The AI Bubble Eats Itself
The AI boom is the biggest bubble in modern history, and when it bursts, the choice will be stark: austerity for AI slop, or something better.
Seventeen times the size of the dot-com boom. Four times the subprime bubble. The biggest speculative rush in living memory. Built on a machine that can’t add two and two without seeing seven.
The hype is not really about intelligence. It’s about faith. The old capitalist prayer that this time, finally, profit can be conjured without labour. The worker can be subtracted, the body erased, and value will still appear on the balance sheet. Marx had a word for this: fetishism. Strip away the sermons about “AI revolution” and you see what it really is. A last roll of the dice for a system that has run out of ideas.
The dot-com boom sold connection; subprime sold homeownership. AI sells intelligence itself. Not the messy kind embodied in people, but a shimmering simulation in silicon. Each bubble promises emancipation, and each ends by tightening the grip of capital. When it bursts (and it will) the question is not if but how the wreckage will be gathered up and who will pay for it.
The AI boom is not the future. It is the sound of the system eating its own tail.
Scale without purpose
AI evangelists repeat the word scale like a mantra. More data, more compute, more GPUs. But scale without purpose is just excess. The bigger the model, the bigger the bill — in energy, in water, in minerals ripped out of the earth. One training run can drain a reservoir, scorch a power grid, or suck up the labour of thousands of poorly paid data-labellers in Nairobi or Manila.
The miracle is always somewhere else. The cloud is not in the sky; it is in the desert, where servers hum and rivers dry. The intelligence sold as immaterial turns out to be one of the most extractive industries on earth.
The cloud is not in the sky; it is in the desert, where servers hum and rivers dry.
The priests of the boom
Every bubble has its clergy. Altman talks like a man entrusted with humanity’s destiny. Musk prophesies between rocket launches and X tantrums. Politicians trail behind them, eager to bless the machine with subsidies and contracts. Trump calls for “patriotic AI.” Starmer calls for “AI for growth.” Different words, same creed: bailouts dressed as strategy.
But can states really bail this one out when it goes pop? That is the gamble. After 2008 there were banks to rescue, mortgages to refinance, assets that could be bought up and resold. With AI, what’s left when the hype collapses? Empty data centres, overvalued chips, half-finished models nobody wants to license. You can’t repackage hallucinations into bonds. The more this frenzy swells, the less salvageable it becomes — and the more exposed the state looks when it is asked to underwrite the wreckage.
And what would that underwriting look like? Another round of austerity (services gutted, wages frozen, communities hollowed out) all so the state can bail out AI slop. Ask yourself if that’s really the future you want. Because that’s what it means when politicians talk about “AI for growth” or “patriotic computing”: not some dazzling new era of prosperity, but your children’s school stripped bare, your hospital waiting list stretched even longer, your pay packet eroded, all in the name of keeping alive a technology that can’t tell the difference between truth and noise.
When the bubble bursts
The financial fallout will be spectacular. The ideological collapse may be more profound. For two decades, digital capitalism has survived on the story that code equals progress, that data equals destiny. Strip away the hallucinations and you see the truth: what keeps capital afloat is not intelligence but belief. The conviction that some new machine will deliver us from the contradictions of the old order.
Machines don’t make value. People do.
But machines don’t make value. People do. And a capitalism that tries to expel people from the process ends up cannibalising itself. The AI bubble is not the future. It is the sound of the system eating its own tail.
What comes next
History tells us what happens after collapse. Dot-com gave us platforms; subprime gave us austerity and surveillance capitalism. Post-AI, the likeliest future is data feudalism: computation owned by a handful of firms, rented back to the rest of us as privilege. Knowledge, once social, enclosed and rationed.
Unless labour intervenes. Unless movements insist that the means of computation — data, chips, energy, infrastructure — are social goods, not private monopolies. Otherwise, the next chapter will be crueller than the last.