Fed’s Williams says AI is reshaping supply in ways not fully understood

New York Fed President John C. Williams has a new item on his list of things keeping inflation elevated, and it is not oil or tariffs. It is artificial intelligence.

Williams said AI is influencing supply in ways that are not yet fully understood. For a central banker whose job depends on reading the economy accurately, that admission matters.

The comment follows a run of late September 2026 remarks. In them, Williams described AI-driven demand as an inflation force the Fed has to watch closely.

A race between supply and demand

The core issue is simple. AI buildouts need specific physical goods: chips, power equipment and the inputs that go into data centers. Demand for those goods is rising faster than producers can make them.

Williams called the situation a race, and he framed it in his own words:

“A race between available supply and surging demand.”

In his September 29, 2026, remarks, Williams named three forces behind inflation. Tariffs were the first, though he said they are no longer having an impact. The second was energy and commodity pressure tied to the Middle East.

The third was AI-related demand for particular goods and services. Williams described the inflationary effect of that AI demand shock as increasingly salient.

Four days earlier, on September 25, 2026, he made a related point. Williams said the Fed cannot afford to look past supply shocks that persist. He also noted that AI’s effect on productivity is not yet visible in the data.

The numbers behind the AI spending boom

The scale of the spending is large. Williams said AI capital expenditures are running in the hundreds of billions of dollars each year.

He also suggested that growth leaned heavily on that spending. Without the AI investment, Williams said, growth would have been approximately one-third weaker during certain periods of 2026.

There is a catch, though. Williams pointed out that much of the AI-related investment is offset by imports. When companies buy chips and equipment from abroad, the money flows out, which limits how much the spending adds to US GDP growth.

On the inflation outlook, Williams laid out a gradual path. Inflation is forecast at 3.5% for 2026. It is expected to fall to just above 2% in 2027 and reach the Fed’s 2% long-run target by 2028.

That path comes with conditions. It depends on energy prices stabilizing and on AI supply and demand moving into better balance. Williams also acknowledged significant uncertainty about how large these inflationary forces are and how long they will last.

Background: a double-edged sword

Throughout 2026, Williams has described rising AI investment as cutting both ways. On one side, the spending fuels demand and supports economic growth. On the other, it creates ongoing supply-side pressure that could make monetary policy less effective if it becomes entrenched.

Supply shocks are a tricky problem for central banks. The Fed’s main tool, interest rates, works by cooling or heating demand. It cannot build a chip factory or add capacity to a power grid.

He has also warned that AI’s expected productivity gains may take substantial time to materialize.

What this means

For markets, the most direct takeaway concerns interest rates. If a senior Fed official sees AI demand as a persistent inflation driver, that may make the central bank more cautious about easing policy quickly.

A forecast of 3.5% inflation for 2026 sits well above the 2% target. Williams’ expectation of a return to target only by 2028 implies a slow glide rather than a quick landing.

Williams tied part of the inflation picture to Middle East energy and commodity issues, and his forecast assumes energy prices stabilize. AI data centers are power-hungry, so energy costs and AI demand are not fully separate stories.