For most of the past two years the AI build-out has been discussed as a capital allocation question. Federal Reserve officials have started discussing it as a price stability question, which is a different and more consequential thing.
Governor Lisa Cook raised the issue in remarks at the Exchequer Club of Washington, and New York Fed president John Williams has flagged it separately, according to Yahoo Finance. The concern also appears in the minutes of the Fed's June policy meeting.
A note on what that does and does not mean. These are individual officials and a theme in the minutes, not a formal position of the Federal Open Market Committee. Cook voted to hold rates steady in June. This is the Fed thinking aloud rather than the Fed acting.
The mechanism
The argument is straightforward. Companies have announced more than $1.5 trillion of AI data centre plans. Building and running those facilities requires memory chips, electricity and copper, in quantities large enough to move the price of all three.
That is a demand shock arriving in inputs that feed directly into consumer goods. This is not a story about the price of GPUs, which most households never buy. It is a story about the components inside things they do buy.
The evidence is already visible. Apple has raised laptop and iPad prices by hundreds of dollars, and Xbox console prices are rising by $100 to $150 from August 1, with memory costs cited in both cases. Memory price increases are working through laptops, tablets, consoles, smartphones and cars, the last of which now contain a great deal of memory.
Readers who followed our reporting this morning on the memory supply squeeze will recognise the mechanism from the other end. High-bandwidth memory for AI is taking an increasing share of wafer capacity, leaving less for the conventional memory that goes into consumer devices. What looked like a semiconductor industry story is now turning up in the inflation data.
Why this is awkward for the Fed
Inflation is running at 3.5 percent year on year, against the Fed's 2 percent target.
The difficulty is not the level but the character of the pressure. Central banks are generally comfortable looking through supply shocks they judge to be temporary, on the reasoning that raising rates cannot conjure more oil or unblock a shipping lane, and that the price effect washes out. That is the standard treatment for an energy spike.
AI-driven demand for memory, power and copper does not obviously fit that template. It is not a disruption to supply so much as a sustained increase in demand, funded by committed corporate capital spending that is planned years ahead and is relatively insensitive to interest rates in the short run. If that demand persists, the price pressure persists with it, and the case for looking through it weakens.
There is a second awkwardness. Monetary policy works by cooling demand, but the demand in question comes from a handful of very large, cash-rich firms building strategic infrastructure. Higher rates are a blunt instrument against that, and would land instead on households and smaller businesses that are not causing the pressure.
The other side of the argument
None of this is settled, and there are reasonable grounds for scepticism.
Capacity is being added. Memory manufacturers have committed enormous sums to new facilities, and if that supply arrives the shortage eases and prices fall, possibly sharply, as they have in every previous memory cycle.
AI infrastructure spending could also slow. It rests on expectations of future revenue that have not yet materialised at the scale implied, and a reassessment by the companies committing the capital would remove the demand pressure directly.
And there is the productivity argument, which cuts the other way entirely: if AI raises output per worker, it is disinflationary over a longer horizon, even while building it is inflationary now.
What to watch
The practical question for anyone tracking rates is whether this line of thinking spreads from speeches and minutes into the committee's actual reaction function.
The signal to watch is not further commentary but the Fed's own projections, and specifically whether officials revise up their inflation forecasts while citing capital spending rather than tariffs or energy. That would indicate the argument has moved from interesting observation to something shaping policy.
Until then the honest summary is narrow: two officials and a set of minutes have identified a mechanism, the mechanism is real and visible in consumer prices, and the Fed has not yet decided what to do about it.
