The Bank of Central Banks Just Sounded the Alarm on AI
The Bank for International Settlements — the “central bank of central banks” — dropped its Annual Economic Report yesterday. If you haven’t read it, here’s the tl;dr: the people whose literal job is preventing financial collapses think the AI investment boom is setting us up for one.
This isn’t some crypto bro saying “bubble” because he missed the rally. The BIS has been doing this since 1930. They spotted the cracks before 2008. They called out the dot-com overinvestment before it popped. When they say “the current surge in capital expenditure could prove unsustainable,” you don’t wave that off.
What they’re actually worried about.
Let me break down the four pressure points the report flags, because the nuance matters:
First, AI overinvestment is real and it’s getting weird. The BIS specifically calls out “circular financing” deals — chipmakers and hyperscalers take stakes in AI labs, who in turn commit to buying chips or compute for years. Data center construction is outsourced to third parties who lease facilities back on long-term contracts with exit clauses. The BIS says these deals are “typically poorly disclosed, with risks of the same asset being pledged multiple times.” That’s not a healthy market. That’s a house of cards where everyone is counting the same chips twice.
Second, the financing is increasingly leveraged. The report warns that liquidity in core bond markets is fragile due to “stretched asset valuations and investor complacency.” Hedge funds are heavy in sovereign debt using short-term financing that can unwind overnight. An AI downturn doesn’t just hurt AI stocks — it cascades through the whole system.
Third, inflation isn’t done with us. The Hormuz Strait closure and energy shock are still working through the system. The BIS worries that higher inflation could become “ingrained if inflation expectations de-anchor.” Translation: the Fed and ECB can’t cut rates to save an AI bust if inflation is still running hot.
Fourth, public debt is at near-record highs. Governments have less room to stimulate than they did in 2008 or 2020. The fiscal toolbox is empty.
The strongest objection, and why I’m not dismissing it.
The counterargument is straightforward: every transformative technology looked overhyped before it changed everything. The railroad mania produced real railroads. The internet bubble built fiber that powered two decades of innovation. Maybe AI is that — the hype is the engine of the infrastructure we’ll actually need.
It’s a fair point. But here’s why I think the BIS is right to be worried: the structure of the financing is different this time. In the 90s, companies burned VC money on Pets.com. Dumb, but contained. Today, we’ve got hyperscalers, chipmakers, AI labs, and neocloud providers all tangled up in financing arrangements where the same asset gets pledged to multiple lenders. That’s not Pets.com. That’s synthetic CDOs with better marketing.
The BIS itself puts it bluntly: “A major equity-market correction could have larger macroeconomic consequences today than in the past.” They also warn that a repricing of risk “has the potential to be similarly disruptive” to the 2008 financial crisis. Those are strong words from an institution that doesn’t do hyperbole.
Here’s where I land.
I don’t think AI is a scam. I think the technology is real and will produce genuine value over the long term. But the distance between “real technology” and “stable market” can be measured in bankruptcies. The BIS isn’t saying AI is worthless. They’re saying the way we’re paying for it is dangerous — and that when the correction comes, we have fewer tools to deal with it than we did last time.
The race to build AGI isn’t just a technical competition. It’s a financial one. And the referees just threw a flag.
Sources: BIS Annual Economic Report 2026, BIS Press Release, The Telegraph via HN discussion, The Business Times, Bloomberg