AI’s memory-chip appetite is pushing up prices. Washington Monthly argues the US could intervene

AI data-centre demand is helping drive a memory-chip squeeze that is reaching other industries and consumers. Washington Monthly argues the US government could use existing powers to steer scarce chips towards non-AI buyers and speed up new supply.

Watch Desk analysis

What happened

In an article published on 30 October, Joel Dodge of the Vanderbilt Policy Accelerator links the AI infrastructure build-out to rising prices for information-processing equipment. He cites a Federal Reserve Bank of San Francisco figure showing those prices up 13% over the past year, and says memory-chip prices have quadrupled in that time.

Dodge says data centres consumed 32% of dynamic random-access memory chips in 2020 and that AI data centres could consume 60% of the world’s memory chips by 2030. He also points to monthly data-centre construction spending rising from less than $900 million in December 2022 to more than $4 billion in June 2026. Those figures frame his argument: AI buyers with deep pockets may be squeezing manufacturers and consumers elsewhere.

Why it matters

The article says the pressure is already showing up beyond data centres. It notes Apple has raised prices by as much as 20% on some laptops and tablets, which the company attributed to memory-chip costs, and that automakers face higher costs or potential price rises. If the squeeze persists, AI infrastructure could make ordinary electronics and vehicles more expensive, while pulling investment and materials towards data centres.

Dodge’s proposed response is intervention rather than simply waiting for new factories. He argues the US could use the Defense Production Act to prioritise non-AI buyers, organise a government-facilitated buyers’ club, or require chipmakers to reserve some output for them. He also proposes using the law to speed domestic capacity expansions and shift production towards standard DRAM. These are proposals in an opinion article, not announced government measures.

Our read

The useful tension is plain: AI infrastructure needs memory, but so do cars, medical devices and the computers people already use. Dodge makes a case for using existing industrial policy tools to manage that competition; whether such measures would ease shortages without creating new distortions is the argument worth having. “Build more” sounds tidy until the extra capacity takes years to arrive.

What to watch

• Whether memory-chip prices and consumer electronics costs keep rising.
• Whether US officials consider any of the proposed allocation or capacity measures.
• When new memory-chip production comes online, and whether it eases supply for non-AI buyers.

Discussion spark: Should the US reserve scarce memory chips for non-AI industries, or would that do more harm than good while AI demand is driving investment?

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.