The Invisible Trillion · The Bill

Whose Money Is It

Follow the $75 billion backward.

The buyers of a mega-IPO, the buyers of investment-grade bonds, the institutions whose appetite makes these deals possible — Vanguard, BlackRock, Fidelity, State Street, the big pension funds, the insurers, the university endowments, the public retirement systems — are not spending their own money. They are pooled vehicles. They hold the savings of tens of millions of ordinary people. The teacher with a state pension, the firefighter with a municipal plan, the family with an IRA full of index funds, the worker whose 401(k) defaults into a target-date fund — those are the people whose money is actually deployed in the buildout, channeled through intermediaries they will never meet and mostly do not think about.

And the exposure has quietly become enormous. The largest handful of tech companies — the ones an earlier era nicknamed the Magnificent Seven — have grown to roughly a third of the entire S&P 500 by value, with Nvidia alone around seven or eight percent of it. When SpaceX enters the indices, it will climb that list fast; if OpenAI and Anthropic follow, faster still.

Which means the default retirement investment for most working Americans — the broad-market index fund, the thing they were told was the safe, boring, responsible choice — has, without anyone deciding it, become something close to an AI-buildout index. The bond side is the same story through a different door: the investment-grade bonds funding the data centers sit in pension portfolios and the bond sleeves of those same target-date funds.

So when someone says the “sophisticated investors” are confident, the sentence is true and also misleading. Underneath the sophistication, the capital is overwhelmingly ordinary people’s retirement money, and most of those people have no idea what it’s funding and no real way to opt out.

That changes what’s at stake if the bet goes wrong. We have a sense of the scale from the last two times. By common estimates, the dotcom collapse erased something like $5 trillion in equity wealth, and the 2008 crisis took roughly $2.7 trillion in retirement assets at the bottom. Both times, the heaviest losses landed on the same cohort: the people closest to retirement, who had shifted toward stocks as their date approached and who could not afford to wait out a recovery that, as we’ll see, can take a very long time. Millions who had planned to retire between 2008 and 2012 worked five or ten years longer than they meant to.

If the AI buildout resolves the way speculative buildouts usually have, that bill falls hardest on the Americans now between fifty and sixty-five — people who taught school, fixed pipes, kept books, ran small businesses, and saved exactly what the retirement industry told them to save, into exactly the vehicles it told them were safe. They did everything right, by the rules as written, and they are the ones most exposed.

I want to be careful not to state that as a prophecy, because it isn’t one. The bubble might deflate gently. The technology might deliver enough real productivity, fast enough, to grow into the valuations — the boom-that-validates is a live possibility, and serious people are betting on it. But the historical current runs the other way, toward real losses for retirement-aged savers, and nothing in the structure of this market gives strong reason to expect an exception.

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