🇪🇸 Español 🇬🇧 English

Jordi Visser / VisserLabs

The SaaS-Pocalypse, Bitcoin Crash and Why Understanding Palantir is Now Important for Both of Them — Jordi Visser (8 febrero 2026)

AIBitcoinMacroMarketsTrading

Complete 61:17 min youtube 2026 Week 06 Feb 08, 2026 🇬🇧 EN

Open YouTube video Back to explorer

TL;DR

The speaker argues that AI is reshaping software and the wider economy, favoring semiconductors, materials, energy, and AI infrastructure while warning that many software companies may lose value or growth.
They acknowledge being wrong on Bitcoin’s price target but continue buying it, including for their children, and say crypto could benefit as AI agents increase demand for stablecoins and financial rails.
They bought Palantir after its decline and expect a prolonged rotation away from US growth stocks, while flagging risks from forced selling, infrastructure bottlenecks, and rapid changes in AI deployment.

Research summary

▶ Plans for the channel and the broad AI thesis

The speaker says a paywall is expected to start this week, with options intended to provide more regular and detailed AI and trading insights. They frame the market as undergoing a long-term transition, citing the PMI moving above 50, panic in software as a service (SaaS), a roughly $90 trillion infrastructure opportunity, and a coming focus on critical minerals. They argue AI affects industries across traditional boundaries and say they do not think sell-side research can adequately cover that reach. On Bitcoin, they admit their $250,000 year-end expectation was wrong and say the price was 76% below that expectation; they describe buying over the past year, including for a strategic reserve for their children, and say Bitcoin is their largest net-worth position. They credit John Ro for having identified $60,000 as a downside target, which the price reached. The speaker says their long-term Bitcoin thesis remains tied to macroeconomic change and AI-driven disruption, not monthly price movements, and connects Bitcoin to programmable money and software while arguing that its trading relationship with software may change.

▶ Software repricing, venture capital, and Bitcoin’s connection

The speaker says the software selloff followed developments in AI, particularly Anthropic releases and Opus 4.5, and argues that people who have not built apps may underestimate the pace of change. They expect some software companies to survive but say many will not, and describe the sharp fall in the software ETF relative to the Nasdaq 100 as historically unusual; they infer that the worst of the selloff may have occurred, while stressing that the disruption is real. They connect Bitcoin and crypto to the 2021 venture boom, when SaaS and crypto both attracted substantial venture funding, and to the subsequent strain from higher rates, Silicon Valley Bank’s 2023 failure, and AI startups drawing capital. The speaker sees private equity, private credit, real estate, and venture investments as facing liquidity pressure, and describes Bitcoin as a liquid hedge for private-market exposure. Their bullish Bitcoin scenario depends on other growth businesses being disrupted by AI while Bitcoin remains a growth asset. They argue that agents could increase stablecoin use and crypto utility; they note that the Genius Act exists and describe the Clarity Act as having roughly 50/50 odds of passing before year-end. In their view, Bitcoin’s larger potential catalyst is disruption across businesses built on code, rather than financial guardrails alone.

▶ AI agents, enterprise software, and the Palantir case

The speaker describes how Claude Code led them to build apps and argues that falling intelligence costs and agentic AI threaten software businesses. They question Microsoft’s competitive position, saying they use several AI models but not Copilot, and ask whether businesses will need Salesforce if agents can query lead data directly. They characterize Salesforce as a forms and workflow wrapper, while acknowledging that established systems of record may persist as data stores. They discuss an OpenAI enterprise diagram as positioning AI orchestration above traditional SaaS, potentially reducing those companies’ role and growth without necessarily putting them out of business. Using a transcript analysis of Jason Calacanis’s proposed “Ultron” project, the speaker distinguishes AI-native startups from large enterprises: they doubt new AI-native startups will use Salesforce and suggest some may be disrupted before becoming mature businesses. They describe Palantir as an AI company suited to the agentic enterprise model. They cite reported revenue of $1.4 billion, up 70% year over year, government revenue of $570 million, up 66%, and commercial revenue growth of 130% year over year and 28% quarter over quarter. The speaker says they bought Palantir after a roughly 35% fall from its peak, seeing value around $130–$125, and says at $100 they believed it could earn more than $1 per share that year. They caution that Salesforce might become an AI company but that this would be a pivot with uncertainty warranting a lower multiple; they also say SaaS valuations remained expensive despite compression.

▶ Software contagion fears and spending risks

The speaker says they are not trying to pick a bottom in software. They argue that fears of systemic contagion overstate the size of the software exposure: after excluding Microsoft, cybersecurity, and Palantir, they estimate roughly $2 trillion remains, comparable in scale to energy and materials combined. They expect continued software pressure and a decade-long rotation of passive investment away from the US and toward materials, Brazil, and German manufacturing, while allowing for near-term short covering and slower mutual-fund repositioning. They dispute the claim that software is simply “dead,” describing competition and repricing as ongoing, and cite a private-credit executive’s statement that the tech portfolio had no red or yellow flags and largely green flags. They see rapid AI adoption as a potential threat to software revenue, especially if Salesforce reports slowing top-line growth, though they say that had not happened yet. They cite Palantir’s customer account that 97% of employees use Foundry daily and that it replaced third-party software functionality and added features within a year, attributing the result to its ontology. They warn AI could compress change into a short period and note that continued software weakness could pull down Bitcoin and Palantir again, particularly in a forced liquidation. They cite combined Big Tech spending of $650 billion for the year and revenue growth of 17% at Microsoft, 18% at Alphabet, 24% at Meta, and 14% at Amazon, while warning that data-center construction may become bottlenecked. They expect AI agents eventually to break the recent correlation between SaaS assets and AI winners.

▶ Weekly markets and the growth-to-value rotation

In the weekly recap, the speaker says the S&P fell 10 basis points, marking a fourth consecutive week with a move under 1%; the Nasdaq 100 fell 2%, the Magnificent Seven fell 4.7%, and the Russell 2000 rose 2%. They say only 7% of small-cap companies had reported through Tuesday, but 79% had beaten estimates, which they describe as tracking toward the best beat-rate performance ever. They report that an equal-weight ETF made new highs during the week’s panic. Their turbulence model, which tracks 100 assets, signaled four days of elevated covariance before a larger S&P drop; they say the market had not yet cleared the signal because the S&P had moved sideways for three months and had not retaken its highs. They say they resumed buying Bitcoin after previously stopping at $92,000, having said that three closes above $92,000 would indicate a bottom; they acknowledge the bottom was about 40% lower and that John Ro was right. Sector performance, in their account, showed staples and small caps stronger while technology and the Magnificent Seven lagged. They expect energy and materials weights, then 3.3% and 2% respectively, to rise over five years as growth exposure shifts toward companies benefiting from infrastructure spending. They cite a 7.89% weekly fall in a growth-versus-value measure and say software’s year-to-date decline contrasted with energy’s gain; they describe a Chevron-versus-Salesforce proxy trade as up more than 60% year to date.

▶ PMI strength, semiconductors, and the $90 trillion buildout

The speaker says the manufacturing PMI broke above 50, as they had forecast, and expects it to move toward 60 over the next three to five years; they say a move close to 60 before year-end would not surprise them. They cite a 4.7-point monthly increase and new orders at their highest since 2012, while acknowledging uncertainty about the drivers, including possible effects from a depreciation bonus. They expect broadening market participation and continued pressure on momentum strategies concentrated in software and the Magnificent Seven. They link rising DRAM prices and strength in semiconductors and materials to a hardware shift that also contributed to software’s decline. They favor equal-weight semiconductors and cite names including Teradyne, Lattice Semiconductor, and Corning, arguing that AI demand may support optical fiber, glass, and components needed for upgrades to vehicles, phones, trucks, and computers. They argue that investors who dismissed AI demand and capital expenditure as a bubble missed memory and related opportunities. The speaker compares the current rotation to the post-dot-com period, when technology fell while energy and materials rose, and says the current infrastructure trade benefits semiconductors, cooling, and power more than the Magnificent Seven. They cite a $90 trillion decade-long shift and say copper and silver supply is insufficient for the buildout; they also cite Micron trading below 10 times that year’s earnings.

▶ Commodities, critical minerals, and Brazil

The speaker compares pressure on incumbent SaaS companies to the retail disruption from Amazon and to energy’s changes after fracking, arguing that incumbents may survive without matching new competitors. They say Truflation was below 1% for the week and regard it as informative, while cautioning it may not move markets. In their view, investors should worry more about disinflation or deflation until gasoline prices rise, despite a conference audience expecting inflation above 3%. Summarizing Jeff Currie’s argument, they say asset-light hyperscalers are putting steel in the ground, which could pressure equity multiples as they resemble asset-heavy commodity producers; he identifies a possible repricing window in 2026–27 as compute prices fall while capital spending rises. The speaker agrees that the AI race depends on energy and many metals, and says commodity prices may be volatile but tradable. They argue that rising capital expenditure could reduce buybacks and coincide with substantial technology IPO supply. They also warn that long-term commitments and remaining performance obligations may become liabilities if bottlenecks or technological changes prevent delivery. They cite US and Chinese intentions to stockpile critical minerals, including China restricting silver exports, and interpret this as evidence that supply chains may be weaker than commonly understood. They mention a US-Brazil deal involving $565 million for rare-earth extraction and describe Brazil as a major mineral opportunity. They expect Brazil’s currency, bonds, equities, and commodities to move together this year, while noting that it is an election year and political chaos is possible. They also recommend reading about gas-turbine lead times.

▶ Edge AI, deployment, and infrastructure constraints

Discussing Elon Musk’s interview, the speaker emphasizes the difficulty of building data centers, power plants, and space infrastructure, and says Musk was projecting satellite deployment within 30 months. They agree with the argument that hyperscalers are building computing capacity faster than physical infrastructure can use it: training remains necessary, but deployment and inference are where value compounds. The speaker contrasts about $6 trillion of data-center capital expenditure with the $90 trillion opportunity they associate with AI deployment and the physical economy. They say edge AI could route around power constraints by putting inference in cars, robots, and factories, and describe AI as shifting from SaaS toward industrial infrastructure. They caution that long-duration capital spending is risky when technology evolves exponentially and that companies may not receive projected revenues. They then describe an edge-AI risk case involving local open-source models: the speaker says the Chinese model Kimi 2.5 could reduce costs by 90% compared with running an agent on Claude, though they qualify that it is a Chinese model. They report that Anthropic briefly restricted a tool’s usage and say this illustrated the vulnerability of relying on closed models. They describe AI’s progression from chatbots to agents and then to decentralized computing on Mac Minis and Mac Studios, saying they ordered two Mac Minis and a Mac Studio for local agents. They identify a potential inference-revenue cliff for hyperscalers if capable models run locally, while distinguishing large enterprises from budget-constrained startups and entrepreneurs. They say an American open-source model catching up would be an important further disruption; they expect smaller firms to adopt open source sooner than government and defense.

▶ AI adoption, crypto utility, and closing conditions

The speaker says open-source AI could pressure hyperscalers but distinguishes adoption by small businesses, mid-market firms, large enterprises, and government or defense, with the latter expected to move more slowly. They argue that forward revenue commitments can be bullish evidence of predictable growth but become a risk amid uncertainty about model costs, disruption, and delivery. They connect the rise of AI agents to Axie Infinity as an example of how crypto-related economic systems could have consequences when agents participate. They cite a report that physicists at a closed meeting said AI can now do up to 90% of their work and may soon push discovery beyond human understanding, noting that the participants had been skeptical. The speaker closes with a reported stablecoin volume record of $10 trillion and argues that AI agents could increase transaction volume and crypto network effects. They say crypto could remain under pressure because it is connected to SaaS through the 2021 cycle, and identify a break in the Bitcoin-SaaS correlation or stabilization in SaaS as a possible catalyst. They note John Ro had called $60,000 and might expect a lower move, and say most technicians would not yet call a bottom. The speaker ends by saying the paywall launch is expected this week, offering help with demonstrations and tool use, and saying they intend to write about Palantir, the “Ultron” theme, and related companies.

Full transcript

[0:00] in my home base of Brooklyn. Uh we got a
[0:03] lot to go through. Uh I just want to
[0:05] basically start with uh for those of you
[0:07] who've been reaching out and been
[0:08] patient. The payw wall will start on 22V
[0:12] this week. There'll be a variety of
[0:13] different choices for people. The goal
[0:15] at the end of the day is both to help
[0:17] you with AI, but also to help you from a
[0:19] trading perspective and hopefully give
[0:21] you some insights that are more regular
[0:23] and more detailed than what you get on
[0:25] on the video. uh where I'm trying in a
[0:28] matter of 45 minutes to just take you
[0:29] through kind of the week but also get
[0:31] you the themes and the secular moves
[0:33] that I think are happening. The reason
[0:34] this week is an important one is because
[0:36] a lot happened this week. We had the PMI
[0:38] breaking above 50 finally. We had SAS
[0:40] going into a complete panic and leading
[0:43] to fears over contagion and I had a lot
[0:45] of people sending me stuff as I uh
[0:47] posted in X. I will always post more in
[0:50] X when I'm receiving things and when I
[0:52] think people are kind of losing their
[0:53] mind. Uh this is going to be a 10-year
[0:56] process guys. Uh we did a lot in let's
[0:59] say a week in terms of the SAS side but
[1:02] what we haven't yet done is just the
[1:05] recognition of the $90 trillion
[1:07] infrastructure in number seven and
[1:08] what's coming the race for critical
[1:10] minerals. There's a new theme that I'm
[1:12] going to focus on a lot. Uh number four,
[1:14] Palunteer, what All-In talked about in
[1:17] SAS late, the drop that happened late
[1:19] last night. I listened to it this
[1:20] morning. It's Saturday. Uh Elon's
[1:22] interview, same thing. And then
[1:24] preparing for the edge. These are all
[1:26] trades and themes which I think you want
[1:28] to be on top of with me because like
[1:30] I've said before, I don't think the sell
[1:31] side can cover this. I don't think they
[1:33] have the capability of it. Uh and that's
[1:35] because AI is like electricity. It is
[1:39] affecting things outside of the
[1:41] traditional silos in the same way that
[1:43] people are trying to pick SAS and trying
[1:45] to stick with silos. There are no silos
[1:47] in a world of AI. There are no silos
[1:50] withund
[1:51] 60 IQs and Einstein IQ's for AI across
[1:55] every single vertical that you can think
[1:58] of. We're going to start with Bitcoin.
[2:00] Um, so the line up top there, which is
[2:04] 250,000, um, that's where I thought it
[2:07] was going to be at the end of last year.
[2:09] Yesterday, um, it was 76% below where I
[2:12] thought it would be at the end of last
[2:13] year. So, I've been wrong. The other two
[2:15] lines there are where I've been buying
[2:16] the last uh, year, partly for myself,
[2:19] partly in terms of setting up a
[2:21] strategic Bitcoin reserve for my
[2:23] children, which I've talked about with
[2:24] Anthony Pompiano many times. Uh
[2:28] so I've been wrong and it's not felt
[2:31] good. These this is the biggest part of
[2:33] my net worth at this point. Um and it's
[2:35] gone wrong. So the question is has
[2:37] anything changed for me? Well, the first
[2:39] thing I want to do is give some kudos.
[2:41] Uh, I've shown John Ro and I've talked
[2:43] about the fact that John Ro and I had
[2:44] different views on Bitcoin and I'll say
[2:46] for this year, but regardless of for
[2:48] this year, what has been true, John put
[2:50] this out um,
[2:53] last year uh, at the end of the year and
[2:56] basically said his target on the
[2:57] downside was 60,000. Well, we hit
[3:00] 60,000. So, if you guys are not
[3:02] following John Rog, John and I do a lot
[3:03] of work together. Um, we are on the same
[3:06] page on literally almost everything for
[3:08] this year. We've known each other for o
[3:11] over 25 years at this point. Uh we've
[3:14] been in the same business. We've both
[3:16] been from the macro side. The great
[3:18] thing is when John and I align, I think
[3:19] there is a tremendous amount of upside.
[3:21] I'll be highlighting again. I use him on
[3:23] all the ideas that I send out through
[3:24] 22V. I just wanted to bring it up and
[3:26] give him credit. If you're not listening
[3:27] to him, if you're not signing up for his
[3:29] work, I think you're making a big
[3:30] mistake. He has been all over this
[3:32] stuff. And I think this world that we're
[3:34] in is meant for John Rog. So I think you
[3:36] should be spending more time. Um
[3:40] there's not only bearish sentiment in
[3:42] terms of the crypto community, which
[3:43] I've talked about and written about many
[3:45] times about how bearish people are, but
[3:47] when you get in stuff like this, um
[3:49] buying Bitcoin is like buying air. Too
[3:51] many people are brainwashed by
[3:53] non-rigorous and unquestioning actors
[3:54] dressed up as the business media, social
[3:56] media, investment strategists trying to
[3:57] sell you garbage at inflated prices. It
[3:59] is why I call BS to them so frequently.
[4:02] You enjoy your Bitcoin. So, the great
[4:03] thing about this is I don't I don't give
[4:05] a crap about what Doug Cass says. I
[4:06] mean, people that are bearish on things
[4:08] when they've just fallen, good for them.
[4:10] Kudos. I'm sure they were bearish at one
[4:12] and 1,000 and 10,000. Uh, this is not a
[4:15] personal thing against him and what he
[4:17] posts. But I do think we've reached a
[4:19] point where when it's beaten down in the
[4:21] ground, I not only want to see the
[4:22] crypto side bearish. I want to start
[4:24] seeing the people who are perma bearish
[4:27] on it, another one, Michael Bur, um,
[4:29] come out and start talking about how
[4:31] things are going to go bankrupt and
[4:32] what's going to go on. So hopefully
[4:34] between John Row calling for 60,000,
[4:36] everyone coming to pound out, me being
[4:39] uh completely wrong and trying to
[4:40] appease the gods and not only say I was
[4:42] wrong, but explicitly say I was wrong
[4:44] and something I'm buying for my kids and
[4:47] just saying to myself, I do like me. I
[4:49] think I'm going to be okay with this
[4:50] one. I really do like it. In fact, I
[4:52] wrote this piece. Now, the interesting
[4:54] thing I want you to pay attention, this
[4:55] was written December 8th. This is my
[4:58] view of Bitcoin. I'm not in this for
[5:00] trading on a on a monthly basis. I don't
[5:02] care whether it's 170,000 right now or
[5:04] 50,000. My belief is on what I think is
[5:07] happening from a macro basis from the
[5:08] disruption of AI. And because most
[5:10] people and especially the two I just
[5:12] referenced have called AI a bubble
[5:14] continuously on the upside faded things
[5:16] like Micron and Nvidia and everything
[5:18] else. I have full faith that when the AI
[5:21] aentic world accelerates which is what
[5:23] is happening right now people start to
[5:25] recognize how the crypto utility phase
[5:28] starts to accelerate. So I wrote this
[5:29] paper if you haven't read it. It all
[5:31] gets through why I believe in this from
[5:33] a macro basis and it hasn't changed. But
[5:34] that was on December 8th. At the same
[5:36] time on December 8th, I wrote this vibe
[5:38] coding is a chat moment for code. Why
[5:40] the great software rerating of 2025 to
[5:43] 2030 has already begun. The reason I
[5:45] wanted to bring this up is not to pat
[5:47] myself on the back for at least getting
[5:48] the software side. The key thing in this
[5:50] is the software rating of 2025 to 2030
[5:53] has already begun. My belief is that
[5:55] this would be a five-year multiple
[5:57] compression that the street would
[5:58] continue to try and buy SAS. There'd be
[6:00] these little bounces and there'd never
[6:02] be this dramatic shift in like a day.
[6:04] That's kind of what happened with
[6:06] Amazon. That's kind of what happened to
[6:07] a degree with the energy side until
[6:10] crude fell violently from fracking. When
[6:12] technology disrupts something, usually
[6:15] you don't build it in and rerate
[6:17] everything off the bat. Maybe it was
[6:18] because software's, you know, price to
[6:20] free cash flow is so high. I don't
[6:22] really know. I'm surprised at how much
[6:24] it's fallen, but it also happened at the
[6:26] beginning of the year. The reason I
[6:27] bring this up is I fully expected and
[6:30] know that Bitcoin is part of what I talk
[6:33] about all the time, which is the reason
[6:35] I think the dollar is going to be weak
[6:37] is because we had a monopoly on code in
[6:39] the United States and that monopoly is
[6:40] now gone. Coding is now ubiquitous.
[6:44] Software companies are forever changed
[6:46] and most of them will die. Some of them
[6:49] will survive. You can I'm going to go
[6:50] through some of the ones that I believe
[6:52] should be surviving now. But when you
[6:54] have a fall like this, and I want to
[6:56] highlight this is December 8th. I did
[6:59] not expect when it was sitting just off
[7:01] the all-time highs to see a complete
[7:03] plunge like we're in Liberation Day. In
[7:05] fact, this is far worse than liberation
[7:08] day. So, this is a fairly historic move
[7:11] how quick it is. Now, Bitcoin is built
[7:13] on code. So, as Jim Biano said, it's
[7:16] programmable money. Yes, it should trade
[7:18] like software stocks, but not forever.
[7:21] It should trade like agentic software
[7:23] stocks in my opinion. So AI fears pummel
[7:27] software stocks is it illogical panic or
[7:29] a SAS apocalypse?
[7:32] This is all based on many many things
[7:34] released by anthropic over the course of
[7:36] the last two months. And I want to
[7:37] reemphasize that the reason I wrote the
[7:39] December 8th piece is because on
[7:42] November 29th, Opus 4.5 was released and
[7:45] everything changed overnight. The clawed
[7:47] moment has been the disaster. I've
[7:49] written multiple pieces about it. I've
[7:51] done every video shown the things that
[7:53] I've been able to build. You've seen the
[7:55] things in X. For anyone that fades this
[7:57] at this point that hasn't built their
[7:58] own app, you have no right to say
[8:00] anything about it because you don't know
[8:02] what you're talking about. I don't care
[8:03] if you work for Goldman Morgan or a
[8:05] hedge fund. If you believe that you can
[8:08] that you understand the part of being
[8:11] able to build apps, the agentic side
[8:13] that is spreading, if you haven't bought
[8:14] a Mac Mini, I think you should go spend
[8:17] some time on critical minerals. That's
[8:18] just my suggestion. Find something that
[8:20] is definitely going to work as opposed
[8:21] to going through and figuring out
[8:23] software companies like Salesforce, like
[8:25] Adobe, and try to figure out whether
[8:27] there all of a sudden people are going
[8:29] to change. There will be survivors and
[8:31] there will be companies that last for a
[8:33] long time. Hell, I'm sure Morgan Stanley
[8:35] still has uh mainframes from IBM from
[8:37] the 1970s. They did when I was there. It
[8:40] doesn't mean they'll get rid of
[8:41] everything, but I do think you're
[8:42] underestimating how fast this is going
[8:43] to go. Anthropics AI agents spark panic
[8:46] over future of Indian IT services. This
[8:48] is not a US thing. This has been global.
[8:50] To put this in perspective, this is the
[8:52] IGV software ETF relative to the NDX.
[8:55] Let's now strip out the growth component
[8:57] to some degree. This is the 30-day rate
[9:01] of change.
[9:02] This is unlike anything we've seen in
[9:05] doubting it. This you can't even find
[9:07] anything in terms of this. When we start
[9:10] when the Fed started to raise rates and
[9:12] we started to repric longduration assets
[9:14] based on inflation going higher, you
[9:16] didn't see anything like this. This is
[9:18] unprecedented
[9:19] and it leads me to believe that we've
[9:21] probably seen the worst of it. Now,
[9:22] here's the overlay of IGV with the VC.
[9:27] This is the uh Thompson Reuters Venture
[9:31] Capital Index. That is the white line.
[9:33] The orange line is IGV. I want to bring
[9:35] this up because this is where
[9:38] I always knew this was the case, but I
[9:40] didn't think this would go this fast. So
[9:43] remember,
[9:45] when you're going through what happened
[9:48] in venture, you have to remember what
[9:50] happened in 2021. 2021 was a massive
[9:54] venture world. Everyone was coming out
[9:56] of stuck at home. We had SAS startups
[9:59] everywhere.
[10:01] And yes, crypto had a huge boom as well.
[10:04] the crypto boom that led to FTX and
[10:06] everything else occurred during the 2020
[10:08] to 2021 period. This is an overlay of
[10:10] Bitcoin to that venture index. And I
[10:14] brought it up. It's obviously the same
[10:15] because it tracks with IGV. Here is the
[10:17] total number just to put in perspective
[10:20] of global VC that occurred in 2021 and
[10:24] 2022 combined. Those two years, I mean,
[10:27] you're basically as big as the four
[10:28] prior years, but you can go back. It's a
[10:31] a huge majority of 2015 to 2020.
[10:37] Here is the crypto VC. So again, in
[10:41] 2021, crypto and SAS were together and
[10:45] SAS is a huge part of the VC world. I
[10:50] bring that up because there's no way to
[10:52] separate that Bitcoin is related and
[10:54] crypto is related. So, one of the
[10:56] reasons in my opinion, as I've said,
[10:57] that crypto is going through its
[10:59] post.com bubble year, the three-year
[11:01] performance through the end of June
[11:03] 2025, this is from Cambridge Associates
[11:05] for Venture was zero.
[11:09] Here's NASDAQ.
[11:11] So, again, the three-year returns, and
[11:14] again, the these are annualized returns.
[11:16] So you're dealing with something that
[11:18] right now crypto or VC has been in
[11:21] trouble and that's the overhang for me
[11:23] for a lot of what's going in the crypto
[11:25] space. You also have to remember in
[11:26] April 2023 or in March of 2023 we had
[11:29] Silicon Valley Bank go under. That was a
[11:32] similar situation. Massive dollars went
[11:34] into the startups. Massive dollars went
[11:36] into Silicon Valley Bank. It was
[11:38] supposed to take down the VC world. It
[11:40] didn't because they got bailed out. This
[11:43] is just the overhang because what
[11:44] happened in
[11:46] November of 2022 was Chat GPT came. So
[11:50] now you got the other disruption. So you
[11:52] not only had the bubble that was built,
[11:54] you not only had the rates that went
[11:56] higher in the VC world go through the
[11:57] Silicon Valley bank, you had Chat GPT
[11:59] and then you had all these AI startups
[12:01] that were going from zero to a billion
[12:03] dollars uh in ARR so quickly that it
[12:06] started sucking up all the money. So it
[12:08] should not be a a question as to how
[12:10] Bitcoin is directly related. The
[12:12] question is now I've been looking for a
[12:13] catalyst. A lot of time what catalysts
[12:15] come from are just falling and getting
[12:18] everyone out. Uh I do believe that
[12:20] Bitcoin is the most liquid hedge for
[12:22] those people that are in private
[12:24] investments of which the VC world has a
[12:27] lot of issues because private equity is
[12:29] under pressure, private credits under
[12:30] pressure and you've seen last year with
[12:32] the endowments the liquidity side. So I
[12:34] think this liquidity private market
[12:36] issue, commercial real estate,
[12:38] residential real estate, all of it is
[12:40] under pressure. Liquidity matters and
[12:42] what Bitcoin is is the most liquid hedge
[12:45] for private SAS if you don't want to do
[12:48] the public companies. But this is the
[12:49] relationship. I think you've lost kind
[12:52] of hope on the SAS trade. It took
[12:54] everything down. This was my belief is
[12:56] that the only way Bitcoin can get to a
[12:58] million is that it needs to be the only
[13:01] growth engine that is not disrupted from
[13:03] AI. It needs to watch everyone else get
[13:05] disrupted. That is what Michael Sailor
[13:07] talked about. That is why he started the
[13:09] Bitcoin side. wasn't because of the
[13:11] government. It wasn't some magic bean
[13:13] thing. He was getting disrupted from
[13:15] Microsoft and he talked about the fact
[13:17] that he was losing from the government
[13:18] and he was losing from Microsoft. Claude
[13:21] Code is the inflection point. If you
[13:23] haven't read the semi- analysis piece, I
[13:25] highly recommend reading it and just
[13:26] going through it. I think they uh they
[13:29] maybe get a little too excited about
[13:31] certain things, but this stuff I agree
[13:34] with. Agents will be the primary method
[13:35] of how organic intelligence interacts
[13:37] with the artificial intelligence. But
[13:38] clog code is also the demonstration of
[13:40] the reverse how showing agents interact
[13:42] with humans. Again, as agents go up,
[13:44] you're going to see massive growth in
[13:46] stable coin usage, which we saw in
[13:48] January, which will be the last slide I
[13:49] show. It means the utility function of
[13:52] crypto is happening. The financial rails
[13:54] are going. We're still waiting for the
[13:55] clarity act. Um it's about 5050 to get
[13:58] through before the end of the year. We
[14:00] already have the Genius Act. I don't
[14:01] think you can stop what's happening. But
[14:03] again, I think the ma the massive bull
[14:05] phase for Bitcoin has nothing to do at
[14:07] the end of the day solely with the
[14:09] financial guardrails. It has to do with
[14:11] the disruption of all businesses built
[14:13] on code and eventually all businesses in
[14:15] general.
[14:17] Here's a chart showing the GitHub
[14:20] commits over time. This is just their
[14:22] illustration of how Claude Code changed
[14:25] everything in terms of people like me
[14:27] going to GitHub. I have personally gone
[14:29] to GitHub about two times before this
[14:33] date trying to do something. I have now
[14:36] built multiple apps multiple times going
[14:39] to GitHub directly from Claude.
[14:42] So this is no no shock that this should
[14:44] be going on. The price of intelligence
[14:46] is collapsing. Asure just signed a deal
[14:48] to train 30,000 professionals on Claude
[14:52] the competitive landscape. I talk about
[14:54] this all the time. I don't know how
[14:56] people can look at Microsoft and not be
[14:58] worried that I'm someone that is AI
[15:00] native. I use five different models all
[15:04] day long. I have I do them on my
[15:06] ThinkPad. I have Microsoft off Office on
[15:09] it. I never repeat never use C-Pilot.
[15:13] Never. So, I don't know how people
[15:15] aren't worried about this. Um why does a
[15:18] company need to standardize Salesforce
[15:20] if an agent is just going to query data
[15:21] on leads on your behalf? Salesforce is a
[15:23] form and workflow rapper. The fact that
[15:25] I've had so many arguments still to this
[15:27] day with people on CRM and Salesforce
[15:31] when I ran businesses for 30 years and
[15:33] have had them come in and pitch for
[15:35] about the last 10 and always ask
[15:37] questions on why I can't customize
[15:39] certain things and how much of a
[15:40] workload it is and now realizing what's
[15:43] coming and I will show you some of that
[15:44] because it's becoming more commonplace
[15:46] and I'm going to do a paper this week on
[15:48] it for 22V.
[15:50] Goldman Sachs is tapping anthropics AI
[15:52] model automate accounting compliance
[15:53] rules embedded enthropic engineers. You
[15:56] don't need all the details on this. You
[15:58] just need to know this is happening.
[16:00] Introducing open a front open eye
[16:02] frontier. There's a lot of things that
[16:03] happened this week that you probably
[16:04] weren't even aware of. Um pretty
[16:06] straightforward in terms of how
[16:08] dangerous is is to your systems of
[16:11] record. I keep hearing well the systems
[16:13] of record. Look how far down the line it
[16:15] is. Where do you think the value is on
[16:17] this? What are these guys trying to
[16:18] disrupt? There's agents along here.
[16:20] There's the enterprise. There's all this
[16:22] stuff. I completely agree with this
[16:24] whole thing. What does this visual say
[16:25] about traditional enterprise SAS
[16:26] companies? All you have to do is go take
[16:28] it in, copy it, and let an unbiased
[16:31] group that just sits in any of the LLMs
[16:34] tell you. This visual signals a
[16:36] strategic ship where OpenAI positions
[16:38] itself as the orchestration layer, the
[16:40] brain, sitting directly above
[16:42] traditional SAS companies, which are
[16:44] relegated to being the data substrate,
[16:46] the records. So, they don't go out of
[16:48] business. But the question is how do
[16:50] they grow and are they just becoming
[16:52] some elaborate database that is
[16:54] protected
[16:56] for traditional SAS center companies
[16:58] that function as system of records. Uh
[17:00] the architect carries several critical
[17:02] implications. So again they're not
[17:05] taking these things to zero. They've
[17:07] just taken the price to free cash flow
[17:08] down to levels that's more
[17:09] representative. We've seen this act
[17:11] before with retail mall stocks when
[17:13] Amazon was disrupting. We saw it with
[17:15] energy companies as well as fracking
[17:16] went through on the all-in podcast. I
[17:18] mentioned it. They have a thing on is
[17:20] SAS dead.
[17:23] I just want to go through one thing on
[17:24] it. Uh I took the transcript from it and
[17:28] I said run the hedge fund analyst skill.
[17:30] I've shown you guys what this is. This
[17:31] is some of the things that uh I I do for
[17:34] 22V for people in terms of of of working
[17:36] on this stuff so they can do this.
[17:39] specifically on the part of this
[17:41] transcript where Jason Calcanis talks
[17:43] about creating Ultron. He's describing
[17:45] how their workload will be done with
[17:47] data gathering and skills for each
[17:48] employee. So I run the skill just on
[17:51] that part. It gives me this eightpage PM
[17:55] research memo. The enterprise agentic
[17:57] stack project ultron and the
[17:58] reorganization of knowledge work. I'm
[18:00] telling you guys, if you don't use this
[18:02] stuff, you're missing out. Here's the
[18:05] visual. I had Gemini build for me. Show
[18:07] me the visual of what he's talking
[18:10] about. Okay, Ultron's in the middle and
[18:12] then you've got your APIs that you're
[18:14] connected to in terms of reading. So,
[18:16] you're getting all your information that
[18:17] way. Even talked about if they were to
[18:20] say, "We're not going to let you use our
[18:21] APIs," then I would get rid of all these
[18:23] things. Now, again, Jason is startup, so
[18:27] I want you to separate startup from
[18:28] enterprise. The startup side, I don't
[18:31] see where those companies that are AI
[18:33] will ever use Salesforce. So they're
[18:35] losing the growth of small businesses
[18:37] becoming big businesses. Just my guess.
[18:40] Um I'm sure $300 billion companies. Yes.
[18:43] But the question is how many of those
[18:44] will ever exist? Are there going to be
[18:46] moes around them? I don't believe so. So
[18:48] there's another issue that comes. If you
[18:49] never get to the point where you're a
[18:51] five-year business and your business
[18:52] gets disrupted after one year, which is
[18:54] what cursor is starting to see in their
[18:56] side,
[18:57] you're separating enterprise from AI
[18:59] native startups. I don't think the
[19:00] startups will ever use it. And for the
[19:02] enterprise side, I think we're going to
[19:03] get into what I think is happening. Is
[19:05] this what Palanteer does? So, for those
[19:08] of you who've doubted Palunteer, it's a
[19:10] defense company. In their most recent
[19:12] earnings reports, there was a shift that
[19:14] happened. So, I asked this, I had it run
[19:17] on this
[19:18] and basically it said yes, this is
[19:20] effectively where it is except one
[19:22] little point. Palanteer's on the top
[19:25] with Ultron. So, again, you can go read
[19:28] this. I'm going to go do a report on
[19:29] this. Palanteer reported record revenue
[19:32] this week, 1.4 billion, up 70%
[19:35] year-over-year. But here's the critical
[19:36] important. US commercial revenue, here's
[19:38] the government, 570, up 66%. Commercial
[19:42] revenue, exactly what we're talking
[19:45] about in terms of the agentic side
[19:47] picking up and companies now going, "Oh,
[19:50] I'd rather build the Ultron model."
[19:54] 130% year-over-year, 28% quarter over
[19:57] quarter. I'm I'm just gonna tell you
[20:01] guys if you're not focused on this from
[20:03] that basis, Palunteer got knocked down
[20:04] significantly this week. Um I think it
[20:07] was $70 off the peak, so 35% fall. It
[20:10] actually got to a level where I bought
[20:12] it. Um and again, I started doing
[20:14] research on it as we went through over
[20:17] the last two months for SAS, figuring
[20:19] that eventually the chart would fall
[20:21] down. I'd shown the chart. I thought it
[20:23] was going to go uh someone reached out
[20:25] next thinking that last week I was
[20:27] negative on Palunteer wasn't negative on
[20:29] it but when you have a panic going on in
[20:31] software and the charts are horrible you
[20:32] want to see it get to a level where it
[20:34] makes sense and down around 130 to 125
[20:37] it started to make sense at 100 in my
[20:39] opinion they're going to make more than
[20:40] a dollar this year becomes a a gimme
[20:43] when you've got earnings growing at over
[20:45] uh 130% in terms of of of uh revenue for
[20:50] their top line relative to the
[20:52] commercial side. So I took this took
[20:55] this and said is this similar? Yes, this
[20:57] open AI frontier diagram is remarkably
[21:01] similar to the ultron palunteer model.
[21:03] So you're getting the way it goes. All
[21:05] three models represent a shift from
[21:07] software as a service to agents as a
[21:09] service instead of human manually moving
[21:11] data between siloed apps. The frontier
[21:14] the reason this is important is
[21:16] Palanteer is an AI company. Salesforce
[21:19] is not an AI company. Can they become an
[21:21] AI company? Sure, maybe that'll work for
[21:24] them. I think you're betting on a pivot
[21:26] and I think the uncertainty on it means
[21:28] you need to have a lower multiple. So
[21:30] all that's happened in SAS, as I showed,
[21:33] their stocks had not fallen. Their
[21:35] multiples had compressed somewhat, but
[21:37] as Brad Gersonner went through in the
[21:39] all-in podcast and he highlighted, it's
[21:41] still expensive. if the uncertainty in
[21:44] five years and he I think he said it
[21:46] well if price to cash flow is still at
[21:48] 15 15 years from now are they going to
[21:50] be able to still be in business I don't
[21:52] know um you just don't know IGV relative
[21:55] to the NDX 30-day rate of change this is
[21:57] a post I put I have no intention of
[21:59] picking a bottom in software so don't
[22:00] read this the wrong way this was on
[22:02] Wednesday but I do think when fears of
[22:04] systemic financial risk from a group ex
[22:07] Microsoft that honestly is not that big
[22:10] so when I said that you're dealing with
[22:11] about 5 point something trillion dollars
[22:13] in the S&P 1500 uh uh software. You take
[22:17] out Microsoft, you're dealing with 2
[22:19] point something. If you take out the
[22:20] Cyber Group and the Palanteer, which I
[22:22] think will be fine during this, uh
[22:23] you're left with about $2 trillion. For
[22:25] people to think that this is a contagion
[22:28] is insane, especially when at $2
[22:30] trillion, you're finally getting that
[22:33] number down to around energy and
[22:34] materials combined. Yes, that's right.
[22:38] Energy and materials are about 5% of the
[22:40] S&P. Let's assume $60 trillion 55
[22:43] trillion at least ex Microsoft you're
[22:46] dealing with a group that's about the
[22:48] same size as energy material. So if
[22:50] that's the rotation that's the risk
[22:52] you're playing. You're getting in front
[22:54] of that and I think that's a mistake.
[22:55] Software has a shock software liquidity.
[22:57] It started to get into private credit. I
[22:59] knew when I heard this I was like well
[23:00] this is ridiculous. Um once the hottest
[23:02] bet on Wall Street private credit is
[23:04] starting to crack.
[23:06] Private credit stocks crash. I mean
[23:08] guys, yes, it has an impact, but here's
[23:12] the number that Bloomberg came out with.
[23:14] When I see software is dead all over X,
[23:16] I mean, there software is dead
[23:17] everywhere after four months of selling.
[23:19] It's just not news anymore. It wasn't
[23:21] news really in December that the mult I
[23:23] got called out on the rerating, but now
[23:26] we've had a sharp fall like it's going
[23:27] to go out of business today. I do think
[23:30] people are underestimating how fast this
[23:32] will continue. But I think the stocks
[23:33] have done enough damage in the near term
[23:35] to where you could see short covering
[23:38] and you'll still see mutual funds rotate
[23:40] and people that are a little bit slower
[23:41] moving uh moving out as people sell US
[23:44] and they buy international that is very
[23:46] negative on software as well because
[23:48] most of the waitings are software. This
[23:50] rotation, this unwind of the passive
[23:53] investment or the MSCI world where 72%
[23:56] is the US is going to be continuous for
[23:58] a decade, guys, as the money rotates
[24:00] into materials and into places like
[24:03] Brazil, as I'll talk about, and into
[24:04] manufacturing in Germany. Distress
[24:06] software loans swell by 18 billion.
[24:08] Whoa. 18 billion. Google backs up 185
[24:12] billion in spending with receipts. I
[24:14] mean they raised their capex
[24:17] by more than 70 billion. So take it as
[24:21] four times this number. But this is a
[24:23] news headline and this one people view
[24:25] as a negative as well. So this is a
[24:27] negative and this is a negative. This is
[24:29] when I think
[24:32] people may have gotten a little too
[24:33] negative. Blue was collapsing this week.
[24:36] The whole private equity world was. So
[24:39] he gets interviewed on or he gets
[24:40] interviewed on CNBC and basically jokes
[24:42] about how it is good. Um but he said
[24:45] this on the earnings call. We don't have
[24:47] red flags in point of fact. We don't
[24:49] have yellow flags even. We actually have
[24:51] largely green flags. The tech portfolio
[24:54] continues to be the most pristine
[24:55] amongst all of our portfolios amongst
[24:58] all of our subsectors.
[25:00] So again I think people I think people
[25:02] treated this as a panic because they're
[25:04] creating narrative after price which is
[25:06] what normally goes on. The narrative is
[25:09] now software is dead. If you've written
[25:12] a piece this week titled software is
[25:13] dead. I literally come on you're you're
[25:17] doing this after some of the charts I've
[25:19] already shown. It's just not worth it.
[25:21] It is going to be up under competition.
[25:22] But this is not new news at this point.
[25:24] This started when Andre Carpathy created
[25:27] the word vibe coding last year. That's
[25:29] when we started to see this multiple
[25:31] compression start. Claude code kind of
[25:34] made it more likely that this is going
[25:36] to happen. And then when you get into
[25:37] the beginning of the year and people
[25:38] haven't made the shift because they
[25:40] thought they could buy software, but
[25:41] then every single report is saying that
[25:43] this is an issue. I can see where mutual
[25:45] funds are like, whoa, whoa, whoa, whoa.
[25:47] Okay, we're starting off the year. Let's
[25:49] rebalance and go through. And everyone
[25:50] just chose the same year. And then you
[25:52] get into hedge fun unwinds and you're
[25:53] into this. As a reminder, the Palunteer
[25:56] earnings you should go through. Uh this
[25:59] is what I'm going to be writing more
[26:00] about. I'm going to do a deep uh report
[26:02] on Palunteer less specific about the
[26:06] company more about the technology and
[26:08] the way people need to think about this
[26:09] in my opinion that they haven't because
[26:11] this is going to feed into the edge and
[26:13] the on-remise enterprise this is a huge
[26:15] huge huge dollar making scenario this is
[26:18] part of the $90 trillion that Jensen
[26:21] Yuang is talking about the $6 trillion
[26:23] that is going to be on the data capex
[26:25] side that is old news and I think that's
[26:27] going to cost people money this year you
[26:29] want to remain in the semis. You want to
[26:31] remain in the places that are going to
[26:33] be part of the 90 trillion dollars over
[26:35] the next decade. That will also involve
[26:38] copper and silver and all of the
[26:40] components of the commodities that have
[26:42] gone higher. But software needs to be
[26:45] rerated because it is a massive part.
[26:48] When you combine software with the Mag
[26:49] 7, guys, you're dealing with a situation
[26:52] of waitings that is just enormous
[26:54] relative to the other side. This is not
[26:56] something that can be done overnight.
[26:58] And if we continue to see it happen at
[27:00] this speed, and remember you're talking
[27:01] to someone who said he thought the
[27:02] Russell uh 2000 would be up over 50 to
[27:05] 60% this year while the MAG 7 is flat.
[27:08] If that occurs, we could be taking down
[27:11] quant strategies. We could be taking
[27:12] down hedge funds because I don't know
[27:13] how you can rotate that fast if it
[27:15] happens this year over five years. I get
[27:18] it. But remember, I started this off
[27:20] saying I like me and the John Candy
[27:22] reference. I was wrong on Bitcoin
[27:24] because I got caught in the software
[27:27] trade. If you're caught in the software
[27:29] trade, I don't know how quickly this is
[27:31] going to go, but I think it can go
[27:34] faster than we all realize because all
[27:35] it's going to take is some earnings
[27:36] reports from Salesforce.com that suggest
[27:38] that topline revenue is now slowing and
[27:41] I think you're going to have an issue.
[27:42] We're not there and everything has been
[27:44] fine so far, but I'm worried about this
[27:46] year in terms of how quickly this goes.
[27:48] So again, Palunteer, you should spend
[27:50] the time on it. um because they're part
[27:52] of that Ultron thing that I said. I'm
[27:54] not going to spend more time on it, but
[27:55] I will say this. This was from their
[27:57] earnings call and this was from a
[27:58] testimonial from a client. Every other
[28:00] software must and so this is a client
[28:02] saying this. Every other software must
[28:04] justify its existence. And so far, they
[28:06] haven't been able to haven't been able
[28:08] to 97% of our employees use Foundry
[28:10] every day. Foundry is our operating
[28:12] system. Not only we're getting rid of
[28:14] our thirdparty software, we've replaced
[28:17] their functionality and then beaten them
[28:18] to the new features all within one year
[28:20] because of the ontology. Ontology is
[28:22] basically the brain. So again, remember
[28:26] this is happening faster and everything
[28:28] with AI goes at onetenth the time. So it
[28:32] is possible that what we saw with Amazon
[28:34] will happen this year alone. So be very
[28:36] careful and that's why I said I'm not
[28:38] trying to pick a bottom in software, but
[28:40] I do think we've done enough. But on
[28:41] Friday, I thought it was interesting
[28:43] that Palunteer was up, Bitcoin was up,
[28:47] Salesforce and Adobe were basically
[28:48] flat. Continue to watch that dynamic
[28:50] because I think you're going to start to
[28:52] see the the things that were thrown out
[28:54] with the bathwater separate. If software
[28:57] continues to go lower, then I think
[28:58] Bitcoin and Palanteer will have another
[28:59] leg lower and we'll get to lower levels
[29:01] until we've been through the force
[29:03] liquidation. If it's a force
[29:04] liquidation, it's going to be a bigger
[29:06] issue. Software PE,
[29:08] we're back down, believe it or not.
[29:13] to lower than when the iPhone came out.
[29:17] That's how fast this has been.
[29:20] Big tech combined to spend 650 billion
[29:23] this year. I want to remind all of the
[29:25] people that wrote the AI bubble piece
[29:27] and the fact that they will not spend
[29:28] this. There will eventually be a a
[29:30] spending change.
[29:32] 650 billion is what came out. And for
[29:36] you saying they're never going to get
[29:37] their revenues, revenue growth.
[29:39] Microsoft 17% year-over-year. Revenue
[29:41] growth 18% year-over-year for Alphabet.
[29:43] Meta 24%. Amazon 14. This is the revenue
[29:47] side. And again, this is with capacity
[29:50] constraints and RPOs for Microsoft,
[29:52] which is their issue. Um, you've got
[29:55] capacity constraints for Alphabet.
[29:57] You've got capacity constraints for
[29:58] Amazon.
[30:00] I personally think these companies are
[30:02] going to have a really hard time getting
[30:04] the numbers because their revenues are
[30:06] growing. But the problem is the data
[30:08] center buildout to me is going to be an
[30:10] issue. It's going to be bottlenecked at
[30:12] some point this year. The rise of AI
[30:13] agents to forever change the way work is
[30:15] done will lead to a break in
[30:16] correlations with investments and assets
[30:18] built on code. So far SAS has hit
[30:19] everything including crypto and agentic
[30:21] winners like Palunteer. They need AI
[30:23] agents to be winners while SAS suffers.
[30:25] Today Palanteer and Bitcoin are up. This
[30:26] is what I was talking about with looking
[30:28] for changes. Recap for the week. All
[30:30] that going on. The S&P was down 10 basis
[30:34] points. fourth week in a row of less
[30:37] than 1% move. The VIX barely bug barely
[30:41] budged. I don't have it there, but
[30:42] obviously we've had a higher V in
[30:44] singleame equities. I think that
[30:46] divergence is going to be there all
[30:47] year. I expect that the VIX EQ, which is
[30:50] up around 40, uh will remain at high
[30:53] levels while the VIX will remain around
[30:55] 20. And I think you're going to see
[30:57] spikes in the VIX EQ uh all year uh for
[31:00] a variety of reasons. Q's down 2%. Mag 7
[31:04] down 4.7% biggest move since liberation
[31:07] daytime. Uh Russell up 2%. Small caps,
[31:12] baby.
[31:14] Oo, this is kind of important. Seven
[31:17] only 7% of small caps reported in
[31:19] through uh through Tuesday, but 79% had
[31:22] beaten estimates on track for the best
[31:23] beat rate performance ever. We'll see
[31:26] what happens, but I think small caps are
[31:28] there. You can't see it on this chart
[31:29] because my head is in the way. I'm not
[31:30] going to move it, but you guys know the
[31:32] drill. This is the equal weight ETF.
[31:35] Made new all-time highs this week and it
[31:36] made it on the day that the world was
[31:39] panicked.
[31:40] I put I I I sent this out um to the
[31:45] Salesforce at 22V. If you guys want to
[31:48] get these warnings, just reach out to
[31:49] the 22V salesforce if you didn't get
[31:52] this. Um Colin and Max can help you out.
[31:55] But I sent this to them uh on uh
[31:57] Wednesday morning. And the as a
[32:00] reminder, this is my turbulence model.
[32:01] It has a 100 different assets in there
[32:03] and it's meant to warn me when we start
[32:06] seeing shaking but the headline index is
[32:08] not moving. So before the S&P dropped
[32:10] 100 and some odd points uh I got a
[32:14] signal and this was the first signal I'd
[32:15] gotten this. So again four days elevated
[32:18] the coariance matrix was shaking. It's
[32:21] like a tremor report to it's like my HRV
[32:24] stuff that I do in Substack. HRV is a
[32:26] warning for cancer. It's a warning for
[32:28] heart things. Uh if you guys aren't
[32:30] reading my substack and you want to
[32:31] remain alive forever uh or at least have
[32:33] a higher probability, HRV is a metric
[32:36] that will tell you ahead of time if
[32:37] there's something going wrong. It will
[32:39] tell you the second that your immune
[32:41] system is going. That's what this thing
[32:42] is. That's why it's called the immune
[32:43] system. It's meant to highlight things
[32:45] when a problem isn't doing and then it
[32:47] gives recommendations. Um it was showing
[32:50] up in the AI side but also in the other
[32:52] assets. The alarming thing you can see
[32:54] that the S&P has been kind of going
[32:55] sideways now for three months. So this
[32:58] warning to me and even though we've gone
[33:00] back up, we haven't taken out the highs
[33:01] yet. I don't think we're out of this
[33:03] yet. That's why I'd be very careful with
[33:05] the software. We have a lot of
[33:06] disruption that's happening. I am and I
[33:09] have stuck my nose into Bitcoin again.
[33:11] So I stopped from 92 uh after I said
[33:14] publicly that if we close three days
[33:16] above 92, I thought that would be the
[33:17] bottom. Uh the bottom tended to be 40%
[33:20] lower. And again, John Rog was right. I
[33:22] was wrong. Um here's what happened in
[33:25] the sectors for the week. This is just
[33:27] for the week. So the S&P 1500 down 10
[33:29] basis points. Staples up big.
[33:31] Discretionary which is mainly Amazon and
[33:34] Tesla. Uh and then all of the tech
[33:36] basically this is the mag 7 guys. And
[33:39] this the industrials energy materials
[33:42] growth good tech bad.
[33:45] Building 90 trillion good
[33:49] spending six billion six trillion no
[33:51] revenues bad.
[33:54] Deal with it. S&P this is the waitings.
[33:57] So again I'm going to highlight this
[33:59] energy 3.3% materials 2%. You come back
[34:02] in five years these numbers are going to
[34:04] be significantly higher. These are going
[34:06] to be significantly lower. If your
[34:08] portfolios, you know, if you run a fund
[34:10] or run a a firm and 70% of your money is
[34:14] in growth,
[34:17] I think the growth better shift to these
[34:18] companies and you better find it. Call
[34:20] me up. I'm happy to help on some of
[34:22] these things. But uh growth verse value
[34:25] uh this is the two-week chart. Uh I'm
[34:29] sorry this is the oneweek chart. Uh down
[34:31] 7.89%.
[34:33] The reason I want to show this so it got
[34:36] whacked here during the belief that we
[34:39] were going into a depression and then it
[34:41] got whacked here when the Fed started to
[34:43] aggressively raise rates and we had to
[34:45] repric long duration assets. Well, this
[34:48] is with rates not going higher. This is
[34:50] from a disruption from AI and it comes
[34:53] at a time where growth had outperformed
[34:56] value for a long long time. Um this is
[34:58] when growth really started to outperform
[35:00] value in a big way because this is when
[35:04] we had the iPhone. I've done
[35:06] presentations where I went Singularity
[35:07] University right around here. I just
[35:09] want to remind you at this point when I
[35:11] wrote a paper called adapt or die
[35:13] related to what was happening with
[35:15] software and that GDP would no longer be
[35:18] a relevant statistic. we'd never have
[35:20] another cycle uh employment uh
[35:22] recession. The size of the mag 7 was 1.5
[35:26] trillion. Before this whole thing
[35:29] started last week, it was up at 22
[35:31] trillion. That's how you made your
[35:32] alpha. That's how everyone in the world
[35:34] made their alpha. There's no other way
[35:35] to make alpha. If you look at VCs since
[35:38] 2013 over this period, they
[35:39] underperformed the NDX. This has been a
[35:41] public equity story. And now I'm telling
[35:43] you my viewpoint is that they are done
[35:48] as the multiples they have and the
[35:49] multiples have to come down. Great
[35:51] companies, the models are fantastic. I
[35:53] use them every day. You've got some
[35:55] problems with Microsoft. You've got some
[35:57] other problems with things, but I just
[35:59] don't think they're going to outperform.
[36:00] So I know Dan Ives and everyone has all
[36:02] these people uh continuing in these
[36:04] things. I have a very different
[36:06] viewpoint on this and I think it's going
[36:07] to be an issue. Uh this is the growth
[36:10] verse value on a 7-day rate of change.
[36:12] the worst in 25 years. This is the
[36:14] Morgan Stanley one which has a sector
[36:17] tilt towards it but still you're to get
[36:19] it in perspective year to date. This is
[36:21] where we are. The S&P 1500 is up 2% even
[36:24] though the S&P is unchanged for the
[36:25] year. This is because the Russell is up
[36:28] software down 18.5%
[36:33] energy up 19. This is why I give you the
[36:36] proxy trade of Chevron versus
[36:39] Salesforce.com that has worked out
[36:41] exceptionally well now up over 60% year
[36:44] to date.
[36:46] ISM look what we got guys. So my PMI
[36:50] calls that I was saying beginning in
[36:52] August when I wrote a paper saying PMIs
[36:54] will break above 50. We rocketed above
[36:57] 50. Um people still don't seem to
[37:00] believe it. We are going higher and as
[37:02] I've said we are headed back up towards
[37:03] 60 as part of this over the course of
[37:05] the next three to five years. Wouldn't
[37:07] surprise me if we get up close to it
[37:08] before the end of this year especially
[37:10] now that we've kind of got the
[37:11] launching. Uh and whether it was because
[37:13] the one big beautiful bill depreciation
[37:16] bonus side has kicked in which is why
[37:17] these guys have raised their capex
[37:19] numbers so big. Uh I don't really know
[37:21] but let's just see what happens. The the
[37:23] main point is this is a big important
[37:26] thing from a quant basis because
[37:28] historically it used leads to big
[37:30] things. The monthly number was up 4.7
[37:34] aside from
[37:36] this here coming out of COVID.
[37:40] There is nothing. You got to go back to
[37:42] 1995
[37:45] 30 plus years to see a jump that big.
[37:47] Wasn't just the headline. New orders
[37:50] also highest since 2012 coming out of
[37:53] the Mario Draghy line in the sand.
[37:58] Here's what happens historically to
[38:00] momentum when
[38:03] this transfer. Now, I'm not expecting
[38:05] that kind of a momentum unwind uh
[38:08] because we didn't have a runup. We've
[38:09] actually been already declining while
[38:11] the PMIs were there. And that's because
[38:13] a lot of this had to do with in my
[38:15] opinion now semis are up big. You've got
[38:17] the materials up big because they went
[38:19] up. so fast that they're already in the
[38:21] momentum side because of how quick. I
[38:24] think factors are going to be very
[38:25] difficult to use, but I think momentum
[38:28] in this is going to have a hard time as
[38:30] this goes with the software names in the
[38:32] Mag 7. So, I think you have to be wary
[38:35] of it. I do think the things that had
[38:37] been working are going to be more
[38:38] challenging, but we're going to have a
[38:39] broadening out. I think size will
[38:41] continue to go that direction. I think
[38:43] growth versus value is still going to go
[38:44] that direction. Uh remember the trigger
[38:47] point I wrote the paper on the PMIs back
[38:49] here. I've been showing this chart. This
[38:51] is DRAM prices. This is the semi. This
[38:53] is really what's been driving this. But
[38:55] this is also what started the IGV fall.
[38:57] Meaning it really started to unwind once
[39:00] we started to get the reality that DRAM
[39:02] prices were going up and this was a
[39:03] hardware shift. This is that in in this
[39:07] point in here IGV versus NDX. So, I
[39:09] wanted to just show that that this
[39:11] collapse here is directly related to how
[39:14] fast we've seen these charts go. There's
[39:17] semis versus IGV. So, everyone that kept
[39:20] trying to pick the top in this, well,
[39:23] you got that one size wrong. Um, here is
[39:27] my favorite that I've pitched, which is
[39:29] the equal weight semis. This is more the
[39:31] analog names and everything related to
[39:33] IGV that has gone into hyperspace now.
[39:35] This has pterodine. This has uh lattis
[39:39] semiconductor. A whole bunch of names
[39:40] which have have gone up. I've
[39:43] highlighted Corning as an optical fiber
[39:46] trade and I showed last week when it
[39:48] moved back down to towards 100 uh that
[39:51] this is you know the same thing to me as
[39:53] Samsung and you should expect this thing
[39:55] to go up violently. Plus, when we do the
[39:57] upgrade cycle for autos, for trucks, for
[40:00] phones, for computers, the glass screens
[40:04] that are necessary, we are going to have
[40:05] a fullborn AI demand cycle with a lot.
[40:09] And then humanoids recording ain't going
[40:11] to stop. That is the optical fiber side
[40:13] from the metatransaction. You're still
[40:15] going to have the glass side. There's
[40:16] Pterodine. I've highlighted that. These
[40:18] are all names that I've highlighted. The
[40:19] only reason I'm highlighting them again
[40:21] is because when I did my work, I was
[40:23] shocked at where they were sitting,
[40:24] knowing what was coming. The reason
[40:26] hopefully you guys watch this is that it
[40:28] is very difficult for a semi-analyst to
[40:31] do this unless they spend time on the
[40:34] demand side of AI. There is no way to do
[40:36] this if everyone thinks there's a bubble
[40:38] in AI. And that's why everyone got
[40:39] memory wrong. That's why they got all
[40:40] this stuff wrong because they were
[40:42] doubting the demand side. They were
[40:43] doubting the capex side. They thought
[40:44] there was a bubble. They've seen this
[40:46] before. Everyone got that wrong. They
[40:48] are not ready for this one now.
[40:49] Salesforce.com verse Chevron. Very early
[40:52] stages. We are still in the very early
[40:55] innings of this. Um the twoe rate of
[40:58] change of the MS value factor which I'm
[41:01] not showing here. You can go to the
[41:02] expost is the highest since the.com
[41:05] bubble. So value ripped higher over the
[41:07] last two weeks. The value bounce from
[41:10] the dot bubble lasted for six years as
[41:14] commodities in the China trade
[41:15] dominated. The sector shifts so far this
[41:17] year are the same as the seven years
[41:18] back. So what I wanted to highlight was
[41:21] here's what happened over the seven
[41:23] years out of the dotcom bubble and into
[41:25] the China
[41:27] infrastructure story. You had tech
[41:30] communication services down violently.
[41:33] The S&P was unchanged during the seven
[41:35] years. The energy sector was up
[41:37] violently. Materials go through the
[41:40] list. Here's what we have going on right
[41:42] now so far this year. It's the same
[41:45] trade. So again, I bring this up because
[41:48] the AI infrastructure trade is big tech
[41:52] spend. We're going to spend 650 billion.
[41:54] I don't think it's good for the Mag 7. I
[41:56] don't think it's good for them, but I do
[41:58] think it's good for the semis and for
[41:59] the people and for the cooling side and
[42:01] for all the stuff we've talked about on
[42:02] power. Nvidia did another presentation
[42:05] this week with the salt. Now they agreed
[42:08] that to build the physical AI to power
[42:10] 90 trillion industry shift in the phase
[42:12] of AI based re-industrialization the
[42:14] biggest opportunity lies where the
[42:16] information and physical worlds meet
[42:17] totaling 90 trillion from that was from
[42:21] Jensen we are living in an era of
[42:22] unprecedented acceleration and this
[42:24] massive wave led by AI is completely
[42:26] changing manufacturing processes from
[42:28] the head of the salt. So again, I
[42:31] wouldn't care about these, but you're
[42:32] seeing the charts. The charts are
[42:34] telling you something historic is
[42:36] happening. And people look at the charts
[42:38] and they go, "This can't go." But in the
[42:41] case of Micron, again, the PE finishing
[42:43] this week is under 10 off this year's
[42:45] earnings. This year's like, forget it. I
[42:49] We're going through this pace that we've
[42:51] never seen earnings grow this fast
[42:53] because we've never seen these kinds of
[42:55] dollars, $90 trillion over a decade. Go
[42:58] read how much China was. We're we're
[43:01] dealing with stuff. We don't have enough
[43:02] copper. We don't have enough silver for
[43:04] this. I just wanted to remind you that
[43:07] if you go back and look what is in the
[43:09] world is causing the retail meltdown of
[43:11] 2017. This is the Amazon story. So this
[43:15] SAS thing, there were people that wanted
[43:17] to buy the retailers back then. The
[43:19] exact same thing. They want to buy SAS.
[43:22] People never never able to see the
[43:25] technology and the damage. They think
[43:26] old incumbents can reinvent themselves
[43:29] with the new technology. They all tried
[43:31] to do it. Has Walmart been able to find
[43:33] a way over the 15 years to be able to
[43:36] have an online presence? Yes. Does it
[43:38] compete with Amazon on that? Not even
[43:40] close. But they have a business that has
[43:41] survived in terms of the brickandmortar.
[43:44] This is also from February of 17 from
[43:47] McKenzie on what happened to resources,
[43:50] what happened to energy post the
[43:52] fracking side. Truflation
[43:57] less than 1% this week. Truflation is a
[44:00] real thing. Like I I hate to tell people
[44:02] that want to doubt it. Um it is a real
[44:04] thing. I I put it in the same camp as
[44:06] the ADP. It's not like it's completely
[44:08] useless. It may not move the market.
[44:10] There may not be anything that ever
[44:11] trades off of it, but there's
[44:13] information in there that's real. Um it
[44:15] is correlated historically. So there's a
[44:16] huge divergence going on. And again,
[44:18] until gas at the pump goes higher, I
[44:20] think people should be more worried
[44:21] about disinflation. and eventually
[44:24] deflation than they are inflation. But
[44:25] the majority of people and I was at a
[44:27] conference are worried about inflation.
[44:29] They expect I forget what the I'm going
[44:32] to say close to 90% expected the
[44:33] inflation to be above 3%. Um most
[44:36] inflation numbers right now are below
[44:38] three. So everyone's expecting inflation
[44:40] to go higher. The jobs market, most
[44:42] economists are still saying AI is not
[44:44] the disruptive force. I I literally
[44:46] don't even know what to say to people
[44:47] anymore on this stuff in terms of this.
[44:50] If you want to get more up to speed and
[44:52] listen to two very, very smart guys on
[44:55] this rotation, I gave you last week,
[44:58] episode one, Smarter Markets. This one
[45:00] starts off phenomenal. Um Jeff Curry,
[45:03] uh, you know, people always um give
[45:06] commodity guys crap because these are
[45:08] long cycles, so they're predicting them
[45:10] and sometimes they're, you know, early
[45:12] or sometimes they're they're just wrong
[45:14] because they're in a commodity space. I
[45:16] think the way Jeff started this out, uh,
[45:18] very very good. He's he's a historian
[45:21] and I'm just going to put Jeff argues
[45:22] this cycle is different because asset
[45:24] light hyperscalers like Amazon and
[45:26] Google are now putting steel in the
[45:27] ground. That should put pressure on
[45:29] their equity multiples as they start to
[45:31] look like asset heavy commodity
[45:32] producers producers. He flags a
[45:34] potential rerating window in 26 and 27
[45:37] and notes compute prices falling even as
[45:39] capex surges. Everything I just showed
[45:41] you is exactly that. Now he's talking
[45:44] about what's happening, but I think this
[45:47] is important. You can go read through
[45:48] all the other things. The AI race is
[45:50] about energy and materials. Completely
[45:52] agree. They require dozens of metals,
[45:55] not just coppers. And true hoarding is
[45:57] unobservable. Okay? And I want you to
[45:59] remember that just because the price of
[46:02] silver goes from 12 back down to 70,
[46:06] there's going to be a lot of swing
[46:07] trading in this. For you traders out
[46:08] there who want to trade things, the
[46:11] river is flowing with the metals. Um,
[46:13] it's flowing with the metals. They're
[46:15] going to trade volatile, but if you can
[46:18] catch the swings, there's a lot of money
[46:19] to be made on it. Google earnings came
[46:21] out. Felix
[46:23] uh put this out, and again, I've talked
[46:27] about this as well, but I think the end
[46:28] main thing here is to look that as these
[46:31] companies expand
[46:33] their capex violently, their buybacks
[46:37] have to suffer. So the forecast in terms
[46:40] of buybacks going down, but remember
[46:42] also we're talking about a race of
[46:45] bringing IPOs to the market in the
[46:47] growth category, trillions.
[46:51] Why do you want to be long stuff that
[46:53] has trillions of supply, isn't going to
[46:54] be buying back, and is spending
[46:55] trillions of dollars on commodities
[46:58] where there's going to be bottlenecks,
[46:59] where they already have RPOS, which are
[47:01] basically like liabilities that they may
[47:03] never get because, as Alon Mus Elon Musk
[47:06] will talk about, we're going to have an
[47:08] edge acceleration at some point here,
[47:10] and it may just completely make the
[47:12] training models less important. Trump
[47:15] launches 12 billion mineral stockpile to
[47:17] counter China.
[47:21] Craig Tinddale, Adam to your um to your
[47:25] expose. China may look to stockpile more
[47:27] copper as part of its. So in response,
[47:30] Trump's doing stockpiling. China's doing
[47:32] stockpiling. China is restricting any
[47:35] exports of silver. Uh I mentioned last
[47:38] week that I was aware that governments
[47:39] were looking into stockpiling critical
[47:41] minerals. Less a week later, the US and
[47:42] China announced their intentions to ramp
[47:44] up stockpiling. Do these other countries
[47:46] do it soon? That's the whole thing with
[47:48] commodities is that they're a global
[47:50] thing. Everyone needs them. What the
[47:52] world sees as an irrational Trump and
[47:54] government might be better interpreted
[47:55] as US critical mineral supply chains are
[47:57] in far worse shape than the Western
[47:58] consensus realizes. Investing critical
[48:00] minerals of which silver, copper are
[48:03] included. Yes, this is not just rare
[48:05] earth. Critical minerals. Oh, the US and
[48:08] Brazil did a deal. Remember I showed you
[48:10] guys EWZ last week. Brazil holds the
[48:12] second largest reserves. This week, the
[48:14] US DFC announced 6 565 million finest
[48:17] rare earth extraction. Brazil is a
[48:19] sleeping giant of minerals
[48:22] and I wrote a paper on this this week.
[48:24] For those of you on the um payw wall,
[48:27] assuming it is launch Tuesday, which is
[48:29] again what I'm being told, uh you can
[48:32] get access to this and start to read
[48:34] some of the stories on it and I will be
[48:36] coming out with more specific names. But
[48:38] in the case of Brazil, uh I believe the
[48:40] currency, the bonds and the equities
[48:42] will be in lock step this year and
[48:45] commodities will be in lock step as
[48:46] well. That is a powerful situation for
[48:49] Brazil even though it's an election year
[48:50] and chaos can ensue during election. Uh
[48:54] Art Burman put this thing out on the gas
[48:55] turbine lead times. Uh I would read it
[48:59] again. We are not done with this whole
[49:01] thing yet. So
[49:03] Elon Musk interview cheeky pint
[49:07] with uh one of the Collison brothers. I
[49:10] think it's Patrick uh and Dwaresh.
[49:14] Again, I say it all the time. That's
[49:16] three hours. You got to listen. This is
[49:18] the main points. I'm going to try to
[49:19] give you kind of the the the I could
[49:21] talk about the data centers. All of the
[49:24] hedge fund people that told me Elon Musk
[49:26] cannot build satellites in space should
[49:28] go listen to this. He's now saying 30
[49:30] months uh it'll be happening. Can you
[49:32] imagine building that many data centers?
[49:34] He's talking about the hyperscalers,
[49:35] that many power plants. Those who have
[49:37] lived in software land don't realize
[49:40] they're about to have a hard lesson in
[49:42] hardware.
[49:44] I I you need to listen to this because
[49:47] he's basically saying this is really
[49:49] hard to do and to believe that these are
[49:51] going to happen, that you're going to
[49:53] get them all done, that there won't be
[49:54] issues, that there won't be problems.
[49:58] AI is about to slam into a hardware
[50:00] wall, which is why space becomes the
[50:02] escape valve. Um, hyperscalers are
[50:07] building the brain faster than the body
[50:08] can use it. Training spend is still
[50:10] necessary, but it's no longer the
[50:12] binding constraint on AI value creation.
[50:14] I could not agree more. Um, where these
[50:17] things have gotten to, we are now at
[50:19] arguably 140 IQ and accelerating by the
[50:22] end of this year. What does that mean?
[50:23] And if we grow at the same pace we have
[50:24] over the course of the last three years
[50:26] since chat GPT was launched. You're
[50:28] dealing with above 160. You're getting
[50:29] recursive self-improvement. You're
[50:31] getting all of this stuff already. The
[50:33] compounding of this to use Moors to use
[50:36] Ray Kerszswwell's singularity. It's
[50:38] going to accelerate. So the question is
[50:41] is this money is the models getting
[50:43] bigger going to work? Hyperscalers
[50:45] assume bigger model more value more
[50:46] training more defensibility for remote.
[50:49] What Elon is talking about training is
[50:50] frontloaded. Inference and deployment is
[50:52] where value compounds. That's what we're
[50:54] seeing now. It's about inference. This
[50:57] is why Nvidia had to diversify and they
[50:59] did the Grock deal back at the end of
[51:01] the year. They need to be part of the
[51:03] inference and the deployment into the
[51:05] embodied AI. Go back to what we talked
[51:08] about. 6 trillion verse 90 trillion. 90
[51:11] trillion is related to the inference and
[51:12] the deployment. That's that's where the
[51:14] value is 90 trillion. the six trillion
[51:17] if you're sitting in the the the vertvs
[51:20] in the the GE vernovas and all that okay
[51:23] that's great but they got to focus on
[51:26] the edge now we got to be involved so
[51:28] you have to do your homework on the edge
[51:29] and where the edge is going to work the
[51:31] on premise the Cisco the routers the
[51:33] there's so many other companies Qualcomm
[51:35] that are going to benefit from this as
[51:36] we do the roll out which is starting
[51:38] this year training is centralizing
[51:40] deployment is decentralizing
[51:43] so this gets back into the
[51:44] decentralization theme. This gets into
[51:46] something very very important. Edge a
[51:49] edge AI is a way to route around the
[51:51] power bottleneck.
[51:52] I'm not going to go through reading all
[51:54] this. You have to listen to itself. But
[51:55] edge AI turns cars into rolling data
[51:57] centers, robots into mobile inference
[51:59] nodes, factories, embedded intelligence.
[52:02] This is capital efficiency, not just
[52:04] technology. Basically, building these
[52:06] gigantic data centers, these massive
[52:08] things, and then sending it from the
[52:11] cloud is not going to work for the 90
[52:13] trillion. So they need it for the models
[52:16] and they need it for the race. They need
[52:17] it to cure cancer. They need it for all
[52:19] of these things because we need to get
[52:21] to that point. But the central AI is a
[52:25] capital goods regime. That's why AI
[52:28] stops looking like SAS and starts
[52:30] looking like the industrial infra
[52:32] infrastructure. This is about getting AI
[52:34] at the edge into these things. So he's
[52:37] basically talking about, you know,
[52:38] longduration debt is now becoming
[52:40] riskier as interest rates volatile
[52:42] because you've locked in terms that may
[52:43] not reflect future reality. Longduration
[52:45] capex becomes riskier when technology is
[52:47] evolving exponentially. And this again
[52:49] gets into if you're making long-term
[52:51] capex. You're assuming you're going to
[52:53] get the RPOS and you're also assuming
[52:54] that there won't be disruptions coming
[52:57] along the lines in terms of capex.
[53:00] You have to listen to Elon and
[53:02] understand why he's doing SpaceX and cap
[53:04] and XAI. It's really important for this
[53:06] buildout. So, one more edge AI bear case
[53:09] playing out. I put out a long thread on
[53:11] X last weekend because I had listened to
[53:14] and written I done a video on Gavin
[53:16] Baker worrying about the scariest bear
[53:18] case for AI infrastructure less than two
[53:20] months ago and it's happening out now
[53:21] and that's because I listened to the all
[53:23] podcast. So, this is the thing if you
[53:25] want to go listen to Gavin Baker talking
[53:26] about on invest like the best. He
[53:28] highlighted this. He actually liked the
[53:30] X thread. It was five half a million
[53:32] people read it. Baker's framework edgei
[53:36] scary space case basically talking about
[53:39] when you get to the point that you can
[53:42] have open source
[53:45] on your computer where you can actually
[53:48] get around using the models that you're
[53:50] paying for and actually have them in
[53:52] there and the all-in podcast started to
[53:53] talk about this with Claudebot
[53:57] and again this stuff moves too fast
[53:59] we've talked about co-work
[54:01] cla code we've talked about all of these
[54:03] different things related But Claudebot
[54:05] was the story that eventually turned
[54:06] into Maltbot and eventually Molt book.
[54:09] Um, so basically all you need to do is
[54:13] read this. Jason goes through his
[54:15] experience with Claudebot. It's an AI
[54:17] personal assistant. And then they talk
[54:19] about Kimmy 2.5 and the open- source AI
[54:22] breakthrough. Basically being able to
[54:23] reduce cost by 90%. So he talks about
[54:26] how he used this connected to Claude,
[54:28] but that it was costing him a lot of
[54:30] money to have the AI agent run
[54:32] full-time. So think of it as
[54:33] electricity. So then
[54:36] you get into Kimmy K2.5 which is an
[54:39] open- source model which you can connect
[54:41] and reduce your cost by 90%. It's a
[54:44] Chinese model so you take it for what it
[54:47] is. But they talked about how you're one
[54:49] terms of service update away from
[54:51] everything breaking with closed models.
[54:52] When anthropic briefly restricted
[54:54] clawbot usage it proved exactly this
[54:56] vulnerability accelerated the shift to
[54:58] local open source alternatives. So
[55:00] that's what ends up happening is if if
[55:03] you get to the point that you don't want
[55:06] to use claude and you want to use these
[55:08] open source, the question is are they
[55:09] good enough? And the answer is they're
[55:11] pretty much Kimmy 2.5 is right around
[55:14] claude. So you're getting to the point
[55:16] where the open source is there. So what
[55:19] this means for way where AI stands today
[55:22] phase one the form factor evolution
[55:24] David Sax said AI has mainly come in a
[55:27] form factor of a chatbot. So chatbot, if
[55:29] you guys have only used it like a Google
[55:31] search as a chatbot, you're already
[55:33] missing out. The next phase is when we
[55:35] start getting into open source models
[55:36] and we start getting into AI agents.
[55:38] That's where we are. That's where we're
[55:40] going right now. Um phase three is when
[55:43] you get to the decentralization Mac
[55:45] Studios with M series. I've talked about
[55:47] the fact that I ordered two Mac minis
[55:48] and a Mac Studio because I'm going to
[55:50] run this stuff. I don't want it on this
[55:51] laptop, which is my ThinkPad, sorry, my
[55:53] Mac. I also don't want it on my ThinkPad
[55:55] which I use with 22V and do most of my
[55:58] work but I do want to use and have AI
[56:00] agents work but I want it to be cheap.
[56:02] So I want it to be a Chinese model
[56:03] sitting on an open source uh on a
[56:06] framework that is completely secure so I
[56:08] don't have to deal with it. That's why
[56:09] the Mac minis are selling and that's why
[56:10] the studios are selling. Please talk
[56:12] about the risk to the hyperscalers from
[56:14] this reality of moving off. The
[56:15] inference revenue cliff is really the
[56:18] big thing. The entire hyperscaler narrow
[56:19] has been built on a core assumption. AI
[56:21] inference at scale requires centralized
[56:23] data center infrastructure.
[56:26] They're highlighting you don't need
[56:27] that. And again, it's moving so quickly.
[56:30] And if their models, the Chinese models
[56:32] keep up, and you have to think about it,
[56:33] that was a shock last year with
[56:34] DeepSeek. The models have continued to
[56:36] keep up. They're not as good, but all
[56:39] it's going to take is a US model to get
[56:42] there because when a US model gets
[56:44] there, then you're going to have all of
[56:46] these things occur. So I just wanted to
[56:47] go through open source is attacking all
[56:49] four. The question is when are we going
[56:52] to get a US open- source model that is
[56:55] good because are AI are large
[56:58] enterprises like in the US the S&P 1500
[57:00] likely to use an open source Chinese
[57:02] model or is it more likely for small
[57:03] businesses and entrepreneurs with
[57:05] budgets? Yeah, certainly no. And that's
[57:08] what Dario Modi talked about. He said
[57:10] he's not losing anything to theirs. So
[57:13] the large enterprises, again, you have
[57:14] to separate this. Large enterprises, AI
[57:16] native startups that are running on a
[57:18] tight budget, not trying to make
[57:21] trillions of dollars to defend. If
[57:23] you're trying to build a business and
[57:24] get up to two million, three million,
[57:26] and you're an entrepreneur, basically
[57:27] what I'm doing, I absolutely want to use
[57:30] an open source model, but I'm also
[57:32] paying for the five models at the
[57:34] highest level. So I have an employee
[57:36] that's $12,000 a year. Um, I like using
[57:39] all five all day long. I like getting
[57:41] their different views. So everything is
[57:43] getting cheaper and cheaper and if they
[57:44] were to raise prices which they're not
[57:46] going to be able to then everyone would
[57:47] shift to open source if the models are
[57:50] close enough. So if an American open
[57:52] source catches up then it gets worse and
[57:56] I would say that is the next thing to
[57:58] look for as a disruption. Um so
[58:02] entrepreneurs and small businesses are
[58:04] using it today. Mid-market and tech
[58:05] firms this is basically using the open-
[58:07] source model. um you're going to see
[58:09] smaller companies that are going to that
[58:11] are going to do it. They're going to be
[58:12] able to compete on that basis. Large
[58:14] enterprises are the ones that are will
[58:16] eventually get off and go to open
[58:18] source. And then the government and
[58:20] defense even longer. So the more secure
[58:22] and the more you have to protect and
[58:23] this is very much a Bitcoin story. Uh as
[58:26] Michael Ser said, you don't find
[58:28] Bitcoin, it finds you. You don't find
[58:29] open source, it finds you. Small
[58:31] entrepreneurs and businesses trying to
[58:32] grow, they care more about revenues than
[58:34] they care about expenses as much as they
[58:36] do revenues. in this case mid-markets to
[58:39] compete if they're realizing how
[58:41] expensive it's going to be to bring
[58:42] Palunteer on they're going to have to go
[58:44] the open source model and go a different
[58:45] direction so again I think that's the
[58:46] way to look at this um
[58:50] the pace is going fast the RPO is
[58:53] liability framework forward revenue val
[58:55] is a bullish signal predictable growth
[58:57] but it's a very negative thing in all
[58:59] these uncertainties that are coming down
[59:01] the road if you don't know what moltbook
[59:03] is go read this um I think it's
[59:06] fascinating what's happening. It is
[59:09] important to understand especially as
[59:11] crypto people I tied this back to Axi
[59:13] infiniti uh which is a phenomenal
[59:15] economic story a disruption that
[59:17] happened basically with crypto but with
[59:20] the form of humans now it's going to be
[59:22] AI agents and you can see where this is
[59:23] going to have huge repercussions if
[59:26] you're doubting AI still and you think
[59:28] it's still just a chatbot this interview
[59:31] at a closed meeting at the Institute of
[59:34] Advanced Study that's in Princeton
[59:36] that's where Einstein and all of the the
[59:38] the Manhattan Project Brains worked. If
[59:42] you remember watching the movie, uh top
[59:45] physicists agree AI can now do up to 90%
[59:47] of their work and may soon push
[59:49] discovery beyond human understanding.
[59:51] They all went into this being completely
[59:53] skeptical. That's how far we've gotten.
[59:56] That even the smartest people on the
[59:57] planet, the astrophysicists, people
[60:00] don't seem to think Elon Musk is in that
[60:02] group even though he has more rockets
[60:05] than anyone. But these are truly the top
[60:08] physicists basically saying the work can
[60:10] be done.
[60:11] I'm going to end with this. Stable coin
[60:13] volume sets a historic record $10
[60:15] trillion. AI agents are starting.
[60:18] Transactions are going to go up. The
[60:20] internet network effects in terms of
[60:22] crypto are going to happen. The
[60:24] financial guard rails are necessary.
[60:26] Fade crypto with your own
[60:30] at your own risk. SAS is connected to
[60:33] crypto because of 2021. We've now taken
[60:36] things down as soon as SAS stabilizes or
[60:38] you see a breaking correlation between
[60:40] Bitcoin and SAS. I believe that will be
[60:43] the trigger point and the catalyst will
[60:44] just be the technical side. John Rog had
[60:47] called 60,000. I'm sure he still thinks
[60:49] there might go lower. Most technicians
[60:52] at this point wouldn't be calling a
[60:53] bottom, but we'll see how the bounce
[60:54] lasts. That's it for me this week.
[60:56] There's a lot to go through. Again, the
[60:58] launch is this week. Um, any of you want
[61:01] more help on the demo side and learning
[61:03] how to use the tools and come up with
[61:04] ideas, great. I will try to get a
[61:06] Palunteer piece done um on the
[61:08] importance of that Ultron theme and what
[61:11] it means and the different names that
[61:12] are associated with that buildout. Uh,
[61:14] but that's it for this week. Thanks
[61:16] guys.

← Back to videos list