The AI industry shows classic signs of an unsustainable bubble. Capital spending on data centers and chips may require roughly $2 trillion a year in revenue just to break even, yet no credible forecast gets anywhere close to that figure — and the hardware itself needs replacing every five years or sooner. Making matters riskier, a handful of tech giants, chipmakers, and AI firms are investing in and buying from one another, creating a circular economy where one failure could cascade. Much of this is financed with debt rather than equity, while public resistance to new data centers grows and corporations increasingly hesitate to fully adopt AI tools given their real costs. Unlike most technologies that get cheaper at scale, AI models keep demanding more resources as they grow — a potentially fatal flaw. Credit agencies like Moody’s have already flagged the risk.
A crash, whether sudden like the 2000 dot-com collapse or a slower unwinding, would likely wipe out trillions in wealth, halt data center construction mid-project, and leave utilities, fiber vendors, and communities holding stranded investments and higher rates.
Yet history suggests such a reset could ultimately help the technology mature. Just as the telecom crash weeded out overleveraged competitors while enabling long-term industry growth, an AI shakeout could force surviving companies toward genuine efficiency and sustainable economics — qualities the current field, flush with capital but light on cost discipline, has yet to prove it possesses.