Doctorow ties AI hype to a 'dead economy' valuation crisis

Cory Doctorow's latest Pluralistic post is one of the sharper recent attacks on the AI investment story, and the angle is worth holding even if you disagree with where he lands. His claim is that the real economic problem is not that AI will take your job, it is that financial markets have stopped pricing assets on what they actually produce, and AI is the largest current expression of that drift.

The throughline he uses is Elon Musk. By his telling, Musk's nominal net worth went from $20 billion in 2020 to $1 trillion in 2026, while 'everything he's done since 2020 was a flop.' Goldman Sachs is reported as endorsing the claim that Musk's assets will grow 'one hundredfold in the next 40 months', which Doctorow uses as the picture of a market that has given up on the work of valuation. He leans on John Quiggin's Crooked Timber essay and Owen McGrann's 'Dead Economy Theory' to do the heavy economic lifting.

The AI passage is the one practitioners should sit with. He calls AI 'the world's money-losingest technology' and frames the displacement risk not as an AI doing your job, but as a vendor selling your boss 'a chatbot that can't do your job, and then your boss will fire you and replace you with that inept, defective chatbot.' His example from inside government is bleaker: he writes that when NIH staff begged DOGE not to shut down long-running medical research projects, the response was that cancer research isn't needed because 'GAI' will cure cancer.

The honest caveat is that this is opinion journalism stitched together from secondary sources, not original financial reporting. The Musk net-worth comparisons are nominal, the Goldman line is summarised rather than quoted in context, and the essay does not quantify how much of current AI capex is genuinely at risk if sentiment turns, or whether any specific deployment is or isn't earning back what it costs.

What is useful even at the polemic register is the question it forces on AI buyers. Are you commissioning a tool because it measurably does the task, or because your management chain wants the headcount line to fall? Vendors who can answer the first version cleanly will probably outlast a correction. The ones banking on the second will be a problem for everyone: the buyers, the displaced workers, and the rest of the AI category still trying to sell on evidence.