AI gains may lift neutral rate, complicating Fed policy

The Fed says AI-driven productivity could lift the neutral rate. Sounds neat. But the chain from code to higher policy rates is full of hinges, loose screws, and assumptions no central banker quoted in the Bloomberg piece bothered to tighten.

Start with this: the story isn’t crazy.

If AI really does raise trend productivity, you’d expect higher potential output, stronger underlying growth, and, yes, some upward pressure on the neutral rate. That’s the clean model version. Economists love that version because the math works. Markets love it because it sounds like a tech boom without the hangover.

The problem is everything that happens between the PowerPoint slide and the policy meeting.

Productivity ≠ Instant Rate Lift

Productivity is slippery. AI can boost measured output in data-heavy services; it can also substitute for labor or reassign work across firms. Those outcomes don’t have the same effect on aggregate demand. If AI raises output while shrinking payrolls, consumption may lag even as output per hour improves. A central bank that treats a headline gain in productivity as unambiguously disinflationary, or as a clean signal of higher neutral rates, is reading one line of the story and skipping the rest.

Measurement is just as messy. Productivity stats capture output per hour — but not all quality improvements or new services show up neatly in GDP. If AI creates better products at the same price, official numbers may understate gains; if it quietly automates back-office tasks, they may overstate them by ignoring displaced workers’ lost pay. The neutral-rate calculation rests on a view of long-run real growth, and the data series feeding that view are noisier than the technocratic tone suggests.

Then comes timing. Big technology shifts almost never deliver a smooth glide path. Think about the early internet or corporate IT: a burst of investment and hype, years of uneven adoption, then a slow burn of diffusion. Monetary policy, meanwhile, works with lags and feedback loops. The Fed could react to early AI-driven productivity signals, misread their persistence, and end up too tight in a short-lived surge or too loose if the gains fizzle.

Here’s what they won’t tell you: “AI will lift the neutral rate” is not a fact, it’s a stacked forecast dressed up as a policy anchor.

Convenient, isn’t it.

Follow the money — and the missing people

If you want to test the neutral-rate story, follow the money. Who actually captures the AI gains — capital or labor?

Right now, big tech platforms, cloud providers, and a small circle of AI infrastructure firms are positioned to hoover up productivity rents. If those gains accrue mostly to capital owners who save a large share of their income, the global supply of loanable funds could rise. More savings chasing relatively scarce safe assets is the classic recipe for lower equilibrium rates, not higher ones.

Corporate profit growth concentrated in a handful of firms also distorts investment dynamics. High returns at the top of the market — in a few hyperscalers and AI chip makers — do not automatically translate into broad-based capital spending. If AI amplifies monopoly-style profits, the result may be more retained earnings and stock buybacks, not the kind of economy-wide investment that would absorb savings and push equilibrium rates higher.

There’s a recent echo here. During the last big tech wave, Silicon Valley sold a story of permanent high growth and transforming productivity. We did get real gains — faster communication, leaner supply chains, new business models — but we also got concentration, uneven wage effects, and long stretches where measured productivity refused to match the rhetoric. Central banks learned the hard way that tech narratives arrive years before reliable data.

Now add the international angle. The neutral rate is global; capital flows and foreign demand matter. If AI raises productivity in the United States more than abroad, capital could pour into U.S. assets, pushing down yields even as domestic growth looks stronger. Or foreign central banks could move faster or slower than the Fed in responding to their own AI stories, creating interest-rate differentials that swamp any clean, model-driven “neutral” level. Follow the money — and you see a knot of cross-border flows that makes a neat domestic narrative much messier.

Then there’s political economy. Policy choices will shape how AI affects demand. Tax policy, retraining programs, unemployment insurance, regional support — these determine whether displaced workers become new consumers in different sectors or long-term slack on the sidelines. The Fed can only set rates; it cannot legislate a labor-market safety net or force Congress to share AI gains with households instead of balance sheets. Yet the headline claim in the Bloomberg piece floats above all that, as if distribution and fiscal choices sit in a separate universe.

A tidy counter — and its weak joints

You could argue — and plenty of optimists do — that AI’s productivity boost will be large, fast, and broad enough to raise potential output decisively; that firms will invest, workers will be reskilled, consumption will adjust, and the neutral rate will climb with the new growth path. That’s a tidy counter to my caution.

Here’s why it doesn’t erase the concern: that scenario requires a string of favorable outcomes across measurement, distribution, investment, international spillovers, and politics. Each link is fragile. The history of past tech booms is not a record of seamless transitions; it’s a record of overshoot, mispricing, and policy scrambling to catch up.

The Bloomberg piece captures a real possibility. But possibility is not destiny, and a central bank’s quiet embrace of the AI narrative is also a bet on who gets the gains, how fast they spread, and whether the political system can turn code into broad-based consumption instead of concentrated paper wealth.