The Hidden Crisis: Private Credit, Hyperscaler Leverage, and the Software Reckoning — Jordi Visser (1 marzo 2026)

[0:00] All right, uh you can tell by
[0:03] 15 things on the list.
[0:05] We got a lot to go through.
[0:07] It's been a uh
[0:10] a historical month from a dispersion
[0:13] basis and a lot's going on.
[0:16] And I think it's going to be important
[0:18] going forward
[0:20] um
[0:21] for people to stay on top of things just
[0:22] because there's a lot of moving pieces.
[0:24] I'm going to try to take you through
[0:26] what's happened this week, how it
[0:27] connects a little bit into prior trends
[0:29] that were in place.
[0:31] But for those of you on the subscriber
[0:33] side who
[0:35] saw the webinar
[0:36] um this week. So for those of you who
[0:39] are not subscribers, I did a webinar
[0:41] this week where I really spent a lot of
[0:43] time just going through what I expect uh
[0:48] has occurred, what is going on in
[0:50] particular with credit, financials, uh
[0:53] the spread that has
[0:55] led from the software unwind, but how
[0:58] this is turning into a much bigger
[1:00] story. And regardless of whether you
[1:03] want to be bearish, or bullish, whatever
[1:05] the case is, to me the risk reward has
[1:07] changed in the market significantly and
[1:09] we have a lot of things to get through.
[1:12] Uh that webinar, I'm going to do
[1:14] probably um
[1:16] more than a couple times a month if the
[1:18] current situation continues. The
[1:20] headline index hasn't moved much, which
[1:22] I'll go through, but under the hood,
[1:23] whether it's the dispersion that's
[1:25] happened, the credit markets which
[1:26] continue to weaken out, financial stocks
[1:28] which are the worst performing sector, a
[1:30] whole bunch of things. I covered it in
[1:31] the webinar. If you haven't subscribed
[1:33] yet, uh you'll be able to get the link,
[1:36] go subscribe, um and you'll be able to
[1:38] go through and see it. But more
[1:40] importantly, I just want to make sure,
[1:42] you know, this is really kind of what
[1:45] the webinar was. So for for these
[1:47] weeklies, for the stuff I do with
[1:49] Anthony Pompliano and any other podcasts
[1:51] I do, I'm really just talking about the
[1:53] bigger picture.
[1:55] Uh that is a long-term thought. That
[1:57] could change along the way. What I did
[1:59] in the webinar this week in the way that
[2:01] I approach things is very similar to
[2:03] what is in here.
[2:04] Uh,
[2:05] at some point there's a reveal where the
[2:07] market starts to, based on the
[2:09] correlations and the trends that are
[2:11] happening and how much it's spreading
[2:13] and the contagion, you get a clearer
[2:15] sense as to what's happening. Uh, these
[2:17] are all of the pieces of the puzzle that
[2:19] I cover in these weeklies, but at some
[2:21] point everything starts to gel together
[2:23] and it becomes a bigger story. It
[2:25] doesn't mean anything is going to
[2:26] happen, but what it does mean is that
[2:28] the risk reward has changed. And what I
[2:29] went through in the webinar is not only
[2:31] what I'm doing in my portfolio and I
[2:33] made some shifts.
[2:35] And remember, I'm long the physical
[2:36] upgrade trade, a lot of the things you
[2:38] guys have seen here, and I still believe
[2:40] in it, but I also believe that the risk
[2:41] reward has changed because of how big
[2:43] the dispersion is
[2:44] and really the contagion that is going
[2:46] on. So, that's what was covered this
[2:48] week.
[2:49] Um, and if it's uh, if we're in the
[2:51] stage, basically during these time
[2:54] periods like in here when energy fell,
[2:57] in here when the S&P fell 20%, in here
[3:00] when we had COVID, in here when the rate
[3:02] hikes were happening, and then obviously
[3:05] uh, in here with SVB, and then last year
[3:08] with liberation day. This is CDX for IG.
[3:11] Right now we are sitting at very low
[3:12] levels. Um,
[3:14] I started warning about the risk when
[3:16] they were at low levels, when I got my
[3:18] turbulence model. And so, it starts with
[3:20] the turbulence model and then it starts
[3:22] to spread. Credit is still tight and the
[3:24] S&P is only 2 and 1/2% off its all-time
[3:27] highs. So, it's not like we've done
[3:28] anything. This line in the sand is 5%.
[3:31] Uh, if we get a CDX move, which I
[3:34] believe it the probability is increasing
[3:37] every day,
[3:38] uh,
[3:39] you're going to see a move that gets 15
[3:41] to 20%, which I talked about at the
[3:43] beginning of the year. And right now the
[3:45] dominoes are shaping up that that's
[3:46] getting closer and closer, or at least
[3:48] the probability is increasing for that
[3:50] to occur. Again, it was February 3rd
[3:53] that I got the turbulence model
[3:54] basically warning signal. I showed how
[3:57] many of these uh events had occurred uh
[4:00] to warn.
[4:02] For those of you who've had uh
[4:05] not going to This is where you go for
[4:07] the subscribing side. So, the AO macro
[4:09] nexus is where all of all of the stuff
[4:12] on 22V's account is for me.
[4:14] If you didn't get the link to the
[4:16] webinar or if you missed the webinar,
[4:18] everyone got the link. Uh just go to
[4:20] contact and send it. It will be up on
[4:22] the website uh Monday or Tuesday. Uh but
[4:25] if you want to see it, just go there. If
[4:26] you want to subscribe, go through it. By
[4:28] the way, the main uh outreach I think
[4:31] and where this is going to be the most
[4:32] hopeful as I go through stuff today is
[4:34] going to be with financial advisors and
[4:36] RIAs. Um we're getting more and more um
[4:39] groupings of people from RIAs in there.
[4:41] And I think this year, because of what's
[4:43] happening in private credit, because of
[4:45] what because of what was happening in
[4:47] private equity, which is also going to
[4:48] be impacted by private credit, uh also
[4:52] the amount of ETFs that were created
[4:53] that have a lot of AI {quote} {unquote}
[4:55] software in them, which I think I just
[4:58] continue to see the people pound the
[4:59] table that this is a mistake in
[5:01] software. Uh
[5:03] I think this is where you at least want
[5:05] to have someone on the other side. I do
[5:07] not think the software uh follows a
[5:09] mistake. In fact, I think it's got to
[5:11] the point where people that are saying
[5:12] that are trying to fight the market,
[5:14] which is never a good thing to do. And
[5:15] as I go through this, I'll highlight it.
[5:17] From an idea basis, the freshest idea I
[5:19] finally did release the chemical side uh
[5:22] with about 15 to 20 names, which again,
[5:24] you can get from 22V. So, the big story
[5:27] this week, at least to start out, that
[5:29] got everyone freaked out a little bit,
[5:30] was the Citrini Global Intelligence
[5:33] side. You've got people
[5:36] going on the other side of it. And then
[5:38] Gavin Baker put something out that
[5:39] basically
[5:41] I'd say covers one of the angles that I
[5:43] agree with. I do not uh believe in the
[5:45] Citrini side. I also believe it is part
[5:48] of the distribution. So,
[5:51] instead of saying whether it can happen
[5:52] or not, which is what we seem to become
[5:56] a society of everything is binary,
[5:57] that's not the way markets work. Is
[5:59] there a possibility that we have an
[6:01] unemployment rate going higher?
[6:03] Absolutely, it is a possibility. Is it
[6:05] going to happen from AI alone? No, it's
[6:07] not going to happen from AI alone. And
[6:09] the reason that Gavin Baker gives,
[6:11] which I agree with, is that it would
[6:13] take a thousand times more compute. Now,
[6:15] whether or not that's the number that it
[6:17] would need, uh actually he may have said
[6:19] 10,000, but I'm not going to pull this
[6:20] down. Uh the main point is we don't have
[6:22] enough compute to replace. But that's
[6:25] not the only issue. And this is the
[6:27] thing I want to make sure you guys
[6:28] realize cuz this gets into the risk with
[6:30] the hyperscalers that I've been debating
[6:32] people on all week. This was part of
[6:34] what I went through in the webinar. The
[6:35] webinar covered
[6:37] everything starts with AI. So, what has
[6:39] changed in AI since November? And this
[6:42] is why I have strong disagreements with
[6:44] the people who are out there promoting
[6:46] like this is not
[6:47] correct what's happening in software.
[6:50] They're also not acknowledging all of
[6:52] the changes that have happened in AI
[6:54] since the end of November. And there
[6:55] have been dramatic shifts that I went
[6:58] through. There's about six of them that
[7:00] I covered that are all new news. And if
[7:03] you don't change your viewpoint or at
[7:04] least say the market is discounting the
[7:06] possibility of something in the future,
[7:08] you're just not doing right by retail
[7:10] investors in particular, but you're not
[7:12] doing right by anyone to sit there and
[7:13] just continue to say the same story like
[7:15] nothing has changed. Unlike liberation
[7:17] day
[7:18] >> [snorts]
[7:18] >> where it was a government action that
[7:20] was then pulled back, that one makes
[7:22] sense. COVID where it was a a
[7:25] uh
[7:26] a virus with tons of money printed and
[7:28] then a vaccine,
[7:30] this one is not something to get way.
[7:32] AI's not going away, which is why I
[7:33] called it a supersonic tsunami.
[7:36] The other issue with uh the hyperscalers
[7:38] is the adoption.
[7:40] Enterprises are not going to adopt AI.
[7:43] And this is the part that I want to make
[7:45] sure I'm clear to everyone. They want to
[7:47] adopt, but there's a lot more issues
[7:49] that go in with big enterprises using AI
[7:52] than someone like me sitting at home or
[7:54] any entrepreneur sitting at home.
[7:57] They I can use open claw. They cannot. I
[7:59] can use AI agents today. They cannot. I
[8:02] can use Chinese model. They cannot. As
[8:04] the year goes on and we realize that
[8:06] agents for the 8 billion people on the
[8:09] world who can afford a computer like a
[8:12] Mac mini or a Mac studio or as these
[8:14] things go on the cloud can use AI agents
[8:17] every single day. More and more gentic
[8:19] stuff is being done, which means the 8
[8:21] billion people have tools that the
[8:22] enterprises don't aren't able to use
[8:24] yet. Data's having worked at a Fortune
[8:27] 500 country 500 company at a very high
[8:30] level, having run an office for the
[8:32] firm,
[8:33] you have to understand that the amount
[8:36] of
[8:37] friction within side there at so many
[8:39] levels will prevent complete adoption to
[8:43] equal all the firings that Citrini had.
[8:45] It will also make the SAS growth side
[8:49] not a story. Uh and I'll cover that
[8:51] more. So, the real friction slowing
[8:53] enterprise AI adoption, um this does not
[8:55] get brought up enough, but I just do not
[8:57] see them being able to adopt adopt at
[8:59] the pace. If that does uh slow down, it
[9:02] has huge implications cuz that means the
[9:04] revenues for Anthropic and all of these
[9:06] companies are going to be in trouble. If
[9:08] you add in the data center delays and
[9:09] they can't get the cloud up and people
[9:11] move off
[9:14] the cloud and they move to devices like
[9:16] Apple Mac minis,
[9:18] you end up with a situation where
[9:19] they're not getting the inference money
[9:20] that they thought they'd get or at least
[9:22] not yet.
[9:23] If you want to watch something that goes
[9:25] through the friction within side
[9:27] enterprises, watch this podcast. And
[9:29] again, for people who subscribe on the
[9:31] link, all the podcasts there's about
[9:33] five or six that I'm going through today
[9:35] that I used for some of the information.
[9:37] Uh there is a recap with the links so
[9:39] that way you don't have to go memorize
[9:40] them or write them down.
[9:42] Um
[9:43] This in that podcast the speakers argue
[9:45] the problem is not the models. The
[9:46] models are good enough. Completely
[9:47] agree. The models are already good
[9:49] enough. Claude is already good enough.
[9:50] The failure stems from poor enterprise
[9:52] data quality and scale issues, not LM
[9:54] capability. This is one of the reasons
[9:56] why I love Palantir. It is a very very
[9:59] difficult situation to get the data
[10:01] organized. So what everyone believes is,
[10:02] "Oh yeah, we'll just pay more money for
[10:04] all of the SaaS companies agents."
[10:07] That's not the way this is going to go.
[10:08] AI simply exposes the long ignored
[10:10] garbage in garbage out problem at scale.
[10:13] Completely agree. Lived it, saw it.
[10:16] McKinsey, these are all of the places in
[10:19] AI risk by enterprise risk category that
[10:21] these enterprises need to solve. They
[10:24] move slow. They make decisions slow.
[10:27] AI's progress from November to now
[10:31] is exponential times whatever you want.
[10:33] It is amazing how much more you can do.
[10:36] The enterprises can't move that fast.
[10:38] Dario Modai was on with Dwarkesh.
[10:41] You don't have to listen to the whole
[10:42] thing, but
[10:44] he again
[10:45] went through the bankruptcy risk. And I
[10:47] just want to give you the quotes here.
[10:49] "If my revenue is not $1 trillion, if
[10:51] it's even 800 billion, there's no force
[10:53] on Earth. There's no hedge on Earth that
[10:55] could stop me from going bankrupt if I
[10:57] buy that much compute." This gets into
[11:00] the debt and the compute purchases
[11:03] which I go through in the webinar are
[11:04] getting and have gotten more expensive
[11:07] due to memory. For those people who want
[11:10] to debate what the memory costs are in a
[11:12] data center, to get this correct, you
[11:14] have to spend a lot of time on where
[11:17] memory prices are right now compared to
[11:20] where they were in September. Go through
[11:22] the numbers and realize how much of a
[11:24] Blackwell and how much that has changed.
[11:26] We are around the level now on the
[11:27] current prices of somewhere 45% compute
[11:31] as a whole is somewhere around 50 to 60%
[11:35] still and especially with the memory
[11:37] cost going up. So, when you go through
[11:39] the numbers and people want to debate
[11:40] whether it's 10 billion for memory, 15
[11:44] billion on a
[11:46] a 50 billion data center,
[11:48] the main point is the numbers have gone
[11:50] up exponentially and these guys are
[11:51] having to hoard compute. If you end up
[11:53] without the power coming through and
[11:55] they've prepaid for all the compute,
[11:57] all of these things fit into the bet. I
[11:59] think the most important thing about
[12:01] this is they're making a bet on revenue
[12:03] coming in
[12:04] and as I mentioned between Chinese
[12:06] models, between
[12:07] a lot of things going off
[12:10] the cloud where they're not going to get
[12:13] this, you have to start paying
[12:14] attention. It AI agent open cloud adopts
[12:16] Chinese models for cost edge over US
[12:18] rivals. This is becoming a bigger and
[12:20] bigger theme. 80% of USA startup US AI
[12:24] startups is from Andreessen Horowitz
[12:26] relying Chinese open source models. Robo
[12:28] pal had Emad Mostaque. If you haven't
[12:31] heard Emad before, he's an interesting
[12:32] guy because hedge fund guy who then went
[12:36] and did a AI company.
[12:38] He always has a lot of fairly extreme
[12:41] views. I agree with most of them down
[12:43] the road, but I think he's on spot with
[12:45] bunch of things in here. He talked about
[12:47] crypto,
[12:48] but most importantly this interaction.
[12:50] Claude scores top in terms of human
[12:52] interaction and naturalness of speech
[12:54] and writing quality. Claude is the best
[12:56] for doing Excel or documents. Could not
[12:58] agree more. His point, the only issue
[13:00] with Claude is that it's too damn
[13:01] expensive. MiniMax,
[13:04] Chinese model which has the same
[13:06] performance as Claude is 20 times
[13:08] cheaper for the same performance.
[13:10] He's an AI So, anyone who's going to
[13:12] argue with me that this is not important
[13:15] because enterprises won't use Chinese
[13:17] models,
[13:18] I agree with you, but the friction will
[13:20] prevent the revenue from the from the
[13:22] enterprises of getting there. Any
[13:24] hackings that happen this year will go
[13:26] through hackings as time go on. I've
[13:28] done the cyber side. We'll slow down the
[13:30] process. I think on-premise is the only
[13:32] solution for the enterprises. I do not
[13:35] believe the cloud is a potential
[13:36] solution for them, which means they're
[13:39] not going to fire as many people.
[13:41] They're going to have to spend more
[13:42] money because the cost of inference is
[13:44] going higher. It causes all kinds of
[13:46] bottlenecks. So, remember, bottlenecks,
[13:48] bottlenecks, bottlenecks, that's what
[13:50] we're seeing.
[13:51] Um
[13:52] Dan Ives put out the Claude events
[13:55] suggested the fears of AI supplanting
[13:56] the software industry may be overblown.
[13:59] I don't have a problem with saying AI
[14:00] supplanting the software industry may be
[14:02] overblown. I think for enterprises that
[14:04] story is correct, but as I go through
[14:06] things,
[14:07] I think you're missing
[14:09] a bigger picture here. And that's why I
[14:11] I think this is really not a uh
[14:15] I don't think this is the right thing to
[14:17] do to totally call this a ghost trade.
[14:21] That there's no reason for this going on
[14:23] and this is purely fear. As I go through
[14:25] this, this is not purely fear. Or this
[14:27] stuff. I I I don't know if this is just
[14:29] an ETF sales pitch or what, but I I put
[14:32] this through and ask uh Claude to be a
[14:35] forensic uh
[14:37] uh accountant and to take a Quantico FBI
[14:40] approach, and I won't show the results,
[14:41] but
[14:43] let's just say that kind of stuff of
[14:44] saying this is a goodbye when nobody's
[14:47] embedding any of the conversation of the
[14:49] the themes that have changed since
[14:51] November. Uh I've mentioned Andreas
[14:53] Steno. I like his work. Um you know,
[14:56] he's been skeptical on software, but I
[14:58] disagree with this part, too. Uh making
[15:00] the rounds that Bloomberg
[15:02] can be replaced with complexity.
[15:04] Uh it can't completely be uh replaced by
[15:07] it. Obviously, some of it can be, but I
[15:09] think the more important thing is what's
[15:11] going to happen to the seats of
[15:13] Bloomberg. Is it a growth side anymore?
[15:15] Instead of saying that AI is a binary
[15:19] thing and it won't replace, everyone
[15:20] should agree with that. We still have
[15:22] taxis. Uber never replaced all the
[15:24] taxis. But, the problem is have taxi
[15:26] medallions increase since Uber came? Is
[15:29] there a problem with the fact that these
[15:32] things aren't going on? This is the
[15:34] issue. So, it reminds me of something
[15:35] I've used in a bunch of papers over the
[15:37] last 20 years, which is this famous rule
[15:40] I'm a big Sherlock Holmes fan in this
[15:41] famous
[15:42] uh scene from one of his books when he
[15:45] asks Watson, "What do you see in the
[15:47] sky?"
[15:50] And he tells him, "I see galaxies. I see
[15:51] stars."
[15:53] "What do you see?"
[15:55] And Sherlock Holmes says,
[15:57] "You're missing the main point. Who
[15:58] stole our tent?"
[16:00] This is the problem with right now just
[16:02] saying software is not going to be
[16:05] replaced with AI. It is a very very
[16:08] loose story.
[16:11] Here's the argument I want to make.
[16:13] Friction. Again, all of these
[16:15] conversations with the suits that are
[16:17] hanging up at a high level.
[16:19] Uh one of my great achievements of
[16:22] leaving Morgan Stanley was not having to
[16:24] wear a tie anymore.
[16:25] Uh occasionally I have one on on the
[16:27] website. There's a photo of me with a
[16:29] tie on. But, trust me, I don't even know
[16:31] where my ties are right now. Um
[16:34] the big thing is this. For any startups,
[16:36] as I said, will they ever use
[16:38] salesforce.com? I see Dario Amodei
[16:41] saying, "Of course they will. Look at
[16:42] these 500 billion, 800 billion-dollar
[16:46] companies." I agree.
[16:48] If you're an 800 billion-dollar company
[16:50] like OpenAI or
[16:54] xAI or any of these, will you use
[16:56] Salesforce? Sure. But, if there's 300
[16:59] million entrepreneurs building bespoke
[17:02] things, if you want to change your CRM
[17:05] because you have another business line,
[17:07] you just do it. I do this all the time
[17:09] on the software that I build for my
[17:10] business. I change it on the fly. That
[17:13] is what is the power of AI. So, the
[17:16] other part with going against this,
[17:18] again, let's assume that of the people I
[17:19] just showed, you tend to agree and you
[17:21] want to step in and buy. Well, there's
[17:23] two arguments you're making. One is,
[17:25] okay, I don't I think this is all fake.
[17:28] Okay, well, the market's telling you
[17:29] it's not. Because this chart, to just
[17:32] jump into this and say, I think we're
[17:34] going back up all the way to the highs.
[17:36] This is the IGV versus NDX. So, I agree.
[17:40] I'm I'm in the same camp. We're at the
[17:42] point now where I think software has
[17:44] done it as done as much as it needs to
[17:46] do. Would I buy a name? No. Because
[17:50] anyone can miss a name and fall 20 to
[17:52] 40% at this point. Do I think there's
[17:54] upside? Yeah, we've moved Salesforce
[17:56] down to the same PE as Ford. Is there
[17:58] upside? Sure. But remember, Ford is
[18:01] unchanged for the last 36 years.
[18:05] If you ask me, there's a big similarity
[18:07] between salesforce.com and Ford. I'm not
[18:09] saying they're going out of business and
[18:11] they're not a zero, but I am saying if
[18:13] you wasted your time trying to pick the
[18:15] bottom of Ford and buy it, you want to
[18:17] go with the flow. Go buy some silver. Go
[18:19] buy uh some analog semiconductors. Go
[18:22] buy some chemicals. Go buy some energy
[18:24] stocks. We're going to need more of that
[18:26] over the next 3 years. We just don't
[18:28] know if the software names are going to
[18:29] be in business in 3 years or which ones
[18:31] will.
[18:32] If you believe the equities are wrong
[18:35] and you're saying this is a ghost trade,
[18:37] then I will go to the argument that I've
[18:38] always heard since I was at Morgan
[18:40] Stanley. I'm from the equity side. And
[18:41] everyone who's watching this, that's a
[18:43] macro person from the fixed income side,
[18:45] knows what I'm about to say. Every fixed
[18:47] income person during the '90s
[18:49] especially, while the equity market was
[18:50] in a raging bull market that made no
[18:52] sense,
[18:53] and the bond people were sitting there
[18:55] both having equity envy, but also
[18:57] thinking everyone on the equity side was
[18:59] an idiot, um and that the sharp pencils
[19:01] were on the fixed income side. There's a
[19:03] reason for it. Bond investors have to do
[19:05] a lot more homework and pay attention to
[19:07] things on a detailed basis, because the
[19:09] fear was they wouldn't get paid. Where
[19:11] equity people, it was trading on the
[19:12] hope. So, the math behind bonds is
[19:15] showing up. So, this is the leverage
[19:17] loan side for technology. So, if you
[19:19] don't believe the software trade, just
[19:20] look what's happened over the course
[19:22] last 5 weeks. So, if you're a financial
[19:24] advisor or an RIA, and you have private
[19:27] credit, and you have technology stocks,
[19:28] and software stocks are a big portion of
[19:30] your portfolio, when the debt and equity
[19:32] are moving the same way, it's a big
[19:34] issue. This was the liberation day. It
[19:36] was barely a budge down. It didn't move
[19:38] that much. This is a huge move. The bond
[19:41] market is matching up right now on the
[19:43] levered loan side, and it's spreading
[19:45] into other places. So, option traders
[19:47] pile into bets against the software
[19:49] exposed loan ETF. This is follow it real
[19:51] time, BKLN.
[19:53] Tons of puts being purchased over the
[19:54] course of the week. Has a lot of
[19:56] software in it.
[19:59] Here we are month-to-date.
[20:02] S&P only down 1%. So, for all the things
[20:04] I'm saying, again, you can leave the
[20:06] market going, it's fine. And utilities,
[20:08] staples,
[20:10] energy, material, industrials, and then
[20:12] down here,
[20:13] basically, you have the Mag 7 and
[20:15] financials. We'll get into that.
[20:18] Uh
[20:19] down 1% for the month, the S&P. Again,
[20:21] you wouldn't know anything was going on.
[20:23] Russell 2000 up 70 basis points. Even
[20:25] the Nasdaq, the QQQ, only down 2%. Not a
[20:29] big deal.
[20:30] The big deal under the hood was we had
[20:33] 150 names. So, if you take this, this is
[20:35] 106 of the S&P 500 were up over 10%, and
[20:40] at the same time, you have basically 46
[20:43] that were down at least 10%. So, 150 of
[20:47] the 500, 30% of the names moved at least
[20:50] 10%. That is why you've had such a
[20:53] massive dispersion. That's why my
[20:55] turbulence model is up. That is never a
[20:57] healthy sign, but we know what it is.
[20:58] It's a rotation.
[21:00] The
[21:01] utility and staple side, just remember
[21:03] this, and I covered this on the webinar.
[21:05] We are at a point where
[21:07] I want to buy things that I know are
[21:08] going to be here in three years. I don't
[21:10] know if Salesforce will still be growing
[21:12] in three years, but I know McDonald's
[21:13] will still be a business in three years.
[21:15] I know that CVS will. I You can go
[21:18] through the list and go through the
[21:20] companies. They are overpaying now for
[21:22] multiples in companies based on who's
[21:24] going to be here
[21:25] in three years.
[21:27] When will that end? I don't know. This
[21:28] has a rounding side to it. Is it
[21:30] possible that we could make a new high
[21:32] and then come back down? Yeah, I think
[21:34] the risk is growing that there's going
[21:36] to be a correction that is big. But
[21:38] outside of the US and outside of the
[21:40] dispersion,
[21:41] MSCI World ex-US looks fantastic. Again,
[21:45] I showed this before.
[21:47] We barely broke out middle last year,
[21:49] but this chart is going parabolic the
[21:52] other way of IGV. And again, this is cuz
[21:55] they don't have software in there. It's
[21:57] low-tech. Here are the monthly numbers
[21:59] for MSCI World. If you're looking for
[22:01] something to buy, this is the place
[22:03] where I would go. Foreign stocks,
[22:04] foreign stocks, foreign stocks, because
[22:06] the software thing is leading to more
[22:08] rotation. We had Nassim Taleb warning on
[22:10] the software bankruptcies.
[22:12] Traders rush to dump software loans. So
[22:14] again, if you think this is just an
[22:16] equity problem, it's not just an equity
[22:18] problem.
[22:19] The private credit story is a real story
[22:21] and people need to start being on top of
[22:23] it. This is not some ghost thing. This
[22:26] is a real thing. This is something that
[22:29] has a lot of the eerie part of the
[22:33] subprime side where retail ends up
[22:35] getting trapped in things and we're
[22:37] seeing
[22:38] debt widen. The question is, will the
[22:41] contagion spread from tech and go into
[22:42] other places? And as I go through this,
[22:45] it's already in the equity market in
[22:46] this. So, watch the private credit side
[22:48] and just be aware of it. And if you've
[22:50] got investors that are in there, you
[22:53] want to stay on top of this thing so at
[22:55] least you have a sense as to what's
[22:56] going on. If this happens, UBS
[22:59] now sees private credit defaults
[23:00] reaching 15%. I highlighted the fact
[23:02] that this was this week. The reason I
[23:05] highlighted is because 3 weeks ago they
[23:07] said it would hit
[23:09] 13% in worst case. They've already upped
[23:12] it.
[23:12] >> [snorts]
[23:13] >> Uh you've got uh hedge fund managers
[23:16] trying to buy things at a discount
[23:18] basically out there saying private
[23:20] credit's going to worsen. Apollo,
[23:24] bad loans were not just software. They
[23:26] were in industries like healthcare,
[23:27] transportation, manufacturing. They had
[23:29] a bad week this week.
[23:32] Um
[23:33] you start treating as volatility and
[23:35] hiding. This is someone who went
[23:36] through, again, the great financial
[23:38] crisis. We have here here uh we have a
[23:41] lot of things going on that remind me of
[23:44] that. It also reminds me of parts of the
[23:46] dot-com bubble. And remember, the
[23:48] unwinds that were happening there were
[23:50] not just in equity. It was in telecom,
[23:54] it was in Enron, there were a whole
[23:55] bunch of things that led that as the
[23:57] tide of liquidity went out, which is
[23:59] what's happening in private credit, you
[24:01] have investors that are trapped in
[24:02] vehicles now. This is becoming a bigger
[24:04] story. They want their money back,
[24:07] but unfortunately, they signed up for
[24:08] something where they can't get their
[24:09] money back. So then you have gates going
[24:11] up. So once you start getting into this,
[24:13] you have to make sure that this is not
[24:14] an isolated problem.
[24:17] Leverage didn't go away, it just moved
[24:18] off the banks' balance sheets. I said
[24:20] this before, I'll say it again.
[24:23] The seeds of the current crisis were
[24:26] always planted in the prior crisis. When
[24:28] you stop the banks from being able to
[24:29] make loans, it leads to private credit,
[24:32] and you get growth in that, and then
[24:34] when the tide of liquidity goes out
[24:35] there
[24:37] in an opaque situation, it's a little
[24:39] bit harder.
[24:40] New credit blow up in London. This was
[24:42] on Friday.
[24:43] MFS is collapsing in London with themes
[24:46] similar to those of the auto lender
[24:48] Tricolor
[24:49] and First Brands Group.
[24:51] Banks such as Banco Santander and
[24:52] Jefferies are scrambling to recoup money
[24:56] with other entities like Apollo
[24:59] and Barclays.
[25:02] The collapse of the UK property lender
[25:03] sent shockwaves through Wall Street.
[25:05] Again, this seems to involve fraud.
[25:08] Tide of liquidity goes out, you start to
[25:10] see who's naked. Apollo holds 20% of MMS
[25:14] MFS senior debt.
[25:17] Bowas Weinstein again, we own a lot of
[25:18] CDS on life insurers and price in past
[25:21] crises they went from trading at
[25:23] Life insurers are becoming more of a
[25:25] story. I'll get into that also in a
[25:27] little bit. Uh here's the chart of
[25:29] credit as we end the month. So, the
[25:31] white line here is the option adjusted
[25:34] spread for technology. Basically, we are
[25:37] approaching levels if you remember that
[25:38] CDX. We are at levels where historically
[25:41] the VIX goes higher, the S&P has a
[25:43] drawdown.
[25:45] The main reason we haven't had it yet is
[25:47] cuz credit hasn't spiked here. If we do
[25:49] see the red line, which is all sectors,
[25:51] which is moving higher and is at the
[25:53] highs basically since liberation day.
[25:55] The green here is junk cash bonds. So,
[25:57] this is the option adjusted spread.
[26:00] We're tracking that way and it's
[26:02] spreading. Here is the IG CDX again.
[26:06] This is when my turbulence model started
[26:08] to give me a signal. I had a lot of
[26:09] people in credit say, "I'm not seeing
[26:10] anything." Well, now you're seeing
[26:12] something.
[26:14] Here's the leverage loan total return
[26:16] index. This is the B cat BKLN, but as a
[26:20] total return index, you guys can see the
[26:21] symbol in Bloomberg if you want to track
[26:23] it. This is the 200-day moving average.
[26:26] Basically, since the Fed pivot in
[26:27] October of '22 when the tightening cycle
[26:31] ended.
[26:32] Leverage loans have been in a bull
[26:34] market. Not so much anymore, and it
[26:37] seems to be just starting. Here is the
[26:39] monthly returns.
[26:41] Normally, if people are trapped this
[26:43] stuff tends to get worse. Now, I will
[26:45] have you go back to a report in 2024
[26:48] from the IMF on the global financial
[26:50] stability report. The sector could
[26:51] experience this is about the rise and
[26:53] risks of private credit. It could
[26:55] experience large and unexpected losses
[26:57] in a downturn.
[26:59] Okay, well, we're starting that right
[27:00] now. Liquidity risk could rise with the
[27:03] growth of retail funds. That is where we
[27:05] are right now.
[27:07] So, this was again
[27:09] from the IMF.
[27:12] Julia LaRoche had on Chris Whalen. Um
[27:14] Chris Whalen was uh one of the people I
[27:16] love listening to during the SVB crisis
[27:19] situation. I thought he was the most
[27:20] rational and uh actually had a a good uh
[27:24] a lot of good insights in terms of both
[27:26] sides. Uh I wouldn't care that much
[27:28] whether he's right or wrong. I think
[27:30] he's been in this a long time on the
[27:31] financial side and I think he does good
[27:33] work. Uh I'll actually be on Julia's
[27:35] show coming up if we can agree on a date
[27:38] in the next couple weeks. But, this is
[27:39] probably the main thing that I took from
[27:42] the interview.
[27:43] I would tell you right now, Julia,
[27:44] probably half of all the managers out
[27:46] there that do private equity are going
[27:48] to end up having to go out of business.
[27:53] I don't want to say it again,
[27:55] but I'll just finish it. Because the
[27:56] results on their portfolios are so poor,
[27:59] they're not just going to be able to
[28:00] raise new capital. So, he's not saying
[28:03] this will be a domino of collapses. What
[28:06] he's saying is they're going to go out
[28:08] of business. Their portfolios may be
[28:09] purchased and they won't be out of
[28:10] business. So, don't take it as the end
[28:12] of the world. I am not a systemic bear.
[28:14] I do not believe this can't be
[28:16] controlled, but I do believe private
[28:18] credit is big enough and has been a big
[28:20] enough issue that has been a worry on
[28:22] everyone I know probably for the last 3
[28:25] years at least just because of the size
[28:28] and how quickly they had grown. If I
[28:30] took you through some of the private um
[28:32] equity managers that have been uh
[28:34] especially aggressive in the last year,
[28:37] and you look at their AUM growth, and
[28:39] you realize that a lot of that is coming
[28:40] from a combination of
[28:42] the insurance side and retail.
[28:45] A lot of these companies bought
[28:47] insurance companies
[28:49] and they use retail books. So again, for
[28:52] RIAs and FAs, if you're not getting the
[28:54] story, go subscribe, call up Mark
[28:57] Wailing at the firm and we can talk
[28:59] about a bigger package and I can come
[29:00] speak to you guys on a regular basis uh
[29:03] to at least keep you up on this so you
[29:04] can talk to your clients.
[29:06] If you haven't followed this story, this
[29:08] is in Connecticut um and this is a
[29:10] problem of an insurance company that
[29:13] basically
[29:15] had to run into a liquidation process
[29:17] and there's a lot of losses happening on
[29:20] the variable
[29:22] annuity side.
[29:23] So, I I just go read it. Um
[29:26] this kind of slipped through the cracks
[29:27] this month, but I'm bringing this up
[29:29] because when people ask me how are the
[29:31] hyperscalers connected to what's
[29:33] happening in the software and the debt
[29:34] side,
[29:35] they are absolutely connected to this.
[29:37] Um 100% connected to it. And the reason
[29:41] is
[29:42] when you go through the details and you
[29:44] realize who has been helping all of the
[29:47] AIs in their SPV off-balance sheet
[29:50] financing for the data centers,
[29:53] you start to get back to the same names
[29:55] that I've highlighted in multiple points
[29:57] already from the software stuff.
[30:00] That is not from the data centers yet.
[30:02] The problem with the data centers is not
[30:04] that AI is not going to be a boom. It
[30:06] is. But everyone underestimated memory
[30:09] going up to the level it did
[30:11] at the same time as the bottlenecks
[30:14] happening that are happening.
[30:16] And the bottlenecks are happening for a
[30:17] variety of reasons on the data centers,
[30:19] which I can go into detail with people
[30:20] if they're interested.
[30:22] You've got all of those bottlenecks and
[30:24] then like I said, you've got competition
[30:26] from the Chinese open source, which has
[30:27] caught up, which is preventing a lot of
[30:29] the revenue that would come to these
[30:31] companies. If they don't get the revenue
[30:32] in as I go back to Dario Modai, this is
[30:35] a levered bet. So if you have a levered
[30:37] bet, and I bring this up because meta
[30:39] auditor Ernst & Young raised red flags
[30:41] on data center accounting.
[30:43] This is a big story. Um and the reason
[30:46] it's a big story is because you're
[30:47] starting to get all of these things.
[30:48] You've got lawsuits going on with
[30:50] Oracle, lawsuits going on with some of
[30:52] the private equity companies
[30:54] by people that are trapped in things.
[30:56] And again, this is part of the doing
[30:57] business. This is not like some
[31:00] horrible story, but the fact that Meta's
[31:02] auditor
[31:03] raised red flags on this at their
[31:05] earnings report is just something you
[31:07] need to pay attention to because in the
[31:09] story, this means it was one of the
[31:11] hardest riskiest judgments the auditor
[31:13] had to make. Such a warning label is
[31:14] rare for a specific high profile
[31:17] transaction and major audit client. Meta
[31:19] owns 20% of the venture. Funds managed
[31:22] by Blue Owl, which we've seen has has
[31:24] sold off sharply, own the other 80%. A
[31:27] holding company called
[31:29] which owns the Blue Owl portion,
[31:31] sold the then record 27.3
[31:34] billion of bonds to investors.
[31:36] And here are the senators asking
[31:39] the Financial Stability Oversight
[31:41] Council led by Scott Bessent to
[31:43] investigate the risks
[31:45] posed by this to the AI system.
[31:48] Um you can't hide from these things. The
[31:51] strange case of Meta. Did Meta pay 270
[31:54] million a year to potentially keep some
[31:55] of its AI assets and liabilities off the
[31:57] balance sheet? Now, this is something
[31:59] you've seen Jim Chanos talk about.
[32:01] You've seen
[32:02] uh Michael Burry talk about. The only
[32:05] reason to me this matters now is because
[32:07] of what has happened with software
[32:09] stocks. Again, I did not
[32:12] for the life of me expect to see
[32:14] software fall off as fast as it is. When
[32:16] I was writing paper after paper about a
[32:18] re-rating in software,
[32:20] I did not expect it to go this fast.
[32:21] That's why the webinar went through what
[32:24] has happened to justify how quickly
[32:26] software has gone down. I thought it was
[32:27] a three-year process. I didn't think it
[32:29] would happen in two months. By happening
[32:32] in two months, it creates the credit
[32:34] cracks because believe it or not, guys,
[32:35] equity is part of the capital structure.
[32:37] That's why you're seeing the bonds come
[32:39] down. If the bonds are coming down, that
[32:40] means the bond investors think there's a
[32:42] risk that they're going to get their
[32:43] money back. Well, one of the ways to
[32:45] hedge that is to go short equity.
[32:47] We're in that cap structure sign. That
[32:48] happened with the fracking situation
[32:50] with oil. It happened with the banks and
[32:52] the mortgage brokers. If it's a game of
[32:54] deleveraging, remember,
[32:56] there is massive leverage in the system.
[32:58] Leverage did not go down since the great
[33:00] financial crisis. It just shifted from
[33:02] the banks to other places, and now we're
[33:04] seeing the side of liquidity go out.
[33:06] We've still got commercial real estate.
[33:07] We've still got private equity in terms
[33:09] of where did Harvard have to sell things
[33:11] off last year? How far down below par?
[33:14] All of these things are in there. In
[33:16] here, fun fact, this was the fifth
[33:18] largest bond issue ever done in the
[33:19] United States,
[33:21] the Meta deal. The accounting privilege
[33:23] of looking somewhat asset light and
[33:24] liabilities a plus with the private
[33:26] credit deal with Blue Owl. Of course,
[33:27] Meta gets to keep the associate debt of
[33:29] 27 billion off its balance sheet. S&P
[33:32] issued a statement clarifying they will
[33:33] not consolidate the Meta Blue Owl debt
[33:35] with Meta's debt.
[33:37] For this accounting privilege,
[33:41] the Meta Blue Owl JV is paying 6.6%
[33:45] to borrow money
[33:46] from Blue Owl funds. Again, insurance
[33:50] companies, retail, and whoever else is
[33:51] in these funds.
[33:53] To be fair, the Meta press release talks
[33:55] about the speed and efficiency in which
[33:56] private credit operates. So, this is the
[33:58] reason we needed to get this done
[34:01] quickly, and private credit allows us to
[34:04] raise the capital quickly and get the
[34:06] deal done quickly.
[34:10] Here is the overlay between private
[34:13] equity stocks and the hyperscalers
[34:15] relative to the S&P.
[34:17] You cannot separate the hyperscalers
[34:20] from the risks that are happening.
[34:22] Here is the hyperscalers relative to the
[34:24] S&P. And again, this is an equal weight
[34:26] of Amazon, Google, Meta, and
[34:30] uh
[34:30] Microsoft.
[34:32] You have the worst
[34:35] month
[34:36] over the last decade except for
[34:40] 2022.
[34:42] So, I leave it to you guys to make your
[34:44] own decision, but I think there is a
[34:46] clear sign of where the risks are
[34:48] associated. You had CoreWeave down big
[34:50] on Friday. Oracle CDS is still at the
[34:52] wides. There is a direct line right now
[34:55] into all of these deals and the fact
[34:57] that the software thing has unwound
[35:00] means there's an issue. Now, if it were
[35:01] only the software stocks and the
[35:03] software debt
[35:05] and the private equity firms, I I
[35:07] wouldn't be completely worried about it.
[35:10] But now
[35:11] on Friday, financial shares walloped by
[35:14] AI. Credit woes hit 3-month low. So, as
[35:16] the S&P is spinning its wheels near
[35:19] all-time highs the financials broke
[35:21] under the 200-day moving average where
[35:23] they closed the week. In the webinar, I
[35:25] went through how bad of a sign this is
[35:27] historically.
[35:29] Uh, again, the last 5 days, the two
[35:32] worst sectors, meaning last week
[35:35] the banks. So, this is not the private
[35:37] equity. This is the banks, commercial
[35:39] banks. So, that includes the KRE, which
[35:41] was down big, the regional banks. It
[35:42] includes other names within inside
[35:45] there. And then this one, consumer
[35:46] finance.
[35:48] Down 8%. This includes Amex, which was
[35:50] down big last week. Again, it's
[35:52] spreading, guys.
[35:54] So, now you've got financials that were
[35:55] hit and you're even getting to the cream
[35:57] of the crop.
[35:59] On Friday Goldman Sachs relative to the
[36:01] S&P had its worst day since
[36:04] the Great Financial Crisis.
[36:08] Here's just their outright move.
[36:11] Worst move since the liberation day.
[36:15] And again, if you want more contagion,
[36:17] and I said this on the webinar, this is
[36:19] one of the things I want my portfolio
[36:21] right now is duration. So, 10-year
[36:23] yields went below four. This is a fairly
[36:25] crowded view that we would have a
[36:27] steepening of the yield curve and rates
[36:28] would go higher. I I talked about this
[36:30] on the webinar as well and we broke the
[36:33] 200-week moving average for the first
[36:34] time since the tightening cycle started
[36:36] to be built into the market. So, you
[36:38] remember how much people have been
[36:39] bearish on Treasuries, it shows up in
[36:41] almost everything. You'll never hear
[36:42] anyone say I I love long bonds,
[36:44] particularly macro people.
[36:46] Uh I think it's a very crowded short to
[36:48] be short fixed income. Here is the major
[36:51] thing that I want people to have in
[36:52] their mind and why this this disruption
[36:55] is so real and so important. This is the
[36:58] game-changer that happened.
[37:01] I wrote this in Substack. If you didn't
[37:03] read it,
[37:04] AI agents change time. We are repricing
[37:08] the capital structure because of time.
[37:11] Time has shifted, guys.
[37:13] We are no longer in a world
[37:16] that moves slow.
[37:17] We are in a world that moves fast and
[37:20] all of this disruption is because AI
[37:22] agents allow things now to be built not
[37:24] just by humans,
[37:26] but by a billion to a trillion bots.
[37:29] Infinite bots running on Chinese
[37:31] software.
[37:33] This is something you just can't imagine
[37:35] would happen. I didn't envision this
[37:36] going on to pace. Open claw was a
[37:38] game-changer. Valuations are falling for
[37:41] a reason. AI is repricing the future. I
[37:44] start going through Bitcoin more and
[37:46] more. If you guys haven't done your
[37:47] homework on Bitcoin,
[37:50] by the time we get through this, if
[37:52] there is an event where the private
[37:53] credit situation does deteriorate, the
[37:55] Fed will have to do something. It'll be
[37:57] another liquidity facility to help. We
[38:01] have a midterm election coming up. We've
[38:02] got lawsuits going on. You've got the
[38:04] Democrats saying there's an issue with
[38:06] the SPVs. If people get spooked out,
[38:10] you're going to have to see liquidity
[38:11] facilities put in place, especially
[38:14] ahead of the midterms. That's when
[38:16] Bitcoin will start to separate itself
[38:17] because at this point, software is going
[38:19] down creating a deleveraging phase on
[38:21] the back of deflation. We need AI, which
[38:24] I'll get into as well. That is the other
[38:25] powerful part.
[38:26] Ben Horowitz was on
[38:30] moonshots.
[38:32] There were so many great nuggets in
[38:34] this. I highly recommend going in there.
[38:36] I'm not even sure they realize how many
[38:38] good nuggets were in there, but one of
[38:39] them
[38:40] was his statement about open claw.
[38:43] He basically said as you go through this
[38:45] that open claw changed the way Silicon
[38:48] Valley views AI.
[38:51] They're looking at this different. This
[38:53] changed everything. He goes through the
[38:55] startups getting more involved. This was
[38:57] something that shocked people.
[39:01] Andrej Karpathy had talked about AI
[39:03] agents being as as much as a decade away
[39:06] and for enterprises, which is where I
[39:08] think everyone made the mistake,
[39:11] it's true.
[39:12] Because there's so many things that have
[39:14] to happen to avoid the mistakes, the
[39:16] hallucinations, all of that stuff. A lot
[39:18] of it has to do with the data cleansing.
[39:20] The enterprises are going to take a long
[39:21] time. But for someone sitting at home,
[39:24] even with the mistakes, you can deal
[39:26] with that. I can deal with them making
[39:28] mistakes, with them doing something I
[39:30] don't want them to do.
[39:32] But I can run a business that way. I
[39:34] just have to make sure that they're not
[39:35] in my files that matter and I have them
[39:37] in a sandbox. This can't be done in
[39:40] enterprise. So the problem is again,
[39:43] if you have too many people, you can't
[39:45] handle the speed. I don't care what goes
[39:47] on. Too many people mean too much
[39:48] friction. Humans mean friction. Everyone
[39:51] has different incentives. So if you go
[39:53] back to the show me the incentives, I'll
[39:55] show you what happens. The incentives at
[39:58] an enterprise are to not blow up. It's
[40:00] to be risk averse cuz they get paid a
[40:02] lot of money
[40:04] at the C-suite level to not make
[40:06] mistakes. AI will not happen quickly,
[40:08] which means in my opinion,
[40:11] there is a high probability the revenues
[40:13] for the open AIs, for the anthropics,
[40:15] for the enterprise level are going to be
[40:16] much slower than what people had
[40:18] expected and the debt or the CapEx is
[40:21] going to either have to slow down or
[40:23] they're going to have to cancel buybacks
[40:24] in the case of the public companies. It
[40:26] doesn't really matter. Founders are
[40:28] excited about productizing open claw
[40:30] instantiations. He talks about how
[40:32] there are just tons of startups coming
[40:35] in where the people are like, "I can
[40:36] build something on open claw." So, this
[40:38] is having an impact. So, what you're
[40:39] going to get is the bot population bomb.
[40:42] There are 8 billion humans on the
[40:44] planet. If we start using agents in any
[40:46] meaningful sense, you'll get to a
[40:48] trillion agents very quickly.
[40:52] That is where this problem comes in on
[40:54] so many levels.
[40:57] Kilo claw, again.
[41:00] Another
[41:01] another ability by
[41:03] this stuff to happen, allowing anyone to
[41:05] deploy hosted open claw agents into
[41:07] production in 60 seconds.
[41:10] Open claw users are allegedly bypassing
[41:12] anti-bot systems. This is where the
[41:14] hacking stuff starts to become an issue.
[41:16] When you've got billions of Einsteins
[41:17] running around and then trillions at the
[41:19] end of the year.
[41:21] AI compresses time horizons. If 3 years
[41:23] becomes uncertain, long duration
[41:25] equities reprice first because their
[41:26] value depends heavily on distant cash
[41:28] flows. Credit cares about survival, not
[41:30] story. Multiples compress before spreads
[41:32] widen, that's what we've seen. Duration
[41:34] shrinks across the capital structure in
[41:36] an AI accelerated regime. Balance sheets
[41:39] matter more than narrative. Stocks price
[41:41] the future, bonds enforce where the
[41:42] future can be funded. That is what is
[41:44] happening and that's why we've gotten to
[41:46] an important thing. This gets back into
[41:48] equities or optionality in the future.
[41:50] So,
[41:51] especially for financial advisors,
[41:54] retail, anyone watching this that
[41:56] doesn't think of equity and fixed income
[41:58] in the same context, there's a big
[42:00] difference particularly in the cap
[42:01] structure. So, I'm writing a paper about
[42:04] how equity is now becoming
[42:06] debt. Meaning, you're at a point where
[42:08] if you don't know if an equity is going
[42:10] to make it, the PE goes down to 10 or
[42:12] 12, like it did with Salesforce.
[42:14] What's the probability of them being in
[42:16] business in 3 years, or at least growing
[42:18] in 3 years? What's the probability of
[42:20] getting money back? Salesforce is a big
[42:22] company. They're not the issue. They're
[42:23] still announcing buybacks.
[42:25] But you have issues for other companies.
[42:27] So,
[42:28] we're making a bet that margins will
[42:29] stay high. This is the risk that I think
[42:31] is going to be an issue this year based
[42:33] on
[42:34] all of the things that I've mentioned.
[42:35] Uh the cost for a data center, the cost
[42:37] for cloud, the inability to give all of
[42:40] the capacity that's needed means if
[42:42] people want capacity for AI agents, it's
[42:45] going to cost more.
[42:46] Um I don't know what's going to happen,
[42:48] but I don't see the data centers being
[42:49] built in time to supply what's
[42:51] necessary. If anything, I do for for a
[42:54] few clients, I do a a monthly data
[42:57] center recap, and the the the terawatt
[43:01] hours, the gigawatts by 2030, they just
[43:03] keep declining based on all of the data
[43:05] center delays. So,
[43:08] you can't price growth in the future if
[43:10] you don't know what's going to happen.
[43:11] Fixed income,
[43:13] you either pay the coupon or you don't.
[43:15] So, that's just the reality. That's why
[43:17] bonds and and equities are just
[43:19] different, and you see where it is.
[43:21] Uh and in stress regimes, it becomes
[43:23] obvious. There's secured debt, unsecured
[43:25] debt, preferred, and then there's the
[43:26] equity. Equity can rally while bonds
[43:28] quietly deteriorate, but when the bond
[43:30] market cracks, equities usually follow.
[43:32] So, watch credit. Credit is smarter.
[43:36] Credit's cracking.
[43:38] Um the other thing that has changed this
[43:39] month in the Substack that I wrote uh
[43:42] about agents and time, it also destroys
[43:44] the Black-Scholes model. I'm not going
[43:46] to go through all the details here. You
[43:47] guys can read it if you want.
[43:50] But basically, all of these things that
[43:52] the Black-Scholes model assumes, we are
[43:54] at a point now where
[43:56] none of that's there. We are trading
[43:57] like prediction markets. A stock can go
[43:59] from being a 25 PE, it can go down to an
[44:03] 18 PE the next day, and then be at a 12
[44:06] PE by the end of the week.
[44:08] That was never possible before. Um now
[44:11] you're getting movements that are
[44:13] jumping faster and out of nowhere. So,
[44:16] you're getting multiple standard
[44:18] deviation moves. And so, we're going to
[44:20] have times where things are quiet, you
[44:21] don't get a lot of movement, and then
[44:23] it's going to go. And market vol
[44:27] this is where it gets counterintuitive
[44:29] may not increase proportionately. So,
[44:31] individual equity vol isn't
[44:33] unambiguously higher, but cuz you keep
[44:35] getting these one-off gunshots.
[44:38] But market vol, this is what we've seen
[44:40] so far.
[44:42] But the dangerous part
[44:45] is that we are probably episodically
[44:47] explosive index moves. That's why I said
[44:51] my whole thesis on this is vol of vol
[44:54] VIX call options or other points
[44:59] you just have to have these in your
[45:00] portfolio to some degree. This is the
[45:02] paper that I'm releasing which is
[45:04] basically going to
[45:06] show people that the leverage in the
[45:08] system is in the equity market. The
[45:09] equity market's the most important
[45:11] thing. I've highlighted this before, but
[45:13] just so we see it.
[45:15] If this were debt to GDP, which we have
[45:17] at the government level at very high
[45:18] levels at over 100%, but this is the
[45:21] equity market relative to GDP, the
[45:23] quote-unquote Buffett indicator. Market
[45:26] cap to GDP is at 220%. Equity is where
[45:29] the leverage is in the system. That
[45:31] leverage is predominantly tech stocks.
[45:35] The 22V team, Dennis DeBusschere
[45:37] strategy team, this is a a must-see uh
[45:40] chart. Uh
[45:42] they broke this down and said, "Let's
[45:44] look at the S&P cash return relative to
[45:46] net income."
[45:48] And again, this is going down because
[45:49] the CapEx is huge. So, they're spending
[45:53] money,
[45:54] and you have this down at levels which
[45:56] historically has been when you've come
[45:58] out of a recession.
[46:00] Or the dot-com bubble here. This is
[46:03] massive amounts of spending, and the
[46:05] reason this gets important as you break
[46:07] it down. First of all, fair value.
[46:12] You can make the argument right now that
[46:13] we're 17% lower than we are just based
[46:16] on history.
[46:18] If you go through
[46:20] the impacts to this of what I put in
[46:23] here, rising memory cost, private
[46:24] competition, meaning from China, data
[46:27] center delays, or the open-source sorry,
[46:29] this is the private competition meaning
[46:30] open AI and XAI and Anthropic relative
[46:33] to the S&P companies. So, they've got
[46:35] competition from very, very big
[46:37] companies. The enterprise friction,
[46:39] meaning the adoption, and buyback
[46:41] productions. So, people keep asking me,
[46:43] will the hyperscalers cut CapEx? I don't
[46:46] think they will. Do I think Anthropic or
[46:48] open AI could? Yeah, I think open AI
[46:50] technically did. they were going to do
[46:51] 1.4 trillion. Now, we're talking 600
[46:53] billion by 2030.
[46:55] Uh didn't hurt them because they're a
[46:56] private company with a valuation, and no
[46:58] one seemed to care.
[47:00] Uh but, you could cancel buybacks if
[47:03] this becomes an issue, and I think that
[47:05] would not be taken well.
[47:07] Um if you go through what has when this
[47:09] has happened before and what has caused
[47:11] it, you get into these things. And
[47:13] again, like I said, it's mainly been
[47:15] times coming out of a recession.
[47:18] If the cash growth fails to keep pace
[47:20] current record high CapEx spending, we
[47:22] enter period often referred to as
[47:23] finance as the CapEx hangover.
[47:26] It's already consuming over 90% of their
[47:27] operating cash flow.
[47:29] The cash flow
[47:31] goes negative when you start to get into
[47:33] a bad situation. A shift goes from cash
[47:36] to debt financing. We're already seeing
[47:38] that. You're crowding out shareholder
[47:40] reserves uh returns. Buybacks disappear.
[47:42] Dividends stagnate. You get a value
[47:45] valuation de-rating. I just want you to
[47:47] read this because this is something I
[47:49] believe investors will stop valuing them
[47:51] as software companies and start valuing
[47:53] them as utilities or industrial firms
[47:55] with a lower multiple.
[47:57] That's what we've seen, guys.
[47:59] You spend too much money and you're
[48:01] being disrupted by something that is
[48:02] structural. If there's a question on
[48:05] whether you're going to be able to do it
[48:06] and no one has proven yet that they can
[48:07] make money off AI agents. Nobody.
[48:09] Uh the hyperscaler pivot to CapEx, or at
[48:12] least relative to their costs, I should
[48:14] say.
[48:15] Um
[48:15] so this is just the reasons they've
[48:16] fallen down to there. So again, I I I
[48:19] talked about it. I think you should just
[48:20] keep it in the back of your mind because
[48:22] even at the S&P level, you can see how
[48:23] this problem is growing. Uh Fed Governor
[48:26] Waller spoke this week. I just want to
[48:29] make sure you highlight that as we're in
[48:31] the situation, remember we do have a
[48:32] K-shaped economy. If the stock goes
[48:34] down, the economy is very levered to the
[48:37] stock market. We have very good PMIs, we
[48:39] have very good growth.
[48:41] I think at the end of the year, the
[48:43] growth will come through. I think the
[48:44] earnings will come through.
[48:46] Profit margins may not come through, but
[48:49] I think at some point if this thing
[48:51] doesn't see a significant change in the
[48:54] month of March. And when I mean
[48:55] significant, I mean where everything
[48:58] kind of tightens up the credit markets.
[49:00] I'm not caring for an equity bounce cuz
[49:02] that's why I think going through
[49:04] software stocks is a waste of time.
[49:06] Remember, the job market is weak and I
[49:08] do expect it to get weaker.
[49:11] Um not in a way that Citrini talked
[49:13] about, but I just think it's much weaker
[49:15] than people realize. If for no other
[49:17] reason, everyone's waiting to get fired.
[49:19] Uh
[49:20] there is an increasingly dystopian tone
[49:22] to commentary on AI. Fed Governor Waller
[49:24] say the thing that with AI is that it's
[49:27] coming at us so fast that it's easy to
[49:28] start seeing jobs that may go away
[49:30] before you see the new jobs that are
[49:31] going to be created.
[49:33] Um
[49:35] I I I tend to agree. Um I do think
[49:38] people should leave and start being an
[49:39] entrepreneur if they can.
[49:41] Uh this got a lot of
[49:44] This got a lot of press. Again, I I
[49:46] cannot believe what circulates. Um it
[49:48] got pressed cuz he laid off 40% of
[49:50] people. I mean, just go look at the
[49:52] stock of the last 10 years. I mean, they
[49:54] over hired. So, in the same way that a
[49:56] lot of the private equity firms have
[49:58] taken in tons of retail money and tons
[50:00] of money and purchased companies like
[50:03] insurance companies,
[50:06] you just get to the point where you got
[50:07] to get rid of people. So, I don't think
[50:08] this is an AI thing as much of it's they
[50:10] needed to do it cuz their stock was down
[50:12] over the course of the last 5 years.
[50:14] Um
[50:15] this didn't get a lot of press, but I
[50:16] just want to say it. I think this again
[50:18] is a read through the lines indication
[50:22] that the reason Dario Amodei is talking
[50:24] so openly and so often about the risk.
[50:27] He's insinuated it's OpenAI, but I think
[50:30] the real risk is that their business is
[50:32] doing well and yet they're showing every
[50:35] sign
[50:37] that they're surprised they're not
[50:38] making as much money. So, I showed last
[50:40] week that their margins have come down
[50:44] by 23%.
[50:47] This week, they dropped their flagship
[50:49] safety pledge and the reason embedded
[50:51] was, we didn't really feel
[50:54] with the rapid advance of AI that it
[50:55] made sense for us to make unilateral
[50:57] commitments if competitors are blazing
[51:00] ahead.
[51:01] So, this is a competition thing to get
[51:03] rid of something that they were built
[51:05] on. That says to me, again, like I said,
[51:08] if my revenue is not $1 trillion or it's
[51:10] even 800 billion, there's no force on
[51:12] Earth that will prevent me from going
[51:13] bankrupt.
[51:15] I I I think this is the issue to pay
[51:17] attention to particularly since
[51:20] now they're in a heated fight with the
[51:22] Defense Department. Uh this was all
[51:24] week.
[51:25] >> [laughter]
[51:27] >> And of course, the President of the
[51:28] United States, I'm directing every
[51:30] federal agency in the United States
[51:32] government to immediately cease all use
[51:34] of Anthropic's technology. We don't need
[51:36] it.
[51:37] Um
[51:38] in the same week, a hacker used
[51:40] Anthropic's Claude to steal Mexican data
[51:42] trove. So,
[51:45] this is going at the Mexican government
[51:48] agencies.
[51:52] I just want you guys to think about
[51:54] what's coming in terms of the cyber side
[51:56] and what it would do to enterprises.
[51:57] Would it speed up their adoption or
[51:59] would it slow down their adoption? That
[52:01] is necessary for Anthropic to have their
[52:02] money because they've already spent the
[52:04] money on the CapEx. So, we get back to
[52:05] the whole adoption. Will a massive hack
[52:09] or fears of hacking growing, which is
[52:11] becoming more and more of a reality
[52:12] because of open claw and because of how
[52:14] many people now have AI agents that can
[52:16] sit on a Mac mini or a Mac studio and do
[52:18] insane 24 hours a day damage?
[52:22] It's going to slow adoption. There's no
[52:24] way it speeds it up.
[52:26] Howard Marks
[52:27] Uh he wrote an article back in I believe
[52:30] it was November, December, may have been
[52:32] October. Is it a bubble? Where he went
[52:34] through the AI and effectively what he
[52:36] put in there is there are definitely
[52:38] signs of a bubble and there always are
[52:39] bubbles in this kind of spending.
[52:42] But I kind of took it as he didn't
[52:44] really know. So,
[52:45] this is the reason why everyone loves
[52:47] Howard Marks. He's a well-thought-out
[52:48] person, but also he did more work.
[52:52] So,
[52:53] his point is by making more powerful
[52:56] tools, analytical tools available to
[52:58] everyone, not just experts, AI
[53:00] democratizes innovation. I'm putting
[53:02] this up here because
[53:05] you guys have heard me speak a lot about
[53:06] the things that I'm I've got highlighted
[53:08] here.
[53:09] It democratizes innovation. I just wrote
[53:11] a substack on this. It allows more
[53:13] people to become inventors, innovators,
[53:15] and entrepreneurs. This is the rise of
[53:16] the 300 billion 300 million in the US.
[53:19] It's the rise of the 8 billion. This is
[53:21] the reason why Bitcoin will be the thing
[53:24] that comes out of this. I'll go through
[53:26] all the reasons why over the course of
[53:28] the the final 10 slides. AI democratizes
[53:30] innovation. It's creates a cycle of
[53:32] self-generating ideas where ideas
[53:34] generate more ideas. This is the whole
[53:36] point about bespoke. This is the thing
[53:37] about building stuff in home.
[53:39] If you talk to someone who said has a
[53:42] view on AI and on software and they
[53:43] don't have open claw and they haven't
[53:45] used it or they can't give you a long
[53:48] reason on why it's so important, I don't
[53:51] think you should listen to their
[53:52] viewpoints on software at least as a
[53:54] growth. No software company is going to
[53:57] be thrown out of Morgan Stanley right
[53:58] now in any meaningful way, but the
[54:00] question is will they get more seats?
[54:02] Will they pay for the AI agents and will
[54:04] there be another Morgan Stanley behind
[54:06] the Morgan Stanley? Will that Morgan
[54:07] Stanley be a crypto AI native company
[54:10] fintech
[54:11] that builds their own software that
[54:12] never gets to 5 billion and builds their
[54:15] own.
[54:16] The democratation supports Roubini's
[54:18] concept that ideas can be shared
[54:19] repeatedly without being used up
[54:21] potentially transforming innovation by
[54:23] allowing more people to contribute to
[54:25] economic growth. There's a 300 million
[54:27] people as opposed to the 10,000. AI's
[54:30] emergence is poised to be the latest
[54:32] example of creative destruction
[54:33] economist Joseph Schumpeter I've I've
[54:35] referred to him. I don't even know if
[54:37] it's every month for the last
[54:39] probably 7 years, but certainly over the
[54:41] course of the last 3 years. The
[54:42] possibility that job displacement may
[54:44] precede job creation
[54:47] such that the unemployment rate rise and
[54:48] participate in the limit. Okay, yeah,
[54:50] that seems like same thing and to be
[54:52] sure the AI transition I'm contemplating
[54:53] could have profound implications for
[54:55] monetary policy.
[54:57] I'm saying the same thing.
[54:59] The most significant thing that
[55:00] distinguishes AI is something we've
[55:01] never dealt with in connection with
[55:02] prior technological developments.
[55:05] Listen to Howard. AI's ability to act
[55:08] autonomously according to Claude AI was
[55:10] and again, he did all this work using
[55:12] Claude.
[55:13] AI was at a level one in 2023, level two
[55:16] in 2024, but now it's level three and
[55:18] the difference is a big one. Something
[55:20] big is happening for Matt Houmer. I went
[55:21] through this. He references it and talks
[55:24] about how well it's written, but also
[55:25] talks about a lot of the things in there
[55:27] that fit into the same piece. So the
[55:29] bottom line is is it a bubble?
[55:32] Is the technology a fad or an illusion?
[55:34] I I with conviction that it's a very
[55:36] real thing with the potential to vastly
[55:38] alter the business world and change much
[55:40] of life as we know it. I think it has
[55:42] the potential to
[55:43] I think its potential is more likely to
[55:45] be underestimated today than
[55:47] exaggerated. Completely agree. As I
[55:49] pointed out in December in every example
[55:51] of sweeping technological innovation,
[55:53] the headlong rush to build
[55:54] infrastructure has vastly accelerated
[55:56] the adoption of the innovation
[55:58] and caused
[55:59] a lot of capital to be malinvested and
[56:01] destroyed. There's no reason to assume
[56:02] this will be different. My entire
[56:05] webinar, everything we've gone through
[56:07] here is saying the same thing.
[56:10] The adoption of the innovation
[56:12] takes longer than people recognize, but
[56:15] the build-out is happening and the
[56:18] friction is growing and if it wasn't for
[56:21] Open Claw and the ability of Chinese
[56:23] models to be used very close to US
[56:26] models, I don't think that altogether be
[56:28] there, but when you add in the data
[56:29] centers and everything else, it's an
[56:30] issue.
[56:31] Um it's important to note that more
[56:33] money is going into inference capex
[56:35] these days and training capex. This is a
[56:37] good thing. This is why it's not
[56:38] systemic. This is why I have it in here.
[56:40] If it was going into pre-training, I'd
[56:42] be more worried, but there is massive
[56:43] pre-training data centers being built.
[56:46] Those are the ones with the massive
[56:47] memory costs associated with which is
[56:49] another thing of the equation that if
[56:51] you compare inference, a lot less cost
[56:55] for memory. If you go into the big data
[56:57] centers with the Blackwells and the
[56:59] heavy heavy compute, you're in a
[57:00] different story. So again, I wrote
[57:02] democratization as the input
[57:03] concentration as the output. How it's
[57:06] re-pricing this.
[57:07] The entrepreneurial labor everything
[57:09] that he just talked about is in this
[57:11] paper if you didn't see it.
[57:13] The one thing that I spent a lot of time
[57:15] on that he didn't is how capital changes
[57:17] in the A of AI of agents. This is a
[57:20] deeper story. It's about duration.
[57:22] Equity begins to behave less like
[57:24] ownership of a franchise and more like a
[57:26] call option on execution. This is the
[57:29] whole thing of prediction markets. You
[57:31] just don't know whether they're in
[57:32] there. A call option can fall 20% in a
[57:34] day.
[57:35] AI rewards velocity and velocity demands
[57:38] different infrastructure.
[57:41] Public markets, again, I have said
[57:43] publicly I will continue to say it. I
[57:46] don't know what public companies will be
[57:48] able to grow in a decade. I don't know.
[57:51] In 5 years, I don't know any of the non-
[57:55] physical ones. So, you guys can take it
[57:58] however you want, but this is where the
[58:00] story for crypto and Bitcoin starts to
[58:02] take off. Traditional banking rails were
[58:04] constructed for human speed.
[58:07] For transactions between institutions
[58:09] operating on a quarterly and annual
[58:10] clock, none of this was designed for
[58:12] where were millions of AI native
[58:14] businesses need capital formation and
[58:16] continuously are transacting.
[58:19] Tokenization enables fractional
[58:21] ownership and capital access. The last
[58:23] 15 years were rewarded the question, who
[58:25] can build the largest moat? The next 15
[58:27] will be a different one, who can adapt
[58:29] fastest when moats are no longer
[58:30] guaranteed time?
[58:33] This is why I wrote it. What does the
[58:34] capital architecture look like when the
[58:36] economy it serves runs 24/7 and time is
[58:39] the scarcest asset of all?
[58:41] This is where you can go listen to Raoul
[58:43] Pal.
[58:45] Emod.
[58:47] You can go listen to Jeff Curry.
[58:49] You can go listen to Ben Horowitz. All
[58:52] of them
[58:53] talked about this. All of them,
[58:55] including Jeff Curry, mentioned the role
[58:57] that crypto plays in this world of
[59:00] speed. All three of these
[59:02] gentlemen talked about it. In the case
[59:04] of Ben Horowitz,
[59:07] AI cannot fulfill its potential without
[59:09] crypto.
[59:11] If AI is going to fulfill its potential
[59:12] like it would help a lot if crypto was a
[59:15] pervasive utility for it. The financial
[59:17] guardrails is one angle which I just
[59:19] talked about, but this is the other one
[59:21] and this is the part that I talk about.
[59:23] With everyone scared about quantum,
[59:25] which is down the road, we are not there
[59:27] yet, and we may not be there for a
[59:28] decade.
[59:30] Nobody is worried about agent swarms. I
[59:32] can say this over and over again,
[59:34] but where do you go if you've got deep
[59:37] fakes and you can't tell when someone is
[59:39] calling you who it is. What do you do
[59:42] when you get a video and you can't tell
[59:43] whether it's real or not, and what do
[59:45] you do with agents hacking into your
[59:47] system?
[59:49] Well, you go to cryptography and you go
[59:51] to the blockchain and on chain. That's
[59:52] where we're headed. If you haven't done
[59:54] your homework on it,
[59:55] you're playing from behind. And oh, by
[59:58] the way, the final positive note for
[60:00] Bitcoin, my thesis has always been that
[60:03] the only time that the people who own
[60:05] the world's wealth, the 800 trillion,
[60:07] will go buy a little old asset with $2
[60:09] trillion, is when they start to doubt
[60:12] everything.
[60:13] So, you're doubting everything with AI
[60:15] as I just showed, but the other thing is
[60:17] you have to doubt the investments that
[60:18] you're in. If I'm right about public
[60:20] companies basically not growing ever
[60:22] again, then being tokenized and becoming
[60:25] these just out there things that are
[60:26] surviving like Ford,
[60:29] where are you going to put your money?
[60:30] Into something that everyone else is
[60:31] putting your money into. Part of it will
[60:33] be gold as it is.
[60:35] But in particular, the biggest negative
[60:37] that could happen is this.
[60:40] The government is fighting
[60:42] with the hyperscalers. And Anthropic may
[60:45] not be a hyperscaler, but basically what
[60:47] they're saying to the world right now is
[60:50] we are taking over AI because AI is too
[60:53] important for national security. It's
[60:55] too important for each individual. If
[60:57] the government gets involved and it
[60:58] becomes regulated,
[61:00] what multiple should a regulated
[61:02] utility-like company have?
[61:04] That is what I imagine right now for
[61:06] multiple compression. The hyperscalers
[61:07] have a debt issue. I think they have a
[61:10] we've gone too big, too fast, too
[61:11] important thing. Uh a lot of stories
[61:14] about Anthropic basically saying they
[61:16] didn't want to be involved in the Iran
[61:17] situation. They were clearly involved in
[61:19] the Venezuela situation. AI is already a
[61:21] military issue, and we already have
[61:23] hackings and involves Claude and
[61:25] Anthropic. I think it is very, very
[61:27] possible that the government gets too
[61:29] involved. The nationalization of AI
[61:31] threatens innovation in the American
[61:32] mind. This is a libertarian
[61:35] paper. I bring this up again because I
[61:37] think this is a story that everyone
[61:39] needs to have in the back of your mind.
[61:40] The most bullish thing
[61:42] for crypto will be if the most important
[61:44] fiat assets are basically in some way,
[61:46] shape, or form controlled by the
[61:47] government. That's it for me this week.
[61:50] Um again, thanks to everyone who's
[61:52] subscribed. Thanks to everyone who's
[61:53] reached out and who's sending me what
[61:55] they've been building.
[61:56] Uh I love seeing all the open cloth
[61:58] stuff. I love hearing what people are
[62:00] doing. The video series on AI and how to
[62:02] use it more and make sure your kids are
[62:04] using it. Uh I'm hoping to get the
[62:06] entire completed series out by the end
[62:09] of next week. That is my goal.
[62:11] Uh and I will see you guys next week.
[62:14] Go check out the webinar.