The Invisible Trillion · The Architecture That Works

The Anticipation Economy

There’s one more layer to the Colossus story, and any honest reading has to include it. Taken purely as a transaction, the rental is excellent for SpaceX. It pays the build off in eighteen months, the margins are reportedly strong, and the cash flow steadied the balance sheet right when it needed steadying. The financial press read it as Musk, once again, finding a way to monetize an asset that looked like it was about to become a liability — and called the move visionary: a pivot from “win the model race” to “become a hyperscaler,” which in 2026 is a posture the market will value at multi-trillion-dollar levels.

All of that is true and none of it contradicts the analysis. The rental is simultaneously a tactical financial win and a confession that the strategy behind the original build failed. The fact that both are true at the same time is the most important thing the deal reveals, because it’s how the whole technology economy currently runs.

We are in an anticipation economy — a thing the standard vocabulary doesn’t have a clean word for. In it, an asset’s price is set not by the cash it produces but by what the next round of buyers will expect it to produce. The valuation floats up on the meta-level, expectations about expectations, several stories above the actual productive activity — and the operators who are best at conjuring assets whose anticipated value runs ahead of their realized value get rewarded most, even, or especially, when the thing underneath isn’t making money.

Musk is the clean case because both things are true of him at once. He has delivered genuinely enormous contributions to several frontier technologies — reusable rockets, satellite internet, mass-market electric cars, real progress on self-driving — and he has become exceptionally good at the financial-anticipation game. The technology isn’t fake because the financial layer runs on anticipation, and the financial layer isn’t erased by the real engineering underneath. The honest reading holds both, and the Colossus rental is a textbook instance of the second sitting on top of the first.

What Keeps It Going

The anticipation economy keeps running because a handful of conditions prop it up:

Index and venture participation. Big institutions hold SpaceX and xAI through index funds and venture allocations, where any single position is a sliver of a portfolio and never gets judged on its own. It’s sized to catch upside with the downside diversified into oblivion. Lose a little if the thesis fails; lose a lot, in relative terms, if you sit out a thesis that works. That asymmetry pushes everyone toward participating no matter what the underlying analysis says.

Capital with nowhere to go. The long stretch of cheap money produced more capital hunting for returns than there are technologies generating returns to match it. The surplus flows toward whatever can absorb it — which means the assets with the most credible-sounding stories about transformational future cash flow. Anticipation becomes the asset class.

Storytelling as a craft. The operators who specialize in this have honed it: stories about transformation rather than optimization, exponentials rather than lines, physical infrastructure — rockets, data centers, tunnels — that signals seriousness, real technical substance laced through the speculative parts, all delivered with enough confidence that being the one to hedge in the room costs you socially.

The cost of walking away. Once enough big institutions are in a trade, getting out while everyone else stays in makes you the one who missed the upside if it works, with no reward for prudence if it doesn’t. The incentives in asset management favor going along over exercising judgment, which is exactly why the people who could puncture the thing don’t.

Tactical wins that launder strategic failures. This is the feature that makes Colossus such a clean example. Each win along the way — a launch, a contract, a partnership — gets presented as proof the big thesis is working, even when, looked at squarely, it’s proof the original thesis failed and is being swapped out. The Colossus rental is a real win: real cash, real margin, real customer. And it quietly papers over the fact that the rental only exists because the Grok-training thesis never produced the demand that was supposed to fill the building. Win and failure, reported as a single piece of good news.

What It Produces

An economy organized around anticipation rather than cash flow throws off a few predictable things, and you can see every one of them in the AI buildout.

It produces buildouts bigger than realized demand can justify, because the capital is there, the narrative blesses the spending, and the running tally of tactical wins provides cover. It produces operators who are spectacular at managing the story and merely adequate at running the business, because the story is what gets rewarded. It produces externalities that land on people outside the game entirely — the neighborhood residents, who hold none of the equity and collect none of the fees and eat the costs whether the game rolls on or falls apart.

And it produces a public conversation that misreads the whole thing, because the business press’s instruments can’t tell the difference between an operator building something durable and an operator stacking tactical wins on a tower of failed strategic bets. From the outside the two look identical, right up until the chain snaps.

When does it snap? That’s one of the hardest questions in economics to answer with a straight face. Previous anticipation economies ran a surprisingly long time — the dotcom era roughly 1995 to 2001, the mortgage-securities era roughly 2002 to 2008, call it six years each. The AI era got going in late 2022; by that yardstick it has years left in it. Or it cracks next quarter, when some lab misses an earnings number, or a rate case rejects a hyperscaler’s interconnect, or a court rules the wrong way on one of the cases from the neighborhood chapters. Nobody knows. The honest answer to “when does this break” is: we don’t know, but the shape is familiar.

And when it does break, the damage falls on the late investors who bought at the top, the employees whose options go to zero, the public infrastructure committed to support it, and the communities that hosted the buildings — while the senior operators, paid the whole way through, generally walk away with their fortunes intact and go find the next thing.

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