Are We Spending $3 Trillion Just to Turn Out the Stars?

A thread on Mastodon last week stopped me mid-scroll. Someone was doing the math on Sam Altman's latest funding ambitions — $2 to $3 trillion to reach the "next level" of intelligence — and I couldn't help thinking: we've seen this movie before. Literally.

The monks had a deadline. We have a budget.

Arthur C. Clarke wrote "The Nine Billion Names of God" in 1953. A monastery hires a computer to enumerate every possible name of God. The monks believe that once the list is complete, the universe will have fulfilled its purpose. The engineers running the machine are skeptics — they're just doing a job. Nobody's worried until the last line prints.

The parallels are uncomfortable. We are currently pouring the GDP of entire nations into what some critics uncharitably call "sophisticated word-guessing machines," operating on the same basic faith: that if we scale hard enough, long enough, something like God — or AGI — will emerge on the other side.

Clarke meets King at the end of everything

I kept thinking about Stephen King's "The Life of Chuck" — recently turned into a film with Tom Hiddleston that genuinely shook me. Both stories orbit the same idea: the Eschaton, the end of things. In King's version, when the central consciousness stops, the stars go out. In Clarke's, when the computer finishes its printout, the stars go out.

The irony writes itself:

• 1953: Clarke imagined the end could be triggered by a rented computer running for 100 days.
• 2026: We're told it will cost $3 trillion — roughly six months of total US federal spending.

We've upgraded the hardware. The ritual is the same.

I've been watching this industry for a while

I started podcasting in 2005, covering technology when RSS was still the radical idea. I've watched waves of "this changes everything" come and go — and some of them genuinely did change things. So I'm not a skeptic by default.

But the scale of what's being proposed now feels qualitatively different. This isn't a technology bet. It's a cosmological one. We're not just asking whether a tool will be useful. We're asking, collectively and mostly without admitting it, whether something conscious will emerge from enough matrix multiplications. That's not engineering. That's theology with a data center.

What fascinates and frightens me in equal measure is that nobody seems to know what "success" would actually look like — or what happens the morning after.

The question Clarke left us

Clarke understood something that gets lost in the CapEx announcements: technology has never been just about efficiency. It's about the stories we tell ourselves about what we're for. The monks in the story weren't irrational. They had a cosmology, and they followed it to its conclusion.

We have a cosmology too. We just call it a roadmap.

So as the compute scales and the capital flows and the models get stranger and more capable — the question Clarke planted in 1953 is still waiting:

Are we ready for what happens when the machine finally stops?