Factories, Financing, and Freeing Energy

The column started with a single X post from @HedgieMarkets laying bare the financial engineering behind the AI boom: massive data centers funded through special purpose vehicles, off-balance-sheet debt, and leverage that would make a subprime lender blush. Meta’s $27 billion Hyperion deal with Blue Owl, Oracle’s $38 billion Vantage play tied to OpenAI’s Stargate, xAI’s $20 billion raise looping in Nvidia GPUs-these aren’t just big numbers. They’re bets on explosive revenues that have to show up in the next 5 to 15 years, or the whole structure risks looking like a house of cards when the music stops.

🦔 Meta is paying $6.5 billion extra in interest to keep $27 billion in AI infrastructure debt off its balance sheet. They're using something called conduit financing, where a special purpose vehicle borrows the money, builds the data centers, and leases everything back to Meta.… pic.twitter.com/GIs0hAu0kM

— Hedgie (@HedgieMarkets) February 12, 2026

On paper, Meta is just a customer making lease payments, not a debtor.

Oracle is doing the same thing at even larger scale: $38 billion for two data centers through Vantage, part of its $500 billion Stargate partnership with OpenAI. xAI raised $20 billion through a similar structure, with Nvidia contributing $2 billion in equity while also being the hardware supplier.

The circular financing is incredibly dizzying. Nvidia invests in CoreWeave. CoreWeave uses that capital to buy Nvidia GPUs. CoreWeave leases capacity to Microsoft and OpenAI. Their revenues support the lease payments. Nvidia reports revenue from the chip sales and marks up its CoreWeave investment. Everyone’s balance sheet looks clean.

My Take

Some analysts are comparing this to subprime-era tactics where firms shifted risk off their books to reassure investors. The structures are legal and the accounting is technically compliant, but the effect is the same: hundreds of billions in obligations that don’t show up as debt in the traditional sense.

The whole thing depends on AI eventually generating enough revenue to justify the infrastructure costs. Moody’s flagged that Oracle’s data centers rely heavily on OpenAI, which won’t be profitable until 2029. If AI monetization disappoints, we’ll find out whether bondholders have secured claims on essential infrastructure or whether they’re functionally unsecured creditors of overleveraged single-purpose entities whose assets are worth less than the outstanding debt.

I keep coming back to the same question with all of this: what happens when the music stops? The conduit structures haven’t been stress-tested. The hyperscalers are profitable now, but these are 20+ year obligations built on assumptions about AI demand that haven’t been proven. If custom silicon undercuts Nvidia demand or the revenue never materializes, a lot of people are going to discover that the risk they thought was somewhere else was actually right where they were standing.

Then came the human side. Google’s voluntary buyouts for staff not fully “all in” on AI, Salesforce slashing support roles almost in half, predictions from Dario Amodei and Sam Altman that entry-level white-collar work could evaporate. John Sexton’s HotAir piece captured it perfectly: AI isn’t just augmenting-it’s replacing at speed, and the MIT economists reminded us that how fast matters more than the tech itself. If change is gradual, people adapt. If it’s a tsunami, pain follows.

The inevitable and fully justified reaction by much of the population against a corporate AI machine that will strip them bare and guarantee many of their children an anxious existence of unemployment and irrelevance is not being a "Luddite."

It's standing up to an invasion.

— Walter Kirn (@walterkirn) February 11, 2026

That pain has voices, and few capture it more raw than Walter Kirn with a cri de coeur for creaturehood-the primal defense of our messy, embodied humanity against a cold algorithm that promises efficiency but threatens soul. Kirn nails the invasion metaphor: AI as an existential force marching on jobs, creativity, relevance. Parents see it clearest, staring at a future where kids might inherit irrelevance instead of opportunity. Yet this cry isn’t a call to smash the looms-it’s a reminder that resistance alone won’t win. The path forward honors that human fire by channeling it into adaptation, turning invasion into evolution.

Here’s where the optimism kicks in. Disruption doesn’t erase human edge; it transfers it. If you can sell a buggy whip, you can sell a Model T. The skill-understanding desire, solving pain points, closing the deal-survives the shift from horse to horsepower. The same goes for writing: If you can write, you can prompt. Clear thinking, precision, structure, iteration-these are the exact muscles that make AI agents and models deliver useful output instead of gibberish. The journalist, the analyst, the copywriter doesn’t vanish; they level up into prompt engineers, workflow architects, enterprise adoption strategists who direct intelligence at industrial scale.

Scale that up further. Those data centers aren’t the Model T rolling off the line. They’re the River Rouge factories building the Model T’s-massive, vertically integrated complexes churning out agents, reasoning engines, synthetic data, frontier models. The moat isn’t in the product; it’s in owning the means of production: compute clusters, custom silicon, power grids, cooling systems. Most of the world won’t build their own factories. They’ll buy the output-tokens, APIs, deployed agents-and the sales opportunity explodes for anyone who can move that abundance into enterprises, governments, and daily life.

This is where policy enters the picture. Capital gains is a sidebar, but a true one: lower the rate, reward the risk-takers funding these factories, make exits more lucrative, draw more private capital into the build-out. Add cutting interest rates-Fed funds still hovering around 3.5–3.75% after 2025 cuts-and borrowing costs for massive capex drop. Cheaper debt service on SPVs and corporate bonds means faster expansion, lower leverage risk, earlier abundance. Both levers protect the branches: the capital stock, the reinvestment, the growth structures. Don’t prune them early with high taxes or rates; let them thicken. Tax the fruit-realized profits, gains, dividends-when it’s ripe. The tree grows taller, the harvest bigger, sweeter, more plentiful for everyone.

Obsessing over the wealth gap often reveals the wrong priority. If the real goal were lifting the floor, we’d cheer when billionaires multiply alongside poverty dropping from 36% to under 10% globally. But when the loudest voices push punitive measures that slow overall growth, it starts looking like they’d rather have the poor poorer as long as the rich are less rich. In the AI era, cheap compute lifts the bottom most: free tutors, diagnostics, automation for the grunt work. Kill the golden goose with branch-pruning policies, and we all lose.

Which brings us to the highest-leverage move right now: energy permitting. Capital and energy are the eternal drivers-always have been, always will be. US electricity demand is surging 35–50% by 2040 from AI alone, yet we take years longer than China to build power plants or transmission lines. NEPA reviews drag on, litigation clogs the courts, interconnection queues back up for years. The Trump administration is pushing hard: Executive Order 14318 fast-tracks AI-related power infrastructure, EPA streamlining Clean Air Act reviews, a voluntary compact getting Big Tech to fund 100% of new generation without hitting ratepayers. The SPEED Act passed the House (bipartisan), ePermit Act digitized processes-now the Senate needs to move.

Wise policy clears every removable stumbling block. Unleash permitting, pair it with private funding, and factories scale at warp speed. Cheaper compute arrives sooner, revenues hit the window before leverage cracks, abundance offsets the job shifts. Government shouldn’t erect barriers to capital and energy; it should get out of the way.

Texas knows this instinctively. We’ve seen energy booms when the path is clear. Protect the branches. Let the factories hum. The fruit will take care of the rest.