Source: BIS Bulletin No. 130 — Aldasoro, Gambacorta, Kharroubi, Rottner — 2026-07-28
Summary
The Bank for International Settlements published a staff bulletin arguing that the AI investment boom is simultaneously boosting economic supply (productivity gains, data-center and chip capital expenditure) and demand (investment spending, wealth effects from AI-driven equity gains), which blurs the cyclical signals central banks rely on to judge whether an economy is running hot or cool. AI-related capex — data centers, semiconductors, related infrastructure — now runs around 1% of GDP in the most AI-exposed economies, increasingly financed through public debt markets and private credit, while the eventual productivity payoff remains large but uncertain and unevenly distributed across countries and sectors. The bulletin's conclusion: this raises real risk of monetary policy missteps in either direction — too loose or too tight — because policymakers can't cleanly separate an AI-driven supply shock from an AI-driven demand shock.
Key Takeaways
- Central banks set interest rates by reading signals like inflation, output gaps, and investment trends — AI's dual effect on both supply and demand at once makes those signals noisier and harder to interpret correctly in real time.
- ~1% of GDP in AI-related capex, in the most exposed economies, is a large enough number that it isn't a rounding error in growth and investment statistics — it's now a macro-relevant force that official monetary policy analysis has to explicitly account for.
- The financing detail matters: AI capex increasingly running through public debt markets and private credit means a slowdown or repricing in AI investment wouldn't just hit tech stocks, it could hit credit markets central banks watch closely.
- This is one of the first instances of an official central-bank research institution formally naming AI as a monetary-policy confounder — a step up from market commentary or think-pieces into the kind of analysis that can actually shape rate-setting discussions.
Reel Script
Hook (~16s, 36 words): The Bank for International Settlements just said something most people missed: the AI boom might be making central banks more likely to get interest rate decisions wrong. Here's the mechanism.
Core Concept (~70s, 160 words): Central banks set interest rates by trying to read two signals: is the economy growing because supply capacity is expanding, which is healthy, or is it overheating because demand is outrunning supply, which risks inflation and usually calls for raising rates. Normally those two forces move somewhat independently, so policymakers can tell them apart. AI breaks that separation. On one hand, AI capex — building data centers, buying chips — shows up as a demand-side spending surge, the kind that would normally signal an overheating economy. On the other hand, if AI eventually delivers real productivity gains, that's a supply-side expansion, the kind that's disinflationary and healthy. Both effects are happening from the exact same underlying investment wave, at the exact same time, and a central bank looking at aggregate data can't cleanly tell which one is dominant — so a rate decision calibrated for one risks being wrong for the other.
Hands-On (~50s, 115 words): The flow worth diagramming: one box labeled "AI capex boom" with two arrows coming out of it. One arrow goes to "demand-side effects" — investment spending, wealth effects from AI stock gains, terms-of-trade shifts. The other arrow goes to "supply-side effects" — productivity gains, though uncertain and uneven across sectors and countries. Both arrows converge into one box: "blurred cyclical signal," which feeds into "central bank policy calibration risk." Underneath, drop in the one hard number: AI-related capex is running near 1% of GDP in the most exposed economies, increasingly funded through debt markets rather than cash.
Takeaway (~22s, 48 words): The verdict here is uncomfortable but real: AI's economic effects are big enough now that getting monetary policy wrong because of them is a genuine risk, not a hypothetical. If you're in finance, watch central bank language for explicit AI caveats — this bulletin suggests they're coming.