When the Bank for International Settlements speaks, it is worth noticing who is talking. The BIS is the bank that central banks use, and in the mid-2000s it warned, correctly, that a build-up of property loans in the US could trigger a financial crisis. This week, in its annual report, it turned that same wary eye on artificial intelligence. The verdict, as reported by The Sydney Morning Herald, was blunt: the AI investment boom now powering the US and global economies bears an uncomfortable resemblance to past "manias" that ended in tears.
The bank's list of precedents is pointed. The canal-building surge of the 1830s. The "electrification exuberance" of the 1920s. The dotcom boom of the late 1990s. Each promised to remake the economy, each drew a flood of investment, and each, in the BIS telling, "ended with an eventual reversal in investment, inducing economy-wide recessions." The scale today is hard to overstate: five US companies, including Microsoft, Meta and Amazon, are forecast to spend around a trillion dollars on AI infrastructure over the next year alone. If those bets do not pay off, the bank warns, "disappointment in returns could trigger a sudden pullback in financing and turn the capex boom into a protracted investment bust."
There is a second worry stacked on the first, and it is about jobs. The optimists' standard comfort is that every past technology destroyed some work but created more. The BIS is not convinced the comfort applies here. "Unlike past general purpose technologies, AI competes directly with human cognitive abilities," it cautions, "possibly narrowing the scope for workers to move up the value chain." It concedes that large-scale labour displacement has not yet happened, but it sees "signs of possible adjustments to come," with analysts reporting more companies planning to swap workers for automation.
Set against this gloom is a very different reading of the same facts. In a report titled, with no subtlety, "Boomsday Not Doomsday," the group Unleash Prosperity argues AI could deliver the single greatest productivity revolution in American history, as long as Washington does not smother it with premature regulation. Co-founder Stephen Moore reaches for the same historical analogies the BIS uses, but flips them: every major invention of the past century, he says, ended up creating more jobs, not fewer. In 1900 nearly 40 percent of Americans worked in agriculture; today it is under 2 percent, and the country grows far more food. "The race is on," Moore says. "Let's win the race."
Both camps cannot be right, and the gap between them is really a gap about timing and returns. The BIS does not deny AI could be transformative; it doubts the money being spent will earn its keep before the financing mood turns. The optimists do not deny the risk; they argue the upside dwarfs it. Lurking underneath is a quieter charge raised elsewhere this week, that some firms are "AI washing," dressing up ordinary cost-cutting as a technological revolution to please shareholders. That matters here, because a boom built partly on narrative is exactly the kind that reverses fastest. The useful question for the rest of 2026 is not whether AI is real. It is whether the spending is.