Chicago Fed President Austan Goolsbee dropped a logic bomb in a recent speech that sent shivers through markets:

If AI actually succeeds in transforming productivity, the Fed might need to raise rates — because it could overheat the economy. If AI ends up disappointing everyone — we're looking at stagflation.

Sounds like nonsense. Overheat and stagflation? The Fed loses either way?

But crack open his reasoning, and you'll see he's not talking about winning or losing. He's talking about the one thing the Fed fears most: uncertainty.

Two Variables

The Fed's decision logic is simple — it tracks two things: inflation and employment.

AI impacts both through two very different transmission paths.

Path A: AI Boosts Productivity, Economy Overheats

1. AI drives major productivity gains
2. Companies expect higher profits, invest heavily in AI infrastructure
3. Hiring surges for AI talent, pushing up wages
4. Consumer confidence rises, aggregate demand explodes
5. Productivity gains can't keep up with demand growth
6. Result: inflation + rate hikes needed to cool things down

The key issue here: AI's "supply expansion" and "demand stimulus" are not synchronized.

Demand gets stimulated first — companies raise capital for GPUs, hire AI engineers, launch AI products. Productivity gains (supply expansion) take time to materialize — organizational change, process redesign, human-AI collaboration workflows. None of this happens by installing ChatGPT.

So in the short term, demand runs faster. That's a classic "overheating" signal for the Fed.

Path B: High AI Hopes Fail to Deliver, Stagflation

1. Companies invest heavily in AI, but productivity doesn't materialize
2. Costs go up (GPUs, electricity, AI talent salaries)
3. Revenue and efficiency gains lag behind
4. Companies start cutting costs and laying off staff
5. Economic growth slows, but cost-pushed prices stay high
6. Result: stagflation — stagnant growth + high inflation

This path brings back an old friend: the Solow Paradox.

In 1987, Nobel laureate Robert Solow famously said: "You can see the computer age everywhere but in the productivity statistics."

In 2025, Fortune reported that thousands of CEOs admitted AI had no impact on employment or productivity. Economists had to dust off the 40-year-old paradox to make sense of the current situation.

AI could replay the early internet's awkward phase — everyone knew it would change the world, but productivity curves barely budged for a decade.

But this time, the stakes are higher. AI's investment cost dwarfs the PC internet era. A GPU cluster costs billions. Electricity bills are tens of times higher than traditional data centers. If all this investment yields only "efficiency improvements" instead of "fundamental transformation," corporate balance sheets will bleed red.

Goolsbee Isn't Alone

San Francisco Fed's Mary Daly, in the same speech series, put it bluntly: "The Fed is studying AI's economic impact, but the uncertainty is too great — policy needs to wait for more data."

Translation: We don't know what to do yet. Let's watch.

Fed Governor Barr poured cold water directly: don't expect AI to lead to faster rate cuts.

A PIIE working paper walks the same line: AI's impact on inflation could go either way — pushing it up or down — depending on whether "demand stimulus" or "supply release" moves faster.

Two Extreme Predictions, One Deep Contradiction

                AI at Scale
                    │
          ┌─────────┴─────────┐
          ▼                   ▼
    Demand Booms          Costs Go Up
    Supply Lags           Efficiency Lags
          │                   │
          ▼                   ▼
      Inflation ↑          Stagflation ↑
    Rate Hikes Needed     Nowhere to Go

Both predictions share the same premise: the Fed cannot accurately forecast AI's deployment pace.

Goolsbee isn't saying "AI is bad." He's saying: we might face a scenario where monetary policy becomes completely ineffective.

If AI succeeds but overheats first, rate hikes are correct (suppress demand). But if AI's success is false and you hike before productivity materializes — you kill growth.

On the flip side, if you hold rates waiting for productivity that never arrives — stagflation.

This is the Fed's two-sided risk.

What This Means for Chinese Companies

Lesson 1: Don't let the "AI productivity revolution" narrative own you

The Solow Paradox from 40 years ago still stands. Companies investing in AI must do the math — it's not "invest in AI = competitive." It's "how do we invest in AI in sync with organizational processes and talent structure to actually improve output?"

Lesson 2: Global interest rates will be more volatile because of AI uncertainty

If the Fed delays rate cuts (or even keeps the hike option on the table) because of AI, it directly impacts Chinese companies' overseas financing costs and currency risk exposure.

The real risk isn't AI success or failure. It's that the entire global economy is making decisions against a giant question mark — one whose answer may take a decade to arrive.

References

  1. Fed's Goolsbee: AI success would be 'lovely,' but Fed would still need to watch for overheating (Reuters/Yahoo Finance, 2025-05-09)
  2. Fed's Goolsbee: AI could produce stagflation if boom disappoints (Barron's, 2025-05-09)
  3. Thousands of CEOs admit AI had no impact on employment or productivity (Fortune, 2025-05-09)
  4. SF Fed Daly: The AI Moment? Possibilities, Productivity, and Policy (2025-05)
  5. Fed's Barr casts doubt on AI as rate-cutting tool (Reuters, 2025-05)
  6. The AI productivity boom is not here (yet) (The Economist, 2025-05-01)