Doctorow calls sovereign AI 'nonsense' — the real fight is over apps and datacenters

Cory Doctorow isn’t buying the sovereign AI hype. The writer and activist, famous for coining ‘enshittification,’ dismissed the entire concept as ‘nonsense’ in a sharp analysis of what nations actually need as the U.S.-led world order frays under Trump. He argues the fixation on building national AI models misses the point entirely. What countries really require is control over sovereign applications and the physical datacenters that run them. Without that, you’re just renting someone else’s intelligence.

Doctorow’s argument flips the current AI arms race on its head. While governments scramble to fund homegrown LLMs, he points out that the real dependency isn’t on model weights — it’s on the software stacks and server racks owned by a handful of American cloud giants. You can train a model in France or India, but if it’s served on AWS or Azure, your digital sovereignty is a polite fiction. The infrastructure layer is where lock-in happens.

The timing is brutal. Trump’s dismantling of traditional alliances has spooked allies into reevaluating their tech dependencies, and AI has become the shiny object everyone wants to own. Doctorow is telling them they’re chasing the wrong thing. A national LLM is a trophy. A national datacenter running open-source apps is a fortress. The difference matters.

It’s a classic Doctorow take — cutting through the marketing to look at the plumbing. And he’s probably right that no one wants to hear it. Politicians love cutting ribbons on AI institutes. They’re less excited about the unsexy work of laying fiber and writing procurement rules that don’t funnel money back to Seattle. The real question is whether any of them will listen before the next crisis hits.