The Invisible Trillion

The Bill

Who’s Actually Paying for All of This

The last two chapters were about the machine — how big the buildout is, how it should be sized, what the right architecture looks like. This one is about the money. Specifically, about a question the technical chapters never quite reach: who is paying for all of this, and what happens to them if it doesn’t work out the way it’s being sold?

Two arguments stand in the way of taking the costs seriously, and the buildout’s defenders reach for both. The first is that the smart money has spoken — the capital markets are pouring trillions into this, and people who manage money for a living don’t do that on a whim. The second is that it doesn’t matter whether it pencils out commercially, because it’s a matter of national survival: the war for the twenty-first century will be fought partly in compute, and falling behind is not an option. Both arguments are serious. Both are partly right. And the best way into both is to start with a single morning in June.

The Largest IPO in History

On the morning of June 12, 2026, about an hour before the opening bell, a Falcon 9 lifted off from Florida carrying a load of Starlink satellites. It was the 650th flight of the rocket. In Starbase, Texas, Elon Musk stood behind a Nasdaq-branded podium, and in New York his executives rang the bell. SpaceX — recently merged with Musk’s AI company, xAI — was going public, and by the end of the day it had made history.

The numbers are worth stating plainly, because they are real and they are staggering. The company sold roughly 555 million shares at $135 and raised $75 billion in the offering itself — the largest IPO ever, nearly triple the previous record, Saudi Aramco’s $25.6 billion in 2019. The stock opened at $150, surged past $160, and at one point touched a valuation north of $2.2 trillion, one of the ten most valuable companies on Earth. Musk, on paper, became the world’s first trillionaire. By any conventional measure, it was a triumph.

Here is the part the triumph tends to crowd out. SpaceX has never turned a profit. It lost close to $5 billion in 2025, and $4.28 billion in the first quarter of 2026 alone — more than four times its loss a year earlier. Strip out the rockets and Starlink, and the AI division it had just absorbed was the wound: roughly $6.4 billion in losses on $3.2 billion of revenue. Morningstar, running the kind of discounted-cash-flow model that has valued companies for a century, put the whole enterprise at $780 billion — call it overvalued by more than half.

And six days after the IPO, on June 18, the company’s bankers were back in the market arranging a bond sale of at least $20 billion, the bulk of it to refinance a bridge loan SpaceX had taken on to buy xAI in the first place.

Sit with the shape of that. A company losing billions a quarter raises $75 billion from the public, then borrows $20 billion more within the week, while its founder keeps roughly 85 percent of the voting power and the analysts who model these things for a living call it overvalued. Michael Burry — the investor who famously shorted the 2008 housing market — took a public position against the trade. None of this means SpaceX is a fraud, or that it won’t deliver. It means we are watching a particular kind of moment: the one where paper valuations get converted into real, banked capital while the window is open.

There is a name for it, and it isn’t flattering, but the name matters less than the pattern, and the pattern is everywhere right now. OpenAI and Anthropic have both filed to go public. Oracle, Nvidia, Alphabet, Amazon — all have rushed into the debt markets to fund data centers, raising hundreds of billions between them.

So the capital markets are, as the defenders say, voting. The votes are coming in at trillion-dollar valuations. The only question is what, exactly, they’re voting on — and whose money is being voted with.

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