YT YouTube2Text

YouTube2Text → Transcripts → The AI Bubble Didn’t Pop. Hedge Fund Leverage Did.

The AI Bubble Didn’t Pop. Hedge Fund Leverage Did. — Transcript

by Jordi Visser · 13,886 words · 2,168 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00We got a lot to cover.
0:02A lot of slides, a lot of
0:05market internals. Hopefully, I'm going
0:06to be able to shed some uh
0:09uh
0:11some experience on
0:14what have you. See, I'm I'm hoodies up
0:16this week because I don't know if you
0:17guys read the news, but the air bubble
0:20has popped.
0:21That is the news that is coming out at
0:23the end of last week. And I have never
0:25received more panic emails from people
0:28I've known over the years in terms of
0:31does this remind you of '08? Does this
0:32remind you of '01? Is this the same type
0:35of situation? And the answer for
0:36everyone is
0:38uh my career began in 1994 with Mexico.
0:41I trade started trading a book 2 months
0:42before the devaluation. I was at Morgan
0:45Stanley. I was a young person involved
0:47in a market going through things I've
0:49never seen before. Then I was part of
0:51'97 and '99 in emerging markets trading
0:54everything through the EM crisis,
0:56including the closure of LTCM was on
0:58both sides of their business, meaning on
1:01the
1:02other side of their trades as they were
1:03doing things in Brazil, and then part of
1:05the shutdown, and then a bunch of them
1:07worked at Morgan Stanley, and it was
1:09like working with the greatest people
1:10I've ever been with because I got to ask
1:12them, "How could you how could what
1:14happened happened occur?" And you notice
1:18LTCM leverage lessons. I'll be going
1:20through those.
1:21I came back to the US. I took over the
1:23S&P. Then I was there for the dot-com
1:25bubble, managing ETF business, index
1:27businesses. I was a part of all of that
1:29on the other side of billion-dollar
1:31transactions every day.
1:33Then I traded a billion-dollar macro
1:35book throughout the great financial
1:36crisis, took money out of the banks
1:38weekly and put it into gold. I have
1:40traded all kinds of things and been
1:42involved as a CIO during COVID. I bring
1:46this up because
1:48this is not a bubble popping. It might
1:50be in the future. It's not now. So, with
1:53hoodies up,
1:54I'm going to give you guys a lot of
1:57different things to go through. Again,
1:59I'm doing this for your information,
2:01your education, and hopefully not this
2:03week. We didn't make money this week,
2:05but over the long term, the goal of this
2:07is to make sure that in weeks like this,
2:09you're paying attention to the stuff
2:11that's actually happening
2:12and not to the noise that's going on.
2:15There is a bubble being popped.
2:18It is not in the equity market, and it
2:20is not in the economy.
2:22This is a true number. This is the
2:25rolling drawdowns of the S&P 500 over
2:27time. Here is the Great Financial
2:30Crisis. Look at that, almost 60% peak to
2:33trough. Here's the dot-com bubble.
2:35Eventually got to 50.
2:37Here is COVID, got close to 40. Here is
2:41when the Fed's raising rates. Here is
2:43the tariff tantrum.
2:45Here we are now. We are 2.7%
2:48off the all-time highs in the S&P 500 as
2:50of the close of Friday.
2:52Guys,
2:54please.
2:56Here's the bubble that has popped. I
2:57will go through this. Hedge fund
2:59leverage has popped. It will never ever
3:01ever ever ever, in my opinion, exceed.
3:03And what we are going through is a
3:05constant deleveraging that needs to
3:08happen. CoreWeave, not a problem.
3:11Hedge funds, big problem. I'll go
3:13through the reasons why. This is
3:15multi-strat leverage, which peaked, and
3:17this is from
3:18the Treasury.
3:20Peaked, started to decline.
3:23You're going to see more of this. I'm
3:24going to go through why.
3:26Hedge funds leverage is in a complete
3:30and utter dismantling.
3:33There will never be a time where you
3:35will see it go up, in my opinion, based
3:37on the things that I'm going to show
3:38you,
3:39because at the end of the day, you have
3:41to provide a sharp ratio that makes
3:43sense. And we are seeing extraordinary
3:45moves. And remember, as I go through the
3:48leverage, the leverage in the hedge
3:49funds
3:50dramatically bigger than anything going
3:52on in AI right now, not even close.
3:56Here's the Coreweave chart. This is
3:57where we closed the week. I just want to
3:59show this because this is the fifth time
4:02the equity price has been here. Every
4:04single bubble I have ever traded when it
4:07has popped, the equity is down massively
4:11in the bubble popping situation way way
4:14way way further than this. By the time
4:16we get to the bubble popping, when you
4:18see the bubble popping. And again, maybe
4:20there will be an AI bubble in a year, in
4:222 years. We definitely are constrained
4:25like I said last week, capital is the
4:27biggest risk. We continue to tap the
4:29markets for more capital. This is not
4:31about non-payments, this is not about
4:33not earnings, this is not about GPU
4:34prices falling. So, we're ahead of the
4:36game on the fundamentals, which means
4:38like there's been in Coreweave where
4:40we've had these panics, we get a bounce
4:41and we get a bounce. Now, the bounces
4:43seem to be getting shorter
4:45and I don't have a problem like I said
4:46on the weekly video for subscribers with
4:49the equity price of of Coreweave
4:51building in a higher risk premium.
4:53But, look at what started this whole
4:55thing. And again, hedge funds were
4:57losing money from July in a historic
5:00fashion.
5:01They got a bounce somewhat in September,
5:04but as I'm going to show you, some news
5:05came out after I did the video last week
5:07that shocked me. I can't believe it
5:09wasn't more of a story. Gartner says
5:11over 3 and 1/2 trillion of AI
5:13infrastructure spending could be exposed
5:14to impairment by the end of 2030. First
5:16of all,
5:18we don't have this number yet.
5:20So, I I just want to make sure you
5:22realize what this is saying. I've never
5:23in my history of trading markets seen a
5:26forecast out to 2030 that anyone cared
5:28about.
5:29We're talking about 4 and change years
5:31from now
5:33could be exposed. And yet, people sent
5:35this to me going up, this is a problem
5:37for Coreweave. Coreweave might be in
5:39trouble then, but we have no idea what's
5:41going to happen. If interest rates go
5:43down, it's not a problem for them. If
5:45the cost of memory and everything goes
5:46down to building a data center, but the
5:48compute demand stays high, you can have
5:50massive endless compute demand. And as
5:52Jim Chanos has showed, it may not be a
5:55good equity business. It doesn't mean
5:57it's going to zero. It doesn't mean
5:58we're going to have an unwind. Here's a
6:00chart of the S&P, again, sitting near
6:02the all-time highs.
6:05Here
6:06this one's also astonishing. So, here
6:09are Friday's returns around the globe.
6:13So, the S&P's down 1%. Big deal. The VIX
6:16has been trading 16. So, we're barely
6:19getting one standard deviation falls. It
6:20had been like 100 days that we had gone
6:22without getting a 1% move. We had two of
6:24them last week. The Nasdaq's down two.
6:26On the same day that that's happening,
6:29look at these.
6:30I mean, again, Brazil's up, Mexico's up,
6:34Canada's up, Europe's up. Every single
6:36one of them is up.
6:38If you go look at the last 5 days, S&P
6:42down, so this is for the week, down 2%,
6:45down 4 and 1/2. Again, Europe's up
6:47across the board. Look at Mexico.
6:49Uh
6:50Nikkei's down. The AI trade is what's
6:53down. Again, the AI trade is down. So,
6:56year-to-date, here's where we they are.
6:57The S&P's up 12, Nasdaq's up 13. Again,
7:01bubbles popping. If people are right on
7:04this, they are the best forecasters I've
7:06ever seen. So,
7:08in January of 2007, just to give you one
7:12example, you can do this throughout the
7:13dot-com bubble. It did not become
7:15obvious what the problems were in the
7:16dot-com bubble until well after the
7:18equity market had started to fall, well
7:20after. In the case of here, we were
7:23getting defaults already. Like, the
7:26housing bubble was unwinding in 2006,
7:292007, way before. Probably 2007 is the
7:33better one. I think March of 2007 is
7:35when I associate as when at the fund we
7:38were at, we were short, we had the big
7:40short trade on as well. We started
7:42seeing these things which we knew were
7:45horrible mortgages that were being
7:47written. As I said in my subscriber
7:49thing this week,
7:50when I go through the core weave
7:52clients, remember the clients of the
7:54subprime mortgages? The story and thing
7:56strippers in Florida were getting
7:59mortgages. So, again, this is not the
8:01same scam that was done. There's demand
8:05for all of this. Here's AIG. You can go
8:07back and look at AIG stock price. I
8:09advise you to go look at Sienna stock
8:10price, which is another AI chart right
8:13now. But, when this thing finally fell,
8:16again, before we got to Lehman Brothers,
8:18which was in here,
8:19all of financials were down massively.
8:22The
8:23credit spreads were widening. Here's
8:25where we finish when people said, "How
8:27do I watch credit? How bad is credit?"
8:29Here is the high yield side
8:33of option adjusted spreads within side
8:35the cash bonds. So, again, are we doing
8:38anything other than what we did across
8:40so many different times here? Of course
8:42not. Is this COVID?
8:44No. Is this
8:47the fracking situation in by the
8:49[clears throat] time we got again to a
8:50serious problem in oil? No. And again,
8:53it's not to say that we can't get up to
8:54those levels. We can. We're not there
8:57yet. And now I'm going to take you
8:58through what this is from my
9:00perspective.
9:01I was heavily involved with Long-Term
9:03Capital Management. For those of you who
9:05haven't read this book, it's probably a
9:07good time to pull it out
9:09and go understand it.
9:11Hedge funds at the end of the day are
9:13leverage vehicles. They are making bets
9:17to some degree on volatility and
9:19correlation amongst a diversification of
9:22assets. For the most part, a multi-strat
9:24in particular, it's across different
9:25types of strategies. Hundreds of them in
9:28some of the largest places with
9:29thousands of investment professionals
9:32hoping that there'd be a low
9:33correlation. The problem is,
9:36since I've been involved with markets,
9:39hedge funds have focused their attention
9:41for the most part, the way that you
9:42really get diversified is by having high
9:45trading activity, which means you depend
9:49on stocks to move
9:52based on information.
9:53So, if you get really good information
9:56and you think companies are going to
9:57beat the quarter cuz you've worked
9:59really hard, you've done all the channel
10:00checks, and you've done things that a
10:01retail investor can't do historically,
10:05and they beat the the quarter, you
10:07should make money.
10:08Well, again, I show this all the time.
10:11Here's Nvidia. The stock price is
10:13effectively unchanged since October of
10:16last year. 1 year, it's effectively
10:18unchanged.
10:20Their earnings have blown away consensus
10:22in a way that nobody could have
10:23expected, and what we've had is a
10:25complete crash
10:28in their PE, which is now down off next
10:31year's earnings to 15, and depending on
10:33where you go, if you take the higher
10:34area, it's probably closer to 10.
10:38That's what's happened is as they beat
10:40numbers, the stock's not reacting that
10:42way. And it was fine when it was just
10:44Nvidia for most of this year, but it has
10:47spread to the entire AI trade, and this
10:49is why when I said back the fireworks
10:51show is over, this is directly what it
10:54was referenced to.
10:55You will never go through what you went
10:57through in the first quarter again, ever
10:59again in public stocks, ever. As far as
11:01I'm concerned, we have destroyed
11:03terminal value.
11:05That terminal value means that any
11:06company's earnings past 3 years, we have
11:09no idea whether they're going to survive
11:10or not.
11:12The hedge fund industry is at all-time
11:14record leverage.
11:15I'm going to move this over so you can
11:16see it. This is from the fret the the
11:20the Fed, and this is through Q1
11:23of this year.
11:24Okay? All time, but more importantly,
11:27look at the leverage effectively from
11:292018
11:31to 2023, it was unchanged. We've now
11:34gone up dramatically and more than
11:37doubled. The leverage provided by the
11:39prime brokers of the banks, another
11:41place where crypto is going to have an
11:43impact,
11:44is at all-time highs. Here it is from
11:48the OFR report.
11:50Breaks it out by category, multis.
11:53These are big numbers. Here's long-short
11:55equity.
11:57Close to two to one.
11:58You're borrowing money. So, hedge funds
12:00make money on leverage.
12:02And there are multiple ways that you can
12:04use when you use leverage. When you're
12:06not taking directional bets in the
12:08market. And again, the net exposure
12:09across the board for multis is very low.
12:12For long-short equity, it's been on the
12:14decline for most of this year, but the
12:15gross has remained high. Now, this is
12:18all through Q2. We know that July must
12:22have had an impact and these numbers
12:23must have come down because the vol has
12:26gone up so significantly as I showed on
12:29the factor side. It just means the
12:31instability within side all parts of the
12:33market and most importantly,
12:35there are very limited hedges for the AI
12:38trade because it is a cross sectors.
12:40There is no S&P put, as I will show you.
12:43Here's the leverage growth as I'm
12:45showing, 15% in hedge funds and then the
12:48BDs, which again, the broker-dealers,
12:50heavily involved with
12:55the prime brokerage, massive growth.
12:58When you're looking for leverage, like I
12:59said, we're talking about 40 to 60
13:02billion in core wave debt over the
13:04course of 6 years.
13:06Hedge funds have money, as we learned
13:08with situational awareness, that can
13:10become an issue in a matter of weeks.
13:14On that point,
13:16we all forget and everyone started
13:18bringing out X jokes, like this was the
13:22issue.
13:23This was just one of the issues in July.
13:26So, yes.
13:28To get a number here, again, we're
13:30caring about CoreWeave whose debt is
13:32spread out over the next seven years,
13:35six years. And at the same time, the
13:38clients they have are the biggest
13:40clients in the world with some of the
13:42best balance sheets in the world.
13:44Companies like Microsoft and Meta
13:47dominating their stuff. Are they not
13:49going to get paid by them? Is that
13:51really going to be an issue? Are they
13:53stripper-related?
13:55Here is what happens.
13:57These are the numbers that were reported
13:59by the Wall Street Journal that everyone
14:00kind of backed into. I don't really care
14:02whether or not this is for illustrations
14:04purposes. I do know one thing. They're
14:06not far off.
14:07So, if there's $11 billion of equity
14:11and $45 billion of positions,
14:15the borrowed capital
14:17is 33 billion.
14:20And as long as you're making money,
14:22there is not an issue.
14:24So, hedge funds are the problem, guys.
14:27Mark my words. They are the problems. In
14:30my world of terminal value being
14:32eradicated in the bank apocalypse and
14:34the pushback that I get, the pushback I
14:36got with Micron last year, this is a
14:38scenario that hedge funds are not being
14:41honest with themselves about how much
14:43the market structure has changed. And
14:46it's not just hedge funds that are
14:47feeling the pain. Mutual funds are
14:48feeling it, too. That's why I did the
14:51demo that I did last week to show the
14:53types of
14:55market structure changes that I'm trying
14:57to help people with on how bad this is
14:59going to be. You don't need as much
15:01leverage to make the same amount of
15:02money as we're going to see. So, the
15:04leverage is going to come down. That
15:05doesn't mean the hedge fund industry
15:06won't be producing the same types of
15:08returns.
15:10I think it's possible they can generate
15:12similar returns. The money might
15:13aggregate even more into the the bigger
15:15places that have an advantage on the
15:17risk management side, but I definitely
15:19think the market structure has changed
15:20making it more challenging to have a
15:22good sharp ratio.
15:24So, Vara, another
15:26multi-billion dollar hedge fund,
15:28$20 billion dollar flagship AI fund, the
15:31losses paired the year-to-date returns
15:33from to 65% through July.
15:36So, again,
15:37they fell 44%
15:39in a month. This was in August. So, this
15:42is not now, but these types of drops, I
15:45don't think people It wasn't just
15:46situational awareness.
15:49This one happened and came out after the
15:51bell last Friday. So, I didn't include
15:54this in my video. So, if you asked me to
15:56name
15:58the 10 best non-multi-strat hedge funds
16:01of this century, Lone Pine would
16:03absolutely be in there based on my
16:04experience with them, based on the way
16:07people talk about Steve Mandel.
16:10Just
16:11an incredible hedge fund.
16:13So, this is not even saying anything
16:15negative about them. They were up 44%
16:17and it erased the gains.
16:19It is the staggering drawdown that
16:21happened
16:23in a quarter.
16:25So, to be down, to give up 44% gains,
16:28assuming they're charging 20% on their
16:31AUM,
16:32that means that they lost over 55%
16:37in gross returns.
16:39That's That's a drawdown that again, I
16:42showed you in the S&P 500,
16:45that's equivalent to the great financial
16:46crisis drawdown by a hedge fund.
16:49One of the greatest hedge funds of all
16:51time.
16:54Here is their returns that I have
16:57basically on public data through Q2
17:00relative to the S&P.
17:02So,
17:03basically, they move a lot with the S&P.
17:06Based on what you're seeing over the
17:07course of the last 3 years, there is a
17:09high correlation to the S&P returns.
17:12This is what happened during the quarter
17:14that they lost 55%.
17:17This kind of stuff has never happened
17:19before in my career. I've never seen
17:21anything like this. This is LTCM-ish.
17:25That's what it is. Now,
17:27uh again, that was through September.
17:31Yes, this month has been a disaster. The
17:33reason I'm bringing this up is not
17:35because Lone Pine is in any way, shape,
17:37or form in trouble. That has nothing to
17:39do with this.
17:41This has to do with when you're
17:42short-stacked and you have that kind of
17:44a drawdown, you're already figuring out
17:46what's going to happen, what your people
17:47are going through, what the investors
17:49are going to say. There's a lot of work
17:50that goes on. And so, to put on risk at
17:53that point,
17:54it is is just very unlikely. And more
17:57importantly, to deal with any more
17:59losses come up, and that's what I would
18:01say is going on right now with inside
18:02the marketplace. You're seeing things.
18:05Jared Kubin is is someone who's great to
18:07watch on the quant side and just to
18:09highlight. But,
18:11he bloody last 21 days, this is through
18:13the middle of last week, and it's just
18:15showing how many names are up violently
18:17in terms of sectors. Here it is here.
18:20Again, what's down?
18:22It's not just the semiconductors.
18:26It is fertilizer and crop inputs. How
18:29often did we hear how easy it would be
18:31that we'd have food shortages and that
18:33uh
18:34these are all hedge fund trades that
18:36people had gotten into. You obviously
18:38have short SAS up here. You have short
18:40software up here.
18:42You just have a lot of different
18:44components that are not just related to
18:46that one thing. And the reason I know
18:48that's the case is this kind of stuff
18:50also doesn't happen. This is the Mexican
18:52stock market.
18:55Trading at 52, well, close to 52-week
18:58lows at the end of the quarter. It has
19:00gone straight up. Here is the monthly or
19:03the weekly return of the Mex Bolsa
19:06relative to the Qs going back over the
19:08last 3 years.
19:11Here is the chart of the Goldman Sachs
19:13crowded hedge fund longs relative to the
19:16S&P.
19:17So, again, this month here
19:21Vara down 44%. Situational awareness
19:23down I don't remember how much they were
19:25down, 60, 70, whatever to get to the to
19:28get bailed out by Ken.
19:30We barely got a bounce here. And this is
19:32the most important thing is the hedge
19:34funds, these are crowded long, so this
19:36includes the fertilizers, includes
19:37everything.
19:38It just never got a movement on it, very
19:40similar to momentum, which I'm not going
19:42to bring up.
19:44Here's what's happened so far this
19:45month, and it's been a straight line
19:47down. So, again, these are the equity
19:49positions that hedge funds are holding
19:52that are dominated by the AI trade
19:55because that's the only thing that
19:56works.
19:57That is about 25% of the market, which
20:00I've talked about. Everyone's in the
20:02same names.
20:03So, it's still up big for the year, and
20:06as I'm going to show you, semis are
20:08still up about 50% year-to-date despite
20:12all of this going on.
20:14Nvidia's PE has collapsed.
20:17This is deleveraging, guys, and this
20:19started in here, and then as I show you
20:22all these hedge funds that had losses,
20:24and then you get these bounces as we
20:26head into hyperscaler earnings and the
20:29midterm elections. So, if you ask me
20:31what the pain trade is, the pain trade
20:33for this is to continue into the
20:35hyperscaler earnings and maybe all the
20:37way into the midterms.
20:39Take everyone out, they will not take
20:41risk on, and then in the end of the
20:42year,
20:43assuming I'm right and there's no AI
20:45bubble and this is just a positioning
20:47shakeout, you get a snapback. Here is
20:50the hedge fund
20:52that that chart now with monthly
20:54numbers.
20:55So, in July,
20:57it was down 10%. So, the crowded hedge
20:59fund names relative
21:03to the S&P.
21:05Hedge fund leverage was down already
21:07when this occurred in 2008. That's the
21:10thing I want to make sure you know. This
21:11does not include leverage. This is an
21:13unlevered number. This is just market
21:15neutral the stocks that are the most
21:18crowded according to Goldman Sachs
21:19relative to the S&P 500. So, these are
21:22market neutral, dollar neutral
21:24scenarios.
21:26There's beta in this to some degree, but
21:28as you can see like over time there's
21:31the movements this is an outsized move
21:33and again this is the Lehman post-Lehman
21:35Brothers move. So, hedge funds were
21:37already delevered at this scenario.
21:39Right here
21:41not delevered as much as they should
21:42have been based on all the metrics that
21:44I got. Deleveraging
21:45has come down. Here are
21:48the rolling four months.
21:51Again you're talking 15% over the last
21:54four months of that chart.
21:57Again, unlevered. So, if you add in the
22:00leverage that we've said, that would
22:01mean I mean a significant amount of
22:03losses.
22:05Here's here, but that really is all
22:07equities. So, I just want to make sure
22:09you know everything I'm showing you
22:10right now is equities. The real problem
22:12came in when it started to spread this
22:14month. That didn't happen in July. Here
22:17is the AI, the Goldman Sachs AI winners
22:19versus losers, which is the orange the
22:21white line and then the other line is
22:23that Goldman Sachs hedge fund. So, you
22:24can see this trade is basically at the
22:27end of the day it's the AI trade that
22:29every hedge fund has on.
22:31Everybody that it's it's the pair trade
22:33that everyone has on the winners versus
22:35losers and because they have a
22:37month-to-month bogey, because they have
22:39a drawdown level, because the majority
22:42of hedge funds don't have five-year
22:43locked up money, they've got short-term
22:45locked up money, they can't have
22:47drawdowns without their business
22:49becoming a risk. And I think this is the
22:51puking of a lot of businesses that had
22:54significant drawdowns that can't handle
22:56it anymore. Here is my thematic
22:58portfolio for the week.
23:00Last week was down 11%.
23:03This is worse than any week during July.
23:06The month isn't as bad as July, but the
23:08week is. That felt a lot like
23:11capitulation, but here's the place that
23:13it started to really become an issue.
23:16So, this is where you really start
23:17getting into the pods. The pods, the big
23:20multi-strats, they're going to be losing
23:22on the equity long short like everyone,
23:24but when credit starts to become an
23:27issue, that's when it really gets bad
23:28because credit starts to impact a lot of
23:31other strategies. I heard biotech was a
23:33major, major problem in September.
23:36Again, the weakening kind of spreading
23:38the contagion within side pods is always
23:41a big story, but when credit has this
23:43kind of week, this is HYG relative to
23:46IEF, which was sitting near all-time
23:49highs at the end of the quarter, which
23:51meant if you were running a credit
23:52strategy, everything was great. And this
23:54is a synthetic credit strategy. If
23:57you're just long HYG, and then you're
23:59short IEF, which is a treasury,
24:03that's theoretically taking out your
24:05rate exposure, and the only thing left
24:06is the credit credit side. And so, if
24:08spreads stay unchanged, you're just
24:10collecting the differential. So, it
24:12basically, since February, went up in
24:15kind of a straight line. Well, this gave
24:17back effectively all of the returns from
24:20the summer. So, in the same way that we
24:22had the July numbers, look what happened
24:24in July? Nothing.
24:26Nothing happened to credit.
24:28Here it is on a a weekly basis. Not
24:32anything like what happened with the
24:34tariff situation, but still, a big week.
24:38Here is the convert index. So, converts
24:42have joined the party. Why? Well,
24:44converts ex- extend from CoreWeave,
24:47uh MicroStrategy. It's a very
24:49concentrated world as well in in in
24:51in converts, but again, if the only
24:54place raising capital
24:56is AI, which is what's been happening.
24:58The size of the capital raises for AI,
25:01and not just SpaceX, not just Anthropic,
25:04not just OpenAI, but Broadcom coming to
25:07the market with over a hundred billion
25:08dollars this year. Google coming to the
25:10market between equities and bonds for
25:11over a hundred billion. We're talking
25:12about massive massive capital raises,
25:15which again the market has,
25:17but if you come to the market, the cash
25:20that people have might be from maturing
25:22bonds. They have to sell something to go
25:24buy things. People don't just have cash
25:27in most cases just sitting around
25:28endlessly. They got to sell something.
25:31Part of that was treasuries. Part of
25:32that ends up being some equities,
25:34whatever the case is, but converts have
25:36now started to get hit. And again, they
25:38were getting hit in July. They do have
25:40an equity one, but this is a very
25:41different fall. This is more of a
25:43waterfall. This was more of a reluctant
25:46fall that immediately bounced back and
25:48and made back most of the fall. Um the
25:51VIX did have a big move higher. Again,
25:53we're still at at at fairly reasonable
25:56levels, but the VIX was up 25 points in
25:59a week. VIX is the vol of the VIX. So
26:02people are buying call options. The VIX
26:04itself closed just below 20, which again
26:06is shocking that we could still be below
26:0820, but that's like I showed you, the
26:10S&P is not moving that much.
26:12Uh the Qs were down 5 and 1/2% for the
26:15week. The S&P, like I said, down 1.95,
26:19and over the last year that's one of
26:22one, two, three, four, five, six
26:25seven moves of a similar move. Um during
26:29the week, 350
26:32of the S&P 500 stocks were up on the
26:34week. So just like in the first chart,
26:37this was a unwind where we had been
26:40watching as equal weight and breadth had
26:42been getting hit
26:43while cap weight had been doing well.
26:46Well, that unwound in a vicious way so
26:48far last week. IWM M about 3%. Again, a
26:52big week but not uh one thing I will
26:54mention, it's been 6 weeks in a row for
26:57small cap and
26:58eight of the last nine with one little
27:01blurb, one little minor move higher.
27:04Uh Bitcoin for the crypto people
27:09didn't budge. Down 2.3%. So again,
27:13everyone who's continuing to say that
27:16or in the past has said Bitcoin's too
27:18volatile an asset to own. For everyone
27:21watching this that's deciding how much
27:22should be in their portfolio
27:24I'm going to say this over and over
27:25again. The majority of my portfolio
27:27moved into crypto. I'm very, very happy
27:30about that. I still have Marvell. I
27:31still have a little bit of Micron. I
27:33still have Nvidia. I still have Eli
27:35Lilly. I still have names like that, but
27:38the size of my AI portfolio is still
27:41significantly down and as I've said, it
27:42will never be as big as crypto again.
27:45Almost certainly
27:47based on my views and what's going to
27:49happen. There may be times where I add
27:51to AI for trading purposes.
27:53But the way Bitcoin is trading and
27:55hanging in there when everything else is
27:56blowing up this is a different kind of
27:59story as an asset class. The move index
28:02rates, you didn't get a big move in
28:04there. So this was really isolated to
28:06the equity market. So now let's go
28:08through
28:10the AI bear talk.
28:13So again, this is something that A16Z
28:17put out and I'm going to keep saying it.
28:19This index people have no idea how to
28:21read it. They just continue to say
28:23things about it. Then they mock out
28:25Jevons paradox. No matter who you talk
28:28to, no matter where it is, if you lis-
28:30listen to Jensen Huang
28:32and I guess people have this viewpoint
28:34that you can't listen to any of the
28:36people involved in it because they're
28:38going to lie to you. This is not 2007
28:41guys and this is not 2000. You actually
28:44need to listen to this cuz here's the
28:45one thing I will tell you with absolute
28:47certainty that occurred through the dot
28:49com bubble. The dot com bubble was built
28:52to have things like your smartphone, to
28:54have your app store, to have all of
28:56those things. That's what it was being
28:57built for. It ended up being right. The
28:59problem was we didn't have any money
29:01coming in to offset it. So, everyone's
29:04trying to find a way to create an image
29:06that the money's not coming in and yet
29:08GPU rental for the old ones, not the new
29:11ones, it just continues to sit there.
29:13And I'll as I'll show you
29:16This is This is a completely different
29:18world. So,
29:20Kuppy, who wrote a piece last year that
29:22I highlighted in terms of the ROIC risk,
29:25and I I bring this up I don't know the
29:28dude and I really don't care. This is
29:29not to pick on people. This is just
29:31basically to say
29:33people believe to compare this 100%
29:37to
29:39in this case
29:41the shell situation.
29:44Shell
29:45and
29:48open anything re- regarding
29:50intelligence.
29:51There is a similarity, which is the
29:54price goes down and the build out
29:56depends to some degree on the prices
29:58being up. The token prices are not the
30:01way to do this. The revenue is
30:04because in the case of oil, guess what?
30:06If oil goes down, well then revenues
30:07have to go down. In the case of AI, that
30:10is not the case. And so far, that has
30:12not been the case as this has gone down.
30:15The revenues can come, as I'll show you,
30:17from so many different places that this
30:19is not a one commodity thing. So, to
30:20compare it to oil
30:23is
30:24honestly, aside from the hey, this price
30:27is going down, I don't really use this
30:30thing that much and I'm not saying he
30:32doesn't. I'm just saying the majority of
30:33people, based on the data that I have,
30:35don't use it.
30:36They don't see any benefit for it. They
30:38don't think solving cancer matters.
30:39There's There's money with that. They
30:40don't think that solving math problems
30:42that go to physics that allow us to do
30:44things to make more efficient that no
30:45one would want to pay for that. They
30:47would like to pay for open source and
30:49cheaper models. It's just not true.
30:52I don't know what's happened to Carl
30:53Quintanilla, but it seems like he's just
30:55having a fun time putting out um
30:57bearish stuff. I mean, almost every It's
30:59almost like he's become a a
31:01bear. Must be some kind of way to get
31:03more people to pay attention. I have no
31:04idea, but
31:06I've been yet presented a credible
31:08framework as to where this money is
31:09going to come from. The reason I'm
31:10bringing this up, and this is not from
31:11Carl, this is from Barclays desk. The
31:13reason I'm bringing this up is
31:15nobody in 2009, so this is after the dot
31:18com bubble, nobody knew
31:21this was coming. Nobody did.
31:23When the iPhone came out, nobody knew
31:26where the revenues were coming from. And
31:28right now, the one thing I know with
31:30certainty is Barclays desk has no clue
31:34as to where the revenues are coming
31:36from.
31:37Where the revenues are coming from has a
31:40huge impact. Do they know where
31:41humanoids are going to be? Do they know
31:43how big consumer agents are going to be?
31:44I added everything here. We don't know.
31:47So, to say in 2030, which again is what
31:50we're getting to, that the revenues
31:51won't be here
31:53is just impossible for them to be right
31:56on or for the market to trade on. So,
31:58you follow the data and the
31:59extrapolation of the most recent
32:01numbers. Everyone else is jumping
32:02forward and saying, "I've been through
32:04this before. This is what it'll look
32:05like. This is too much being built."
32:07It's just not the case. So, here's
32:09Andrej Karpathy this week
32:12saying what I completely know, and the
32:14reason why I don't listen to macro
32:16podcasts anymore because I'm listening
32:18to people speak about something they
32:20know nothing about because they don't
32:22use it.
32:23So, he's talking about the fact that
32:26there's two groups of people.
32:28Most people on the planet have little or
32:30no exposure.
32:31The
32:32remaining of them, and not a high
32:34percent, the remaining people are casual
32:36users using it for chatbots.
32:38There are very few frontier professional
32:41users. So, how could anyone
32:44be able to say with certainty
32:47what's going to go on? This is far more
32:48complex than oil.
32:50This is more demand right now than the
32:53dot-com bubble ever saw on the
32:55build-out. So, again, I can keep showing
32:58you guys this, but this is coming from
33:00Andrej Karpathy.
33:02He's been honest throughout the process.
33:04He's the one who got it wrong. Agents
33:06weren't going to be here a year ago.
33:07They They're not only here, it's gone
33:09through. So, the AI debate
33:12the people on X that are bearish is
33:13being conducted by billions of people
33:15who have barely experienced the
33:16technology, millions who are beginning
33:18to understand its economic power, and a
33:21few thousand who are already living in a
33:23future the rest of the world cannot yet
33:25see. That is what exponential innovation
33:27is. Aaron Levie, if you want to know
33:29where all the tokens will get consumed
33:31in the future, it's agent swarms. We
33:33only started having agent swarms in the
33:36last 2 months. That's why we've had
33:38consumer agents, and that's why I keep
33:39highlighting. We are at a step-up
33:41function in demand right now, which is
33:44why you're seeing more and more interest
33:46in GPUs at this current time. DRAM
33:50prices made new highs after
33:51consolidating for weeks.
33:55All we're seeing with prices on demand
33:57is going higher. And that's coming
33:59because there's more money showing up in
34:02terms of demand. They're booked out as
34:04I'm going to show.
34:05Uh Patrick Collison talked about agents.
34:08I'm not going to spend all this time on
34:09it, but thanks to personal agents, we
34:12will probably get somewhat more rational
34:13in how we engage with companies.
34:15Basically, you're getting to the point
34:17where companies are starting to see the
34:21ability to use AI agents, and that's
34:23coming across. I just listened to a pod
34:26uh
34:26an interview with AI engineers this
34:28week, where they estimate at this point
34:30that agent workflows at companies, agent
34:33workflows, are only being used by about
34:352.2 or 2% of the companies.
34:382% of the companies. Now, over 50% are
34:41using
34:43coding agents. But, using agents for
34:45your workflows, which is when the job
34:47situation starts to deteriorate, which
34:49I'll talk about later,
34:51that is a different story. So, when
34:54you're thinking in bets again, and you
34:55listen to people,
34:57the prospect of a bet makes us examine
34:59our beliefs. The the whole point of this
35:01is
35:02in almost every case with people saying
35:04they're bearish, and especially people
35:06that have been bearish for over a year,
35:08which is almost everyone. I challenge
35:10you to go into X, find someone who's
35:12bearish right now, who wasn't bearish a
35:14year ago, 2 years ago, 3 years ago on
35:16the bubble trade. And not people that
35:18aren't taking a short-term, I think the
35:20market is overdone here and going down.
35:22That is the reason why I wrote the
35:23mid-cycle slowdown in May. That is the
35:25reason why I sold that of my the
35:27majority of my positions. That is the
35:28reason why I wrote the fireworks show
35:30was over, and we will never see that
35:31again. We will never see what we saw in
35:33the first quarter again. So, I'm being
35:34rational with the fact that to make
35:36alpha, you need a gen to compute.
35:38Everything else to me is up for risk.
35:41So, unfortunately, when you're listening
35:43to people say there's a bubble, by
35:44definition, they can't be Alexander
35:46Wang. He hasn't been through one.
35:48You can't listen to him talk about it.
35:50You can't listen to most people in AI
35:52talk about it
35:53because they haven't been through it.
35:54Now, Gavin Baker has, and Brad Gerstner
35:57has, but people choose to ignore them.
36:00The reason I bring this up with what's
36:02going on is this people should listen to
36:06this week.
36:07This people should listen to this week.
36:09This is Tibo
36:11from OpenAI. These are all ones I
36:13listened to this week. This one
36:15absolutely you should listen to. Dylan
36:17Patel
36:18absolutely has the most and best
36:21information of any research place on the
36:23street, because he has set up his
36:24business to be vertically integrated,
36:26meaning he covers only the AI situation.
36:30He covers the data centers. He covers
36:32the semiconductors. He covers Anthropic
36:34and OpenAI and models all this stuff
36:36out.
36:37They spoke this week. Go listen to them.
36:39Go hear how things are. This was on Alex
36:42Kantrowitz, who's great.
36:44The problem is for people, Alexander is
36:4729 years old. Tibo is in the early 30s.
36:50It's always good when you can find
36:51someone who's so mysterious that you
36:53can't even find his proper age. Dylan
36:55Patel is 30.
36:56When you're listening to people say the
36:58AI bubble has broken, they're not
36:59wearing a hoodie like me. They're
37:01telling you I've been through this. You
37:02don't know what you're missing.
37:04These kids don't matter. They've never
37:05seen this. The reality is
37:08these guys have built good businesses
37:10and their research has been dead on.
37:13Dylan Patel founded the company,
37:14SemiAnalysis, if you don't know this
37:16about it.
37:17Has 90 employees. It's in 11 different
37:19countries. In 2025, they had $20 million
37:21of revenue. Now they have $100 million
37:23in revenue. 90 employees, guys.
37:26This is the way AI is changing the world
37:29for real. So he not only knows this from
37:32his business being uncovering it, unlike
37:35the public banks,
37:37he's spending millions of dollars on AI
37:40and he's openly talking about it. So
37:42it's vertically integrated versus Wall
37:44Street.
37:48Exponential AI needs a new system. Why
37:50this wins. This is what my belief is is
37:54that all public companies are at risk.
37:56That's why I did the Stripe versus
37:57Citibank thing during
37:59the video last week.
38:01You're going to see this over and over
38:03again for me because we went from the
38:05industrial economy to the digital
38:07economy.
38:08From an investment standpoint, you only
38:10wanted to invest in the digital economy
38:11equity. Well, now no more private
38:14equity, no more private credit, no more
38:15VC, no more bonds, and no more public
38:17stocks.
38:19No more.
38:20You'll get some returns in equities for
38:22the next whatever 3 years, but the
38:24returns will be dwarfed in terms of the
38:26tokenized digital economy. That's what
38:29we're going through. The tokenized
38:30digital economy, places like this
38:32business.
38:33He has a hundred million in revenue.
38:37It's a big business, guys. That means
38:39his business I I I don't know where they
38:41would put it right now, but I do know
38:43this. You're probably talking about a
38:44market cap, I don't know if he went to
38:46market, billions of dollars.
38:48That's how fast these businesses are
38:50growing. And what he talked about was a
38:52year ago he was spending a hundred
38:54thousand dollars on AI. As of August it
38:57was up to 10 to 11 million.
39:00So, again, you're really talking about
39:03someone who's doing the work on it, but
39:05is also using it against his revenue
39:08that's come in. So, the question is
39:10how should you think about it? So, I
39:12wrote this. I went through it. People
39:15are not looking to make money right now.
39:16I'm just telling you the thing that I
39:18put out last week or earlier this week
39:20when the cost of intelligence collides
39:21with the demand for intelligence.
39:24AI agents and Vera Rubin are the place
39:26you want to be investing in for the next
39:2812 months within side the agentic
39:29compute. And I put together those 15
39:31names. Most of them are in the optical
39:33side as to who benefits.
39:35This week has been a blood fest, so you
39:38can go in and you're getting things at
39:39much lower levels. If you want to wait
39:41and wait for stability and wait for the
39:43hedge funds to be done liquidating, that
39:45might take another month. Um I don't
39:49know. It could end quicker than that,
39:50but based on how bad the numbers have
39:53been, would not surprise me to see more
39:55of this until we get into the hyper
39:57scalar numbers. Uh I did want to make
39:59sure that if you guys are interested in
40:00Jev, I had a lot of people that reached
40:02out. What is Jev? This video was
40:03fantastic, had over a million views uh
40:06as of last week. Very simple. It'll be
40:08in the uh
40:10all of these podcasts because there's so
40:12many of them. I still have I think two
40:13more to show you guys. I listened to at
40:15least seven or eight this week.
40:17All on the AI chip trade.
40:21Again,
40:22there's no way to forecast the
40:24exponential. So, you have to deal with
40:25the fact that you're going to have these
40:27ups and downs within side something
40:28that's going to go on. That's why in
40:31this week's AI video of what I did, I
40:33went through this to kind of give people
40:35the ability to dip jab. I know I did a
40:36lot of over the last 2 weeks, a lot of
40:39videos, a lot of stuff came out. The
40:41Vera Rubin report, remember you can go
40:43get this off the website. You can do
40:45this yourself. I did the Agentech
40:46consumer inflection. You guys can go do
40:48this on your own. This is the time to be
40:50doing your homework on these names if
40:52you're not in the position the hedge
40:54funds are. So, if you're sitting at
40:55home, even if you've had a drawdown in
40:57terms of things going, your entry point
40:59is getting better
41:01assuming there's not a bubble. And
41:03that's my job here is to basically go
41:04through and say we're not in a bubble
41:06phase, but we are in a deleveraging
41:07phase. It is a very, very difficult game
41:10to know when the hedge fund pain is gone
41:12because when you're dealing with $14
41:13trillion of
41:15of
41:17positioning there and the growth has
41:20been enormous, it could take a while
41:23even if the the fundamentals are still
41:25strong and that's part of the leverage
41:26thing.
41:27I also included the technical side.
41:29Again, you guys that are new to it, I'm
41:31showing you this if you didn't go
41:32through it because there's a lot to use
41:35each week on this stuff. All of the
41:37technical sheets, all of fundamental
41:38sheets, all of the stuff in AI you want
41:41to be on top of. And again, I go through
41:43the different ways the resources and the
41:46stuff that are there. I give you the
41:48prompt on it so you can go through it.
41:50And that's basically what it does. It's
41:51taking 100 names, it's going through,
41:53you're going through the Jeff process
41:55which again, go look at the video I put
41:57on the subscriber site. It will show you
41:59how you can do this. And effectively,
42:01you are doing what hedge funds are doing
42:03and that's the other reason why the
42:04market structure is changing. The
42:05competition for hedge funds in terms of
42:07the speed to ideas is gone.
42:10I am about to list something and put
42:12money into something with my crypto. So,
42:15the tokenized world is beginning, guys.
42:17The market structure has forever
42:19changed. It is a deleveraging event.
42:21You're going to hear me talk about this
42:23more and more. Tokenization is the final
42:25part of the deleveraging stage for hedge
42:27funds. So, hedge fund risk managers
42:30I actually don't want to talk to the
42:31risk managers. Hedge fund senior people
42:33that want to reach out and talk, I'm
42:34happy to go through at least my thoughts
42:36on it so you guys can go through it.
42:37But, tokenization is going to change the
42:39way covariance matrices work and the way
42:42that margin is going to be provided. So,
42:44again, I go from 100 names down to 15
42:47using this process. You guys can set it
42:49up yourself.
42:50And that way you are effectively doing
42:52what hedge funds do, taking thematic
42:54research, whittling it down, combining
42:56it with fundamentals and technicals and
42:58earnings progression, getting an answer
43:00within seconds.
43:03Watch that video. This is the one that
43:05you need to watch this week aside from
43:07the other two that I'm going to go
43:08through. But, to get back to it of his
43:10side in there, AI capex is already
43:12bigger than most people think. He says
43:14this could be close to $2 trillion or 5
43:16to 6% of GDP next year.
43:19The big debate in whether revenue can
43:21catch up the infrastructure. He says
43:23OpenAI and Anthropic alone could
43:25potentially reach 600 to 700 billion.
43:27The reason I wanted to show this,
43:28they're going to finish combined this
43:29year somewhere around 170 billion.
43:32All he's saying
43:34and again, this is the the thing to me
43:36that I just don't think people
43:37understand.
43:39So, everyone was going through the
43:40Anthropic thing this week. So, Anthropic
43:42is on pace for 90 to 100 billion for
43:44this year. Last year they finished the
43:45year at nine.
43:47At nine.
43:49So, they're 10 times again and 10 times
43:52the year before that and 10 times the
43:53year before that. And right now he's
43:55saying combined for OpenAI and Anthropic
43:58maybe 600 700. So, let's take the bottom
44:00end, 600.
44:02Well, that would mean between the two of
44:04them they're doing about three times.
44:06So, he's talking about a rapid
44:07deceleration.
44:09This is already built in
44:11he's saying 600, 700, and that's
44:12basically what David Sacks and Gavin
44:15Baker and Chamath gave as their answers
44:18about
44:19a month and a half ago. Sorry to the
44:21guys who are focused on the way oil
44:24trades in the dot-com bubble they went
44:26through. If you're not in Silicon Valley
44:28and you're not actually talking to the
44:29people heavily involved. Brad Gerstner's
44:31an investor. He was at an investment
44:34at the investment day with Anthropic
44:36this week. He saw everything. He spoke
44:37on TV.
44:39I
44:40I I just don't even know. Um the biggest
44:42revenue opportunity may not be today's
44:44token pricing. Jordan says people are
44:45too narrowly focused on cloud cloud
44:47subscriptions. This is a critical thing.
44:50And per token API revenue. If frontier
44:52AI systems can perform knowledge work
44:54across coding, robotics, self-driving,
44:56and drug discovery, health care, chip
44:58design, and other industries,
45:00the revenue pool could be much larger
45:01than today's software-like products.
45:04Again, this is the part of what I said
45:06at the beginning. When you have
45:07exponential growth, nobody has any idea.
45:09The Barclays desk is not including any
45:12of this in where the revenue's coming
45:14from.
45:15They are doing basic human linear
45:17garbage work. End of story.
45:20You didn't predict
45:22the dot-com the the post-iPhone era. And
45:25trust me, if you did, you would have
45:26made a fortune and not be working
45:27anymore cuz you just would have gone out
45:29and bought the Mag 7. The Mag 7
45:31dominated the world. We are going to see
45:33a similar thing in AI
45:35companies. I don't think it'll be OpenAI
45:38and Anthropic. I have no problem if
45:39people say to me that I don't think
45:41they'll ever be worth 10 trillion.
45:42Great. So, if they're worth one and a
45:44half right now to two, and next year
45:47they're still worth one and a half to
45:48two,
45:50they've won.
45:51That's the part I don't understand that
45:52people don't get. If they just stay at
45:54the same level,
45:55they've won.
45:57These companies are up so much from
45:58where they where they were before. AI
46:00could supercharge productivity in
46:02industries and flat head count and
46:05with flat head count and I think that's
46:06what we're going to see. So again, I
46:08bring this Barclays desk up cuz I just I
46:10think it's so stupid for people not only
46:12to write it, but then for a reporter to
46:15put it out there. It's only one reason
46:16to put it out there.
46:18To be bearish. That's it. Like what's
46:20the purpose for you guys reading that?
46:22It means nothing. They can't predict
46:25what this is going to be. Go look at
46:26their trading desk and see if they got
46:28Micron going from $10 a share to $80 a
46:31share. The answer is no. I guarantee you
46:33they were off by a factor of five at
46:35least. So
46:37you can't predict this stuff. No one
46:39can. And I'm not saying I can. I'm just
46:41saying that the path continues to move
46:43at this pace and unless we see a change
46:45in the revenue growth, unless we see a
46:47change in and I don't mean the revenue
46:49growth of a oh my gosh, in August it was
46:52slightly lower than July. That means
46:54they've lost it. That is complete
46:55garbage in terms of the Anthropic and
46:57Open AI stuff. Looking at a quarter
46:59return for a company and saying that
47:01that's the end of it is just ridiculous.
47:03Especially when we just launched
47:04Grokbot, we just launched Muse, Meta.
47:07You're going to have more and more
47:08people using other parts of AI where
47:11revenues are going to be coming through.
47:12Here's the other problem I have and
47:14again, we look at this buildout. Canals.
47:17We're comparing this stuff 5 years, 20
47:20years.
47:2220 years.
47:2317 years.
47:257 years.
47:27We're near two of this and yet we're
47:29saying the bubble is here. We're saying
47:31there's been 10 trillion, this is a
47:33bubble. We haven't even started the bulk
47:36of the CapEx. It's just starting.
47:41If there's a bubble,
47:43I'm guessing.
47:45In fiber it took 7 years of buildout. In
47:47this case we're saying in year two while
47:49everything still looks positive.
47:52It's just insane.
47:54And again, we're doing everything out to
47:552030. I've never seen forecast in my
47:57career to reiterate out to 2030. We
48:00we've been dealing with what's the
48:01economy going to be in a quarter? Is
48:03there going to be a recession next year?
48:04Now we're talking about 2030? How many
48:07things can be different than what's
48:08expected in 5 years with exponential
48:10innovation? That's why I'm bringing this
48:12up all this 2030 stuff. This got brought
48:14out again. Bains has AI needs 6
48:16trillion. They're going to be short by
48:18800 billion. I mean, this is insanity.
48:22Out to 2031 you're giving me a forecast
48:24on how short you're going to be?
48:26So again, here's here's CoreWeave.
48:29Uh trading this is before Friday. This
48:31is after Thursday.
48:32Um
48:33again, trading in this range they've
48:35been in. Here are their clients
48:37Microsoft, OpenAI, Meta, Anthropic,
48:39Mistral
48:40Jane Street. Like
48:42are they going to default? I I mean,
48:44maybe down the road, not in the next
48:46couple years.
48:48Uh we cared about this last week. We got
48:52this. It's only two trill the two
48:54trillion could be 1.9 uh blah blah blah.
48:57Again, Jared Cuban came out and
48:59basically did the math after everyone
49:00was going through it saying, "Oh,
49:02they're coming in light." And he just
49:03went through it and I love what he said,
49:05"You have to do the work and run the
49:06math."
49:08Let me clear some things up.
49:10100 billion end of year seems pretty
49:11straightforward. I I there's no way to
49:14look at it and not be able to come up
49:15with at least 90.
49:18More importantly, we're talking about
49:20whether they come in at 90 or 100 when
49:23as open as a 16Z put out
49:26the labs have added more revenue in 2026
49:28than all of public software combined and
49:30we're calling them a failure or slowing
49:32down.
49:34Um Broadcom, this was a headline by the
49:36media likens many AI investors to casino
49:39gamblers.
49:40It's not what he said.
49:42Gavin Baker volatility on misunderstood
49:45news flow. Everything I've shown you
49:47today is misunderstood news flow plus
49:50the the the chest beating of the bears
49:53coming out seeing the bubbles pop then
49:54weak at the S&P is two two and a half
49:56percent off the all-time highs. We had
49:59deep seek day. We had liberation day. We
50:00had Kimmy. We had blah blah blah
50:02everything out and then a mistranslation
50:04of Hak Tan's comments yesterday because
50:07he basically did not say what
50:09they said he did. He did not say people
50:11were gambling on AI. He actually said
50:14that OpenAI and guessing whether they
50:17had a it was 70 billion or 50 billion
50:20and jumping to the gun and making
50:22decisions off that. That's gambling.
50:24That is hedge funds.
50:26That is AI agents trading. Make sense to
50:28me.
50:29Um you get all these jumps and all these
50:31moves. You also had CXMT this week and
50:34the fact that they were coming out with
50:35stuff. So all memory names were going to
50:37get hit. So you had all of this stuff
50:39happening and again I I feel bad for the
50:41people who do all this research and then
50:43something comes out and everyone
50:45immediately sells the stocks and then
50:47they get called up. I go through this as
50:49well.
50:50And Elon Musk basically came in. He got
50:52involved and he said, "Yeah, problem is
50:54RAM supply." So
50:56you know, whatever. And remember, here
50:58are the facts.
51:00Four weeks ago, Micron estimates signed
51:02agreements represent more than 35% of
51:04its revenue through 2030. So remember
51:05the 2030 forecast for what CapEx will be
51:08and how this will be the biggest bubble
51:10in history
51:11in five years. How we have a situation
51:15where Bain says we're not going to get
51:16the revenues in five years. How Gartner
51:19says the neoclass could blow up in five
51:22years.
51:24Here are the
51:25contracts. So the these are not forecast
51:28by those guys that have no no facts to
51:30them. That's just their guess, okay? And
51:33that's a fact. That is just their guess.
51:35Micron has signed agreements of its
51:37revenue through 2030.
51:40ASML
51:41AI chip demand is stretching into 2030.
51:43It already expects supply shortages
51:44through 2028. Nvidia-backed Lumentum
51:47sold out of AI parts till 2029.
51:50You have to literally go through and
51:54just realize what I'm showing you.
51:57Pe- papers writing how in 5 years there
52:00could be problems at the same time that
52:03real business is being constructed that
52:05is sold out until then.
52:08Amazon CEO Andy Jassy told him AI
52:10computer sold out beyond 2028.
52:16Nvidia just guided for 70%
52:19revenue growth. This is from a month ago
52:22or a month and a half ago.
52:2570% revenue growth.
52:27It was expected to be I think 40.
52:30It It came in at 70 for next year.
52:34Jensen Huang guided to 70%, which is
52:37going to be higher than that just based
52:39on his track record.
52:42Uh
52:43>> [sighs]
52:43>> Meta Muse just came out.
52:45So again, as people are saying there's
52:47not They're forecasting where revenues
52:48are going to be
52:50and we're getting new progress. As
52:52progress comes out, as we break cancer
52:54research, as we do robotics, and as I'm
52:57going to show you everything is
52:58accelerating, revenues will accelerate,
53:00too. As the progress gets better, it
53:02gets easier for people to pay them to
53:04use it. Plain and simple.
53:06I don't know why this is hard to
53:07understand.
53:11This is the video I did uh on Thursday
53:13putting the If you haven't seen it, go
53:15watch it. Um again, it's a lot of it is
53:17is in this or some of it is. This is
53:19where we closed on Friday. This is the
53:21most bizarre thing to me. I'm listening
53:23to bubble talk and how this is ending
53:25and this is like I said,
53:27do you know how many AI bubble
53:28prickings we've had in this period? Like
53:31it's non-stop. Here's where semis are
53:33after all of the panic of July and all
53:35of the panic of this month.
53:38The SMH is up 50% year-to-date, slightly
53:41more than last year. I think the most
53:43important thing is the we've
53:46If you've ignored semis, you're ignoring
53:48basically the mag seven now. Let's
53:50assume that it started in 2019 when
53:52machine learning and the cloud and
53:54everything really started to to to
53:56become more of a a thing. The
53:58transformer papers, everything.
54:01All you had to do is invest in
54:02semiconductors. You've had seven of the
54:04last eight years effectively be 40% or
54:06greater.
54:08And yet people are saying the bubble has
54:09popped. You're just missing the
54:10opportunity. Regardless of what the
54:12charts look like of how fast things went
54:14up this year,
54:15Oh God. And here's the same thing in the
54:18S&P. And I just wanted to show this for
54:20people who have joined here to
54:23avoid the news, send this to your family
54:25members, send it to everyone.
54:26Compounding is why I got involved in the
54:28markets. My father was a construction
54:31worker, didn't get didn't finish high
54:32school, left school in the ninth grade,
54:35taught me handicapping, taught me math,
54:37taught me everything to go through. But
54:38what he really taught me was, I don't
54:40want to be a construction worker based
54:42on your hearing, based on your back,
54:44based on the electrocution times, based
54:47on me needing to siphon gas. I didn't
54:48want to do that. My mother handed me a
54:50book and it was on compounding. This is
54:53what the S&P 500 has done. If you were
54:55in the last 10 years, you got 17%
54:57annualized. If you were in the last 20,
54:59you got 13. The last 30, 12. The last
55:0140, 13. This is not cherry-picking
55:03numbers. This is just using the last
55:05decade, 20 years, 30 years, 40 years, 50
55:08double digits in every case. Back to 90
55:11years. This is what's happened, guys.
55:14It's been getting better recently. But
55:16this is what happened. Let's just take
55:19the last 50.
55:20So if you put $6,000 in 50 years ago,
55:23I'm sorry, if you put $10 in
55:26every month, just $10 for 50 years and
55:30ignored everything in the papers, the
55:33value of that $10 each month for 50
55:35years, $10 is now $478,000.
55:39And this has got in a 2 and 1/2%
55:41dividend to just use it in there.
55:44Guys, compounding is your friend. And
55:46the S&P 500 and all investments,
55:49effectively equities represent
55:52innovation. So, if you believe AI is
55:54going to continue to have an impact on
55:55the world for now, you want public
55:57stocks, but you also want to involve in
55:59tokenized things as soon as you can. I
56:01don't think VC and any of that stuff is
56:03going to be able to monetize because
56:05they're cherry-picking out of the
56:06billion of winners, and I don't think
56:08you can do that anymore. Here's the log
56:10chart of the S&P 500. Are there periods
56:12of time where it goes sideways? That's
56:14why I just showed you what the returns
56:16are. But the reality is when you're at
56:17all-time highs, you're at all-time highs
56:19for the last 90 years. Could there be a
56:22correction that takes us back 20%? Yes,
56:25we've seen a bunch of those since 2019.
56:28It shouldn't be part of it. This is the
56:29thing that got me going. When I read
56:31this in a book my mother gave me, a
56:33penny doubled every day from day one to
56:36the end of the month, by the time you
56:38get to day 31, there's $10.7 million.
56:41That's why I got interested in not being
56:43electrocuted.
56:46This is the chart of where we are. I
56:48show this time and time again. And the
56:50one thing I know is that this is
56:52incredibly disruptive to hedge funds and
56:54leverage
56:55because it makes certainty gone.
56:58Hedge funds invest in certainty, end of
57:01story. They
57:03trade their book. They get out of it. I
57:05was in one, had my own
57:07for 20 years.
57:10It's about short-term arbitrage. It's
57:13about arbitraging something in the
57:15future, but not carrying
57:17long positions for long periods of time.
57:19We are entering a period right now of
57:21complete uncertainty, so leverage has to
57:23come down.
57:25For all the negative things you've heard
57:26on OpenAI and Anthropic, here's where
57:28they were 2 years ago
57:30valuation-wise versus today. And I'm
57:33this easy. Anthropic is more like a bit
57:35a peer. We're talking about companies
57:37that have gone from less than a hundred
57:39billion dollars in Anthropic's case,
57:41less than
57:42five billion dollars
57:44to two trillion, and we're saying
57:45they're not successful, and you're
57:46trying to pick the top of this. It's
57:48just moronic.
57:50It really is. SpaceX, how many people
57:53continue to say that SpaceX blah blah
57:55blah blah blah.
57:56Here's the valuation. These are just the
57:58facts. If you invested in them, you're
58:00doing great. Here's what the S&P did
58:02over the last year, 16%.
58:05Boring. Public equities are in trouble
58:08on a comparable basis to things that are
58:10moving on the hyper the the exponential.
58:13The exponential are AI trades. I gave
58:15you Anthropic and OpenAI, but this could
58:16have been Cursor. This could have been
58:18so many different companies, OpenRouter,
58:20that have been purchased. SemiAnalysis
58:23goes from a zero-based company or 200
58:25million to 2 billion, whatever you want
58:27to say. If you do 20 times their
58:28revenue.
58:30Somewhere within here at this point
58:31because of the growth, you have to
58:32understand that public companies just
58:34can't grow this fast with AI. And you
58:37don't know where the revenues are going
58:38to be coming from because you're only
58:39focused on a few companies. That's not
58:41the case.
58:42Very important interview, listen to
58:43Zuckerberg and Alexander Wang. First of
58:46all, we had heard a lot about the
58:47relationship. Um this was a great
58:49interview on on Cheeky Pawn.
58:52I highly recommend um listening to the
58:54entire thing. It's about
58:56two hours. Um
58:58he talks about the compute and the spin
59:01not working out. He talks about the neo
59:04clouds and why they don't have as it is
59:06easy as they do. He talks about the fact
59:09that GPUs are basically higher than they
59:12were a year ago. The value of even an
59:14H100, the old ones, at this point is
59:17higher than it was last year. This is
59:18him saying this when talking about
59:21all of this. So, again,
59:24if GPUs remain productive longer than
59:26expected, then they can generate more
59:27revenue from the same hardware with an
59:29AI. This is just someone who's giving
59:31you the positive readout. If GPU prices
59:33are going higher, it's very hard to be
59:35bearish on CoreWeave and not bearish.
59:37You can be bearish on the equity. You
59:39can be bearish on the bonds.
59:41To say it's an AI bubble bursting is
59:43just very hard when the GPU prices are
59:45going up. OpenAI's GPT-6 may lead to
59:49robotics breakthrough as soon as next
59:50year. So, Unitree uh
59:53which as I was told by Emad Mostaque,
59:56who I I was at dinner at this week and
59:59shocked me. I didn't realize Unitree had
1:00:01only sold uh
1:00:03about 11,000
1:00:05robots this year.
1:00:08Humanoids, robots. Um
1:00:10So, when the
1:00:13I This is really a big deal because it's
1:00:15a Chinese company, very important to the
1:00:16government,
1:00:18saying OpenAI's latest advances have
1:00:21accelerated prospects for a GPT moment
1:00:23in embodied intelligence. Think about
1:00:26what that means. It's two things. For
1:00:27all of you who say the Chinese are at
1:00:29the same level, which they're not, this
1:00:32is about Astra.
1:00:34Okay?
1:00:35We are two models ahead of Astra right
1:00:37now, which they haven't released, and my
1:00:39guess is, especially since we're not
1:00:41seeing as many Chinese
1:00:43new models come out,
1:00:45distillation may be over and the
1:00:47benefits of OpenAI and Anthropic, which
1:00:50was talked about by Dylan Patel, if you
1:00:51listen to the podcast,
1:00:54may not be released. This was also
1:00:56talked about on the Zuckerberg one. We
1:00:58might be reaching a point where we never
1:01:00see, or at least for a long time,
1:01:02the models better than Astra and Opus
1:01:045.5. That may be it for us. Because
1:01:07there may be a reason now to not have
1:01:10them come out to avoid distillation, and
1:01:12that would create a really big gap
1:01:14between the Chinese and the US, and this
1:01:17whole open-source thing would be in
1:01:18trouble because one of the major reasons
1:01:20for open-source is catch staying close
1:01:22is you release the model and then
1:01:25there's distillation. Well, if we're not
1:01:26doing that and if we're leading to
1:01:28robotics, why wouldn't they keep the
1:01:31models for robotics to themselves at
1:01:33this point? So, Frontier Model
1:01:34Intelligence, as was said by the
1:01:38Semianalysis guys, reaches a point where
1:01:40their revenues are not going to be about
1:01:43consumers paying for it or even honestly
1:01:45some enterprises paying for it. The
1:01:47majority of the big enterprises are
1:01:48going to continue to pay for Frontier
1:01:51models.
1:01:52But, the reality is most of their
1:01:53business going forward may be from
1:01:55biotech. It may be from
1:01:57from nano
1:01:58technology and materials. It may be from
1:02:00robotics. It may be from
1:02:03That's probably where it's going to be
1:02:05and that revenue hasn't even started
1:02:07yet.
1:02:08Astro makes a robotics breakthrough
1:02:09closer because it potentially solves the
1:02:11hardest remaining problem in humanoid
1:02:12robots, moving from a robot that can
1:02:14perform a program task to one that can
1:02:16understand a goal, reason through
1:02:18unfamiliar situations, and adapt actions
1:02:20to the physical world. This is how
1:02:22important it is for the humanoid
1:02:25explosion. So, again, personal agents,
1:02:28agent swarms, solving math problems, and
1:02:30robotics are all together. Do not get
1:02:32bearish AI compute with all of this
1:02:34hitting right now, guys, as I have my
1:02:35hoodie on.
1:02:37Brad Gerstner, Jensen Huang, Clark Tang,
1:02:40great thing to listen to if you want to
1:02:43feel better about AI.
1:02:46From the guy who's been right the entire
1:02:48time.
1:02:50Uh Frontier Labs are roughly a third of
1:02:52the business, not the whole story. He's
1:02:54basically saying, "Why is so much
1:02:57pressure Why is everyone focused so much
1:02:59on Open AI and Anthropic?" And he goes
1:03:01through the math of where
1:03:04the 10 trillion
1:03:06What it includes, like where the
1:03:08revenues are going to come from. And one
1:03:09of the things he says right off the bat
1:03:10is you have to understand it's 500
1:03:13billion a year just for new compute the
1:03:15way it's always been.
1:03:17So, again, to listen to him go through
1:03:19this and to break down
1:03:21from someone who's been right
1:03:23consistently, and he talks and makes
1:03:25jokes about it, whose business is
1:03:27booming and he's involved with every
1:03:29company in there, to listen to him
1:03:31answer all the questions. He doesn't
1:03:33care about where CoreWeave's debt is.
1:03:35That's why he came out and said,
1:03:36"They're fine. Their cost of capital is
1:03:38actually lower." Which again, then
1:03:39people laugh at. They don't even go do
1:03:41the work, which arguably is true.
1:03:45So, again, as you go through this and
1:03:47you don't figure it out and you don't
1:03:49think about it, you're just responding
1:03:51to things. And I get it when prices are
1:03:53going down, your mind runs to places
1:03:55that don't make sense. I'm just telling
1:03:57you hedge funds are doing this.
1:03:59This is a hedge fund deleveraging event.
1:04:01Now, the Korean market was involved in
1:04:04the first part of the year. Retail got
1:04:05in. It led to hedge funds going down.
1:04:07This is another leg of hedge funds, but
1:04:09remember, hedge funds that are near
1:04:11closure don't have a stack anymore.
1:04:13Hedge funds that have been beaten down,
1:04:15like the ones I showed, that are flat
1:04:17for the year and now could be down 10 or
1:04:1915% if they didn't get out of their
1:04:21positions, they're going to take things
1:04:23off because they don't want to be down
1:04:2440 and they've already seen a down 50.
1:04:27So, you start running into issues and
1:04:29primes start getting worried and people
1:04:31start coming in and investors are going
1:04:33through it.
1:04:34It's not a fun story, especially when
1:04:36it's happening across the board because
1:04:39that makes the primes even more worried.
1:04:41So, again, the frontier labs are a
1:04:43slice, they're not the engine. Bucket
1:04:44one, classical compute. I'm just giving
1:04:46you the three that he mentions here. You
1:04:48should really go through it on your own.
1:04:51I'm going to try to start to speed this
1:04:52up a little bit cuz this is a long
1:04:53video, but there's a lot of important
1:04:55things in there and hopefully you guys
1:04:56are getting it from it. Uh this is
1:04:57talking about his track record where
1:04:59he's basically been right throughout the
1:05:00entire time of doing the interviews in
1:05:03terms of how much compute would be
1:05:04spent, how things would do, what would
1:05:06happen to GPU prices. All of that's been
1:05:09there. And then Brad on the
1:05:11inevitability of AI. this is the most
1:05:13important exchange in the conversation.
1:05:15Brad is pushing back on Jensen's own
1:05:17answer to Ezra Klein, where he gave
1:05:18something like there's two or three
1:05:20years of visibility before the bubble
1:05:21question becomes live. Brad reframes it.
1:05:25This is not a CapEx cycle with a visible
1:05:27end. It's a fundamental re-architecture
1:05:29of global compute permanently.
1:05:32Huang's one-word answer was forever. So,
1:05:36I completely agree with this and this is
1:05:38the reason why
1:05:40the dot-com bubble was a mismatch
1:05:43between the demand not being there while
1:05:46the supply was being built on leverage
1:05:49by the companies that had nothing to do
1:05:52with the demand.
1:05:53This was about the fiber companies
1:05:55putting up the rails. This is very
1:05:57similar to crypto. Crypto didn't have
1:06:00the users, they didn't have the demand.
1:06:01Now, they're going to have the demand
1:06:03from the AI agents, which is why I say
1:06:05it's the same thing to me as the end of
1:06:07the dot-com bubble for crypto and why my
1:06:09positions are in crypto.
1:06:11His point, which I completely agree
1:06:13with, there is no end to this build-out.
1:06:16There is no end. And because the
1:06:17progress is driving the demand and the
1:06:19progress keeps accelerating, as I've
1:06:21shown you, I cannot make it any clearer.
1:06:24Token use has gone higher as progress
1:06:27moves forward. As the models get better,
1:06:29they're able to do more things that
1:06:31generate revenues.
1:06:33That is not a guess, so you guys know.
1:06:35That has been a fact. That's why
1:06:37listening to people saying they know
1:06:39what the revenues are going to look like
1:06:40in 2030 when they didn't even get
1:06:42Micron's revenue is
1:06:44literally ridiculous to listen to these
1:06:46people, including Gartner Group. It is a
1:06:49joke.
1:06:51Here have been the revenues
1:06:53for Nvidia relative to the consensus.
1:06:55And what Brad Gerstner was showing on
1:06:57here is where it was just a year ago.
1:07:00This was the consensus, again.
1:07:04This is what the consensus was for Look
1:07:06how far off the street was.
1:07:09And yet the stock price is unchanged as
1:07:11I said.
1:07:12Think about it. If you're a hedge fund
1:07:15and you thought the street thought that
1:07:17for 2029 it would be 305 and now it's at
1:07:20982
1:07:22a trillion dollars and you
1:07:25watch this happen and made no money.
1:07:27Of course you have to get out of things.
1:07:30Of course you do.
1:07:31This is the only thing right now and
1:07:34this is more of a slowing thing.
1:07:37I think what the market is going through
1:07:39is that the capital coming to the market
1:07:41is just causing the market to have
1:07:43trouble dealing with this stuff.
1:07:46As more money needs to be raised and
1:07:48people go I'd rather I got to make sure
1:07:50I have money for open AI. I have to make
1:07:51sure I have money for Anthropic. And
1:07:53whether it's mutual funds or this
1:07:55they have to raise cash.
1:07:58And I think all of this stuff that's
1:07:59been going on
1:08:01for SpaceX, for the bonds, for
1:08:03everything, the way that rates have
1:08:04moved, it's all been related.
1:08:07There's plenty of capital in the world
1:08:08for the AI build out. If it ends up
1:08:10being 20 trillion and not 10 trillion,
1:08:13there's plenty of capital but it will
1:08:14have to come from somewhere. If the
1:08:16revenues are coming in, people would
1:08:18rather own bonds at some point in
1:08:21CoreWeave than they would the Treasury.
1:08:23Absolutely. Someone this week said would
1:08:25you rather buy the equity or the bonds
1:08:26and I said I'd rather buy the bonds.
1:08:29That's where my mentality is at this
1:08:31point. I just think at this point with
1:08:32the panic that's going on and what
1:08:34people are doing if I had to have a
1:08:35choice, I'd rather have the debt at this
1:08:37point at these levels. When Genius
1:08:39Failed.
1:08:40So again, we're going out to 2030, we've
1:08:43lost all certainty. I watched as LTCM
1:08:46had trades on like Brazilian Telebras
1:08:49preferreds versus ordinaries that traded
1:08:51at a spread. The preferreds traded at a
1:08:53higher price
1:08:54but every day the privatization was
1:08:56going to happen, that was all going to
1:08:57converge to zero, the spread. So if
1:09:00you're trading at a 5 to 10% spread
1:09:04and 6 months from now you're going to
1:09:05have a privatization event, that should
1:09:08collapse to zero. So, if you put that
1:09:10trade on along with a bunch of other
1:09:11spread trades on, which is what LTCM
1:09:13did,
1:09:14you think you're going to make money.
1:09:16But, here's the problem. This is the
1:09:17lessons from leverage. And I want to
1:09:19make sure that for those of you this is
1:09:21even for people managing money at a
1:09:23freaking hedge fund who've never thought
1:09:24about this stuff.
1:09:26Lessons from leverage.
1:09:28Your fundamental thesis can be correct.
1:09:32But, the timing.
1:09:34If the AI trade, if all of Nvidia, if
1:09:37all of a sudden next year Nvidia is up
1:09:38300% and goes back to where it was
1:09:40before, nothing changed. It's just the
1:09:43timing you got wrong. You were in it,
1:09:44you couldn't wait because you've got
1:09:46investors who are measuring you on a
1:09:4812-month basis. The volatility. We've
1:09:50had a dramatic shift in the Vixy Q.
1:09:53Single name vol has exploded. Index vol
1:09:56has barely budged. You have no hedges.
1:09:59So, you can be wrong on volatility of
1:10:01the markets and the relationship of
1:10:02volatility to the index. The financing.
1:10:05You can have rates move higher. That can
1:10:06cause a problem on things. You've got to
1:10:08borrow more money. Maybe the banks can't
1:10:10lend as much or maybe your cost of
1:10:12capital is going up and your investors
1:10:14are like, "Hey, I can now invest in
1:10:15Libor plus 400. I don't need you guys
1:10:18anymore."
1:10:19Crowding. There can be too many people
1:10:21in the trade, which I think is one of
1:10:22the major problems, while
1:10:25liquidity is dropping off, which again,
1:10:27we've seen. Correlation. You can have
1:10:29breaks in correlation for periods of
1:10:31time. Well, we're definitely seeing that
1:10:32right now where the Mex Bolsa is going
1:10:34higher at the same time that the Qs are
1:10:36going down. So, if you've been sitting
1:10:38there riding a winner of Okay, I'm going
1:10:39to be long Qs. I need some shorts on it.
1:10:41Oh, the Mex Bolsa is trading at 52-week
1:10:43lows. That's a good short. Blah blah
1:10:45blah. Liquidity. Obviously, this was a
1:10:47major one with LTCM. Once it went
1:10:49through, they couldn't get out of it.
1:10:51And then the position sizing and I think
1:10:54situational awareness. Gone through it.
1:10:55That's why I did this stack on moral
1:10:57philosophy. I've done presentations
1:10:59mainly for mutual funds at this point.
1:11:01This is the way I see this going. We
1:11:02went from a normal distribution curve
1:11:04where we knew that 5 years from now Ford
1:11:07would still be in business. Again,
1:11:09unless there's some serious recession,
1:11:10but in this case with growth going, we
1:11:13are getting a weird world where we've
1:11:15got losers, think retail mall names,
1:11:17winner, think Amazon, and then we have a
1:11:19bunch of names right now like bank
1:11:21stocks that are sitting in the middle
1:11:23here that I think will gradually go into
1:11:25here leaving very few AI winners, and I
1:11:28think over time more of these will come
1:11:30over here.
1:11:32The losers will be have no growth. They
1:11:35will have their growth eradicated from
1:11:37AI. And that's just the fact. Here are
1:11:40the hyperscalers. Remember,
1:11:42arguably the most important space aside
1:11:45from Nvidia and their GPUs is what the
1:11:48hyperscalers stock price is doing. If
1:11:50capital's the issue,
1:11:52their stocks made new all-time highs as
1:11:54a group. And again, this is Amazon. This
1:11:56is Meta. This is Google. This is
1:11:58Microsoft. This is
1:11:59those four names.
1:12:02I did want to bring this up. Uh one one
1:12:04data point um which I think is important
1:12:07and becoming more important. Consumer
1:12:08confidence made the lowest level since
1:12:102013.
1:12:12Yes, the lowest level. We are below when
1:12:15there was a pandemic and millions of
1:12:16people were dying.
1:12:18We're lower than that. That's how bad
1:12:20things are in the way people view it as
1:12:22we go into the midterms. Since 2014,
1:12:26the unemployment rate here was 200 basis
1:12:28points higher than it is today.
1:12:30To show the importance of this from the
1:12:32jobless situation, jobs is what drives
1:12:34this historically. This is the
1:12:37unemployment rate, and this is now that
1:12:39consumer confidence number inverted, and
1:12:41this is back to 1970.
1:12:43It is the jobs market that drives this
1:12:46historically. Now, we've seen the jobs
1:12:49market slightly weaken, but this year
1:12:50has been fine. But like I've said
1:12:53before, the jobs market is far weaker
1:12:55than economists will tell you.
1:12:57This is inside the number of that
1:13:00consumer confidence. This is the jobs
1:13:01plentiful versus hard to get inverted
1:13:04again. So, hard
1:13:06hard to get jobs
1:13:09versus plentiful,
1:13:10you've got, again,
1:13:13the It says the unemployment rate should
1:13:15be closer to six, just like it was back
1:13:16in 2013.
1:13:18Uh Warren Pies put this out uh last
1:13:21week. And again, financials and tech
1:13:23payrolls are shrinking. Aside from
1:13:26nursing jobs,
1:13:27health care jobs, we are negative now
1:13:29for whatever it's been, 18 months.
1:13:32And he's just highlighting that this is
1:13:34a dramatic change in terms of what has
1:13:36gone on in the past. And we're entering
1:13:39a new period where the agent workflows
1:13:41are about to be adopted. So, I do think
1:13:43you're going to have a labor kind of
1:13:45situation. You can view how that
1:13:47matters. It's not a recession type
1:13:49thing. It's a
1:13:51profit margin type thing, but it's also
1:13:52a politics thing. It could be a Fed
1:13:54thing.
1:13:55Uh we'll just have to see. And remember,
1:13:58another way that we've seen this is
1:14:00we're right now looking at uh earnings
1:14:02growth in the S&P at 29%. So, up here.
1:14:06And this is the year-over-year jobs
1:14:10creation relative to it. So, to finish
1:14:12up, very short week in crypto. I did
1:14:15this video. I continue to go through
1:14:16learn the stack. For those people who
1:14:18have joined, remember to watch the
1:14:19videos. I'll have another one come out
1:14:21this week on It's been a busy week,
1:14:23obviously, with the video I did on the
1:14:24AI side. I will continue to go through
1:14:26it. Um last week, we went through the
1:14:29Oracle data and interop middleware and
1:14:31the settlement and smart contract rails.
1:14:33For those of you interested, especially
1:14:35in tokenization, I think you should
1:14:38sign up if you haven't already uh
1:14:40particularly on the hedge fund and
1:14:41mutual fund side and really start to get
1:14:43these. Uh I think you should be doing
1:14:45your homework now. As you're getting
1:14:48as you're watching the horrible trade in
1:14:50the AI thing, crypto is going to be a
1:14:52very different thing next year. Um
1:14:55privacy is the theme that I covered last
1:14:57week, very important. Here is the
1:14:59month-to-date. So, the index is down 7%
1:15:02Bitcoin's down 2 and 1/2 since August
1:15:05again, still crushing it since June.
1:15:08Same thing as AI has worsened, they have
1:15:11done well.
1:15:12Uh still bull market here. Did see some
1:15:16weakness in the names on the 50-day.
1:15:17Only 20 names above the 50-day, but in
1:15:20terms of the 200-day and what's rising
1:15:21in the slopes, good. And again, a
1:15:24correction within side a very large
1:15:26market. Bitcoin is hanging in here. I
1:15:28don't think we're going to see a move
1:15:30higher in Bitcoin until we see some
1:15:32stabilization in the hedge fund
1:15:35scenario. Uh but hopefully it'll be able
1:15:37in here. It would not surprise me if it
1:15:39kept meandering back down towards the 75
1:15:42area.
1:15:43Uh just in there. And the final piece of
1:15:46information, which I will cover in a
1:15:47paper this week, related to crypto.
1:15:51I think it is shocking
1:15:53this number that their net interest
1:15:55income could reach $107
1:15:58billion according to them next year.
1:16:00This money, as I showed last week with
1:16:02Stripe, is about to come out of the
1:16:04market. All right. Long one, an
1:16:06important one. Um I need to keep you
1:16:08guys happy. I gave you a hoodies up
1:16:11uh week.
1:16:13I'll see you next week. And if big
1:16:15things are happening during the week, uh
1:16:16I will keep you up with one, two, three,
1:16:19who knows how many videos I might have
1:16:21to do this week because of all the news
1:16:22going on. Stay calm.
1:16:26Don't worry about the bubble, but for
1:16:27hedge funds,
1:16:29this is the end of leverage in hedge
1:16:31funds uh as a rising situation. See you.

About this transcript

This page contains the full transcript of The AI Bubble Didn’t Pop. Hedge Fund Leverage Did. by Jordi Visser, generated from the public captions YouTube serves with the video. The transcript has 13,886 words across 2,168 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.