“Socialism is an alternative to capitalism as potassium cyanide is an alternative to water.” --Ludwig von Mises

🦔China’s AI stock bubble already burst. The STAR50, an index of China’s AI and chip companies, fell 40% from its July high in under four months and wiped out about $480 billion. Robot maker Unitree, which retail buyers oversubscribed 8,000 times at its listing, has halved since August. My Take Small investors carried this rally, and a lot of them bought because state-backed funds owned the same companies, so it felt like Beijing had signed off. Those state funds sold into the rally to cool it down while small investors kept buying. I don’t blame them much. When the government’s own money owns a stock, assuming it’s safe is reasonable, and that trust made the losses hurt more. US AI stocks look cheap next to that. The Nasdaq 100 trades around 30 times earnings, and the STAR50 is still near 100 even after the crash. The weak spot here is debt and new stock, with Meta selling $35 billion in shares this week and credit spreads at 17-month highs. If US AI stocks slide, the pain shows up in index funds and 401(k)s, since that’s where most Americans hold this trade. Chinese chipmakers will also end up in a price war at home (one Shanghai fund manager says it’s coming), and cheaper Chinese chips and models make it harder for American labs to cover their compute bills. Hedgie🤗 reuters.com/world/china/chin…

Holy fucking shit

And yet, Blackstone is trying to sell its data centers. Why? $BX

I've made this point before, but "a 5% tax will not make any difference to the way they live" is an argument against levying such a tax. A tax that doesn't cause the taxpayers to consume less will cause them to reduce investment, hurting growth. Some reduction in private investment is unavoidable when levying almost any tax, but we should not levy taxes narrowly targeted to reduce investment as much as possible and consumption as little as possible.

If Prop 40 is enacted, Zuckerberg will be on the hook for about $10 billion. @Meta will allow him to sell up to 20% of his new class C shares. Because these shares are non-voting, Zuckerberg will not lose any voting rights. He will keep control of @Meta. In return, Zuckerberg commits to staying in California for at least 8 more years. This destroys the argument that somehow billionaires “cannot pay” a wealth tax because they would “lose control” of their companies (which would be bad, according to them).

T at Apollo: AI circularity extends to token spending: Ramp data shows 80% of top AI spenders are VC-backed (vs. 13% median), meaning compute is paid for with venture capital, not organic earnings. If the AI payoff lags, VC returns take a hit... and token demand weakens with it.

“I don’t see a systemic problem" in AI, Bessent told investors at a closed-door event. 👇🏼 ft.com/content/5b7c2fdf-2300…

The irony of this proposal is that it essentially calls for the Fed to do the very thing it was supposed to when it was first created: supply reserves to banks as needed to achieve an “elastic” currency. You’d think they’d have gotten it down by now!

More on the "crowding out" problem of AI: "A popular view holds that the AI capital boom is powering a broad US investment revival. The numbers say otherwise. In the three years after ChatGPT launched, US data center investment tripled and broader technology investment grew 50%. Yet total productive investment held flat as a share of GDP, because many investments unrelated to technology, software, and R&D fell relative to GDP. The tech build-out has shifted the composition of US investment rather than ramping up its total." - McKinsey For my next report, I've been buried in "Global Electricity Undersupply" (learn more about our reports here: sageroadresearch.com/collect…). Electricity supply exemplifies AI's crowding-out downside. Last year saw global electricity demand increase 3%. The IEA expects demand to only accelerate over the next two years, hitting 3.6% demand growth in 2026 and 3.8% in 2027. Data centers are only a sliver of that growth equation, currently accounting for only 1.5% of global electricity consumption. Even considering current AI buildout expectations, that share of consumption is projected to reach just 3 to 4% by 2030. The drivers of increasing global electricity demand extend far beyond data centers, from transportation to heating and cooling to industrial electrification. For just one indication of how electrification can boost global productivity, just consider AC in Europe. Only 10% of European households have AC. Europe suffered at least 35,000 excess deaths this summer from back-to-back heatwaves, according to government figures. Yet, as tragic as that is, it is not the only concern. Extreme heat is also an intensifying economic drag on Europe. As Allianz calculated in a report earlier this year: “Workers’ hourly output declines by $1.30 for every degree between 30 and 35 Celsius…Heat waves cost between 0.3% and 0.5% of European GDP overall, and more than 1% of GDP in heat-prone southern regions.” AI is compromising the ability of the globe to meet electricity demand increases outside of AI, generating electrical equipment shortages and inflation. For one, gas turbines are reportedly sold out through 2030, with prices spiking 195% since 2019. For another, delivery lead times for high-voltage transformers have ballooned from about a year in 2020 to upwards of two to four years, with prices climbing to nearly four times 2021 levels, according to Reutersreporting. Much the same is true for circuit breakers, “with high-voltage circuit breakers recording an average lead time of 151 weeks in 4Q25, an increase of more than 130% YoY,” according to Wood Mackenzie. And then there’s switchgear, with “medium-voltage switchgear lead times currently sitting at 40 to 80 weeks, more than double pre-pandemic averages, while large low-voltage switchboards lag at 35 to 62 weeks compared to the 20 to 30 weeks once considered standard,” according to DEI Power Solutions. Far more to come in the report! McKinsey link: mckinsey.com/mgi/unconventio…

Insane. "Taking the next three years of projected earnings for the S&P and summing them, we can see that its rise has never been higher as far back as we have data for, running at a 40% annual pace." Earnings growth has NEVER BEEN HIGHER. Without it, a 10-year yield still over 5% would be a much more serious headwind for multiples.

Today's Post Get Out of the Phone Box polemicpaine.substack.com/p/… On why short-term trading now belongs to the machines and the edge has moved to the medium term. Algos trading correlations they don’t understand. Why the economists’ letter is still the best contrarian indicator going, from Brexit and Trump to Milei. Why wars don’t back test. Alex Karp on the neurodivergent and why sideways thinkers may beat the consensus machines. AI capex boom, a smorgasbord nobody’s sure will be paid for.

This chart shows the key risk to the AI trade, and it's missing from the latest revenue announcements. The AI boom is fundamentally driven by a top 1% of companies that drive 80% of spend. Within that top 1%, 80% are venture-backed. That's all fine if the AI trade works out, and if, as some prominent investors say, everyone wins. But if it doesn't work out, then we have a cascading effect of large customers simultaneously pulling back their AI spend.

Probably better to see some noticeable productivity acceleration on a macro level before pitching 5x the growth rate ever seen in the history of man in just 3yrs. xitter.cf/scaling01/status/21112…

So far, AI has not produced an economy-wide acceleration in productivity growth. Adjusted for shifts in utilization rates, total factor productivity growth was zero in Q2.