US Commerce Secretary Howard Lutnick said the country is not in favour of Apple buying memory chips from Chinese manufacturers. He said other solutions have to be found to the AI-driven chip shortage.

Why memory suddenly became a problem

Memory chips were for a long time the most ordinary component in the technology industry. They were cheap, plentiful, and their price fell steadily.

AI reversed that. The systems that train and run models need extraordinary quantities of fast memory. Data centre demand has absorbed a large share of manufacturing capacity.

The consequence reaches every other product that uses memory: phones, laptops, games consoles. Queuing for the same production lines, those products now compete with data centres.

Apple's bind

Apple sits in the middle of this. The company builds hundreds of millions of devices a year and every one of them contains memory.

There are only a handful of large memory manufacturers worldwide, and a significant part of their capacity has gone to data centre orders. Chinese manufacturers offer an alternative source in that picture, on both price and delivery time.

The secretary's statement makes that alternative politically expensive.

Pressure, not prohibition

The distinction matters. The statement is not a regulation; it is neither a ban nor a sanction. It is a declaration of preference.

Statements of this kind work like this: they set no direct rule but change the company's risk calculation. A company that enters a supply relationship it knows the department opposes exposes itself to regulation that may arrive later.

Large companies usually price that uncertainty as a cost, and more often than not they step back.

What the other solutions could be

The alternatives the secretary points to are limited in practice:

  • Korea and Japan — the existing main sources, but their capacity is already committed.
  • New production in the US — even where investment decisions have been made, plants take years to reach output.
  • Using less memory — a design change on the device side, which means giving up product specifications.

None of the three resolves the shortage in the short term. Which means the cost of the restriction will ultimately show up in device prices.

The broader pattern

On its own this is a small story, but it belongs to a pattern. Supply chain decisions in AI are increasingly made on political rather than commercial criteria.

For companies this creates a planning problem they are not used to. Choosing a supplier can no longer be done on price, quality and capacity alone; which country that supplier sits in, and where relations with that country are heading, now enter the calculation. That makes supply chains more expensive and slower.