The Invisible Trillion · The Bill

At What Cost, and to Whom

The two arguments don’t cancel. They’re answers to different questions. The financial one is about how the buildout is financed and who collects — and it points to late-cycle extraction that will hurt late investors and host communities. The geopolitical one is about whether to build at all — and it says yes, some of it, urgently. Both can be entirely true at the same time. What they jointly imply is not stop. It’s build it differently: site it away from homes, make it pay for the grid it needs, aim the capital at the uses that actually matter, and stop pretending the current pattern is the only one available.

But that’s the prescription. The question this chapter promised to answer is the cost, and the cost is real and itemizable. It falls on the host communities from the earlier chapters — the noise, the air, the bills, the cascade. On the ratepayers across PJM and beyond, tens of billions a year in transfers from households to the hyperscalers whose infrastructure their rates subsidize. On the late investors who bought SpaceX at $135, or will buy the next IPO, or hold the bonds.

On everything the capital and political attention didn’t go to — the housing, the healthcare, the climate infrastructure, the other problems no one had bandwidth for. On the air and water and land around the facilities. And on democratic legitimacy itself — the slow corrosion, across thousands of towns, of the belief that ordinary people can shape the decisions that reshape their lives.

The benefits are real too — the genuine capabilities, the security, the productivity gains. But they flow to a different set of people than the costs do. They accrue to the operators, to the investors who exit at the right moment, to the users paying subsidized prices for premium tools, to the broad economy capturing the macro gains. The costs accrue to the communities, the ratepayers, the late investors, and the institutions whose legitimacy this pattern erodes.

So the buildout is not, on its face, bad for the world. Summed across everyone, it’s probably net positive — the technology likely justifies some scale of it, the security benefit is real, the useful applications will last. The problem is the distribution. The people bearing the costs are not the people getting the benefits, they are not being compensated, and the legal machinery that might have ensured a fairer split was, as the neighborhood chapters showed, deliberately dismantled. An outcome can be positive in aggregate and still be one that decent people, asked honestly in advance, would not have chosen — and the people paying for it were never asked.

It helps to put two faces on it. A woman in Brittany Heights watches her home lose its value to a hum that never stops. A retired teacher in Phoenix watches a third of her 401(k) evaporate when the bubble resolves. A child in Boxtown grows up breathing turbine exhaust. Different people, different states, different mechanisms — and the same buildout, with the benefits flowing somewhere none of them live. They will never know they’re characters in the same story.

Summed across all of them — the host communities, the retirees, the ratepayers, the late investors — what you get is a generational transfer of wealth from a broad base of ordinary Americans to a narrow tier of operators, founders, and the intermediaries who structured the deals. Economists have a clinical phrase for this: late-stage extractive financialization. It isn’t a slur; it’s a description, and it puts the 2026 AI buildout in a lineage with the railroad bubbles of the 1870s, the dotcom bust, and the mortgage crisis of the 2000s. The specifics will differ. The structure — overbuild, inflate, write down, transfer — tends to rhyme.

All 19 chapters of The Invisible Trillion · Library · All courses