La IA está cambiando el tiempo: quien no pueda competir, tendrá que computar

[0:00] OK. A few
[0:03] notes, because people
[0:06] asked. First,
[0:09] this may be my last
[0:10] week in
[0:11] Maine. Not
[0:13] sure yet. I don't know if
[0:16] I'll be recording from here
[0:17] or Maine next week. But I get
[0:21] asked a lot of
[0:22] questions, especially
[0:23] about Maine, and I want
[0:24] to tell those who get
[0:25] to know me
[0:26] through the podcast. I am in
[0:31] Me because
[0:33] September 11, 2001—we are
[0:35] approaching the 25th
[0:36] anniversary of September 11.
[0:40] Many people know that
[0:41] my best
[0:42] childhood friend and the groom at
[0:44] my wedding
[0:44] was killed on September 11th. I
[0:46] then returned to
[0:47] Morgan Stanley. I was already a
[0:49] partner there.
[0:50] The next day, when
[0:51] the markets were closed, I
[0:52] was on the trading floor.
[0:53] Together with other
[0:54] partners, we
[0:55] decided how we
[0:56] would work further.
[0:57] It was very emotional
[0:58] , of course, and then I got a
[0:59] call. I
[1:00] made an offer to
[1:02] buy a house
[1:03] in Maine because I took
[1:04] my oldest daughter there for
[1:05] camp. I fell in love with
[1:08] this place, it
[1:08] reminded me of my father;
[1:10] As the son of a laborer and a
[1:11] driller, I simply became
[1:12] attached to
[1:13] this place. I didn't
[1:17] like the coast where
[1:19] I grew up in New
[1:20] Jersey, and the Hamptons were
[1:21] definitely not my
[1:22] style, so I
[1:23] fell in love with Maine. I
[1:25] bought a house in
[1:26] Kennebunkport,
[1:27] moved there, and
[1:28] commuted from there for about 15 years
[1:29] .
[1:30] Now I've moved further
[1:32] north to
[1:32] hide from the people
[1:33] who have discovered
[1:35] Maine. By the way, people
[1:36] asked about
[1:37] the painting in the background. This is a
[1:38] painting dedicated to
[1:40] Sean Christopher
[1:41] Caton, who died
[1:42] that day. I bought it
[1:44] after moving to
[1:46] Maine. The bottle of wine
[1:47] behind me is a
[1:48] bottle from Napa Valley,
[1:50] as a reminder of
[1:51] 9/11. This week
[1:52] I turned it around.
[1:54] There is a commemorative inscription on it.
[1:55] If you want to
[1:55] take a look, there's a
[1:56] basketball
[1:56] signed by
[1:57] Bernard King. There's
[1:58] my Knicks cap.
[2:00] Sean and I went to
[2:01] a lot of games, and there's
[2:02] still a picture of us in the back at a bar
[2:04] in the Bronx. It was
[2:06] many hours
[2:07] after the Yankees won the
[2:08] World Series
[2:09] when Wade Boggs rode a
[2:11] police horse.
[2:12] Maine is a special
[2:13] place for me. I
[2:14] come here
[2:15] every year, and since you
[2:16] asked, I thought I'd
[2:17] explain. And now for something else
[2:19] , for those of you who have
[2:20] applied.
[2:21] First of all, I will never WhatsApp
[2:22] you
[2:23] offering deals or
[2:25] anything else. I would never
[2:27] do that at X.
[2:28] If you write to me on
[2:29] LinkedIn, I may or
[2:30] may
[2:31] not respond because I get hundreds of
[2:33] messages. The same goes for
[2:34] my
[2:35] email.
[2:36] So don't take it
[2:37] personally if I don't
[2:39] respond. I do
[2:40] n't write to anyone first. So,
[2:42] if you are not subscribed
[2:43] to the newsletter, you are
[2:44] not receiving anything from me
[2:45] . I say this
[2:46] because there are SHI-
[2:47] bots everywhere, and people impersonating
[2:49] me at
[2:50] every turn. I
[2:51] tried to get rid of
[2:52] them for a while, but
[2:53] nothing worked. Another thing
[2:54] : people send
[2:56] me questions
[2:57] like, "What do you think
[2:58] about this expert from
[3:00] Goldman Sachs, Morgan Sachs, or Morgan
[3:01] Stanley? They have a different
[3:03] perspective on XYZ." Friends,
[3:04] if you
[3:05] learned nothing from watching
[3:06] this, then know this: I don't
[3:07] care what anyone
[3:08] thinks. I take into account
[3:09] everyone's opinions, but I don't
[3:10] depend on anyone's opinions
[3:11] . I do my
[3:13] work myself. I gather
[3:14] all possible
[3:15] information and then
[3:17] make a decision
[3:18] by analyzing
[3:18] the probabilities and
[3:19] trying to find something that
[3:21] others disagree with.
[3:23] I usually try
[3:25] to be a bit of a
[3:25] contrarian, but
[3:27] at the same time I look for ideas with a
[3:28] good
[3:29] risk-reward ratio.
[3:31] I will always think
[3:32] independently. I will always
[3:33] be flexible in my
[3:35] views. Today I'm going to
[3:36] break it down with you because
[3:37] some of you say I'm
[3:38] only
[3:39] positive. And if that's what
[3:40] you believe
[3:41] as a subscriber,
[3:42] and if you bought
[3:43] assets at the peak of the market
[3:45] because you subscribed and
[3:46] decided, "Oh, this is a '
[3:47] bull' market." I
[3:48] will show you how to
[3:49] use all of
[3:50] my transcripts and
[3:51] give you tips on how to
[3:53] process them so you
[3:54] can get the
[3:55] information you need.
[3:56] I just
[3:57] speak from the
[3:58] heart every week, friends. I
[3:59] will repeat one thing
[4:00] constantly: we
[4:01] are in a
[4:02] structural "
[4:02] bull" market for
[4:03] demand for
[4:04] computing
[4:04] power. It is
[4:05] insatiable, and will
[4:06] remain so until I hear
[4:08] people I
[4:09] respect on Wall Street
[4:10] begin to acknowledge:
[4:11] the demand for computing will
[4:12] never
[4:13] end. Just like
[4:15] with food: if
[4:16] the number of people on the
[4:17] planet grows
[4:18] by a billion every year, we will
[4:20] never keep up
[4:21] with the needs. The same
[4:22] goes for
[4:23] computing
[4:23] power.
[4:24] It is impossible to get
[4:25] enough
[4:26] computing
[4:27] power for all
[4:28] the billions of agents
[4:29] that appear on the
[4:30] network and are just
[4:30] starting to work. This is an
[4:32] investment in
[4:32] computing
[4:33] power. So
[4:35] this entire video today
[4:36] will be
[4:37] dedicated to them to some extent. And
[4:40] so I'll get back to the
[4:41] agenda for
[4:42] this week. My
[4:43] job is to help
[4:45] your business, whether it's an
[4:46] RIA or an
[4:47] asset manager, help
[4:48] with research,
[4:49] independent thinking
[4:51] , and also prepare
[4:52] you to help your
[4:53] kids adapt
[4:55] to this world as it
[4:56] rapidly approaches
[4:57] and accelerates
[4:59] over the next 24
[5:00] months.
[5:01] Cryptocurrencies are going to be a
[5:02] big part of
[5:03] the discussion, and if you
[5:04] don't know anything about it
[5:05] or don't
[5:05] understand where it's
[5:07] going, you're definitely
[5:07] going to struggle in the
[5:08] next 12 months,
[5:09] and in 24 months,
[5:11] you're definitely going to have to
[5:12] adapt. So,
[5:13] we'll go through the
[5:14] resume, the
[5:15] computing
[5:16] power, the
[5:16] token factories,
[5:18] Coreweave, Google, SpaceX, their
[5:19] comments, the
[5:20] AI ​​regime change as it
[5:21] relates to my
[5:22] work, and perhaps
[5:24] most importantly, two
[5:25] points: number six and
[5:26] number nine. Number
[5:28] six: we're
[5:28] back to
[5:29] work, guys. Summer
[5:30] is over. I'm
[5:31] going back to New
[5:33] York. I spent
[5:34] a lot of time working
[5:35] . I tried
[5:35] creating prompts. I
[5:37] created a new
[5:38] website. I've done
[5:39] a lot with AI
[5:40] agents, and I
[5:41] hope I was able to
[5:42] give you something useful
[5:43] . But I'm starting to
[5:44] focus on
[5:45] predictions for 2027,
[5:47] because next year will be the
[5:48] year of consumer
[5:49] agents. This year was
[5:50] about agents and
[5:51] the discovery of the agent
[5:52] economy. Next
[5:53] year is about consumer
[5:54] agents and their
[5:55] needs, namely
[5:56] financial
[5:56] safeguards, which is where
[5:57] crypto comes in.
[5:58] Number nine: be
[5:59] entrepreneurs, not
[6:01] academics.
[6:01] Academics will continue
[6:02] to cost you
[6:04] money. Anyone who
[6:05] tells you
[6:06] stories, analogies, and
[6:08] other nonsense,
[6:09] especially economists
[6:10] —I will explain why it is not
[6:11] worth your time. The S&P
[6:13] rose 40 basis
[6:14] points in a week. IWM
[6:17] rose 1%, Q's rose 1%,
[6:21] JVAI thematic portfolio outperformed
[6:23] the market again, rising 3% for the
[6:25] week. We have reached
[6:27] new all-time
[6:28] highs in
[6:28] most major
[6:29] indices, as well as around the
[6:31] world. Revenue,
[6:34] and I'll say it again.
[6:36] Earnings growth
[6:39] of 51% for the S&P.
[6:40] Sales growth, which
[6:42] is perhaps the more important
[6:42] indicator, was 15%
[6:44] year-on-year.
[6:47] I don't remember a time
[6:48] when we were in a mature
[6:49] phase of the economic
[6:50] cycle and global
[6:51] earnings per share (EPS)
[6:52] were around 47%
[6:53] year-over-year.
[6:54] Everywhere you look,
[6:55] EPS figures
[6:57] are rising rapidly. This is the
[6:58] most important moment
[6:59] . If you believe those
[7:00] who say this is an
[7:01] isolated "
[7:02] bubble" for just
[7:03] a few
[7:03] tech
[7:04] companies in the US, then that's
[7:05] not true. The whole world
[7:06] is showing rising
[7:07] earnings per share, and that's
[7:08] because times
[7:09] are changing. You
[7:10] will often hear me
[7:12] talk about this. This
[7:13] will make you
[7:14] think. For
[7:15] my Substack readers, a
[7:16] very important article is coming out soon
[7:17] .
[7:19] The time is coming
[7:20] when I will be writing
[7:21] articles about once
[7:22] a month, as I
[7:23] am moving into the
[7:24] crypto sphere, and most of
[7:25] the materials
[7:25] are being transferred to the site
[7:27] for subscribers to
[7:28] my crypto
[7:28] analytics. So, this is the
[7:31] MSCI World Bank index. It
[7:35] moves in a
[7:36] parabola upwards. This is a
[7:37] reflection of what is
[7:38] happening thanks to
[7:39] AI. All of these graphs are a
[7:40] reflection of what's
[7:42] happening with AI. This is a
[7:43] reflection of what is
[7:44] happening around the
[7:45] world. AI is
[7:47] destroying all
[7:47] academic
[7:48] research you
[7:49] know; Everything you
[7:51] believed in is crumbling, and the
[7:52] reason is that
[7:54] productivity is increasing.
[7:55] More and more
[7:56] smart people at the Fed,
[7:58] whom I read, and at the IMF,
[8:00] are starting to
[8:01] realize this. Although
[8:02] again, these should
[8:04] be young people.
[8:05] There is a
[8:06] productivity boom.
[8:07] However, this needs to
[8:08] be considered in
[8:09] context.
[8:10] S&P's earnings forecast for this
[8:12] year has increased by 30%
[8:13] year-over-year.
[8:15] This is our hiring.
[8:17] Zero. Economists
[8:19] say the
[8:20] labor market is strong. They
[8:21] want the Fed to
[8:22] raise rates. This is
[8:23] all nonsense.
[8:26] Given how things
[8:27] change over time,
[8:28] we need to hire
[8:29] a lot of people
[8:30] as we
[8:31] grow profits at a
[8:32] record pace,
[8:33] not as we emerge from a recession
[8:34] . So why aren't we
[8:37] hiring people? When
[8:39] I worked at Morgan Stanley,
[8:40] if profits were
[8:41] growing quickly, we hired
[8:41] people. This is what you
[8:43] did. You invested
[8:44] more money in
[8:44] more
[8:45] people. Now that is no longer the
[8:46] case. And the reason is that
[8:48] corporate agents are appearing.
[8:50] They haven't started
[8:51] working fully yet
[8:51] , but they're getting there
[8:53] . This doesn't
[8:53] mean we're going to
[8:54] lose millions of
[8:55] jobs, but it does
[8:56] mean we're already seeing
[8:57] job cuts, outside of
[8:59] healthcare. We are already
[9:00] seeing small
[9:01] job losses.
[9:02] This is productivity
[9:03] . We don't need so
[9:05] many people. We
[9:06] see
[9:06] new businesses being created
[9:07] .
[9:09] Startups are emerging. We have more and more
[9:10] business
[9:11] formations. Gordy has a
[9:12] business. I no longer
[9:15] work for a
[9:16] Fortune 500 company, nor do I
[9:17] work for a company with
[9:18] 100 employees.
[9:20] Started my own
[9:21] business. So,
[9:22] consumer agents
[9:22] will appear
[9:23] next year. These are
[9:24] the tokens needed for
[9:26] them. This is what they are
[9:27] paid for. This is what
[9:28] they eat. This is what
[9:29] they need to
[9:31] survive. They
[9:32] work 24/7/365. They
[9:35] compress time. So
[9:37] when you say "year
[9:38] over year" for
[9:39] anything now, it's a
[9:40] human concept.
[9:42] Agents are awake.
[9:43] Agents don't go to
[9:44] the restroom. Agents don't
[9:45] go to lunch.
[9:46] Agents don't go to the
[9:47] pantry to flirt.
[9:49] Agents don't do any of this
[9:50] . So, all that
[9:51] remains is
[9:52] that token consumption
[9:54] is rapidly increasing, and time
[9:55] is shrinking. So, in the
[9:57] case of
[9:58] year-over-
[9:59] year figures, the reason
[10:00] that profits are growing
[10:02] so fast and that
[10:03] nominal GDP is
[10:04] now growing is because
[10:05] the math is
[10:06] different here: you need
[10:07] to focus on
[10:08] profit margins and
[10:09] revenue growth.
[10:11] Time is running out.
[10:13] Now we actually
[10:14] do twice as much as we did
[10:15] before in one
[10:16] year. Over time, it will be
[10:17] three, four, and
[10:19] five times more. And
[10:21] the reason is that
[10:22] agents work
[10:23] constantly. This does not
[10:25] mean that
[10:26] earnings will become
[10:27] parabolic. This
[10:28] means that profits
[10:30] continue
[10:30] to outpace prices
[10:31] because much of
[10:32] it is done by
[10:33] private companies.
[10:34] They are going to
[10:35] take away shares from
[10:36] public companies.
[10:37] We will take some
[10:38] time for this. If you don't believe
[10:40] that the computer is
[10:40] insatiable. Now you
[10:42] all know what
[10:43] programming agents are,
[10:44] at least you should
[10:45] . If not, and you don't
[10:46] understand what I just
[10:47] said about working
[10:48] 24/7/365, then we'll finally
[10:49] move on to their
[10:50] actual
[10:51] deployment. They have
[10:52] n't even been
[10:53] deployed yet. We simply
[10:54] have the opportunity to
[10:55] use them.
[10:56] Corporate agents
[10:57] are becoming a permanent
[10:58] digital
[10:58] workforce. This will happen
[11:00] next year.
[11:01] Consumer agents
[11:01] will appear in the second
[11:02] half of next
[11:03] year. This is your Jarvis.
[11:05] These are assistants who
[11:06] order things for
[11:07] you. It's like your
[11:09] Amex assistant who
[11:11] does everything for you. It's
[11:14] coming, and they'll be
[11:15] working
[11:16] while you sleep. They
[11:17] will book your
[11:18] trips. They
[11:19] will order the goods
[11:20] for you. Everything
[11:20] will happen
[11:21] instantly, which will begin to
[11:22] reduce friction in
[11:24] the economy. Then you
[11:26] move on to
[11:26] autonomous driving,
[11:27] and for that you need a
[11:29] model of the world. For
[11:30] this, we need
[11:31] significantly more
[11:31] computing
[11:32] power. We
[11:33] need significantly
[11:34] more
[11:34] computing
[11:35] power for
[11:35] humanoids. This all
[11:36] appears every year.
[11:38] These
[11:39] innovations used to take
[11:39] decades, but now
[11:41] they happen
[11:42] so quickly thanks to
[11:43] intelligence. The numbers
[11:44] are staggering.
[11:45] We are now at
[11:46] this stage. This
[11:47] year we needed
[11:48] 10–100 times more
[11:49] capacity for
[11:50] coding agents
[11:51] than we did for chatbots last year
[11:52] . That's why everything
[11:53] went parabolic.
[11:54] Look at this
[11:55] step function.
[11:57] 10 more times, then
[11:58] 10 more times. This will never
[12:01] end,
[12:02] guys. The amount of
[12:03] calculations required is
[12:04] infinite. We are
[12:06] moving from
[12:07] systems that provide
[12:08] instant answers, like
[12:09] chatbots, to systems
[12:10] that plan, reason,
[12:12] and model
[12:12] outcomes before
[12:13] acting. This is very difficult.
[12:15] This is Demis Hassabis.
[12:16] We need many times
[12:17] more
[12:18] computing
[12:18] power. That's why
[12:19] Jensen Huang
[12:20] is saying—and he said
[12:21] this just earlier
[12:22] this year—that these
[12:23] will be
[12:23] AI factories, the
[12:24] new infrastructure of
[12:25] intelligence. We
[12:26] need...This is what we
[12:27] need to produce.
[12:28] That's why there are
[12:28] futures on
[12:29] computing
[12:30] power. That's why
[12:31] you keep hearing, "Oh,
[12:32] we need
[12:32] computing
[12:33] power." We
[12:34] need them
[12:35] endlessly to
[12:36] provide for the
[12:37] billions of new workers
[12:38] on the planet. Jensen
[12:40] Huang says
[12:40] computing is the new
[12:42] economy. He said
[12:43] this at a Morgan Stanley event
[12:45] in March. Nvidia this
[12:47] week: Computing for
[12:48] AI factories is becoming an
[12:49] investment
[12:50] asset class. We know this is a
[12:52] fact because Apollo,
[12:53] Blackrock, Blackstone, Brookfield,
[12:55] Goldman Sachs, KKR are creating
[12:58] AI infrastructure funding platforms
[12:59] to raise over
[13:00] half a trillion
[13:01] dollars in outside
[13:03] capital to
[13:04] build what
[13:05] Nvidia needs. They
[13:06] don't want to
[13:07] use
[13:08] their own balance anymore, and everyone
[13:09] has signed up for it.
[13:10] Morgan Stanley will facilitate
[13:11] infrastructure
[13:12] initiatives worth
[13:14] $1.5 trillion.
[13:15] There is capital. Last
[13:17] week, Alphabet raised $25
[13:19] billion. Intel
[13:20] announces a $
[13:22] 15 billion stock offering
[13:23] for a company
[13:24] that was
[13:25] worth less than $100
[13:26] billion a year ago, and they
[13:28] raised the amount to $20
[13:29] billion, with the stock
[13:30] trading above the
[13:31] offering price.
[13:33] China is in
[13:34] the same race, which
[13:35] means that computing costs
[13:37] are rising
[13:38] worldwide. You can
[13:39] read the little
[13:40] yellow lines, but
[13:41] China and the US are
[13:42] competing. The government
[13:43] is still trying to
[13:44] make sure we
[13:45] don't lose to China.
[13:46] Therefore, data
[13:47] center bonds are exempt
[13:48] from key
[13:49] securitization rules to make it
[13:51] easier for data center owners
[13:52] to sell asset-
[13:53] backed securities
[13:54] , potentially
[13:55] opening the door
[13:56] to new debt
[13:57] sales. If you
[13:59] are looking for delays in
[14:00] data centers,
[14:01] they are insignificant. We are still
[14:02] moving
[14:03] forward in everything
[14:04] related to data
[14:05] centers, despite what the
[14:06] media says.
[14:07] Delays exist. I've
[14:08] already shown you
[14:09] my work on this.
[14:10] This is already
[14:11] built into the current
[14:12] workflow. There are
[14:13] no
[14:14] surprises here. All this is already
[14:15] happening. There will be
[14:17] some delays, but
[14:18] the capacity is still
[14:19] being put into
[14:20] operation. It's not
[14:21] as fast as it could
[14:23] be, but
[14:24] honestly, it doesn't
[14:25] matter because we're
[14:26] short on memory anyway.
[14:29] So all these delays are
[14:31] even useful, because we
[14:32] don't buy too much and
[14:33] don't use up
[14:34] resources prematurely.
[14:37] Delays allow you
[14:38] to stretch income over
[14:39] time, even despite
[14:40] commitments.
[14:42] This week's Core Weave—this is
[14:43] very important for
[14:44] further development.
[14:46] You shouldn't believe
[14:47] everything they
[14:48] say. You can
[14:48] check it yourself, but
[14:50] I have a bunch of data
[14:51] that confirms what
[14:52] they say. I think
[14:53] the most important thing is
[14:54] to contrast this with
[14:55] the ideas of Michael Burry
[14:57] and Jim Chanos,
[14:58] who I mentioned, because they
[14:59] never say, "
[15:00] You know what? We were
[15:01] wrong." They
[15:02] just keep
[15:03] saying, "No, that's
[15:04] complete nonsense."
[15:05] They never
[15:05] change their minds,
[15:06] at all. There are no
[15:08] options like "
[15:09] 70% probability". They
[15:10] always say it
[15:11] will happen this way.
[15:12] If you're paying for this,
[15:14] I hope you're
[15:15] also paying for my
[15:16] content, where I
[15:16] take the time
[15:17] to explain why the demand is
[15:18] insatiable right now.
[15:20] Contracts for the A100
[15:22] have been signed until
[15:24] 2029. They were introduced
[15:26] back in 2020. They say
[15:28] everything is completely
[15:29] sold out.
[15:30] Physical
[15:30] infrastructure itself has
[15:31] scarce value.
[15:32] So I said it's worth
[15:34] investing in what's in short
[15:35] supply. Previously, they
[15:36] considered a realistic
[15:37] timeframe of two to three years.
[15:39] Now they're talking about
[15:40] six to nine years.
[15:42] New generations
[15:43] will disappear because we are not
[15:44] building enough
[15:46] data centers to replace the
[15:47] old ones, as demand
[15:48] exceeds supply
[15:49] . It's like going to a
[15:51] supermarket: if there's
[15:53] nothing there but
[15:54] canned sardines
[15:55] , you'll instantly fall in love with
[15:57] sardines. Even if
[15:59] it was the last thing
[16:00] you wanted. That's exactly what's
[16:01] happening now:
[16:02] the demand is too high.
[16:03] We have too many
[16:04] agents buying
[16:05] or wanting and
[16:06] needing too
[16:07] many tokens. You
[16:08] can
[16:09] read the
[16:10] rest of the materials yourself.
[16:11] The reality is
[16:12] that at every
[16:13] level, whether it's
[16:14] pricing, whether it's what
[16:15] we hear from
[16:16] them, whether it's from SpaceX or from
[16:17] Google, what we see from
[16:18] Google. Regarding Google,
[16:21] an article and podcast
[16:23] I listened to this
[16:24] week about the AI
[16:25] brain drain from
[16:26] Google on the All-In podcast.
[16:28] They talk a lot
[16:30] about
[16:31] Google's reorganization, calling it a
[16:31] change in
[16:32] capital allocation. I
[16:33] will elaborate
[16:34] on this in a little more detail. Um, the race between the
[16:36] two leaders, according to
[16:36] David Sachs,
[16:38] I completely agree.
[16:39] AI infrastructure
[16:39] can be a more
[16:40] profitable business
[16:41] than
[16:42] model development. So
[16:42] think about it this way.
[16:44] Google is making a decision
[16:45] similar to what
[16:46] Meta and SpaceX did:
[16:47] maybe we should
[16:48] sell some of our
[16:49] computing
[16:50] power
[16:51] because we can't
[16:52] use it all
[16:53] right now, and our
[16:54] investors are angry
[16:56] because free cash
[16:57] flow is going negative
[16:58] . This forces you to
[16:59] do business
[17:00] that guarantees
[17:01] return on
[17:02] invested
[17:03] capital (ROIC) instead of
[17:04] chasing
[17:05] AGI. The problem is that
[17:07] talent wants to
[17:08] chase AGI. So, you
[17:09] see Jeff Dean
[17:10] or Demis Hassabis
[17:11] maybe leaving. We
[17:13] have reached an important
[17:14] tipping point, and this is
[17:15] manifested in the fact
[17:16] that if talent leaves
[17:17] , multipliers
[17:18] shrink, it's
[17:19] simple. I don't see any
[17:20] other way out
[17:21] unless Google finds
[17:22] some other business,
[17:24] but the new businesses
[17:25] they could create
[17:26] , in my opinion, would have to
[17:27] be related to
[17:28] advanced models,
[17:29] or at least that's my
[17:31] opinion, that's more likely
[17:32] . Brad Gerstner
[17:33] gave a
[17:34] huge, important
[17:35] speech about this.
[17:37] Again, Nvidia has
[17:38] funding limits,
[17:39] which is why they
[17:40] supported all these
[17:41] companies. Um, he
[17:42] was talking about all these
[17:44] different things. Advanced
[17:45] labs are willing to
[17:46] overpay for
[17:47] computing due to
[17:48] scarcity, four or
[17:49] even five times the
[17:50] market
[17:51] price, to get
[17:52] scale
[17:53] . This gives them a
[17:54] competitive advantage
[17:55] because others
[17:56] cannot obtain them.
[17:57] So buyers
[17:57] are concentrated in one
[17:58] hand because they have
[17:59] income and want
[18:00] to provide themselves with
[18:01] computing. And that's what
[18:02] he's talking about.
[18:03] The situation is very
[18:04] concentrated.
[18:05] No one disputes the
[18:07] cyclical nature of
[18:08] Anthropic and OpenAI. At the same time,
[18:09] Nvidia itself provides a lot of
[18:10] credit to
[18:11] ensure that these
[18:12] people can
[18:13] buy their chips.
[18:14] It's a vicious circle.
[18:16] That's reality,
[18:17] guys. And, unfortunately, AI is
[18:19] not at all what
[18:20] you have seen before.
[18:22] Demis Hassabis wants to
[18:23] leave Google, but he's
[18:24] staying, even though it's
[18:25] predicted he
[18:26] could leave within a
[18:27] year. Again, this
[18:28] brings us back to the topic of
[18:30] AGI. Moonshots
[18:31] broke it down in detail in two
[18:33] podcasts this
[18:34] week. One with Emad
[18:36] Mostak, the other with a
[18:37] guy named Kush.
[18:41] I think Alex said
[18:42] it best. Those who
[18:44] cannot compete in
[18:45] advanced models
[18:47] move to
[18:47] computing. This is what
[18:50] Meta, xAI, and Google have done,
[18:52] and their models lag far behind the
[18:54] leading edge. Advanced
[18:56] models will be
[18:57] important.
[18:58] Open source models
[18:59] will also be
[19:00] important, but Google
[19:01] may win in
[19:02] distribution. So,
[19:03] they
[19:04] cover that as well. They are
[19:05] talking about
[19:06] Sergey Brin's position. You
[19:07] can listen to it
[19:08] yourself. I'll
[19:09] summarize a few
[19:10] points at the end, but the main
[19:11] point is that
[19:12] Google's real business
[19:13] is the cloud and
[19:14] infrastructure. If
[19:15] that's the case, what
[19:17] multiplier should they have, if
[19:18] not through Gemini? They
[19:19] may lose in the
[19:20] prestige race of
[19:21] models, but still
[19:23] earn a lot from
[19:24] AI infrastructure;
[19:25] they could become the “
[19:26] OPEC” of computing
[19:27] while OpenAI, Anthropic, XAI, and
[19:29] others compete in the
[19:30] intellectual
[19:31] realm.
[19:34] Google's employer brand
[19:35] has deteriorated, and that's where Kush
[19:36] comes to the rescue
[19:37] . He says that for
[19:38] his MIT cohort and
[19:40] young developers,
[19:41] Google is no longer a
[19:42] dream place. The dream places
[19:43] now are OpenAI, Anthropic,
[19:45] XAI, and Frontier Labs. Unfortunately,
[19:48] this is the reality. So
[19:50] , these companies are
[19:51] different. This is SpaceX
[19:55] . This is a private
[19:55] company that is
[19:56] preparing to go
[19:57] public. This is another
[19:58] private company. This is
[20:00] Frontier Labs. This used to
[20:03] be the case with Google and others.
[20:04] They invented much
[20:06] of what is modern
[20:07] AI, but the spiritual
[20:07] center of AI
[20:08] has moved elsewhere
[20:09] . This is an important
[20:10] dynamic, and one of the
[20:11] reasons why
[20:12] I
[20:12] continue to
[20:13] argue about hyperscalers: it's worth
[20:14] betting on those
[20:15] who get capital,
[20:16] because the demand for
[20:17] infrastructure is
[20:17] insatiable. These guys
[20:19] are proving it, because Google is
[20:20] losing talent
[20:21] due to a lack of
[20:21] computing
[20:22] power. If
[20:24] they had enough
[20:25] computing
[20:25] resources, they could
[20:26] do both
[20:27] advanced models
[20:28] and business. But they ca
[20:30] n't. They don't have
[20:30] enough
[20:31] computing
[20:32] power. Therefore, they
[20:33] have to
[20:33] direct these
[20:34] limited resources
[20:35] somewhere else.
[20:36] Google has allocated
[20:36] approximately 200
[20:37] billion in
[20:37] capital expenditures
[20:38] this year for
[20:39] computing and
[20:39] infrastructure. This
[20:40] capital yields a
[20:41] documented 30%
[20:42] return on
[20:43] investment (ROIC). This is a
[20:44] tax advantage
[20:45] due to accelerated
[20:45] depreciation. This is almost
[20:46] guaranteed demand,
[20:47] as the entire industry
[20:48] suffers from a lack of
[20:48] computing
[20:49] power.
[20:50] Model development,
[20:50] on the other hand, requires
[20:51] tens of billions,
[20:52] has high
[20:53] variability, and
[20:54] no guarantees of
[20:54] profit. Again
[20:55] , this brings it all back
[20:57] to the argument that
[20:58] was there. So, I
[21:00] asked them to take
[21:01] a transcript
[21:06] of Brad Gerstner's
[21:07] All-In podcast, and then—successful
[21:09] moonshots. So, these
[21:10] answers are here: if
[21:12] you guys have
[21:13] n't done it yet, gather
[21:14] all the smart people
[21:15] around one table and
[21:16] ask them questions.
[21:18] What do Google's personnel changes
[21:19] say about
[21:20] incentives for
[21:22] multistaking computing professionals? That's
[21:23] how I like to break
[21:24] these things down, and then I
[21:25] move on to the
[21:26] market disruption side.
[21:28] Synthesis for all
[21:29] large companies:
[21:30] durable assets
[21:31] of scale are capital,
[21:32] physical
[21:33] infrastructure,
[21:34] distribution, and
[21:35] regulatory position.
[21:36] The rational end
[21:37] goal, toward which Google is either
[21:38] groping or
[21:39] moving purposefully, is a large
[21:40] company as a
[21:41] manager of capital
[21:42] and infrastructure
[21:43] landlord for an
[21:43] ecosystem of mission-
[21:44] oriented
[21:45] structures that it
[21:46] partly owns.
[21:46] Companies that
[21:48] resist this,
[21:48] trying to keep
[21:49] everything in-house, will
[21:50] lose talent and
[21:51] be left with
[21:52] costs. I fully
[21:54] support this view.
[21:58] SpaceX, SemiAnalysis
[21:59] published this. Elon
[22:01] shocked the world again
[22:02] during
[22:03] SpaceX's first earnings report,
[22:04] announcing 6-8
[22:05] gigawatts of additional
[22:07] data
[22:08] center capacity in 2027 alone
[22:09] . Perhaps much more than
[22:10] 10 gigawatts.
[22:12] It sounds impossible,
[22:13] but we believe it. We
[22:14] see SpaceX
[22:15] moving towards a
[22:16] rate of about 10
[22:17] gigawatts per year in 2027
[22:19] . This is why SpaceX has
[22:20] value for the
[22:21] company. And again
[22:23] , it's worth
[22:23] considering the other side
[22:25] as a call option, but
[22:26] Elon Musk has already proven
[22:27] that he can build data
[22:28] centers faster than
[22:29] anyone. And gigawatts are
[22:31] very expensive,
[22:33] as we can see from the
[22:34] prices of
[22:35] computing
[22:35] power. 80% of their
[22:38] output is intended
[22:40] for space data
[22:41] centers. So, again
[22:43] , he does
[22:44] a lot for Terrafab, which is why
[22:46] I
[22:47] mentioned it. Terrafab is just
[22:50] confirmation of what is
[22:51] happening.
[22:52] Just look at
[22:53] this thing he
[22:54] showed, in terms of what the
[22:56] scale of all this would look like.
[22:57] I mean, you're
[22:58] talking about
[22:59] him building something. Elon
[23:01] Musk knows where
[23:02] the computing
[23:03] side is, and he's
[23:04] basically telling you, "
[23:05] I need to
[23:06] build this, which
[23:07] looks like a
[23:08] Star Trek or
[23:10] Star Wars situation, and it
[23:11] takes a
[23:12] huge amount of
[23:13] land. He needs
[23:15] it for the chips, and
[23:16] specifically for the memory.
[23:18] SpaceX has an
[23:19] exclusive deal with
[23:21] Nvidia, putting Vera Rubin GPUs
[23:22] into orbit. And again
[23:24] , that was part of the
[23:25] reason why when I
[23:26] talk about Nvidia and
[23:27] break it down, he
[23:27] said, "We think the
[23:29] Vera Rubin architecture is the
[23:29] best
[23:30] architecture." We're
[23:31] working exclusively
[23:33] with Nvidia. But
[23:34] he's building Terrafab. ​​So again
[23:36] , while you're
[23:37] listening to this, and I'm picking on
[23:38] people who are
[23:40] skeptical, I want to
[23:41] remind you. So for RIA, this is the
[23:43] same
[23:44] one-
[23:45] page document that I'm
[23:46] sending out to people right now to help
[23:48] them understand
[23:49] why my research is
[23:50] focused on
[23:52] figuring out what's
[23:53] going on in AI
[23:54] that I don't think
[23:55] can be covered. Wall
[23:56] Street. You're definitely not
[24:00] going to get
[24:01] research
[24:02] put together like this
[24:03] . Know what's
[24:05] changed and whether it
[24:07] matters. To provide
[24:08] all of this information,
[24:09] today's video
[24:11] is a hundred slides.
[24:12] I'm taking you on a
[24:13] journey to educate
[24:14] you through
[24:15] data-driven material
[24:16] . You can
[24:17] download this
[24:18] anywhere. You can
[24:19] bring your own
[24:19] agents to work with
[24:20] it. You can
[24:21] verify everything I
[24:22] say. You can
[24:23] get a transcript. I
[24:24] do this every week.
[24:25] Every week I come in with
[24:27] over a hundred slides of
[24:28] new information. I
[24:30] listen to
[24:31] podcasts nonstop. And that
[24:32] way you can come in
[24:33] and see what's
[24:34] changed in
[24:35] real time. You can't
[24:36] get that
[24:37] from Wall Street.
[24:38] Know where it matters
[24:39] and when the
[24:39] tipping point is.
[24:40] Not only do I give you
[24:42] names, but I also spend
[24:43] time every week—as I'll
[24:44] show you soon—
[24:44] telling you
[24:45] what I think the
[24:47] risk-
[24:47] reward ratio is in the
[24:48] market. And if you don't
[24:49] believe me, I'll show
[24:50] you how you can you can
[24:51] do it
[24:52] yourself, and I'll give you
[24:53] a hint on how to
[24:53] do it. The
[24:54] agent question is important. I
[24:56] give you
[24:57] prompts—one of which I'm
[24:58] giving you this week—
[24:59] workflows,
[25:00] Excel files with all the
[25:01] information,
[25:02] guides on how
[25:03] you can get started right away,
[25:04] and
[25:05] hopefully things
[25:06] like
[25:07] Jensen Huang's knowledge base that
[25:08] not
[25:09] only helps you but also your
[25:10] kids, which is what I've heard
[25:11] from them. But
[25:13] next year,
[25:14] starting in September, the
[25:15] next constraint
[25:17] is not physical, it's
[25:18] financial.
[25:20] Fuses,
[25:21] consumer agents.
[25:23] Settlements,
[25:24] stablecoins.
[25:26] Tokenization—it's
[25:27] going to affect all of your
[25:28] portfolios. The bridge between
[25:30] traditional
[25:30] portfolios and
[25:31] assets on the network,
[25:32] your
[25:33] clients' assets are moving
[25:34] onto these rails. You
[25:35] should, of course,
[25:36] learn all of this
[25:37] before it happens.
[25:38] Go to Mark.
[25:40] Mark will guide you.
[25:42] So, I've said many times
[25:46] in this
[25:47] piece that
[25:47] the economic structure is
[25:49] changing. The
[25:52] market structure is changing and the
[25:53] income structure is changing. So
[25:55] earnings history, economic
[25:56] history, market
[25:58] history -- all of that becomes
[25:59] useless. That's what this is all about
[26:00] .
[26:01] There's a
[26:03] regime change. So if you
[26:04] listen to economists
[26:04] who say the
[26:05] labor market is too
[26:06] strong and we need to
[26:07] raise rates.
[26:07] Inflation is too
[26:08] high, we need to
[26:09] raise rates.
[26:10] We're seeing a
[26:11] disconnect here that we've never seen
[26:12] before. On the
[26:14] wall side, on the
[26:14] AI-agent market structure side
[26:16] , we're seeing a
[26:17] wall of
[26:18] technical momentum coming up,
[26:19] while the NDX wall is barely
[26:21] moving. Earnings are
[26:24] rising rapidly.
[26:25] Companies can
[26:26] grow earnings
[26:27] because the times
[26:28] are changing. This
[26:29] portfolio is designed
[26:30] to demonstrate all of
[26:32] that. And again, you're
[26:33] dealing with something that has
[26:35] grown a lot,
[26:37] has consolidated
[26:39] , and now has bounced back and is
[26:41] up 50%. I gave
[26:44] you this as a
[26:45] list of 25, 10, 100 names,
[26:46] and then Morgan Stanley gives
[26:48] you different ways to play
[26:50] on that. As for
[26:53] why, Why
[26:54] can't Wall Street embrace it?
[26:55] Research firms
[26:57] were built
[26:57] vertically. They're
[26:58] built on
[27:00] silos. There's a
[27:01] reason I
[27:02] left Morgan Stanley in
[27:03] 2001, and it wasn't
[27:04] just because of 9/11, it was
[27:05] because 9/11 made me
[27:06] realize that life is
[27:08] too short, and
[27:09] if my job as a 30-
[27:10] year-old man
[27:11] who's now laying off
[27:12] people by profession
[27:13] is what my
[27:14] life is going to be, I need to
[27:15] move on. They
[27:17] can't embrace everything all at
[27:19] once. That's impossible.
[27:22] AI is not a sector. It
[27:25] embraces
[27:25] industrials,
[27:26] materials, chemicals,
[27:27] energy,
[27:28] power,
[27:29] semiconductors. Marvell
[27:32] is not the same
[27:33] semiconductor as
[27:35] Infineon or Nvidia or Intel. They're
[27:37] all different, guys, and
[27:38] you know that. So
[27:39] my job is to show
[27:40] you the structure that's
[27:42] changed,
[27:43] silos don't see
[27:44] through. The speed
[27:45] at which we're moving is
[27:47] at least five times faster than the speed at which we're moving.
[27:49] of what the world used
[27:50] to look like, what
[27:51] academics do,
[27:52] any
[27:52] backtesting,
[27:54] anything about earnings,
[27:55] every time we
[27:56] see earnings like that,
[27:57] it's a bubble. Okay,
[27:58] but we're not
[27:59] hiring, and
[28:00] hiring usually happens
[28:01] before a bubble. We
[28:03] hire too much during
[28:04] bubbles. The
[28:05] synthesized
[28:06] averages destroy
[28:07] the signal. Too many
[28:08] people, slow
[28:09] decision-making, the
[28:10] friction that comes with
[28:11] big companies, that's what's
[28:12] going to kill them
[28:14] compared to
[28:15] AI-native companies. I'm an
[28:17] AI-powered company. Everything
[28:18] you see right now
[28:19] is built
[28:20] with AI, and it
[28:22] took me a
[28:23] while to build it. That's
[28:24] just the easiest
[28:25] way to get it across to
[28:26] you. The organizational
[28:27] structure that makes a
[28:28] firm great at
[28:29] sector coverage
[28:30] is the structure that
[28:31] makes it
[28:32] slower at
[28:33] sector regime change. It's the
[28:34] same
[28:35] property,
[28:36] viewed from two different
[28:37] angles. So if
[28:38] you pay for
[28:39] research, I'm giving
[28:40] you something, that's at
[28:41] least
[28:41] different from what
[28:42] you're getting
[28:44] from other people. That doesn't
[28:45] mean it's better,
[28:47] but it's different, and I
[28:48] guarantee you there's
[28:49] something
[28:50] new and new work coming out every week. That's
[28:53] proof that the NASDAQ is
[28:55] up
[28:57] 19-20% year to date. The 100-
[29:00] name portfolio is up 46%, the 25-
[29:02] name portfolio is up 74%, and the 10-name portfolio
[29:05] is up 94%. And those 10 names are
[29:07] really good 10 names.
[29:08] They're highly
[29:09] correlated.
[29:10] In fact, if you
[29:11] need those 10 names, I
[29:12] showed them in
[29:13] this
[29:14] week's subscriber webinar: the
[29:14] weighted average
[29:15] return is
[29:16] 103%. So I showed
[29:18] you 94%. The median
[29:19] return is 65%
[29:21] . The best is up
[29:23] 249%. The
[29:24] worst is up 16.5%. That's
[29:27] Nvidia. It's outpacing
[29:28] the S&P by 13%. Eight
[29:30] of the ten are up
[29:31] more than 40%.
[29:32] The average return is
[29:33] about 7.9 times the
[29:35] S&P's YTD. The
[29:36] median is five
[29:37] times. From a
[29:38] valuation perspective right now
[29:39] , the average
[29:40] forward
[29:41] PEG ratio
[29:42] is 0.8. The median is
[29:44] 0.81. That's true for
[29:46] these 10 names.
[29:47] Despite the
[29:48] 103%
[29:49] YTD growth, it's all
[29:50] due to
[29:51] rapid
[29:52] earnings growth. The median
[29:54] earnings growth forecast for
[29:55] next year
[29:56] is 40%. Seven of the
[29:57] 10 names have
[29:59] earnings growth expectations of over 30%.
[30:01] Why would you look for
[30:02] other names? This
[30:04] will continue because
[30:05] the demand for
[30:06] computing
[30:07] power is
[30:07] insatiable.
[30:09] Multiples
[30:10] could get squeezed. And
[30:13] when we go through this again
[30:14] , my job
[30:15] as a
[30:16] risk-
[30:17] reward thinker is to get
[30:19] more negative
[30:20] when it looks like it's
[30:21] time to
[30:22] consolidate as we
[30:23] navigate the
[30:24] AI ​​slowdown
[30:25] cycle on June 8th. So
[30:29] here's a tip, guys.
[30:31] Take my
[30:32] transcripts as much as you
[30:33] want. I've done everything
[30:34] since May
[30:35] just to give
[30:36] you a little something, and you just
[30:37] take this prompt.
[30:39] It'll be uploaded
[30:40] over the weekend. Its
[30:41] job is
[30:41] to analyze a
[30:42] timeline of
[30:42] weekly
[30:43] market commentary transcripts
[30:44] to determine how the
[30:46] speaker's long-term market confidence,
[30:47] short-term
[30:48] outlook, and
[30:48] investment
[30:49] risk appetite have changed over
[30:50] time. It's an
[30:50] unbiased thing that
[30:51] you can do
[30:52] yourself. It's
[30:53] a prompt that
[30:54] processes the data, and I'll
[30:55] show you
[30:55] the results. You can
[30:56] do this not only with
[30:57] mine, but I'm
[30:58] sure there are other
[30:59] podcasts that you
[31:00] listen to every week.
[31:01] Measure their tone.
[31:03] All it has to do
[31:05] is convey... So here's
[31:06] how it works: on a
[31:07] structural basis, I'm
[31:09] always
[31:10] bullish. On a
[31:11] tactical basis, it
[31:13] changes. So I
[31:14] think the demand for
[31:15] computing is
[31:15] insatiable. I want to
[31:16] stay in this
[31:17] deal for the next 3-5
[31:19] years. But I want
[31:20] to be tactical and
[31:20] make sure that my
[31:21] risk lower when the
[31:22] tactical level is
[31:23] low and the
[31:24] risk-
[31:25] reward ratio is low.
[31:26] Selective de-
[31:27] risking in May.
[31:28] So that's the
[31:29] risk position that indicates the
[31:30] current stage of the cycle.
[31:31] I said we're
[31:32] in a
[31:33] parabolic
[31:33] exhaustion phase in memory.
[31:34] Selective de-
[31:35] risking. You can
[31:37] go through defensive,
[31:38] selective risk consolidation,
[31:39] and neutral
[31:40] consolidation. Those are
[31:42] just a few of those
[31:42] stages. That's what I
[31:44] was talking about on May 17th. I was
[31:45] gradually getting out of
[31:46] assets. In fact,
[31:47] as far as I can
[31:48] tell, I got out of
[31:49] everything. Definitely
[31:50] got out of Micron completely
[31:51] . So if people
[31:52] call me and
[31:53] say they bought on
[31:55] May 17th or did
[31:56] something else, I
[31:57] try to give you a
[31:58] risk-
[31:59] reward assessment of what's
[32:00] going on and
[32:02] try to
[32:02] focus on
[32:03] preserving or
[32:04] rotating capital with a
[32:05] real reduction in the
[32:07] beta of the portfolio. I
[32:08] talked about
[32:09] the need to
[32:09] reduce
[32:10] leverage. The
[32:11] de-risking
[32:12] goes from
[32:13] partial to almost
[32:14] complete on
[32:15] overheated stage.
[32:16] So, again,
[32:17] if you don't believe
[32:18] what I'm saying, here's what
[32:19] it's going to give you. So it
[32:22] takes all these numbers and I
[32:23] overlay them on my
[32:24] thematic portfolio.
[32:26] So for those of you,
[32:27] especially the
[32:28] asset managers that I
[32:29] talk to regularly,
[32:30] it's a way to track and
[32:32] prompt and see
[32:33] how I'm doing. So I'm
[32:34] starting to get more
[32:35] cautious. This is a yellow
[32:37] light. It doesn't mean I
[32:38] have to sell everything right away
[32:39] . It means
[32:40] I'm getting more
[32:41] cautious. I'm reducing my
[32:43] Micron holdings. The red
[32:44] zone is when it's a signal
[32:45] to stop all
[32:46] investment right
[32:47] now. I'm starting to get
[32:48] more worried. I'm
[32:49] seeing some signs.
[32:51] Here's what happened in
[32:52] those meetings. Each
[32:53] of these dots is one
[32:54] of the videos. These are
[32:55] transcripts. Then I
[32:57] start to get back
[32:58] to neutral
[32:59] . And then the
[33:01] first time I
[33:02] start to say I'm
[33:03] buying assets is
[33:05] when Micron fell below 900
[33:06] , this is here. And now you
[33:07] start to see the
[33:08] green dots where I I
[33:09] 'm starting to get
[33:10] more optimistic.
[33:11] So back then I said
[33:12] we need
[33:13] consolidation. I
[33:14] was showing you
[33:15] consolidations, price and
[33:17] timing. Even here I was
[33:19] still saying we're
[33:20] never going to be able to
[33:21] guess the bottom. So those of
[33:23] you who want me to
[33:24] try to guess this
[33:26] or that, nobody can
[33:27] do that, and
[33:28] certainly not me. And I'm not
[33:29] trying to
[33:30] do that. What I want to
[33:31] do is reduce
[33:33] risk here and then
[33:34] add back in
[33:35] when I think there's been
[33:36] enough
[33:37] panic and
[33:38] consolidation. So I'm
[33:39] looking for a combination, and that's what
[33:40] makes them
[33:41] green here. As you're
[33:42] starting to hear from
[33:43] me, I think we're
[33:44] closer to
[33:45] capitulation. I see the
[33:46] VIX going up. I see it
[33:48] going on. I hear
[33:49] about it. Then the
[33:50] situational awareness comes in
[33:51] and suddenly
[33:53] I'm giving a very positive
[33:54] signal about things and
[33:55] saying, I believe this is
[33:56] it. The
[33:57] risk-reward ratio has
[33:58] now changed. You
[33:59] can
[34:00] use that,
[34:01] create your
[34:01] own. And throughout
[34:03] this
[34:03] consolidation I was
[34:04] creating a
[34:05] consolidation guide to
[34:06] try to get
[34:07] people in a certain position
[34:09] . Two, three weeks ago, the
[34:10] week before
[34:11] situational
[34:11] awareness, I
[34:12] put out a piece
[34:13] called "
[34:14] Sifting Through the Wrecks."
[34:15] To give you
[34:16] a chance to look for
[34:17] things. So here's what
[34:19] happened. We've already
[34:21] played through over 50% of
[34:22] the decline. It was a
[34:24] situational
[34:25] awareness day. So,
[34:26] again, my goal
[34:28] is not to pick
[34:29] tops and bottoms for you. It's
[34:30] really to help
[34:31] you navigate to
[34:32] stay with the
[34:33] structural trend,
[34:34] minimize losses
[34:35] and maximize
[34:36] profits. Plain and simple
[34:37] . That's the
[34:38] game. You maximize
[34:39] profits, you
[34:40] try to find
[34:41] things with a
[34:42] risk-reward ratio
[34:43] in your favor where you
[34:44] feel like everything has
[34:45] cleared up, and that's why the
[34:45] 50-day rate of
[34:46] change and some of the
[34:47] tools that I
[34:48] showed in the subscriber webinar
[34:49] that I'm working on are
[34:50] algorithms for
[34:51] identifying time
[34:52] risks. So I take what I
[34:53] just showed
[34:54] you and I incorporate all of that
[34:55] into an algorithm so that,
[34:56] I hope to give you a
[34:57] score that shows you where I
[34:58] think the risk is
[34:59] . For those of you who have
[35:00] never been on here,
[35:01] I have a
[35:02] website now. I just
[35:03] wanted to show you this
[35:04] for one reason,
[35:04] okay? This is Visser Labs.com,
[35:06] great. Here's my
[35:07] logo. This is linear
[35:08] motion. This is
[35:09] exponential motion
[35:10] . Academic here.
[35:12] Entrepreneur here. The
[35:13] bottom line is, when you
[35:14] go to my site,
[35:15] I want you to
[35:16] go there. It
[35:17] took me about an hour to
[35:18] build. It took me
[35:19] about a
[35:21] day to work
[35:22] with Network Solutions to get
[35:24] it up and running, where I
[35:25] secured the
[35:26] VisserLabs.com domain. But it
[35:30] literally took me about an
[35:31] hour to build the site. So I'm telling you
[35:34] this because when Network
[35:35] Solutions did the score,
[35:37] it got a 100 out of 100.
[35:39] If you're
[35:40] using AI and you
[35:41] 're using it
[35:42] right, all I
[35:43] did was type in
[35:44] the queries to build
[35:46] the site. It's not
[35:46] perfect, it still
[35:47] needs some tweaking,
[35:49] but it works,
[35:50] guys. You can
[35:51] build
[35:52] things like this, and it comes back to the same thing
[35:53] . My
[35:54] job here is to
[35:54] help you
[35:55] prepare for the
[35:56] future, not just
[35:57] in terms of AI, to
[35:58] navigate
[35:59] it, to deal with the
[36:00] parabolas and the
[36:01] bubbles and everything
[36:02] that happens, but to
[36:03] make you
[36:03] smarter about
[36:04] using it. And
[36:05] to prepare you for
[36:06] it. The next
[36:07] constraint is financial
[36:09] . Cryptocurrency for
[36:10] AI Agents. I'm telling you
[36:11] , mark my
[36:12] words. If you want to
[36:14] come to me in
[36:15] a year when I'm wrong,
[36:16] it'll be because of this.
[36:17] So, one company
[36:18] that reported—Figure.
[36:21] I talked a lot about
[36:22] Figure with Anthony
[36:23] Pompliano this
[36:24] week. I highly
[36:24] recommend
[36:25] watching
[36:26] this week's video. I think we
[36:27] touched on some interesting
[36:28] topics, but more importantly,
[36:29] my brain has
[36:30] been working on
[36:31] a lot of different
[36:32] things related
[36:33] to time today, which is what I'm
[36:33] focused on right now.
[36:35] I'm spending a lot
[36:36] more time in the
[36:36] crypto community. I'm
[36:38] speaking in a lot more
[36:39] , um, places.
[36:40] I'm speaking at the
[36:41] Pomp event. I'm
[36:42] speaking at the
[36:44] Bitcoin Policy Institute event in
[36:45] Washington. There's a lot
[36:46] going on. I've probably got
[36:48] four
[36:49] podcasts coming out this week where I'm
[36:50] speaking. We're at a
[36:52] point now, guys,
[36:53] where I'm going to be doing
[36:53] more crypto stuff,
[36:55] and you're going to see
[36:55] more content. So
[36:57] go to the site.
[36:58] Go to the
[36:59] webpage, but Figure
[37:00] is a perfect
[37:01] example of how that's starting
[37:02] to happen. Figure just
[37:04] released its
[37:05] earnings report.
[37:07] Consumer lending volume grew
[37:08] 4. It was 4.3
[37:09] billion, up 132%
[37:10] year over
[37:12] year. The
[37:14] tokenized on
[37:15] -chain marketplace, Figure Connect, is expected to be
[37:16] close to 70% of
[37:18] volume. We're building a
[37:19] modern
[37:20] capital market in a
[37:20] paradigm shift in
[37:22] capital markets toward
[37:23] tokenization and
[37:23] standardization. That's
[37:25] happening now,
[37:26] guys. That's why I
[37:27] built it. This is my
[37:29] ecosystem,
[37:30] correlated with
[37:31] Bitcoin. These are
[37:34] year-to-date numbers, so they're
[37:35] normalized. If I
[37:36] didn't normalize them,
[37:37] it would be a direct
[37:38] overlay.
[37:39] Except what you
[37:40] see is that the ecosystem is
[37:41] trading better than
[37:41] here. I've added a
[37:43] few more names here. I work
[37:44] with a bunch of crypto people.
[37:45] If anyone wants to
[37:46] reach out and wants to be a
[37:47] part of this
[37:48] process, I
[37:48] like to
[37:49] do that. I need
[37:50] , I have 10
[37:51] verticals. I have
[37:52] about four or
[37:53] five names in each. There are
[37:54] about 45 names in the index
[37:56] . Bitcoin is the
[37:57] S&P 500
[37:58] of the future. I'll
[37:59] keep
[38:00] saying this to anyone
[38:01] who wants to listen. Figure
[38:03] is part of it. People
[38:06] will be running around on
[38:07] the internet.
[38:08] Bot traffic will exceed
[38:09] human traffic by 1000
[38:10] times within 5
[38:11] years. Elon Musk on
[38:13] Cloudflare's prediction:
[38:14] Internet traffic from AI
[38:16] agents
[38:17] will significantly exceed
[38:17] human usage.
[38:19] Guys, go
[38:20] make money on this.
[38:21] Transactions will
[38:22] happen.
[38:23] Stablecoin
[38:25] card spending is up another 16%
[38:27] . 16%. You'll want to
[38:30] do an ARR on
[38:31] this.
[38:35] US banking regulator
[38:36] opens up
[38:36] national licenses
[38:37] for Bitcoin and
[38:38] Crypto Firms. The SEC is about to take
[38:40] action while
[38:40] Congress debates the
[38:43] Transparency Act. Just like
[38:44] AI, there’s no
[38:45] stopping crypto,
[38:46] guys. This is going to continue
[38:48] even if the
[38:49] Transparency Act doesn’t
[38:50] pass. So if you’re
[38:51] waiting for this act
[38:53] or thinking there’s going to be a
[38:54] sell-off on the news,
[38:55] I’m telling you: in a year’s time, we’re going
[38:57] to see a
[38:57] volume explosion in this space.
[38:59] We don’t know where
[39:00] all the markets are going to be, but
[39:01] I do know that
[39:02] the volume is going to be high. The
[39:03] financial
[39:04] fuses are going to be
[39:05] on. I wrote
[39:06] this and I posted it
[39:07] on Substack this week and
[39:08] on my
[39:09] subscriber site. I
[39:11] put it out there because
[39:12] I think it’s really
[39:13] important. It’s
[39:14] about
[39:15] the intervention in the yen.
[39:19] It’s the nexus of everything that’s
[39:21] going on between AI and
[39:22] the macro economy:
[39:24] debt, deficits,
[39:25] Japan’s debt, US debt. It’s
[39:28] all laid out here. This is to
[39:30] understand: I
[39:31] think we’re going to look back
[39:32] on this
[39:33] week as a moment of
[39:34] realization that there
[39:35] was an
[39:36] intervention. dollar-
[39:37] yen and Kevin Warsh has
[39:38] actually lost
[39:39] people's trust. And Louis
[39:41] Gav released this
[39:42] piece this week.
[39:43] Like me, he gave it to
[39:45] everyone. He does
[39:47] n't usually do that. I don't
[39:48] do this for my
[39:49] paid content,
[39:50] but that's how
[39:51] important this
[39:52] week was for me. And
[39:53] then he goes on
[39:54] to something equally
[39:55] important on the same
[39:56] topic that I wanted to
[39:57] show you. But this is
[39:58] the most important thing: he
[39:59] said, "AI and
[40:00] stablecoins can
[40:01] change, not just
[40:02] destroy, the dominance of the
[40:03] dollar." Gav Cow
[40:04] raises the
[40:05] provocative idea
[40:06] that a "
[40:07] computational dollar
[40:08] " will replace the petrodollar.
[40:09] "There's that word again—
[40:10] computation." The
[40:12] insatiable demand for
[40:13] computation and the
[40:14] insatiable race for both
[40:15] computation and the
[40:16] capture of
[40:17] the crypto landscape. AI
[40:19] is becoming more
[40:20] than just a
[40:21] technology story.
[40:22] Computers
[40:23] are becoming an
[40:24] industrial resource, a
[40:25] geopolitical asset,
[40:26] and perhaps a pillar of the
[40:27] financial system.
[40:29] This is the other side of a
[40:30] debt world, where
[40:31] we just
[40:31] keep doing
[40:33] whatever it takes to
[40:34] keep up the debt
[40:35] and the spending. It's like "
[40:36] doing everything we can"
[40:37] with the limited
[40:39] firepower of the yen. That's
[40:40] why it's so important:
[40:42] whether it's controlling the
[40:43] yield curve or
[40:44] raising the FEMA cap
[40:45] and them stepping in—it does
[40:47] n't really
[40:48] matter. Scott
[40:50] has it tough.
[40:51] He didn't create this
[40:52] problem, but they're
[40:53] trying to jump out of it
[40:54] , and the problem is
[40:55] they're not
[40:56] making any money
[40:57] and they're trying
[40:57] to spend their way out of
[40:58] it. So
[40:59] the debt and the deficit aren't
[41:00] coming down
[41:02] fast enough, if
[41:03] at all.
[41:05] Wall Street sees
[41:06] the anxiety in the
[41:06] bond market, and we're
[41:07] seeing it. The
[41:09] bond market just doesn't
[41:10] want to recover.
[41:11] We had a week of
[41:13] unexpectedly low inflation
[41:14] . Retail sales
[41:15] missed
[41:15] expectations. The jobs
[41:16] report
[41:18] was also weak the week before, but
[41:19] long-term
[41:20] yields are still
[41:21] not falling. They're not
[41:23] rising, but they're not falling either.
[41:25] And the yen hasn't
[41:27] changed much since
[41:28] intervention. And this is
[41:31] despite the fact that now we
[41:33] have 31% for September
[41:34] for Warsaw and 80% for the
[41:36] Bank of Japan. These
[41:38] figures were about
[41:40] 50 to 5 0 before
[41:40] the release of
[41:42] salary data. Now
[41:43] we are here, but we still don't
[41:44] see any change. I
[41:46] believe this is a mistake
[41:47] by the markets, and you should
[41:48] buy assets that
[41:49] protect against
[41:49] depreciation of money.
[41:51] You should buy
[41:52] gold, silver,
[41:53] Bitcoin, but despite this,
[41:54] until we see a
[41:55] decline in yields,
[41:56] nothing will happen.
[41:58] So, Warsh's attacks
[42:00] continue, and he
[42:01] will speak in
[42:01] Jackson Hole. This is
[42:04] a message from
[42:05] Bill Dudley. Warsh
[42:07] hinted that the PCE index
[42:08] could be dropped
[42:10] as the
[42:11] Fed's primary inflation gauge as early as
[42:12] January. Who knows? After
[42:14] January, we might
[42:15] talk about
[42:16] strategy. He said,
[42:17] "I suspect the
[42:18] working group has something
[42:19] to add." Why did I want
[42:21] you to read this?
[42:23] This is an academic
[42:24] question. That's exactly the
[42:25] problem. This is what
[42:26] Kevin Warsh is about… that
[42:27] whole article I
[42:29] wrote about how there is
[42:30] no such thing as an academic Fed
[42:31] . You can't
[42:32] focus on
[42:34] academic Fed. Bank of
[42:35] America, after the
[42:36] CPI release, still forecasts
[42:38] three
[42:39] rate hikes this year.
[42:42] They hold on confidently.
[42:44] At least Kan-Row
[42:45] understands. Okay, here's
[42:48] what the Fed says. This is a graph of
[42:51] the Fed's statements
[42:52] overlaid on a
[42:53] leading
[42:54] indicator, namely the
[42:56] Labor and
[42:57] Inflation Surprise Index, which has now
[42:58] declined. We'll see
[43:00] where this goes, but that's the
[43:01] reason why we still have
[43:02] n't seen rates come down
[43:03] . So the question
[43:05] is, will Warsh provide
[43:06] any guidance or
[43:07] do anything useful?
[43:09] Two-year rates
[43:10] fell only because the
[43:12] probabilities decreased. They did
[43:13] break through the 50-
[43:14] day moving average
[43:15] after staying
[43:16] above it for
[43:16] one of the longest
[43:17] periods in the last
[43:18] decade. Let
[43:19] 's see where things
[43:20] go. But that's exactly why
[43:21] I
[43:22] wrote it. The academic
[43:23] Fed, secular
[43:25] regime change, is much more
[43:26] powerful. Usually
[43:27] reflects
[43:27] dramatic changes in
[43:28] monetary, possibly
[43:29] fiscal policy,
[43:30] and innovation. I
[43:30] believe we are in the
[43:31] early stages of
[43:32] one of them. This is what
[43:33] AI trading is. In this
[43:35] article, despite
[43:36] your favorite Oh, it's
[43:37] inside my other
[43:39] article. Despite what
[43:40] your favorite economist might tell you
[43:41] ,
[43:42] looking through a
[43:43] historical lens and
[43:44] assuming that the old
[43:45] connections still hold,
[43:46] something has changed. AI is
[43:47] already disrupting the
[43:48] job market, and the rise of AI
[43:49] agents is just
[43:50] beginning. If
[43:51] you look at the
[43:53] labor market in context,
[43:54] it is very weak. This
[43:55] brings us back to
[43:56] the graph I was
[43:57] showing you: S&
[43:58] P earnings versus annual
[44:00] labor rates.
[44:02] So, again,
[44:03] research firms
[44:04] are built
[44:05] vertically.
[44:06] Scientists rarely
[44:07] change their minds.
[44:08] This is what
[44:09] entrepreneurs do. This is what
[44:11] traders do.
[44:12] I hope that's how
[44:14] you perceive my
[44:15] work. I
[44:16] like to be
[44:17] as "bearish
[44:18] " as I do "bullish" if
[44:19] I think it
[44:20] will save me money.
[44:21] You cannot escape
[44:22] this disagreement.
[44:24] So, if you're
[44:24] a scientist,
[44:25] shed some light on this and
[44:26] tell me why this
[44:27] isn't a fair
[44:28] view of things. The
[44:28] labor market is very
[44:30] weak, while we
[44:31] are increasing profits
[44:32] by 32%. We've never
[44:33] seen
[44:34] anything like this before. And
[44:35] we are not coming out of
[44:36] recession. Here's a little more
[44:38] about the labor market. I
[44:39] showed this
[44:39] last week. I
[44:40] just want to
[44:41] emphasize this again.
[44:42] I don't see
[44:43] anyone else doing it. This is the 6-
[44:44] month rate of change in
[44:45] aggregate weekly
[44:46] wages,
[44:47] which takes into account
[44:48] weekly hours,
[44:48] wage growth, and
[44:49] the number of
[44:50] employees hired
[44:51] . A much broader and
[44:52] better way to
[44:53] look at the
[44:54] wage market. We
[44:55] have the weakest 6-
[44:57] month rate of change since
[44:59] 2012, that is, in 14
[45:00] years. Here is
[45:02] job creation now
[45:03] for the
[45:03] non-farm
[45:04] sector excluding
[45:06] healthcare and leisure
[45:07] . I don't like to exclude
[45:09] indicators, but I want to
[45:10] show this because
[45:11] these areas are associated with
[45:12] older people,
[45:14] baby boomers, who
[45:15] have money. So,
[45:16] healthcare and
[45:17] travel—that's great,
[45:18] except for one thing.
[45:19] These are the jobs that
[45:21] cannot be replaced by
[45:22] AI. So, you have a very
[45:23] difficult situation. Here
[45:25] again is the data on
[45:25] non-farm
[45:26] jobs in
[45:27] annual terms. So,
[45:28] we are here because
[45:29] AI agents have appeared.
[45:34] These are the weekly numbers
[45:35] from ADP. So, here we
[45:37] had a jump. This is the
[45:38] moment when
[45:39] the Fed's rhetoric changed. This is the
[45:41] moment when you
[45:41] saw the rate
[45:43] hike rate
[45:44] in December built into the forecast. Well
[45:45] , the ADP and
[45:47] nonfarm payrolls numbers are
[45:48] only
[45:49] getting weaker every week. We are
[45:51] not correcting anything. Here is the
[45:53] labor force participation rate, which
[45:54] continues
[45:55] to decline. If we
[45:56] lose
[45:57] labor supply, these are 10-
[45:58] year rates
[45:59] imposed on this
[46:00] schedule. Why are
[46:01] n't salaries increasing? I saw
[46:03] someone write about it
[46:03] this week.
[46:04] Look at that,
[46:05] a little growth. Are
[46:06] you kidding? So, this
[46:09] Atlanta Fed payroll thing—is it a farce
[46:10] ?
[46:12] Hourly
[46:13] wages, which we just
[46:15] saw from the
[46:16] employment data, fell to
[46:18] 3.2%. The Atlanta Fed
[46:19] showed the same
[46:20] level as before
[46:22] 2020. As well as
[46:23] hourly pay. As well as
[46:24] the level of
[46:25] voluntary redundancy
[46:26] . Give me
[46:27] something that shows...We
[46:28] know that if we're not
[46:29] getting
[46:30] wage growth that would
[46:31] substantially offset
[46:32] the lack of
[46:33] increased
[46:34] hours, and the
[46:35] fact that we're losing
[46:36] labor force participation and
[46:38] not hiring, I don't see any
[46:39] reason why the Fed should
[46:40] raise rates,
[46:41] especially three times. I do
[46:42] n't even understand what this is
[46:43] . Um, if
[46:44] their mandate
[46:44] is about the
[46:45] labor market and inflation.
[46:46] So, let's
[46:47] move on to the second
[46:48] part—
[46:48] inflation data. We are
[46:49] going to
[46:50] follow Bill
[46:51] Dudley on this PCE. I
[46:52] think Kevin Warsh is
[46:54] right. Here's the PCE, all
[46:55] on the same scale,
[46:56] guys. Core PCE, and
[46:58] here are the other core
[46:59] measures, plus...this
[47:02] doesn't include Trueflation. I
[47:04] excluded them simply
[47:05] because they
[47:06] are far
[47:07] down, below the two. All
[47:08] these other indicators,
[47:10] whether it's the "sticky" core
[47:12] CPI from the Atlanta Fed,
[47:14] which is 2.5%.
[47:15] Core CPI, which
[47:17] is 2.6%. The
[47:18] adjusted average of the
[47:19] Dallas Fed is
[47:20] 2.2%. They
[47:22] just changed the
[47:23] calculation method, now it's 2.4%,
[47:24] or something like that.
[47:25] Cleveland, 2.7%. Only the
[47:27] core PCE, this
[47:28] point here, is up there
[47:29] . This
[47:31] doesn't seem like the best
[47:31] metric to
[47:32] use. Here are the
[47:33] inflation swaps
[47:34] relative to the core
[47:36] CPI, they are lower. Why are
[47:37] n't they talked about
[47:38] as important? I took
[47:41] all those indicators that I
[47:42] just showed on the
[47:43] previous screen
[47:44] —not the swaps, but the ones that
[47:45] came before that—and
[47:46] combined them into one
[47:47] index. That's what this
[47:49] number is. This was an
[47:50] average calculation
[47:51] of those four. This is 2.5.
[47:53] This is the core PCE. This is what
[47:55] it looks like over time.
[47:56] Two things are worth noting
[47:57] . First,
[47:59] during the period from 1990
[48:00] to 2020, the white line
[48:02] consistently remained
[48:04] above the orange line.
[48:09] So if you
[48:09] normalize that, and
[48:11] again, in
[48:12] context, that
[48:12] value should be
[48:14] lower, based on
[48:15] history. Something
[48:16] has changed in this
[48:17] calculation. Kevin,
[48:19] if you want me to
[48:20] work on the Fed,
[48:21] give me a call. I'd be
[48:22] happy to help your
[48:24] guys with that so we
[48:25] don't make decisions
[48:26] based on the opinions of
[48:27] people on Wall Street or
[48:29] academics. Real
[48:30] rates, the white line
[48:31] here, the real two-year
[48:32] yield—right
[48:33] here. Here is trading on
[48:35] currency depreciation. These are
[48:36] gold, silver, and
[48:37] bitcoin. And a dollar.
[48:39] The dollar has already
[48:39] reacted.
[48:40] Today he
[48:41] fell again. This is the data to
[48:42] date.
[48:43] So, it's already here, at
[48:44] new lows.
[48:45] Gold has moved, silver
[48:47] has moved, Bitcoin hasn't yet.
[48:48] This is the green line here.
[48:49] He is still in place.
[48:51] Eventually, he too will
[48:52] move. I don't know when.
[48:53] We'll see. Scientists
[48:55] ask, "When has the
[48:57] evidence shown this to have
[48:58] happened under
[48:59] similar conditions?" Markets
[49:00] ask, "What
[49:00] will happen before
[49:02] everyone else
[49:02] figures it out?" That's
[49:03] where the
[49:04] money is made. Entrepreneurs
[49:05] are asking, "What is
[49:06] possible now that
[49:07] wasn't possible before?" This is a
[49:09] world where AI rules. This is the
[49:10] world we
[49:11] talk about at Vesper Labs. And
[49:13] finally, in July,
[49:14] China added 20 tons of
[49:15] gold. This is the largest
[49:17] monthly purchase
[49:19] since October 2023.
[49:20] China's central bank is
[49:21] also adding gold to
[49:23] Hong Kong. Finally,
[49:24] the Bank of Korea is resuming
[49:26] gold investments
[49:27] after 13 years. Korea is
[49:28] back in the game.
[49:30] AI is becoming
[49:30] more than just a
[49:31] technological
[49:32] story. The computer
[49:33] becomes an industrial
[49:34] resource, a
[49:34] geopolitical
[49:35] asset, and perhaps the
[49:36] basis of a monetary
[49:37] system. Thank you, Louis
[49:38] Guevay. Thank you all,
[49:40] guys. Maybe
[49:41] I'll see you
[49:41] next week from
[49:42] New York, or maybe I'll still be
[49:44] here.
[49:44] I hope I'm still
[49:46] here. But I don't know. Have a
[49:47] great week, see you
[49:49] next time.