How to think like a Financer
Andro
This is an essay on the financing of artificial intelligence masked as an essay on internal Western struggles masked as one on Chinese capital markets masked as another on non-cooperative games, which, ultimately, is all a mask for an essay on the financing of artificial intelligence. One of the distinct advantages of London that a starry-eyed, homeless New Yorker might witness is that it is the true destination for international finance. In London you meet the so-called ‘oligarchs’, whereas in New York City you might encounter an Italian proxy for a Russian who, in fact, is a proxy for a noviop, who is in turn a proxy for Putin, who is in turn a proxy for — anyway. Matryoshkas within matryoshkas, and so on. And then, of course, who could leave out the Qataris or Koreans one might meet in London, buying up the land around Whitehall? New York City is all too local and thus it is a credit to J’accuse contributor ‘Franz Pokorny’ when he observed the following:
“The future of humanity hinges on convincing a handful of people in London and SF that there exist pleasures more sublime than merely subsisting as white rajahs of the repugnant conclusion; the right’s perpetually aggrieved style is generally a block to this work of change.”
Leaving aside the commentary on the Online Right, Mr. Pokorny properly notes London and San Francisco are the premier poles of the 21st century. Given this is an essay on financing, however, where does that leave New York City? Is London financing the AI boom? The answer is simply No. All the major public businesses involved in the Intelligence Trade work on the New York Stock Exchange and NASDAQ, and Wall Street banks structure the deals that fund much of the physical infrastructure (Rule 144A private placement anyone?). London, by contrast, is the one launching 500-million-pound Sovereign AI funds to back domestic startups — although, to be fair, they are the prime intermediary for Gulf sovereign wealth, helping institutions like Abu Dhabi’s MGX before they can get their hands on OAI directly. If there were any serious downturn, the New York-domiciled private credit backers and equity in the hyperscalers and Nvidia would face the brunt of the pain.
Nevertheless. Nevertheless. It is deeply curious that Nick Bostrom, Toby Ord and that whole lot formed the ‘Future of Humanity Institute’ at Oxford in 2005 before shutting down in 2024 as a result of Nick Bostrom channeling Mike Enoch and Richard Hanania on a listserv in 1996 (really we could call it cypherpunk excess, a la Tim May). When I ask Americans in the Intelligence Trade about this fact, they’ll often say Effective Altruism is an American product. As Yudkowsky as Apple Pie. For all of Mr. Yudkowsky’s situational awareness, however, the first institutional form of Effective Altruism, the first ‘Center’, was founded in 2012 also at Oxford by an Australian and a Scot. Not much unlike William Paterson founding the Bank of England. What I mean to highlight is the intra-Anglosphere battle being waged here: the fact that America acknowledges London as the conduit for nefarious, Eastern or Continental regulations and as the ‘Fridge guarding the snacks’—namely, Gulf $$$. This cannot stop Donald Trump, who descends from way up in the Hebrides, from making these deals himself, or with the help of the Kushner Crime Syndicate, the Indian-American bag holders of California, and champions of the International Math Olympiad from the years 1990 to present. This is serious stuff: we’re talking trillions in investment, here. Most Americans, especially of the ‘Right-Wing’ (whatever that means in 2026 I leave to the reader), conceive of London as a moribund, bankrupt city led by Elijah Muhammad. To that end, perhaps they read William McAskill as that poor African American woman protesting the Apollo launches, but it’d be much more fitting for them to understand London’s importance as the intellectual and networking hub for the Anthropic-verse and the Titans who run it. When you think London, think Leopold Aschenbrenner and not Andy Burnham dancing to old rhapsodies of smack addiction. A few, select Americans grasp this. They travel to places like Oman for ‘alpha’.
Before I sound too much like a loon asserting some sort of necessary opposition between American capital and British/European nanny-statism, we must first turn to China, the true threat to freedom and the Yudkowsker way of life. One argument — a tidal wave of reasoned discourse billowing out from the mysterious “Harem Financer” on twitter — might go that the US deliberately engineered the AI investment boom in order to reverse capital flows that had previously been going strictly to China, and thus also the broader developing world, over the past four decades. Donald Trump sent America down this path of rabid Keynesianism, whether for green, Dark Woke purposes or military expansionism, and he now intends, at long last, to get the upper hand on the Chinese with all the redirection into American industrial supply chains. The argument would continue that AI buildout is the precise inverse of the carbon-credit and globalization trades of the past forty years. The new trade functions to pull capital from China, Korea, and Japan into industrial outputs such as ball bearings rather than services of emerging-market assets. The key to understanding this is that China is a closed capital account, meaning their government restricts the movement of money across borders for national investment purposes. China caps the amount of Yuan that residents can convert into foreign currency, and, likewise, limits how much Foreign Direct Investment can be made and therefore how much can be pulled out of Chinese securities. Beijing likes it this way because they can secure monetary policy control, prevent capital flight, insulate the banking system, and manage the exchange rate accordingly. Whenever China loosens controls here and there, they reverse course suddenly and sporadically, though normally as outflows spike, because there is simply no way to simultaneously desire global currency status and seek out national financial control.
The fact of China’s closed capital account makes clear the difficulty the Middle Kingdom has in resisting the pull of America’s AI buildout. There has been recent controversy over open-weighted models (viz., GLM 5.2 and Kimi v3) and the foundation labs (viz., OpenAI & Anthropic), as demonstrated by Nvidia’s Jensen Huang tweeting for the first time (cosigned by American tech giants and the French business Mistral AI) and Alex Karp performatively (and profitably!) sperging out on CNBC’s Squawk Box (could one of Mikka, Tim Chapman, or the Marquis schedule the next Goggle Box to focus the viewer-actors on Squawk Box?). As I’ve said ‘on main’, much of this controversy boils down to the fact that the frontier models are plateauing in terms of capability. Progress on the difficult-to-game benchmarks is crawling with each subsequent release (though, to be fair, this is partly a result of goalpost-shifting) and thus open-source models are catching up in practical terms. The San Franciscans have invented terms like ‘Big Model Smell’ to describe some ineffable quality that proprietary models have that the open-source models do not. In the end, it does not matter if these open models are good enough for what people and firms are doing day-to-day. Instead of selling their tastefully scented inference at absurd rates, Anthropic & OpenAI will be in a knife fight with anyone who has cheap electricity and GPUs. The recent open-source rollouts have happened alongside the largely staged Fable and GPT 5.6 Sol fiasco which was a watershed moment. The sovereignty stuff doesn’t matter much to me, and is mostly a game of PR for these characters. The set of people and organizations who simultaneously believe A) their data is valuable enough to safeguard and B) are sufficiently paranoid they don’t trust a contract with some provider is likely small.
But back to China—these open-source AI models generate virtually no profit if the models have been commoditized and cannot effectively compete with the IP-protected, P/E-supporting models that backstop American equity markets. The FOSS (free-and-open software) battles of the ‘90s present a kind of parallel. The goal there was to build free labor pipelines for multinational software infrastructure and thereby spread American political influence through the internet. It would be deeply ironic for China to push open-source AI to challenge the US since the US wrote that playbook in the first place. It’s a fun attempt to pop the bubble, but $7 Trillion in FDI has already hit the US (linking a public McKinsey report in J’accuse ought be a capital offense).
Critically, China is not the only entity caught up in this pull. Imagine for a moment you were an American as wealthy as the Dragon. Well, I suppose that’s priceless wealth, so we’ll set our aims a tad lower—let’s call it a handful of billions. A paltry sum. China is a very closed capital account. The EU is not much of an open market either. America now has deflation like everywhere else because all the liabilities are owed to the old; nevertheless, you stick it out in the Land of Freedom. The US target federal funds rate is roughly 3.5%, certainly a risk-free rate if you’re an upper-middle-class owner of a plumbing firm in Minnesota that does well and has a well-trained CFO, but if your eyes are on the Dragon’s treasure then you’re likely constructing your own conception of the Consumer Price Index (CPI) because you know the state isn’t accurately depicting price levels and because the state exists to confiscate capital. Your CPI-weighted basket is not going be the ‘Generally Accepted Accounting Principles’-accepted version given to asset managers with fiduciary published by the Fed and Treasury. Instead, you’re likely to estimate your actual basket of goods likely runs something like 9% against that 3.5% Fed Funds rate. Quick math gives you a 5.5% negative real yield. That loss is an indirect payment to the American hegemonic system which one can contrast with Xi’s property and capital confiscations over the past fifteen years.
If this is all Greek, or Jewish, or even Chinese, to you then I recommend the following on Xi’s maneuvers over the past decade-and-a-half:
1. Dissertation on Xi’s anti-corruption movement
3. Caixin on the nine Tomorrow Group financial institutions seized (must translate)
5. Evergrande and presale losses
Hopefully we still have Mr. Star’s attention. The 3.5% yield rate is not all loss, though. It still pays real cash now and can be utilized in the precise sorts of asymmetric bets one witnesses in the Office of Strategic Capital lending some $200B in the coming two years of the Trump Administration. Everyone from San Francisco to Washington D.C. to your Gen-X mother (curious if Britpoppers also say this) complain government does not work well because it doesn’t pay high-enough salaries. Assuming that is true, the Office of Strategic Capital has no problem pushing competitive salaries on the unassuming assholes who are the ‘Vice Presidents’ of ‘Goldman Sachs’. Conservatives who believed themselves to be MAGA are flabbergasted, yet again, at the sight of this trade—many concerned they’ll be arrested—whereas the money-lustful see it clearly: Feinberg, that old hack, is running a sweaty shop to make the same, grimy plays on which his brand is built. To hell with Equity Capital Markets! Let’s get lending. As for the more politically aware associates and VPs, they’re reading that Julius Krein’s (founder of American Affairs) roommate, David Lorch, is running the whole operation. It’s been planned from the beginning. The spirit of ’16 lives on.
There exist parts of the Western power industrial supply chain that have not had back orders and bids for decades, but now suddenly do. If the Office of Strategic Capital and aligned private markets investors truly wish to make America Dynamic again, bring industry home, and maintain re-entrenchment they need not only all that foreign direct investment but also the ability to show to that FDI that they can provide cheap power. Enter Stargate, Limited Liability Corporation. The AGI data centers operate here as the flywheel. This will inevitably force a consensus between the rival energy lobbies of gas, coal, nuclear, and green in favor of new coal and nuclear. Here’s the young founder of Valar Atomics and Doug Wilson-acolyte, Isaiah Taylor, informing a genuinely serious Donald Trump that his policies have enabled precisely this. Standing to the right of Isaiah is the Secretary of Energy, Chris Wright, who once drank fracking fluid to show it was safe. Made in America. Will Larry Ellison or Dario Amodei drink data center water to show the same? Made in Yudkowsky. We’d be remiss to ignore the irony that US firms used to export the same nuclear modules folks like Isaiah work on to China while expansion faced harsh regulations in America.
Once this is all off the ground, and it’s happening as we communicate via this exalted publication, the initial public offerings are being prepared. How can we see multiple $1T+ IPOs as anything other than the AI world soaking up the excess money supply from intentional moves of the American Government under Donald Trump, such as the CARES Act, so that they will not chase real goods like the CPI index I delineated above? This is the regime of demonetization. The price fluctuations in Gold, driving the Fox boomers to ecstasy, confirm that the demonetization of commodities is functioning as such. Previously, China, much like Korea, Japan, and Germany, has operated as a vehicle for absorption of the cost of cheap power and production so that Western capital and pensions can remain solvent. The West generally captured some 85% of value-added export markets in some sectors until Xi’s rise undercut that whole arrangement. After all, the great pitch was that if you seek to maintain a broad power industry while ensuring pension funds and capital do not go bust propping up corporate debt—if you want YYY/kilowatt-hour—someone will have to hold the bag. Better to let the productive world and old people do it than Nick, aged 35, at this point.
If we’re going to do this properly, we’ll need to sidestep the Benthamites, the Nu-Lubbavitchers, and the Sriram Krishnan-industrial complex by returning to a truly great American from the state of West Virginia, John Nash, best known for his Equilibrium. Nash diagnosed that good money holds its value over time whereas bad money loses it under inflation. He treated inflation targeting as the fundamental conceit of the central banker that monetary policy does not control inflation. The ideal is zero. Nash proposed a benchmark comprised of an industrial consumption price index built from international prices of copper, silver, tungsten, and similar commodities (to which, notably, he thought crude belonged) as an alternative — but by no means a return — to specie. This benchmark was not necessarily to be fixed: as states evolve toward stable domestic price indices their currencies would drift toward a stable comparative value, and the asymptotic result would be ideal money achieved without hard-adopting this commodities basket as the standard of value. With plural competing currencies—the euro, dollar, yen, pound, franc, kronor—this system would produce pressure for quality and lower inflation with the leverage being the option to leave one state-money for another. Discipline by capital flight back-hand slap! Without this stable unit, we are unable to separate value creation from monetary absorption. The Cantillon effect can only be stopped if there is an immutable, scarce benchmark keeping everyone accountable. Nash’s target was the Cold War regime and for all this writing he was thrown into a sanitarium.
Let’s take Nvidia’s revenue, hyperscaler capital expenditure, and the labs’ valuations which all get quoted in the thing whose quantity is now in question as a policy variable. Nash’s basket refuses this outright. One would instead price the buildout in concrete materials and energy, thus reframing the trade as a large bid on the components of the index itself. This forces a distinction which the sponges heretofore have elided. The initial public offerings absorb claims via paper and retails bidding for a share certificate with none of it chasing the real good at-hand. The capital expenditure then spends for turbines and switchgears and gas interconnects. Demonetization functions on this absorption of claims but fails on the actual expenditure side which explains the gold rally. Anyone who tells you the buildout is disinflationary has only looked at the first half here. With the industrials index sitting as a public and legible entity it permits raw comparison. The open-source models like Kimi are a comparable index but for inference. Once a free reference point establishes a quality level and a serving cost anyone can verify on hardware, the frontier price becomes clean, legible as a premium over that point. The ‘Big Model Smell’ can be analogized to a central bank’s assertion of the superiority of its money while declining to publish a proper, true index.
The Financer’s final questions is: what unit, whose residual? Denominate the whole thing and the answer will be readily apparent. Hyperscalers gain 30%, but the basket gained 40%. Good job, sucker, Yudkowsky has enough money to bid on an Aella sexy-time contract. Luckily he will also lose that trade.
I could carry on, but I figure the point has been made. This is how the top pimp at the Harem is reading the AI Bubble. It’s the only plausible response to the sweeping force of Meritocracy.