Canals in the 1830s. Railways in the 1840s. Dotcoms in the late 1990s. Each one was a real technological breakthrough. Each one also pulled in far more money than the eventual returns could justify. Now the Bank for International Settlements is putting AI on that same list — and warning that the unwind, if it comes, could be ugly.
The BIS, which functions as a kind of central bank for the world's central banks, used its latest annual economic report to flag what it called "AI exuberance" as one of the defining risks facing the global economy right now. Not because AI isn't real or useful — the report is fairly clear that it probably is — but because the amount of capital chasing it has detached itself from any realistic near-term payoff.
The numbers behind the warning
The scale here is hard to overstate. The five biggest "hyperscalers" — the small cluster of tech giants building the data centres, chips, and infrastructure that AI runs on — are expected to spend more than $1 trillion combined between 2025 and the end of 2026.
That spending has been funded partly through equity markets, where AI-linked stocks have been trading at record valuations, and partly through a wave of corporate debt issuance, with tech firms taking advantage of borrowing costs that are close to their lowest levels this century. SpaceX's enormous IPO earlier this month — reportedly valuing the company in the tens of billions — became something of a symbol for just how much investor appetite currently exists for anything touching AI infrastructure.
Three booms that all ended the same way
What gives the BIS report its weight isn't the warning itself — plenty of analysts have made similar points — but the specific historical comparisons it draws on. Each of the episodes below involved a genuine leap forward in technology. Each one also ended in a sharp reversal.
According to the BIS, the common thread running through all three is straightforward: a real technological breakthrough attracted capital in excess of what commercial returns could ever justify. The eventual correction wasn't just a market event — it dragged down the wider economy with it.
Why a correction today could hit harder
One detail in the report stands out: the BIS argues that an AI-driven market correction today could do more damage than past episodes, for a fairly simple reason. Households now hold a much larger share of their overall wealth in equities than they did during previous booms. A sharp drop in AI-linked stocks wouldn't just be a Wall Street story — it would flow more directly into household balance sheets than the dotcom crash did.
There's a second layer of risk sitting underneath that. Because so much of the AI build-out has been financed through debt rather than cash on hand, a disappointing set of returns wouldn't only hurt shareholders. It could also create losses for the bondholders and lenders who financed the spending in the first place — the kind of dynamic that turns a sector-specific wobble into something closer to a financial stability problem.
Investors are already nervous
The BIS isn't a lone voice here. Senior figures at major asset managers have made similar points in recent weeks, with one prominent investment chief at a large European insurer pointing to SpaceX's decision to launch a sizeable bond sale so soon after its IPO as a sign that AI-related markets have already drifted into bubble territory.
Markets have responded with noticeably more volatility since that IPO, compounded by growing expectations that the Federal Reserve may raise interest rates further — a combination that tends to punish highly valued, growth-dependent stocks first.
It's not just AI — energy is in the mix too
The report didn't isolate AI as the only pressure point. The BIS also pointed to lingering disruption from the near-closure of the Strait of Hormuz during the recent US-Iran conflict. Roughly a fifth of the world's oil and liquefied natural gas exports normally move through that corridor, and the resulting disruption is still feeding into inflation, which the BIS expects to remain elevated for some time.
Put together, the picture the BIS is painting isn't a single AI bubble in isolation — it's an economy carrying several pressures simultaneously: a possible AI investment bust, persistent energy-driven inflation, growing financial vulnerabilities, and government finances that are weaker than they'd like heading into all of this.
So is this 1999 all over again?
Not necessarily — and the BIS itself is careful not to claim that. It explicitly leaves room for AI to deliver real, lasting productivity gains over the next decade. The dotcom comparison isn't a prediction that AI is fake; it's a reminder that being right about the technology and being right about the price you paid for it are two very different things.
What the report is really doing is naming a risk that's been sitting in plain sight: an enormous amount of capital, much of it borrowed, betting on returns that haven't shown up yet. History suggests that bet either pays off faster than markets expect, or it doesn't — and when it doesn't, the unwind tends to be longer and messier than anyone wants it to be.
Either way, it's no longer a fringe concern. When the institution that advises the world's central banks puts a warning like this in writing, it's worth more than the usual market chatter.