---
title: "Fed officials say the AI build-out is now an inflation problem"
description: "Governor Lisa Cook and New York Fed president John Williams have both flagged AI infrastructure spending as an inflation risk, and the concern appears in the June FOMC minutes. More than $1.5 trillion of data centre plans is bidding up memory, power and copper."
category: "Markets"
category_url: https://boursel.com/category/markets
author: "Kenji Nakamura"
published: 2026-07-19T19:52:00.000Z
updated: 2026-07-19T19:52:00.000Z
canonical: https://boursel.com/article/fed-officials-say-the-ai-build-out-is-now-an-inflation-problem
tags: ["federal-reserve", "inflation", "artificial-intelligence", "semiconductors", "monetary-policy"]
---
# Fed officials say the AI build-out is now an inflation problem

Governor Lisa Cook and New York Fed president John Williams have both flagged AI infrastructure spending as an inflation risk, and the concern appears in the June FOMC minutes. More than $1.5 trillion of data centre plans is bidding up memory, power and copper.

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](https://finance.yahoo.com/economy/policy/articles/fed-flags-fresh-inflation-threat-182000339.html).
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.

## Sources

- [Fed flags a fresh inflation threat that could rattle markets](https://finance.yahoo.com/economy/policy/articles/fed-flags-fresh-inflation-threat-182000339.html)