6:36] . So, Samsung
[6:38] posted a
[6:39] preliminary
[6:40] quarterly profit
[6:41] before mill sales, which
[6:42] grew 19 times.
[6:44] Micron shares fell that day.
[6:46] Samsung shares also
[6:47] fell that day. Here is
[6:49] the graph, and that's the first thing
[6:50] I want to show.
[6:51] I saw
[6:52] three separate YouTube videos this week
[6:54] about the market where
[6:55] they said it was a
[6:56] news failure. You see,
[6:57] there is good
[6:58] news. They
[6:59] exceeded
[7:00] expectations and the stock
[7:01] sold out.
[7:02] The problem is that we have
[7:03] already sold the shares.
[7:04] I don't agree with the
[7:05] fact that when there
[7:06] was already a correction, you
[7:07] already had a weakening,
[7:08] whereas we
[7:09] remain
[7:09] unchanged during
[7:10] this time, when you
[7:11] get good
[7:12] news, you have to
[7:13] sit back and say that
[7:14] this is a bearish trend. Is
[7:15] this something that showed
[7:16] that the unwinding and
[7:17] deleveraging is
[7:19] still ongoing? Yes. But
[7:20] if you ask
[7:21] me what
[7:22] the range is going to be and where
[7:23] the support is going to be going
[7:24] forward, I'll tell
[7:25] you that I think this is the
[7:26] high from February, this is the
[7:28] low from here, and this
[7:29] will be around the
[7:30] lower end of
[7:31] the range. And I think
[7:33] we'll be significantly
[7:34] higher by the end of the year
[7:35] because of what I'm going to
[7:36] show you. The problem is that
[7:39] everyone is looking for
[7:40] correction. So if I
[7:42] told you that
[7:43] over the course of 10 days,
[7:45] over the course of two days
[7:46] after that, that was the
[7:48] rolling 10-day rate of
[7:49] change at Samsung, okay,
[7:51] that was the highest
[7:52] since the middle of the
[7:54] panic when we
[7:55] shut down the world, 25-26%. It
[7:57] was 21. The only other
[7:59] time it was
[8:00] this big on a 10-
[8:01] day basis was
[8:02] right after Lehman Brothers
[8:04] . So this is a
[8:05] speed crash. It didn't
[8:06] feel that way. You
[8:07] don't think it
[8:08] looks like it, but
[8:09] within a bull
[8:10] market, that's a big
[8:11] drop, especially
[8:12] when you've had these
[8:13] types of increases that
[8:14] were much
[8:14] bigger. So, this is the
[8:16] moving average of the 10-day
[8:17] shift. You had an
[8:18] indicator of 35. So, this
[8:20] was a correction
[8:21] within a bull
[8:22] market. Here's where we
[8:23] are
[8:24] based on P/E for
[8:25] next year. So,
[8:26] everyone who wants to
[8:27] overshadow this, I
[8:28] completely agree. If
[8:29] you believe in
[8:30] the cyclical nature of memory,
[8:31] you should absolutely
[8:32] sit back and ignore
[8:34] it. If you
[8:35] believe demand
[8:37] will remain
[8:38] very high
[8:39] and prices
[8:39] will remain
[8:40] high
[8:41] for the next
[8:42] three years, that is
[8:43] too cheap and
[8:44] suggests a
[8:45] much higher P/E than
[8:46] usual. If you
[8:48] guys look at
[8:49] this and pay attention
[8:50] to my PEG ratio,
[8:52] when we look at
[8:53] earnings growth
[8:54] relative to P/E, they're already
[8:55] not factoring in that it's going
[8:57] to slow down at these
[8:58] levels. They have already
[8:59] factored in a bearish
[9:00] outcome in terms of
[9:01] multipliers.
[9:03] That's how ridiculous
[9:04] these numbers are. And that's why,
[9:06] when people say it's business
[9:07] as usual,
[9:10] Samsung's operating profit in 2026 could
[9:11] exceed
[9:12] $217 billion.
[9:15] That's what
[9:16] analysts think. OK.
[9:18] The company could
[9:19] generate more
[9:20] operating income
[9:21] this year than its
[9:23] combined operating
[9:24] income over the past 40
[9:25] years. Don't treat
[9:27] this as a
[9:28] one-time event
[9:29] where AI just does
[9:30] it and everyone buys
[9:31] excessively and then we'll have a
[9:33] problem. This is nonsense
[9:34] . Hmm, we're not even
[9:36] in the
[9:36] first inning of
[9:37] token demand. I
[9:39] will keep
[9:40] repeating this because
[9:41] no one believed in the
[9:42] demand for tokens until
[9:43] we finally got
[9:44] to January of this year. And
[9:45] I'll talk about in some of
[9:46] Dylan
[9:47] Patel's podcasts from Semi analysis
[9:48] how much
[9:50] his spending has changed and where he's
[9:51] been from last December
[9:52] to today. This
[9:54] should give you an idea
[9:55] of ​​someone who is very
[9:56] knowledgeable about
[9:58] growing businesses, who
[9:59] understands how
[10:00] to use AI, and
[10:01] his team does
[10:02] , and how he
[10:03] went from a
[10:05] $100,000 budget to
[10:06] $11 million
[10:07] in six
[10:08] months in terms of
[10:09] what
[10:10] his annual
[10:11] revenue costs would be. So, the valuation of the
[10:13] Korean company
[10:14] Cosby falls to its
[10:14] lowest level
[10:15] since the global
[10:16] financial crisis. This is
[10:17] exactly what you see
[10:19] at the summits. Just kidding.
[10:22] Um, here's Salesforce.com. Everyone
[10:27] who is
[10:28] pessimistic about
[10:30] memories are the same
[10:31] people who wanted to
[10:32] buy Salesforce when
[10:33] it was falling. So Salesforce
[10:37] now has a 10p to
[10:39] 2028, 11
[10:39] P/E multiple to
[10:41] next year. So
[10:43] , again, we have a
[10:44] P/E ratio of 4 for
[10:45] Samsung, their earnings
[10:47] will grow
[10:48] 524% this year, and next year
[10:49] it will be 43.
[10:51] They are not even close
[10:51] to
[10:52] hitting their estimates yet.
[10:54] They far
[10:54] exceeded them. And yet
[10:56] , somehow,
[10:57] these shares are trading
[10:59] at 11 pence. This is a
[11:01] problem. That's why
[11:03] I've been saying since the beginning of the year
[11:04] that you need a
[11:05] short position on
[11:06] abundance. Anything
[11:07] built on code
[11:08] that can be
[11:09] reproduced by
[11:10] a computer. And the long
[11:11] position is about
[11:12] the deficit. Everything that
[11:16] shoppers spend on
[11:17] that is physically limited
[11:19] by line people. That's
[11:23] all. People create
[11:25] it, computers
[11:26] create it. This topic
[11:27] will continue
[11:28] to pervade because it is
[11:29] about
[11:30] exponential
[11:31] demand. So again
[11:32] , I posted this on
[11:33] June 8th. I started
[11:34] talking about
[11:36] leaving Micron in
[11:37] mid-May.
[11:38] Again, I left well before
[11:39] reaching the summit,
[11:40] but my goal
[11:41] was not to reach the summit.
[11:42] My goal was to
[11:43] be in a
[11:44] position where I
[11:44] would probably be
[11:45] buying
[11:46] back some of the shares in the
[11:47] near future,
[11:47] especially after the
[11:48] SKH Heinix deal that
[11:49] happened today,
[11:50] just because I hear
[11:51] people are
[11:52] pessimistic. And I
[11:53] believe that in 6
[11:54] months we will be
[11:55] much higher than we are
[11:56] now. Now that
[11:57] we've had some
[11:58] pullback, and also
[11:59] because we've had
[12:00] another
[12:01] earnings report, and I'm
[12:02] no longer worried about some of the
[12:03] things that I
[12:04] think could have been a
[12:05] problem in
[12:06] terms of bottlenecks, it
[12:07] looks like we're just
[12:08] going to be in a situation where
[12:09] things are
[12:09] going to be
[12:10] slow.
[12:12] Mainly because there
[12:13] are restrictions
[12:14] set by people
[12:15] that prevent this from
[12:16] becoming a bubble.
[12:17] Believe it or not, it's the
[12:19] supply side,
[12:20] TSMC, which won't be
[12:21] rebuilding,
[12:22] the fact that gas
[12:23] turbines and all that makes things
[12:24] very difficult
[12:25] . We
[12:26] find solutions
[12:27] to every
[12:28] energy problem, and
[12:29] they find
[12:30] ways to do it.
[12:31] So I just think that
[12:32] will keep prices
[12:33] higher because we
[12:34] won't be able to
[12:35] get the capacity in time.
[12:36] Economics, that is,
[12:37] navigating the
[12:38] slowdown of the
[12:39] average
[12:39] AI cycle. I
[12:41] still believe in changing the
[12:42] second derivative, but
[12:43] now I think
[12:44] that's the consensus. That's what
[12:46] we did, and that's
[12:47] part of it. So you're
[12:48] not only paying attention
[12:49] to what I hear
[12:50] from people, but you also
[12:51] want to see what
[12:52] damage we've done. Here's
[12:53] what happened. We have already
[12:55] returned 50 or almost
[12:57] 50% of the
[12:58] technology growth from Opus 4.5.
[13:01] And the reason I'm
[13:02] excited about Opus 4.5 is
[13:03] because that's when the
[13:05] changes happened. That's
[13:06] when people started to
[13:08] recognize it. That's when
[13:09] people started
[13:10] investing. And I want
[13:11] you to look at
[13:12] this because when I
[13:13] look at this from
[13:14] February and then up,
[13:16] it looks a lot like the
[13:17] token index, which
[13:18] I'll come back to
[13:19] later. And I'm not
[13:20] worried about the
[13:20] token index, but
[13:21] I think people
[13:22] have started
[13:22] using the
[13:23] token index for
[13:24] something it should
[13:25] n't be
[13:25] used for. I
[13:26] think it
[13:28] should drop to zero over time. Um
[13:29] , and more importantly, I
[13:30] don't think it
[13:30] should grow. I
[13:31] never think it
[13:32] should grow,
[13:33] and technically it doesn't,
[13:34] which I'll also
[13:35] talk about. When these
[13:35] stories appear,
[13:36] it's usually a good
[13:37] sign. Shares of
[13:38] managers with
[13:38] systematic
[13:39] long short
[13:40] positions fell 2.1%
[13:41] last week through
[13:42] Thursday after
[13:42] falling 3.1%
[13:43] over 5 days. This
[13:44] was their worst
[13:45] performance since December 23.
[13:46] By most
[13:46] indicators I've
[13:47] seen, they've given up
[13:48] the position. And these are again
[13:49] managers with
[13:49] systematic
[13:50] long short
[13:51] positions. So, these are
[13:51] quantitative managers.
[13:52] They gave up the position
[13:53] for about a third of the
[13:54] year. They are still
[13:55] ready for the year. This is not
[13:56] some big panic.
[13:58] I talked about this
[13:59] feeling a little bit, uh,
[14:00] in a webinar I
[14:01] did this week
[14:02] for subscribers,
[14:03] about this feeling that feels a
[14:04] little bit like 2007.
[14:06] Hmm, and that it
[14:06] will end
[14:07] around the same
[14:09] time as there, which is in the
[14:10] first 10 days. I still
[14:11] think there might be
[14:12] a little left, but
[14:13] I also think
[14:14] mutual funds have proven
[14:15] themselves to be
[14:16] lower than I thought, just as
[14:17] they
[14:17] bankrupted a good
[14:19] portion of that
[14:20] SKH Heinik deal. Hmm,
[14:20] fundamental
[14:21] managers also
[14:22] fell and we know that the
[14:23] pod indices had to
[14:24] close, as I
[14:25] said, because of
[14:26] how high
[14:27] V was. Here's the
[14:28] de-leveraging on the
[14:29] retail side
[14:30] . Although we
[14:31] only had about a 23%
[14:33] correction on the
[14:35] SKH Highix and Samsung side
[14:36] to about 30, er, 38,
[14:38] and again, it's
[14:39] back to Opus. The
[14:41] retail side,
[14:42] using
[14:43] double
[14:44] leverage, returned 62%.
[14:46] So I think
[14:47] retail
[14:48] was washed up. Morgan
[14:49] Stanley published this
[14:50] on their
[14:51] AI beneficiaries and
[14:53] you can see the
[14:54] massive outflow that
[14:55] happened and we had three
[14:56] weeks of pretty big,
[14:57] uh, outflows in terms of
[14:59] net flows.
[15:00] Jeff DeGraff, I
[15:01] like to
[15:02] combine these things.
[15:03] It was wonderful. Nearly
[15:04] 87% of
[15:05] S&P
[15:06] 500 semiconductor issues registered and
[15:08] were oversold on a 14
[15:09] -day basis. He
[15:10] uses these
[15:11] lines. You can
[15:12] see that they
[15:13] were all very good
[15:14] bottoms, except for this one here.
[15:15] Everyone else, uh, except for these
[15:17] two little things
[15:18] here. So, can
[15:19] it go lower?
[15:20] Of course, it can. But
[15:21] at least you
[15:22] get it from that
[15:23] side too. Here's what happened on
[15:24] my side in
[15:25] terms of the technical
[15:26] sheet I
[15:27] posted. This was
[15:28] from May 29th, the week just
[15:29] ended.
[15:30] Look at these
[15:31] numbers. Percentage 50
[15:32] days, 100, 91, 98.
[15:33] Look at the
[15:34] RSI averages. Average RSI.
[15:36] Everything was above the 50 level.
[15:38] That's where
[15:39] this
[15:40] week's readings came in. Look
[15:41] how low these
[15:42] numbers are. Percentage above the
[15:43] 50-day. We
[15:44] really sold out here.
[15:46] One index is still
[15:47] higher, but the RSI averages are
[15:48] now there and we still
[15:49] have constructive
[15:51] trends. None of the
[15:52] technological
[15:52] models have changed.
[15:54] I will look into this as well.
[15:55] The rating is good. So,
[15:57] this is a good entry point.
[15:59] Hmm, I'm showing this, uh,
[16:00] again, I showed how to
[16:01] do this, and I'm going to
[16:02] wrap up this video for
[16:03] you guys in
[16:04] terms of how you
[16:05] can do this
[16:06] yourself
[16:07] using Claude
[16:08] Co-work, just by updating
[16:09] the files. But this was
[16:10] day 20 of my
[16:11] thematic
[16:12] portfolio of 100 names.
[16:13] 73 of the 100 names were above the
[16:15] 20-day mark on
[16:16] May 8. We got down
[16:18] to 21 on Wednesday when I
[16:19] did a webinar for
[16:22] subscribers to
[16:23] basically update the stock
[16:24] to give people
[16:25] an idea of ​​where we
[16:26] are. Uh, I
[16:27] like to see
[16:28] growth at the end of the
[16:29] week. We went from
[16:31] 21 to 31. That looks exactly like an
[16:33] oversold level. I
[16:34] mentioned last
[16:35] week that when we
[16:36] go down, or I
[16:36] mentioned this week,
[16:38] when we go down
[16:39] below 40, I'm really
[16:39] interested in, uh, the
[16:40] percentage above the 50-
[16:41] day, and
[16:42] most importantly, above the 200-
[16:43] day. This is
[16:43] back to the
[16:45] same level it
[16:46] ended at on the 26th
[16:47] . But these levels, the 50-
[16:48] day moving average, are
[16:50] still at 76, and the 200-
[16:51] day moving average,
[16:52] which is critical,
[16:54] i.e. the structural
[16:55] situation of the bull
[16:56] market, is still at
[16:58] 86 out of 100 items.
[16:59] We saw a massive
[17:01] inflow of shares on SOXX
[17:04] and SKHEX today,
[17:05] offering over seven
[17:07] times the number of
[17:08] shares subscribed.
[17:09] They also
[17:10] traded
[17:10] higher. Here are
[17:12] the facts about memory
[17:13] over the last
[17:14] three months, uh,
[17:15] just in terms of
[17:16] what's been happening. The
[17:18] memory chip crisis
[17:19] will last until 2030.
[17:21] This was back during
[17:22] its
[17:23] earnings report on March 16.
[17:25] If you thought
[17:26] something would change, here's
[17:27] what he said
[17:28] today. Okay, the
[17:30] memory chip shortage that's plagued the
[17:31] markets for computers,
[17:32] cars, and other
[17:34] devices is likely
[17:35] to continue beyond
[17:36] 2030. So, we have
[17:37] moved on to the period
[17:39] up to 2030. Customers
[17:40] sign
[17:41] long-term
[17:41] contracts. I have
[17:42] n't seen this before, but
[17:43] remember that
[17:44] Micron Inc.'s long-term supply agreement.
[17:45] worth over
[17:46] $100 billion,
[17:46] claims
[17:47] the company has no
[17:47] idea when the
[17:49] RAM crisis will end.
[17:50] That was less than a month
[17:51] ago. In fact, it
[17:52] was less than three
[17:53] weeks ago. And
[17:55] Dylan Patel's memory capacity
[17:56] will only grow by
[17:57] 20-30% per year for the
[17:59] next three
[18:00] years. This concerns
[18:01] human resources.
[18:03] Demand is doubling. This is an
[18:04] exponential
[18:05] demand for memory. Memory prices
[18:07] will continue
[18:08] to rise. This is from an
[18:09] interview that I
[18:10] will review, which
[18:11] came out today.
[18:15] Remember, Elon Musk
[18:17] is building a terraab because
[18:18] he says there won't be enough
[18:19] memory or chips
[18:21] for his humanoids
[18:22] and the demand he
[18:24] sees by 2030. I'm
[18:28] going to talk
[18:29] about this further. I show
[18:30] you these things because it is the
[18:31] reality of where
[18:32] we are right now.
[18:34] So I want to make this very
[18:35] clear, because I've had a
[18:36] lot of people
[18:37] , and I think it's because a
[18:39] lot of people on Next have
[18:40] said that it's just
[18:41] like fracking. This will
[18:42] explode. The only thing that is
[18:44] similar to fracking is
[18:45] the number of people who
[18:47] say it's just
[18:48] like fracking. So, here's the
[18:49] normal cycle to AI
[18:53] . And I say AI because
[18:55] this is a typical
[18:57] demand cycle for hardware-
[18:58] driven demand
[18:59] . So, hardware-
[19:01] oriented demand
[19:02] means physical things.
[19:04] So if demand
[19:05] is driven by physical
[19:06] output, and memory is a
[19:07] physical output, it's
[19:08] very difficult to significantly
[19:09] distance demand from
[19:11] capacity. If
[19:14] the demand is now digital,
[19:16] and it's tokens, and it's
[19:17] growing solely based on
[19:18] the performance of the
[19:19] models, but we
[19:20] need more
[19:21] memory, the capacity
[19:22] can't grow
[19:23] faster. The only way
[19:27] to solve this problem
[19:28] is through
[19:29] algorithmic
[19:30] efficiency, and at
[19:34] this point trying to
[19:35] do that to
[19:36] bring it to market
[19:37] for everyone is
[19:38] simply unlikely.
[19:39] Will this be a happy
[19:40] event? Will
[19:41] it happen eventually? I am
[19:42] sure that will be the case
[19:43] . I don't think it's
[19:45] going to reach a new level
[19:46] anytime soon,
[19:47] especially when I
[19:47] consider what's
[19:48] next. So you have to
[19:49] remember that the difference is
[19:51] this.
[19:52] This is the old world. This is a
[19:53] new world. This is
[19:54] exponential
[19:55] growth. This becomes
[19:56] important when you
[19:57] start dealing with
[19:58] consumer agents,
[19:59] fully
[19:59] self-driving
[20:01] vehicles, and
[20:01] humanoids. All of
[20:02] them have a higher demand for
[20:03] tokens than what we
[20:04] have now. Just like
[20:06] in Chat. We
[20:07] had no problems with Chat.
[20:08] We
[20:10] only had a problem when we
[20:11] entered this period
[20:12] , when it became obvious
[20:13] that we were starting
[20:14] to reach the part of the
[20:15] growth curve that was out
[20:16] of control
[20:18] . That's why, if
[20:20] you compare it to the
[20:21] oil market, it's ridiculous.
[20:23] We had a parabolic
[20:25] or exponential
[20:26] movement on the supply side
[20:27] . There was no movement on the demand side
[20:29] . So, again
[20:30] , for anyone who
[20:31] wants to compare these two
[20:32] , for any
[20:33] smart oil guy
[20:34] or smart person
[20:35] who keeps telling
[20:36] me this is the same as
[20:37] this, I've seen this
[20:38] before. You have
[20:39] n't seen this before because
[20:40] the demand isn't coming
[20:42] from people. The demand
[20:43] comes from
[20:44] computers. It
[20:45] comes from
[20:45] artificial
[20:46] intelligence agents. This has
[20:47] nothing to do with
[20:48] people. People
[20:49] move linearly. It is
[20:51] very important that
[20:52] people start to
[20:53] realize this. This is what a
[20:55] bull market is all about. That's
[20:57] why I do this.
[20:58] My next
[20:59] webinar will be on
[21:00] July 15th, next
[21:02] week. I will specifically
[21:03] look at names in the
[21:04] portfolio that I
[21:05] believe have reached levels
[21:07] that make sense. I'll
[21:08] go over the
[21:08] concentrated basket
[21:10] , all 10 names, and
[21:11] give you guys
[21:12] the details about it. That's what the
[21:14] concentrated basket has done now
[21:15] . So, we have reached the 50-
[21:17] day moving
[21:18] average. We did
[21:19] that back on March 2nd and
[21:20] then we had the
[21:20] next uptrend, we
[21:21] hit the RSI, you guys
[21:22] can go back and
[21:24] look and when the RSI
[21:25] drops here the
[21:25] risk/reward
[21:26] will change. So while I
[21:27] thought there was a
[21:29] mid-cycle slowdown, we
[21:30] actually went sideways
[21:31] during that
[21:32] period. Now we
[21:33] can
[21:33] reboot and
[21:34] continue. I have
[21:35] significantly reduced my
[21:36] position in many of
[21:37] these things, the largest
[21:39] of which is Micron. I
[21:39] left some things
[21:41] unchanged. I transferred
[21:42] some things to silver
[21:43] and bitcoin which didn't
[21:44] work out too well. I'll
[21:45] go back now and look at
[21:46] some of these names and
[21:47] pick out the ones that fit
[21:48] into the concentrated
[21:49] basket. So, that's the
[21:50] whole point of helping you
[21:51] guys, not just from
[21:52] one side, but from the
[21:53] other. If you
[21:54] guys follow
[21:55] me on X, I
[21:56] posted this the
[21:57] same day I was
[21:58] hosting a webinar for
[21:59] subscribers. Nvidia
[22:00] peaked on May 14
[22:01] , three weeks before the peak of
[22:04] AI infrastructure trading.
[22:06] So, again,
[22:07] if I go
[22:08] back and look at
[22:09] mid-May, and
[22:10] actually go past that, that's when
[22:11] we
[22:12] saw the peak of the rate of
[22:14] change and when it started to
[22:15] slow down. Well,
[22:17] Nvidia peaked
[22:18] earlier and more than a
[22:20] month before SMH and the peak of
[22:21] memory names.
[22:22] Today
[22:23] is expected to
[22:24] close above the 20-
[22:25] day moving
[22:26] average for the first time since June 2, keeping the
[22:27] 200-day indicator and
[22:28] MACD buy signal at their
[22:29] lowest levels in a
[22:31] decade. This is one
[22:33] of 10 titles. So, in the
[22:34] consensus basket,
[22:36] Nvidia is one of the names.
[22:37] So again, for
[22:38] RAAS and for anyone who is
[22:39] new to this, if
[22:40] you want to get in touch,
[22:41] here is the website where you
[22:42] can find it. You
[22:44] can always
[22:44] reach Mark
[22:45] Weiling at
[22:46] 22V. He will help you with
[22:48] anything you
[22:49] need, especially
[22:50] regarding RAIA. This is where I
[22:51] want to look at the
[22:53] token spending index because as I
[22:54] showed earlier,
[22:55] if you go through
[22:57] the flows and spending
[22:58] on the previous
[22:59] chart that I
[22:59] showed with everything
[23:01] related to that,
[23:01] including the AI ​​topics, I
[23:03] highlighted back here in
[23:04] mid-May when
[23:05] Nvidia peaked that
[23:06] this would be a growing
[23:07] risk not because it's
[23:09] negative but because the
[23:10] narrative is starting
[23:11] to change, and that's what
[23:12] we've seen. The narrative
[23:14] has changed. So,
[23:15] let's look at
[23:16] this. The first growth in the
[23:18] token index was the
[23:19] shock of the capabilities of
[23:20] coding agents.
[23:21] Then we had OpenClaw.
[23:23] So OpenClaw led to
[23:24] more
[23:25] users immediately
[23:26] starting to
[23:27] use
[23:28] agents that happened to
[23:29] be AI native or
[23:30] Chinese. OpenClaw was a
[23:33] big gamer, but
[23:34] this was a difficult
[23:35] level. So most of the
[23:38] token usage went to
[23:39] a lot of
[23:40] open source stuff.
[23:41] Then businesses
[23:42] finally realized
[23:43] they were missing something.
[23:44] So FOMO (fear of
[23:45] missing out),
[23:47] and everywhere, whether it's Anthropics,
[23:48] ARR, everything started to see the
[23:50] growth of
[23:52] maximizing functionality from there.
[23:53] Token maximization
[23:54] has become a buzzword. Everyone
[23:56] was on the side of
[23:56] research spending
[23:58] . Then they saw the
[23:59] account shock and the
[24:00] token index plummet.
[24:02] Not because everyone is
[24:03] saved, partly
[24:04] because of
[24:05] graduation. All these
[24:06] different things I was
[24:07] talking about. This is when
[24:08] usage starts
[24:09] to be driven more by
[24:10] people
[24:10] using
[24:11] open source
[24:11] software, which doesn't
[24:12] stop companies
[24:13] using
[24:13] artificial intelligence.
[24:14] College students are
[24:15] stopping by. People at
[24:16] Morgan Stanley, Goldman Sachs,
[24:17] places that I know, well,
[24:19] 20% of the people on vacation
[24:20] either work
[24:20] remotely or they
[24:22] don't work on
[24:22] Fridays. In
[24:23] any case, you won't have
[24:24] as many people
[24:25] using
[24:26] these things as there
[24:27] were here. And the shock of the
[24:28] bill began.
[24:29] The problem is that
[24:30] all of this analysis and everything
[24:31] that goes through it
[24:32] doesn't take into account
[24:33] the amount of implementation
[24:34] that is still
[24:35] happening. We
[24:36] are undergoing mass
[24:37] implementation. This is not an
[24:39] indicator of making
[24:40] or losing money, as I
[24:41] will show with the example of
[24:42] Dylan Patel. So,
[24:44] again, the
[24:45] exponential
[24:46] usage of tokens
[24:47] due to memory shortages
[24:49] is increasing day by day
[24:50] . I've heard people
[24:51] argue with me
[24:52] that no one is
[24:52] going to
[24:53] use the
[24:54] most expensive stuff. This is a
[24:55] pile of garbage. If you
[24:58] want to listen to
[24:59] someone who talks
[25:00] about this, and I think
[25:01] always does well
[25:02] in
[25:03] interviews, this came out
[25:04] today. Dylan
[25:05] Patel, or probably
[25:06] Thursday, late. Uh,
[25:09] Dylan Patel about the
[25:10] wisdom tree in
[25:11] Europe. Good
[25:13] interview, I think. The
[25:14] main points
[25:16] I have here: Patel
[25:17] argues that the
[25:17] debate over AI ROA
[25:20] fails to take into account how
[25:21] quickly serious
[25:22] users are
[25:22] increasing
[25:24] token spending, and he
[25:25] reiterates that
[25:26] semi-analytics spending
[25:27] has gone from less than $100,000 to $
[25:30] 100,000 in November,
[25:31] which is very similar to my
[25:32] story: he was paying
[25:33] everyone in the firm for a
[25:35] $200,000 Claude model.
[25:36] Once Claude Code started
[25:38] working after Opus 4.5
[25:39] and building things, that's when
[25:40] they
[25:41] moved to the
[25:42] token spending side and
[25:44] everyone started
[25:44] spending there. By January,
[25:46] it was already
[25:47] $4 million a
[25:48] year. Today, that's
[25:51] about
[25:52] $11 million a
[25:53] year, and peak
[25:54] week was the
[25:55] equivalent of $14
[25:56] million. This is for a
[25:57] company with 90 people.
[25:59] Companies are spending
[26:00] budgets on
[26:01] AI, but the
[26:02] right
[26:02] answer may be to
[26:03] cut other
[26:04] costs on SAS or
[26:04] legacy systems,
[26:05] not cut AI.
[26:06] The right
[26:07] model choice depends on the
[26:08] workload.
[26:09] Cheaper models for
[26:10] fixed
[26:11] workflows, the
[26:11] smartest models
[26:12] for
[26:13] assistants. In this
[26:14] part, I have to
[26:15] listen to it. For
[26:16] assistant-style work, the
[26:17] smartest model
[26:18] can be generally
[26:19] cheaper because
[26:20] it uses
[26:21] fewer tokens, fewer
[26:22] iterations, and less
[26:23] human time. That's
[26:24] why, when you look at the
[26:25] labor market, we have
[26:27] weekly hours. If
[26:28] you can complete
[26:29] a task in a tenth of the
[26:30] time, but
[26:30] spend
[26:31] more money doing it, then that's how
[26:32] it should
[26:33] work. So the
[26:34] gist of it is that
[26:35] we're at
[26:36] a point where
[26:36] businesses
[26:37] are trying
[26:37] to figure out what to do,
[26:38] how much they should
[26:39] use for
[26:40] workflows, which
[26:41] will be much
[26:42] cheaper, and how much
[26:42] they should
[26:43] use for
[26:45] assistant-style work. You don't
[26:46] hire all the people from the
[26:47] most expensive colleges
[26:48] with the best resumes.
[26:51] The team at the firm
[26:52] is divided into many
[26:53] different areas. You
[26:54] hire younger
[26:55] people who are cheaper
[26:56] than people with experience
[26:57] in the field.
[26:58] This is how AI will work
[26:59] . He considers that
[27:00] memory is no longer a
[27:01] conventional commodity
[27:02] cycle. If you
[27:03] don't believe me, he
[27:04] gives more
[27:05] technical reasons. He
[27:06] expects memory prices
[27:07] to continue
[27:08] to rise until AI
[27:09] becomes mainstream
[27:09] , and consumer
[27:10] devices like
[27:11] smartphones and
[27:12] laptops are
[27:12] forced to absorb the
[27:13] higher prices. Essentially,
[27:14] he's saying the same thing as
[27:16] Samsung. We have
[27:17] a shortage when our
[27:18] phones, cars,
[27:19] PCs, we don't have enough
[27:21] memory. So demand is
[27:22] not just coming from
[27:23] hyperscalers for
[27:25] data centers. It comes
[27:26] from everyone. He is
[27:27] optimistic about
[27:28] processors. There's
[27:29] something going on with the processors.
[27:30] I think this is
[27:31] also worth
[27:32] listening to. As far as
[27:34] power goes, I
[27:35] think he's the best
[27:36] at what he
[27:36] does. I
[27:37] have used
[27:38] his services a lot. I
[27:39] referenced
[27:40] him a lot. This was another
[27:41] great performance.
[27:42] Essentially, what he
[27:43] said in the
[27:44] brief as I
[27:45] listened to it and discussed it is
[27:46] that we
[27:46] find a way to
[27:47] get gigawatts,
[27:48] even if it
[27:48] means using
[27:49] diesel engines from
[27:50] trucks, boats,
[27:51] trains,
[27:51] converted to
[27:52] generate
[27:53] electricity. You
[27:54] have to listen to this
[27:55] to understand that
[27:56] if you thought
[27:57] we were going to run out of
[27:57] computing
[27:58] resources because people
[27:59] were going to shut down
[28:00] data centers,
[28:01] that's not true. He went
[28:03] through it, and obviously
[28:04] I'm sure he's an
[28:05] investor, but if not,
[28:07] he obviously has
[28:08] access. You're hearing more and more
[28:10] about Anthropic
[28:12] being
[28:13] cash flow positive or
[28:14] at least achieving
[28:15] operating
[28:16] profit. Again
[28:18] , this just
[28:19] means that
[28:20] profitability is
[28:21] coming in and RAIC
[28:22] shouldn't be there, from a
[28:23] fear perspective.
[28:25] But I made it about the
[28:26] air pocket of
[28:26] capital spending
[28:27] on AI. If you guys were
[28:29] n't reading this because of
[28:30] the holidays, I would go
[28:31] read it. I spent
[28:32] some time, I
[28:33] spent a lot of time
[28:34] on consumer agents,
[28:35] but I really spent
[28:36] time on this
[28:37] Mark Zuckerberg town hall that
[28:38] supposedly acknowledged that
[28:39] Meta Zentic AI progress was
[28:40] slower than
[28:41] expected. And I
[28:43] argue that if
[28:44] they eventually
[28:45] sell any of
[28:46] their capital
[28:47] capacity, it will be
[28:48] because they cannot yet
[28:49] get their
[28:51] consumer agents
[28:52] to work the way
[28:53] they want. It's
[28:55] coming, guys.
[28:57] And if you want
[28:58] to listen, it came out the
[28:59] day before. Meta or two
[29:01] days ago.
[29:03] Meta's CTO, Andrew
[29:04] Bosworth, was with Alex
[29:05] Kanantrowitz. This is a
[29:06] great interview and
[29:07] very easy
[29:08] to listen to. The industry is
[29:10] moving beyond
[29:11] one giant
[29:12] monolithic model. I
[29:13] think he's talking
[29:14] about
[29:15] open source and
[29:16] models at the edge of
[29:17] possibilities.
[29:18] Consumer AI has been
[29:19] slow because
[29:19] products are still
[29:20] too complex to
[29:21] use. This is a
[29:22] really
[29:23] critical piece,
[29:24] especially for
[29:25] mutual funds and
[29:26] long-term
[29:26] investors, to
[29:27] spend time
[29:28] understanding how
[29:29] consumer agents
[29:30] will change the world, which is a
[29:31] dramatic shift.
[29:32] Whatever happened to
[29:33] coding agents,
[29:34] as far as agent language is concerned
[29:36] , it's much
[29:37] more. This is much more
[29:39] important in
[29:40] importance than opus 4.5.
[29:42] So think about it this
[29:43] way. All we've
[29:44] done is increase the
[29:45] coding capabilities
[29:46] , but we haven't found a
[29:48] way to go
[29:49] through all of that yet. It is
[29:51] incredibly important
[29:52] to understand what this
[29:53] means. Consumer AI
[29:55] has been slow
[29:56] because the product level is still
[29:57] immature, not because
[29:58] the model is weak. For
[29:59] consumer
[30:00] AI to work,
[30:01] it needs to disappear into
[30:02] everyday life,
[30:03] like map searches,
[30:04] messaging,
[30:05] or the camera.
[30:07] I'm using
[30:08] my Grock in my Tesla this week. I
[30:10] talked to him during the
[30:11] trip, throughout the
[30:13] entire trip. I
[30:14] can't do it with a
[30:15] phone the same way.
[30:16] It's not that good until
[30:17] we get to the
[30:18] point where you
[30:19] talk to it like
[30:20] you would
[30:22] with Alexa, where you can
[30:22] sit back, cook, and
[30:23] ask it
[30:24] questions about the recipe.
[30:25] So you don't have to
[30:26] reach for your
[30:27] phone, you don't
[30:28] have to look at
[30:28] it. Until we
[30:29] reach this
[30:30] point, which will come
[30:31] soon, it is
[30:32] very difficult
[30:33] to form a habit.
[30:34] But once that
[30:34] happens, billions of
[30:36] users will start
[30:37] having agents
[30:38] working for them,
[30:39] and they will start giving
[30:40] them tasks, but they
[30:41] need context,
[30:42] trust, time, permissions.
[30:44] This is extremely
[30:45] difficult, and
[30:46] frontier models are not capable of
[30:47] it. The token curve
[30:49] is changing because of all
[30:50] this. Frequency increases,
[30:51] context length
[30:52] decreases,
[30:53] multi-step
[30:54] thinking through everything
[30:55] you need here.
[30:57] Everything will be there. Every
[30:59] thing that is the next
[31:00] function. When
[31:01] the problem of AI interface
[31:03] for consumers is
[31:04] solved,
[31:04] the use of tokens
[31:05] should increase dramatically. AI
[31:06] is moving from something that
[31:07] people
[31:08] call upon occasionally to something
[31:09] that runs in the background
[31:10] . So it exists
[31:12] before the advent of mobile
[31:13] data for the iPhone, but
[31:14] usage
[31:15] is limited by friction.
[31:16] You need to
[31:17] go back and
[31:18] think about what
[31:18] happened to
[31:20] bandwidth and what
[31:20] will happen to
[31:21] consumer agents
[31:22] . The problem is that what
[31:23] happened to the
[31:24] bandwidth
[31:25] , again, and
[31:26] getting it
[31:27] was very
[31:28] slow. Once
[31:29] we get to
[31:30] the phone and we're
[31:31] getting closer and I'll
[31:32] show you not only that
[31:33] this is all meta-
[31:34] side, but you
[31:35] need to look at
[31:36] what happened to
[31:37] Apple stock because Apple
[31:38] is also a part of this
[31:39] through the phone. So once
[31:41] that happens,
[31:42] they will start
[31:43] consuming a huge
[31:44] amount of tokens.
[31:48] So the coder agents
[31:49] were the first visible
[31:51] token explosion, and
[31:52] they will consume
[31:53] another thousand tokens.
[31:56] That's why once we
[31:57] got to the
[31:59] coding side of Opus 4.5,
[32:00] everyone had to
[32:01] adapt.
[32:02] Consumer agents,
[32:04] larger
[32:04] user base,
[32:05] smaller at first.
[32:07] Goldman reportedly
[32:08] estimated that AI agents
[32:10] could increase
[32:11] token consumption
[32:13] by about 24 times, to
[32:14] about 120
[32:15] quadrillion per
[32:18] month. FSD robo-taxis are
[32:22] not tokens in the sense of an
[32:24] API, but huge
[32:25] video computations. Um,
[32:27] Elon Musk was talking about
[32:28] getting to the point
[32:30] where it would fully
[32:31] work. We
[32:36] need to
[32:37] work a lot, a lot, and that's why
[32:38] I
[32:39] looked at a lot of visual
[32:41] VLMs, not LLMs. Humanoids—
[32:44] this is again a
[32:45] combination of VLM, but
[32:46] the step function is
[32:47] huge in
[32:48] terms of what you'll
[32:49] need in terms of
[32:50] computational
[32:51] resources. If you have
[32:53] n't watched it yet,
[32:54] you should. We have a
[32:56] much higher
[32:57] demand for computing.
[32:58] So, again,
[32:59] encoding agents and
[33:00] then
[33:01] consumer agents. That's why
[33:02] we're in the first
[33:03] inning now. This
[33:04] only started at the beginning of this
[33:06] year. And you can
[33:07] look at
[33:08] Dylan
[33:09] Patel's answer. Only
[33:10] five months have passed. We
[33:11] are in the
[33:12] early stages of
[33:12] what will
[33:13] take us five
[33:14] years to achieve
[33:15] . If we do it
[33:16] on a time
[33:17] basis only, we barely
[33:17] made it to the first
[33:18] inning. It
[33:19] will just continue.
[33:20] The demand for computing
[33:21] is exponential
[33:22] because each wave
[33:23] changes the unit of
[33:24] consumption. It
[33:25] grows
[33:25] exponentially
[33:26] because each
[33:27] opportunity threshold creates a
[33:28] new surface of
[33:29] consumption. That's why
[33:30] people can't keep up with it
[33:31] . We are used to
[33:32] seeing linear changes.
[33:34] We think of
[33:35] memory in a linear
[33:36] structure. They
[33:38] wrote a positive
[33:39] article on meta. Uh, I'm
[33:41] more interested in this
[33:42] consumer
[33:43] agent thing and the
[33:45] growth potential. And in particular,
[33:47] I just want to
[33:47] raise this issue.
[33:49] This is Apple's stock price
[33:50] when we hit
[33:51] all-time
[33:51] highs. We
[33:52] actually closed
[33:53] at all-time
[33:54] highs yesterday based on
[33:55] the close. This
[33:56] was when they
[33:57] made their
[33:58] announcement about
[33:58] Siri artificial intelligence.
[34:00] It has reached a
[34:00] historic
[34:01] high. She
[34:02] sold out. It was an
[34:03] event that sold
[34:04] news. This was something that
[34:05] looked bearish to me
[34:06] , except for
[34:07] the fact that we
[34:08] broke through the 200-day
[34:09] mark. So, we
[34:10] trade down not for one
[34:11] day, but for two days. We have a
[34:12] very weak rebound.
[34:14] Then we go down, and
[34:15] on this day, this very
[34:16] day, they announced a
[34:19] 20% price increase because of
[34:20] memory prices. We know that
[34:22] fewer phones will be sold. So
[34:23] what's going on?
[34:25] It is returning to
[34:25] historical
[34:26] highs. When I
[34:27] start seeing this kind of
[34:28] activity and reading
[34:30] about consumer
[34:31] agents coming in and these stocks being
[34:32] expensive based on the
[34:33] PEG ratio, I
[34:34] start to think there's a lot more to
[34:36] this story
[34:37] than what's
[34:38] going on.
[34:39] Now, if you don't
[34:40] believe what I
[34:41] just said, in
[34:43] terms of all the things about the
[34:44] future, you don't
[34:45] believe in humanoids,
[34:46] you don't believe in
[34:47] consumer agents, you
[34:48] think it's going to take
[34:49] years, you
[34:50] don't think Meta
[34:51] is doing any of the things
[34:52] I said for these
[34:53] reasons, I'm just
[34:54] tired of hearing that there's too much
[34:55] capacity
[34:56] . When Microsoft said it all in its
[34:58] latest
[34:59] earnings report, strong
[35:00] customer demand
[35:01] continued
[35:01] to exceed
[35:02] capacity. Google's allocation of
[35:03] computing
[35:04] resources
[35:05] follows a
[35:05] robust ROIC system,
[35:06] meaning they
[35:07] only allocate to
[35:08] places where they can
[35:09] generate revenue. Um,
[35:10] Tranium is largely
[35:11] sold out. Tranium 3 is
[35:12] almost fully
[35:13] subscribed with a significant
[35:13] portion of Tranium already
[35:14] reserved. Meta
[35:15] stated that it had
[35:16] limited
[35:16] computing
[35:17] resources and
[35:18] underestimated the demand for
[35:19] computing
[35:19] resources. Again,
[35:21] this is from the latest
[35:22] earnings report. So, it
[35:23] wasn't that long ago. The
[35:24] order book
[35:28] they have
[35:30] is 627
[35:31] billion for Microsoft,
[35:33] 462 billion for Google, 225
[35:35] billion for Amazon and
[35:37] 638 billion for Oracle.
[35:41] There is demand, guys, it
[35:42] already is. Um, if you don't
[35:44] believe that, let's
[35:45] just look at the
[35:46] prices of
[35:47] GPUs. As
[35:49] Dylan Patel said in
[35:49] an interview, he said
[35:50] that processors could
[35:51] reach a price point where
[35:52] they would
[35:53] actually start working in
[35:54] GPUs
[35:55] . We simply don't have enough of them
[35:56] . We still
[35:57] have them, we still
[35:58] see a need for them
[35:59] as we build
[36:00] data centers,
[36:02] and that
[36:02] will continue.
[36:03] So we simply don't
[36:04] see the need for
[36:05] rent. Uh, this is what
[36:10] bullish oil curves should look like.
[36:14] Again, the
[36:15] rent curve. This is from
[36:16] a semi-analysis, guys
[36:18] just
[36:18] using
[36:19] silicon data, according to
[36:21] our data, the
[36:22] compute market continues
[36:23] to grow. As
[36:26] for the
[36:28] capex build-up,
[36:30] the latest
[36:31] semi-analysis forecast
[36:32] is 11.1
[36:33] trillion, well
[36:35] above
[36:36] Goldman
[36:37] Sachs' 7.6 trillion, and almost everyone else
[36:39] doesn't have numbers that
[36:40] extend to that.
[36:42] Here are their numbers. So,
[36:43] again, you
[36:44] get this big
[36:45] step function.
[36:47] This is g, the second derivative
[36:48] comes into play. We won't
[36:49] get a doubling.
[36:50] We will get
[36:51] slow growth
[36:52] each year, but
[36:53] the numbers are huge.
[36:54] The costs are huge and
[36:55] it's just going to take all
[36:56] this time. JP Morgan
[36:57] published something
[36:59] this week. Again
[37:00] , there are several
[37:01] ways to look
[37:02] at this. Uh, everything is bullish.
[37:05] So when you
[37:06] look at things and
[37:07] realize that this is all
[37:08] bullish and we're in a
[37:09] race around the
[37:10] world, it's
[37:11] just not
[37:12] slowing down,
[37:13] guys. Um, regarding
[37:14] the use of tokens
[37:15] , I think that was a
[37:16] good article. So,
[37:17] Aaron Levy
[37:18] published it.
[37:19] Great post on
[37:20] how to think about
[37:20] open source AI and
[37:21] application layer AI.
[37:22] Two things
[37:23] will always be true in AI.
[37:24] Advanced intelligence is
[37:25] likely
[37:25] to remain at the
[37:26] forefront.
[37:27] Again,
[37:27] think about what
[37:28] Dylan Patel said
[37:29] about solving
[37:29] completely new
[37:30] use cases and will
[37:31] often
[37:31] be used
[37:32] to orchestrate and
[37:33] schedule
[37:34] any type of complex
[37:35] workflow. I do
[37:35] n't know how this can
[37:37] be a shock to people.
[37:39] The best intelligence
[37:40] will always be used
[37:41] by businesses
[37:42] for at least
[37:43] part of what
[37:44] happens. And in
[37:45] many cases
[37:46] it will be cheaper. And that's exactly where the
[37:49] difference lies. At the same time,
[37:50] as
[37:50] use cases become
[37:51] mature and
[37:52] predictable and
[37:53] turn into
[37:53] workflows, yes,
[37:54] you will
[37:55] use
[37:55] lower or close
[37:56] models. Yes, I'm already
[37:57] doing that. Like most
[37:59] people. This is not something new
[38:00] . If you want
[38:02] to read the story, it's
[38:03] by Jesse Zang. And
[38:04] again, we
[38:05] run 90% of our
[38:06] workloads
[38:07] on
[38:08] open source models, not
[38:09] open AI or
[38:10] man-made. This is
[38:10] consistent with
[38:11] most companies
[38:12] developing
[38:13] hyper-productive
[38:14] applications.
[38:14] Businesses won't
[38:15] reach this figure for
[38:16] some time yet. They
[38:17] will still have the
[38:18] best models. They will still
[38:20] try to
[38:21] focus on these
[38:22] things. But ultimately,
[38:22] as they become
[38:23] more complex to
[38:24] achieve,
[38:24] enterprise spending
[38:25] as a whole moves in the
[38:26] opposite
[38:26] direction than spending by
[38:27] firms
[38:27] using
[38:28] AI.
[38:28] Open
[38:30] source models just fell to
[38:31] 11% of enterprise spending
[38:33] from 19%. The article
[38:33] argues that
[38:34] popular native
[38:35] open source is
[38:35] devouring enterprises
[38:36] , which is the right
[38:37] direction in
[38:37] the long
[38:38] term, but
[38:39] wrong in
[38:39] current
[38:40] spending data. Closed
[38:41] models dominate the
[38:42] early stages of
[38:43] detection. Open
[38:44] source wins in the
[38:45] later stages of
[38:45] production. More
[38:47] about
[38:48] token budgeting. The same from
[38:50] semi-analysis,
[38:51] Dylan Patel's group. And
[38:53] again, you
[38:54] can see that everyone
[38:56] is spending more on
[38:57] AI. This does not change.
[38:59] We don't see any
[39:00] decline. Palantir
[39:01] speaks its own language,
[39:03] but I think it's
[39:04] worth reading. Uh,
[39:05] they published it
[39:06] this week. He
[39:07] tried everywhere
[39:08] to prove that you want
[39:09] to own your own
[39:10] things before
[39:10] anthropogenic and
[39:11] open AI
[39:12] steals them. Uh,
[39:13] obviously he's
[39:14] saying something. This name is
[39:15] still in my
[39:16] portfolio of 100 names.
[39:18] I still
[39:19] own it too. Um, it's been a
[39:20] long time. I...I
[39:22] bought it
[39:23] for about 130. It's still
[39:24] worth about 130. Uh,
[39:25] and I bought it
[39:26] back, I think, on the
[39:29] first quarter earnings report, which
[39:30] was around
[39:31] February, I think
[39:32] early February, when
[39:34] I was on a business trip.
[39:35] So, it's
[39:36] consolidating, and it
[39:37] looks like
[39:37] consolidation, but I
[39:38] agree that a lot of
[39:40] places are taking
[39:41] Palunteer seriously and
[39:42] really doing what they
[39:43] say. So it's worth
[39:44] a look. Um, I'm not
[39:46] going to read all of
[39:47] it because you guys
[39:48] can review,
[39:49] but I think there are
[39:50] a lot of good
[39:51] points here that are
[39:52] important. 5.6 was released
[39:53] this week. Uh, I
[39:55] hardly
[39:56] used it. Very
[39:57] impressed. Many of
[39:58] the reviews were
[39:59] excellent. I will take my
[40:00] time with this. Uh, this
[40:02] story
[40:03] interested me. I didn't
[40:04] spend
[40:05] enough time on it. Um,
[40:06] but basically we're
[40:08] starting to get
[40:09] to that point with
[40:10] artificial intelligence
[40:11] where the
[40:12] conscious side comes in. Um, I
[40:14] think that's very
[40:15] interesting. If there's
[40:16] anything useful there, I'll
[40:17] include it, but I haven't
[40:18] spent enough
[40:20] time. I listened to
[40:21] this week's episode of Moonshots
[40:22] , in which I
[40:23] had nothing to
[40:24] offer you.
[40:25] Um, and they talked
[40:26] about it, and I watched
[40:28] some videos and stuff,
[40:29] but nothing major.
[40:31] Um, I'm coming back here
[40:32] because this is the start of something
[40:34] that is mandatory for all of you
[40:35] . Um,
[40:38] I very rarely say that. I
[40:39] know most
[40:41] people listening, uh
[40:42] , have views on
[40:43] Bitcoin. They have
[40:45] views on
[40:45] cryptocurrency. They are not
[40:47] the same. Traditional
[40:48] financiers, I'm
[40:49] sure, are more
[40:50] negative, and crypto
[40:51] people watching
[40:52] this video are more
[40:54] positive. Well, here's the
[40:55] thing. You've already
[40:57] passed the point of
[40:57] ignoring this. And the
[40:59] AO macrolink was
[41:00] two things. This was the
[41:01] turning point
[41:02] that artificial
[41:03] intelligence would have in
[41:04] impacting the
[41:04] traditional world. How will
[41:05] it affect
[41:06] the economy, government
[41:07] decisions, how will it affect
[41:09] profits, how will it
[41:10] affect stocks. All
[41:11] this has already happened. Each
[41:12] of these things and the reason
[41:14] I decided to
[41:15] do it has already happened.
[41:16] Once we got
[41:17] to consumer agents, the
[41:18] next step
[41:19] was that financial
[41:20] barriers changed, and
[41:21] that would be one of the
[41:22] most important things
[41:23] in the history of the world. We have
[41:25] been working in this
[41:26] banking system for a very, very long time.
[41:28] Swift has been around for a very,
[41:30] very long time. This is all
[41:31] ending. Everything
[41:32] changes. Everything
[41:34] is transforming. And
[41:35] you can make
[41:36] and lose money on this. And that's the
[41:38] next
[41:39] step, and that's why I'm
[41:40] launching a new
[41:41] YouTube channel, starting
[41:42] with the crypto side
[41:44] , spending
[41:44] at least 20 of the 30
[41:45] minutes that I plan to
[41:46] do on what's
[41:47] going on
[41:48] inside
[41:48] cryptocurrency, so that
[41:49] people can learn
[41:51] how to talk
[41:51] conventionally in
[41:52] cryptocurrency, how to
[41:53] talk about
[41:54] tokenization, how to
[41:55] talk about
[41:55] stablecoins, how to
[41:56] talk in a new
[41:57] payment system.
[41:59] So, Matt Hogan,
[42:00] who I really love and
[42:02] I've been on a panel with him before
[42:03] , posted
[42:06] this and it just
[42:07] piqued my interest. It
[42:08] was on Wednesday. I did,
[42:10] and it has to do with
[42:11] a speech Scott
[42:12] Bessett gave, which he
[42:13] seems to have given at the
[42:18] Economic Club of New York. In
[42:19] this speech, he
[42:20] said: "Digital
[42:21] assets,
[42:22] stablecoins, tokenization, a
[42:23] new payment
[42:24] system will help
[42:25] shape the future of
[42:26] money." He mentioned
[42:28] that El Muhammad
[42:29] Alerian calls this
[42:30] speech extremely
[42:31] important. I
[42:34] posted this
[42:35] today. I think
[42:36] everyone, whether
[42:37] they believe in
[42:38] cryptocurrency or not,
[42:39] should read both
[42:40] Scott
[42:41] Bess' impeachment and
[42:42] Muhammad Alan's New
[42:43] York Times article about the importance
[42:44] of this speech to Wall
[42:45] Street. You can no longer
[42:46] not have any
[42:47] opinions about
[42:47] cryptocurrency. You
[42:49] can't just
[42:49] call it
[42:50] speculative and walk
[42:51] away. This will disrupt
[42:53] business and public
[42:54] actions. We will have
[42:56] tokenization. So the point is to
[42:57] do
[42:58] your homework
[42:59] now. If you don't
[43:01] believe me and
[43:01] ignored all this,
[43:02] then read these
[43:03] articles. Read
[43:05] Scott
[43:05] Besson's speech. Here is, uh, an
[43:07] article by
[43:08] Muhammad Alan.
[43:10] Scott Besson
[43:11] formulated a new
[43:14] US economic doctrine. This administration
[43:16] will be in power for the
[43:17] next two and
[43:17] a half years. His
[43:19] main point
[43:19] is that
[43:20] Wall Street and
[43:21] economists have not
[43:21] fully grasped
[43:22] how massive
[43:23] this shift is. Instead,
[43:24] the United States now
[43:24] views
[43:25] economic security
[43:26] as inseparable from
[43:27] national security
[43:28] . This means that
[43:29] trade, investment,
[43:30] technology,
[43:30] standards,
[43:31] supply chains, tariffs,
[43:32] exports, payment
[43:33] systems, and the role of the
[43:33] dollar become
[43:34] instruments of
[43:35] strategic power. The
[43:36] old structure
[43:37] of globalization
[43:38] is over. This is
[43:39] really critically
[43:40] important. Regarding AI, they
[43:42] have developed an AI action plan
[43:43] . They
[43:44] focused on
[43:44] passing the
[43:46] Clarity Act. It has not
[43:47] been accepted yet. A
[43:47] brilliant law, uh,
[43:49] was passed
[43:49] last year regarding
[43:50] payments in digital
[43:51] assets. The best thing,
[43:53] in Alan's opinion, is that
[43:54] cryptocurrency
[43:55] is transforming from a
[43:56] speculative show
[43:57] into a tool of
[44:00] US financial governance.
[44:01] The future of money will be
[44:02] shaped by digital
[44:03] assets,
[44:03] stablecoins
[44:04] , tokenization, and
[44:05] new payment
[44:05] systems. And the US
[44:06] should not allow
[44:08] this infrastructure to be built
[44:09] elsewhere. They
[44:11] said the same thing about AI. So
[44:13] if you're having doubts
[44:14] now that Bitcoin
[44:15] and the cryptocurrency
[44:16] market are not in a
[44:17] positive mood,
[44:19] and as I'm going through, it's
[44:20] very reminiscent of where we
[44:21] were with artificial
[44:22] intelligence a year and three
[44:23] months ago when I was
[44:25] buying Micron below 100
[44:26] and people were telling me
[44:27] nothing was
[44:28] happening. In
[44:29] May, I wrote
[44:30] an article about logical
[44:31] inference. Crickets best in
[44:32] Stockman seems to be
[44:33] looking at dollar-
[44:34] based stablecoins
[44:35] as a way to
[44:36] expand the dollar. Don't
[44:38] fight digital
[44:39] dollars. Adjust
[44:40] them. For Bitcoin, the
[44:43] consequences are different, but
[44:44] free, and I will
[44:45] understand them. As an investor
[44:48] , is it important to do your
[44:49] homework
[44:50] on digital
[44:50] assets,
[44:52] stablecoins, and their
[44:53] future fit? I just
[44:54] literally walked around
[44:55] and asked a
[44:56] law professor about this issue
[44:58] . What will happen to the
[45:00] financial system
[45:01] when money, collateral,
[45:02] payments, settlements,
[45:03] and ultimately artificial
[45:04] intelligence, agents,
[45:05] commerce become
[45:06] programmable? This is
[45:07] where we really
[45:08] are at the
[45:09] moment. And the reason
[45:10] I want you
[45:11] guys to do this
[45:12] homework is because
[45:13] I
[45:14] also want you to
[45:15] go back after
[45:16] reading Muhammad
[45:17] Alan, after
[45:18] reading Scott
[45:19] Bessen, and then
[45:20] go back to Marc
[45:21] Andre and his article
[45:22] from January 21, 2014
[45:23] called “Why
[45:24] Bitcoin Matters
[45:25] .” Eventually,
[45:26] major products,
[45:27] companies, and industries emerged
[45:28] that commercialized
[45:29] it. His influence became
[45:30] profound, and later
[45:31] many people
[45:32] wonder why his
[45:33] powerful potential
[45:33] was not more
[45:34] apparent from the
[45:35] beginning. What
[45:36] technology am I talking about?
[45:37] Personal
[45:38] computers in 1975,
[45:40] the Internet in 1993, and,
[45:42] I believe, Bitcoin in 2014
[45:44] . People didn't
[45:45] perceive the Internet
[45:46] as something important,
[45:47] that is, they didn't perceive
[45:48] it as a reality,
[45:49] until the
[45:50] iPhone actually came along. That's
[45:52] when everything became
[45:53] real. We were in a
[45:54] bare-bones
[45:54] technology market during the
[45:55] dot-com
[45:56] bubble. So,
[45:57] Bitcoin, we are now 12
[45:58] years ahead. That
[45:59] was 14 years
[46:00] ahead. He's
[46:01] talking about
[46:02] personal
[46:02] computers. You guys
[46:03] can come back
[46:04] and see when they were
[46:05] accepted. In
[46:07] conclusion, there is no shortage of
[46:08] regulatory topics and
[46:09] issues that
[46:10] will have to
[46:10] be addressed within
[46:11] Bitcoin, as almost
[46:12] no country has
[46:13] adopted a regulatory
[46:14] framework for
[46:16] banking and payments for a
[46:17] technology like
[46:18] Bitcoin. Well, we got
[46:19] through it, guys. The
[46:21] US government, the leader of the free
[46:22] world, not only
[46:23] won the election on
[46:25] this, is implementing a
[46:26] regulatory framework, but is
[46:29] doing so simultaneously
[46:30] with artificial
[46:31] intelligence. But I
[46:32] hope I have given
[46:33] you an idea of ​​the
[46:34] huge potential of
[46:35] Bitcoin. Bitcoin offers
[46:37] ample opportunities
[46:38] to imagine how the financial system can and
[46:39] should work
[46:40] . This is
[46:41] the most important thing.
[46:42] To rethink how the
[46:43] financial system
[46:44] can and should
[46:45] work in the
[46:45] Internet age, and to become a
[46:46] catalyst for
[46:47] transforming that
[46:48] system in a way that is
[46:49] more
[46:50] powerful for both
[46:50] individuals and
[46:51] businesses. Andre
[46:55] told investors
[46:56] why Bitcoin
[46:57] mattered, even before the
[46:58] government understood it. Wesson is now
[47:00] telling politicians
[47:01] why this is important.
[47:01] Agent commerce
[47:02] with artificial
[47:03] intelligence is
[47:03] an accelerator.
[47:04] Once AI agents
[47:05] start
[47:05] making
[47:06] transactions,
[47:06] consuming computations
[47:07] , managing subscriptions
[47:08] , paying bills, and
[47:09] moving them, the
[47:09] payment system
[47:10] becomes the operational
[47:11] layer of the agent
[47:11] economy. In 2014,
[47:12] Andre saw
[47:13] the internet's need for
[47:14] native
[47:15] value transfer
[47:16] . In 2026, Betton
[47:18] realized this. Guys, this is an
[47:20] obvious decision.
[47:21] Spending time on this is an
[47:22] obvious decision.
[47:23] Jeremy Grantham
[47:24] goes on a tirade about
[47:25] Bitcoin, saying it's
[47:26] worthless.
[47:27] He's again...I do
[47:29] n't know how to
[47:30] say it anymore. People don't
[47:31] pay attention. They
[47:32] said the same thing about
[47:33] AI. These were
[47:34] hallucinations.
[47:35] I was on trips last year
[47:36] , hearing
[47:36] about how these
[47:37] hallucinations
[47:38] would always be presented.
[47:38] It was garbage. This
[47:40] will never be
[47:40] accepted. This will never
[47:41] pass. So I
[47:42] wrote this article on
[47:44] May 15, 2025. A
[47:45] twist of logical
[47:45] conclusion where
[47:46] real-time assets
[47:47] meet
[47:48] real-world capabilities. We are
[47:48] just beginning with the
[47:49] future growth of
[47:50] AI agents. This is
[47:51] May 2015.
[47:54] AI-powered phones and PCs and embodied
[47:56] AI, such as
[47:56] self-driving
[47:57] platforms and
[47:58] humanoid robots
[47:59] . The demand for logical
[48:00] inference is poised
[48:01] to become parabolic.
[48:02] The implications for
[48:03] semiconductors,
[48:04] memory, networking,
[48:04] power, and
[48:06] computer peripherals are
[48:06] enormous. If the
[48:07] first
[48:08] quarter earnings, this is the first
[48:09] quarter of last
[48:10] year, told us anything
[48:10] , it's that the
[48:11] scalable
[48:12] inference infrastructure has
[48:13] officially arrived.
[48:14] Crickets last
[48:15] year when I wrote this
[48:16] article and I'm sure
[48:17] cryptocurrency is
[48:18] getting
[48:18] crickets now. Now,
[48:19] if you guys
[48:20] want to read
[48:21] the stuff from
[48:21] March of this year,
[48:22] why the confusion when
[48:23] consumers become
[48:24] agents, the
[48:25] floodgates of claws are open? Well, that
[48:26] turned out to be right
[48:27] for the agency
[48:28] side. Tokenization is
[48:29] when ownership becomes
[48:30] programmable. You
[48:31] guys can
[48:32] look at it for
[48:33] yourself. That's all that's coming
[48:34] next, guys. I've
[48:35] been writing about this for the
[48:37] past year and a half
[48:38] , all at the
[48:39] Substack Stripe sessions and the future
[48:40] agency economy of
[48:42] commerce. The reason
[48:43] this is on Substack for
[48:44] all of traditional
[48:44] finance
[48:45] to be aware of is because
[48:46] most people don't
[48:47] pay attention to
[48:48] cryptocurrency. So what
[48:50] I do on the
[48:51] subscriber wall, what I
[48:52] do in what I
[48:53] do with 22V clients,
[48:54] I don't talk
[48:55] about cryptocurrency at all. For
[48:56] those of you who are
[48:57] interested in
[48:57] cryptocurrencies, I
[48:58] know you believe in this
[48:59] . I know you
[49:00] understand that. But all of
[49:01] these things, this is from
[49:03] June 30th, Stripe Sessions and the
[49:04] upcoming Agentic Commerce.
[49:05] They show you
[49:06] what's going on. This is
[49:08] already happening
[49:09] publicly. Well, these
[49:10] guys aren't public,
[49:11] but it's there. That's why
[49:12] I showed it. This will be
[49:14] part of the video. This
[49:15] will be the basic
[49:16] structure around
[49:17] this. The blue line here
[49:19] is the 40-name index, 40
[49:20] names to create
[49:22] the crypto index that I
[49:23] created. They are divided
[49:25] into eight verticals,
[49:27] which are actually
[49:27] sectors. If you
[49:29] go through them, there are 34
[49:31] tokens represented
[49:32] in the private world
[49:34] or the crypto world, and there are
[49:35] six that are now
[49:36] public
[49:37] companies. Things like a
[49:39] circle with equal
[49:40] weight. This is an overlay
[49:42] with Bitcoin. So, ultimately, I
[49:44] will continue
[49:45] to say that Bitcoin is the S&P
[49:46] 500 of the future
[49:47] of tokenization. You
[49:49] can believe it, you
[49:50] can doubt it,
[49:51] you can do whatever you
[49:52] want, but for
[49:53] all of you
[49:54] doubters,
[49:55] especially RAAS and FAS,
[49:56] whose money is invested in
[49:57] public companies,
[49:58] real estate and all these
[49:59] different things, the
[50:01] capital structure of the world is
[50:01] changing. And Scott
[50:03] Bessant told you
[50:04] that it is changing, and
[50:06] it will be represented by this
[50:07] . This is where you
[50:10] will find
[50:11] growth assets. This is where
[50:12] you will find things that
[50:13] grow. This is what I'm going to
[50:16] put together for the next
[50:18] YouTube video, which
[50:19] will actually be like this,
[50:20] but it won't start
[50:21] from the AI ​​side. This will be where
[50:23] AI
[50:24] meets
[50:25] tokenization of
[50:26] stablecoins and
[50:27] new financial
[50:28] barriers. This is what
[50:29] I will go through. This is what
[50:30] I will show you.
[50:32] Now, to conclude,
[50:33] Bitcoin is still a bare
[50:34] market. Technically, I
[50:35] showed you guys
[50:36] last week. I
[50:37] saw it
[50:38] trading well again
[50:39] this week. We
[50:40] have reached the minimum.
[50:41] I had
[50:42] divergence signals. We are still
[50:43] there. The 50-day
[50:44] indicator is the
[50:44] next limit,
[50:45] as far as I understand.
[50:46] We will exceed it, and
[50:47] then we will
[50:48] worry about the 200-
[50:49] day. Uh, that's exactly what I mean
[50:50] . Dodge,
[50:51] which I mentioned before,
[50:52] and for those who don't
[50:53] care about Dodge, Dodge
[50:54] is the energy of
[50:55] retail.
[50:56] This is what worries me.
[50:57] I want it to
[50:58] break above the 200-day
[50:59] mark as well. We have
[51:00] n't even surpassed the
[51:01] 20-day mark yet. We have
[51:02] surpassed Bitcoin.
[51:03] So there is no energy yet
[51:04] .
[51:05] Retail doesn't
[51:06] feel
[51:06] comfortable with this.
[51:07] Michael Saylor did
[51:08] something that everyone said
[51:09] if he ever
[51:10] did, Bitcoin would
[51:11] fall hard. He
[51:12] sold some of the
[51:13] Bitcoin, about 3,500
[51:14] coins. That's when
[51:16] the announcement appeared.
[51:18] This is what happened to
[51:19] Bitcoin. Oh, you know
[51:21] what? Closed higher than
[51:22] when
[51:23] this
[51:23] announcement appeared. So,
[51:25] we had a positive
[51:26] reaction to the bad
[51:27] news, as I
[51:27] mentioned last
[51:28] week. So, I'm
[51:29] going to wrap up the
[51:30] last three slides
[51:31] here. I wrote an article
[51:32] about the
[51:33] Fed's academic stance versus its
[51:34] inflation target going
[51:35] forward. If
[51:35] you haven't had
[51:36] the chance to
[51:37] do this, I highly
[51:37] recommend it. I think
[51:38] this is critically
[51:39] important as we
[51:40] approach what I
[51:41] believe is the
[51:42] only month the
[51:43] Fed will actually raise
[51:45] rates. I think
[51:46] if they're going to
[51:48] raise rates and
[51:49] he wants to gain
[51:50] confidence in the Fed, why
[51:51] not do it, uh, in
[51:53] July? Uh, it's about
[51:55] 50%. But to do that,
[51:57] I think there will be some kind of
[51:58] wink,
[51:59] wink, nod.
[52:00] While he said he wasn't
[52:02] going to make
[52:03] any predictions about the
[52:04] future, I think
[52:05] something is likely to
[52:06] come out because I do
[52:07] n't think he
[52:08] wants to disappoint the market in the first place
[52:09] ,
[52:10] especially with his, uh,
[52:11] relationship with
[52:12] Donald Trump,
[52:13] who just
[52:13] invited him. I think there's a
[52:15] lot to
[52:16] read here about the impact of
[52:17] artificial intelligence
[52:19] on the Fed. And all I'm
[52:20] using are
[52:20] Kevin Warsh's words from
[52:22] his
[52:22] confirmation hearings and from the
[52:23] interview he gave to the
[52:24] ECB forum last
[52:25] week. I also
[52:26] posted this on
[52:27] Substack. If you guys are
[52:29] n't getting my substack
[52:30] , subscribe to
[52:31] it. For the Trady people, I
[52:32] think now
[52:33] more than ever, people are
[52:34] reading what I write
[52:35] about cryptocurrency. I
[52:36] linked Bitcoin to this
[52:37] art
[52:37] of weaning off the Fed and
[52:39] why it's important. And
[52:40] the main reason why
[52:41] this is important is this
[52:42] diagram right here. So this is a
[52:43] rate cut, the white
[52:45] line here. So, we were
[52:48] expecting 150 uh and,
[52:53] well, we were supposed to have a
[52:55] rate cut here
[52:56] and now we have a
[52:57] rate increase here.
[53:01] So, we've moved from a
[53:04] rate cut position, we've made
[53:05] a cut here, we've
[53:06] made a cut here,
[53:07] and then the rest of it
[53:08] moves from
[53:09] future cuts
[53:10] through December of this year
[53:11] to a hike. This is Bitcoin
[53:15] , superimposed on this.
[53:17] So as they
[53:18] became more
[53:18] hawkish as we
[53:19] started to see
[53:20] the end of the
[53:21] down cycle and
[53:22] the fear that that was
[53:22] happening, now we
[53:23] also have a green
[53:24] line here, this is
[53:25] growth versus
[53:26] cost, which
[53:26] was actually the
[53:27] software side
[53:28] . It
[53:29] broke, and I think the
[53:31] Fed side will
[53:32] break. One or
[53:33] two things will happen. I don't
[53:34] think if the Fed
[53:35] raises rates, it's going to be all over in
[53:36] one go
[53:37] , and I think
[53:38] it's going to be 25. I think you're
[53:39] probably going to see a
[53:40] decline in
[53:41] long-term
[53:42] rates because of the confidence
[53:43] in Fred. If they do
[53:44] n't do it, and I
[53:45] don't think they
[53:46] will, then I'll
[53:47] stick to the
[53:48] fact that they're
[53:49] not going to do anything,
[53:50] and I'll
[53:50] stick to my
[53:51] article. And I think
[53:52] at this point you
[53:53] might have a
[53:54] tailwind for Bitcoin. And that's
[53:59] all for this week,
[54:00] guys. Thank you again
[54:02] everyone. I will be in Maine
[54:04] for the rest of the summer and
[54:05] will keep you
[54:07] posted. See you
[54:08] next week.