BIS warns AI buildout is shifting from cash flow financing to debt, with private credit rising fast
- BIS Bulletin (Aldasoro, Doerr, Rees, Jan 2026) says AI-related capex, mostly in the US, now accounts for a substantial share of GDP growth and is projected to keep rising.
- Big tech firms historically funded IT investment from operating cash flows, but the report says AI capex needs are now so large that firms are shifting to debt financing, with private credit playing a rapidly growing role.
- BIS calls current macro and financial stability risk from the AI boom "moderate," but says the boom's sustainability depends on AI firms hitting high earnings expectations.
- The report flags that equity prices have run far ahead of debt market pricing for the same AI firms, a gap BIS treats as a warning sign of tension between stock market optimism and credit market caution.
- AI investment categories tracked include data center construction and equipment, IT manufacturing facilities (chips/hardware), other IT equipment, and software spending.
Hacker News 의견들
BIS already flagged AI financing sustainability as one of the biggest risks to the global economy back in their June report, this is basically the follow up.
Only two scenarios in Graph 2, high growth and medium growth. Isn't medium growth for the next 4 years actually the worst case anyone's willing to model?
That's kind of how financial reports work, they're written for people who fill in the blanks themselves. Sounds a lot like the pre-GFC ratings agency models that never priced in a national home price decline.
At this point anything below medium growth crashes the whole economy anyway, we'd have way bigger problems than a bad BIS report if that happens.
If growth doesn't materialize, this plays out just like the dot com bust, except this time earnings are the load bearing piece. If those fall, everything else crumbles.
The report is written from the perspective of people actually participating in the buildout. Anyone betting on sub-medium growth wouldn't be buying hyperscaler bonds right now regardless.
Other reports say this investment wave dwarfs the 2000 internet buildout and even the 19th century railroad boom, curious why the numbers vary so much between analyses.
In inflation adjusted dollars it's bigger, sure, but as a share of GDP it's actually pretty modest, comparable to the Apollo Program. The economy's just so much bigger now that 1% of GDP for a few years beats 10%+ of GDP for decades back then.
I'd rather see this capital deployed than sitting idle on corporate balance sheets, GDP growth is GDP growth.
We didn't have to burn it on hardware with such a short useful life though, this much money could've built the US high speed commuter rail instead.
Is this really investment or is it just money being set on fire?
I have yet to see real profit evidence from AI outside firms whose profit literally is the AI or its infra. Duolingo is the textbook test case and its stock is down 70% with flat profit despite going AI native.
Translation firms get paid for accreditation that a trained human did the work, not just accuracy, so AI doesn't obviously boost their margins much.
Costco example tracks: either AI commoditizes the product so users skip the wrapper and go straight to the LLM, or it just adds cost without a clear revenue win.
Ad tech AI (Google, Facebook targeting) has been printing money for decades, that's proof AI monetizes fine, just not in the chatbot form everyone's hyping now.