The real (economic) AI apocalypse is nigh (pluralistic.net)

Cory Doctorow argues we’re staring at an economic AI apocalypse driven not by runaway superintelligence but by a speculative bubble: a handful of monopolists have exhausted organic growth and are convincing investors their losses are justified by a pivot to AI. That investor story depends on replacing workers with costly “foundation models” and repurposing remaining staff as “humans in the loop.” But unit economics are dire—each generation of models and GPU-heavy data centers costs more, GPUs are being used as loan collateral despite rapid obsolescence and failure during long training runs, and circular accounting (e.g., Microsoft/OpenAI tokenized server credits) inflates financials. The WSJ and industry estimates show the sector would need trillions in revenue to justify current investments; other research finds AI has produced little measurable gain in worker earnings or productivity. The significance is systemic: concentrated exposure to a rotten growth narrative could trigger a rapid withdrawal of capital, mass unemployment of workers who’ve already been displaced, and huge social and fiscal costs. Doctorow’s solution is to “puncture the bubble” and prepare institutions for the fallout—governments to consider job guarantees, universities to absorb skilled talent, and the community to focus on cheap GPUs, open-source models, and pragmatic deployments rather than hype. The takeaway: AI is a powerful set of tools, but today’s economic model around it is unsustainable and poses broad macroeconomic risk.