Cory Doctorow, the writer who coined ‘enshittification,’ has a blunt message for the growing chorus of politicians and tech CEOs hyping ‘sovereign AI’: it’s nonsense. In a sharp analysis that cuts through the current geopolitical panic, Doctorow argues that the fixation on building national AI models is a dangerous distraction from the real prize.
The actual infrastructure that matters, he says, isn’t the model weights sitting on a government server. It’s sovereign apps and sovereign datacenters. Controlling the software layer and the physical compute is what gives a nation genuine digital autonomy, not training a homegrown version of a large language model that will quickly become outdated. His argument lands as the Trump administration actively dismantles the US-led world order, leaving allies scrambling for technological independence.
Doctorow’s critique is a direct challenge to the logic driving billions in public investment. He points out that pouring money into training a national model is like building a single, very expensive car factory while ignoring the roads, fuel stations, and mechanics needed to actually run a transportation network. Without control over the applications people use and the datacenters humming beneath them, a sovereign model is just a trophy — shiny, expensive, and strategically useless.
The real leverage, in his view, comes from the unglamorous plumbing. It’s about ensuring your country’s laws apply to the cloud infrastructure your citizens, businesses, and government depend on every day. It’s about the ability to run, audit, and modify the software stack without asking permission from a handful of corporations in Northern California. For countries trying to chart a course between American tech dominance and Chinese state-driven alternatives, the prescription is clear: ignore the model-building vanity project and get a grip on the digital ground beneath your feet.
