What Has China Said It Will Do About America’s New AI Chip Restrictions?

The Chips Stay Overseas. Why Are Washington and Beijing Fighting Over Who Uses Them?

Why China Calls America’s New AI Chip Rule ‘Economic Coercion’

Beijing Has Promised “Necessary Measures” And Left Open What They Might Be.

China says it will take “necessary measures” to protect its interests after the United States tightened controls on access to advanced AI chips. At a briefing on Tuesday 17 November 2026, foreign ministry spokesman Guo Jiakun accused Washington of economic coercion. He gave no timetable and named no tariff, export restriction or American company that Beijing intended to target.

Guo was more specific about what China objects to. The American rule can reach a deal between a Chinese business and a data centre in another country, even when the chips stay inside that building. Washington says overseas servers can give Chinese AI developers access to computing power that US export controls were designed to restrict. Beijing says Washington is interfering with China’s right to trade and cooperate with other countries.

What Did China Say Washington Had Done Wrong?

Guo called the US action an example of “long-arm jurisdiction”. Consider what that means for a Chinese developer renting time on servers in Malaysia. The customer is Chinese, the facility is Malaysian and the contract pays for a computing service. The American connection is the controlled technology inside the servers. Washington says that connection, together with who will use the computing power and on what scale, can bring the arrangement within its licensing rules.

That reach is central to Beijing’s objection. Guo accused the US of suppressing China’s high-tech industries and disrupting supply chains. He also said the decision ran counter to an understanding the two governments had reached. His remarks did not identify a particular agreement provision that the US had breached.

There is a practical distinction here for any business trying to judge its exposure. The US has issued a rule that operators can examine against their contracts. China has stated its grounds for opposing that rule. Guo’s announcement alone gives an American exporter no Chinese duty rate to calculate, product list to check or deadline to meet.

Why Can a Chip That Stays in Malaysia Concern Washington?

An AI developer needs access to processors for the time it takes to train a model. Owning them is optional. A data centre can buy the hardware, install it overseas and sell access to many customers. An export check made when the equipment arrives can establish where the servers are. It cannot settle who will be renting them a year later.

The US rule addresses that change of customer directly. An overseas data centre might receive covered equipment for one purpose and subsequently begin supplying a significant volume of computing power to a company headquartered in China. Under the new provision, that later decision can itself trigger a licensing requirement. The location of the customer’s ultimate parent company can also matter.

For operators already providing covered services, the rule sets a concrete deadline. Those wishing to continue must obtain a US licence within 90 days of the rule’s effective date or face future export restrictions. Applications under this provision will be assessed individually. For a developer planning a long training run, that creates an immediate commercial question. Can its provider commit to supplying the computing time it has promised?

Why Does the US Rule Name Five Countries?

Brazil, Indonesia, Malaysia, Singapore and Thailand feature in a separate part of the US chip licensing rule. Washington identifies these destinations as presenting a risk that advanced semiconductors could be diverted. The provision introduces licensing requirements for covered exports, re-exports and transfers within those countries, subject to the rule’s exceptions.

The list answers a narrower question than the one Beijing raised. A supplier arranging a shipment needs to check the destination. A data-centre operator also needs to consider who can use its machines once they arrive. The remote-access provision can apply to a qualifying facility elsewhere, which is why the five names alone give an incomplete picture of the rule’s reach.

Washington has also drawn distinctions between uses of the same technology. The rule has specific provisions for commercial systems used exclusively to rank or recommend products, services or content. Its definition excludes systems that generate new text, images, audio or software in response to a user. An online retailer’s product-ranking tool and a general-purpose AI model can therefore raise different licensing questions, even if the same company operates both.

What Could Change When Trump and Xi Meet?

Donald Trump and Xi Jinping are due to meet on Thursday 19 November in Shenzhen, where China is hosting the APEC summit. Guo’s wording leaves Beijing room to press its objection in those talks. It gives no indication that China has set Thursday as a deadline for action or that either leader has agreed to change the American rule.

The next consequential document could come from either capital. A Chinese notice naming a product, company or effective date would tell businesses what Beijing’s promised measures mean for them. A US licensing decision would answer a different question for an overseas operator and its Chinese customer. While attention is on what Xi might say in Shenzhen, those companies already have a decision to make about the computing time they can sell and buy.

Sources

Next Reads

Previous
Previous

Why Is the Trump–Xi–Putin Summit Being Compared With Yalta?

Next
Next

Why Is Trump Meeting India, Japan and Australia Before His Talks With Xi and Putin?