The collapse of AI hype won’t free us. It’ll bury us.

There’s a persistent fantasy among reformers that when the generative AI bubble pops, we’ll be left with a surplus of infrastructure we can reclaim. Data centers. GPUs. Purpose-built tooling. The thinking goes: if Big Tech falls, maybe we can use their scraps to build something useful.

But that’s not how collapse works. The cloud won’t fall into our hands. It will fall into the waiting arms of the carceral state, the defense industry, the surveillance supply chain.

When AI’s promises no longer attract capital, they’ll attract procurement. Inference pipelines won’t be mothballed, they’ll be nationalized. Whatever isn’t privatized will be securitized. The data centers will hum. The tooling will stay. The models will be repurposed for compliance, for borders, for predictive profiling.

We will be told it’s pragmatic. That it’s better for these tools to serve government than disappear entirely. That they’ve already been built, so we might as well use them. But “use” in this context means control. It means surveillance. It means digitized punishment.

This is not the end of AI. It is the end of illusions about what happens next.

Collapse does not create voids. It creates vacuums. And power rushes in.

There will be no civic reclamation of AI’s ruins. Not without a fight. Not without a fundamental rethinking of what infrastructure is for and who it serves. Because this was never just about automation. It was about legitimacy. Bureaucratic authority. The performance of order.

And when it lands, it won’t be on the billionaires who built it. It will land, as it always does, on the people least protected by the systems it was sold to replace.

Drafted while listening to Ryuichi Sakamoto — Solari