Export Controls

Export controls are laws with which a state determines which products its companies are allowed to sell to which countries. Some goods may be exported freely, others require an official permit. And for some countries a complete ban applies. The reason is usually national security: the state does not want technology to end up in foreign weapons or intelligence services. Classically, this concerned tanks, explosives, or encryption technology. For a few years now, however, computer chips for artificial intelligence have been at the center of attention, that is, the specialized processors with which AI systems compute.

Chips as a lever of world politics

Modern AI systems are not created through clever programming alone. They require enormous computing power, often tens of thousands of specialized chips working in parallel for months. These chips come almost entirely from a handful of companies, above all the US corporation Nvidia. They are manufactured predominantly in Taiwan, using machines from the Netherlands and Japan. This chain therefore has only a few bottlenecks, and whoever controls one of them controls a great deal.

That is exactly why the USA has been relying on export controls against China since 2022. The idea: without access to the fastest chips, China falls behind in building very large AI systems. Because many machines and components originate from US technology, the rules also reach beyond the country’s own borders. A Dutch machine manufacturer can thus be affected by US requirements.

For investors and companies, this is a concrete risk. For years, China was a billion-dollar market for Nvidia, which shrank considerably due to the rules. Conversely, manufacturers who build alternatives domestically benefit. News about new or relaxed controls therefore regularly moves stock prices.

How thresholds and permits are drawn

Export controls work with lists and thresholds. For AI chips, authorities define technical limits, for example how many computing operations a chip can perform per second and how quickly it exchanges data with other chips. If a product exceeds these limits, its sale to certain countries requires a permit or is prohibited. In addition, there are lists of individual companies with which business is fundamentally forbidden.

These thresholds are a game of cat and mouse. Nvidia deliberately developed weakened chip variants for China that fell just below the thresholds. US authorities then tightened the limits, prompting new variants to emerge. Such rounds have occurred repeatedly since 2022, and each adjustment takes months.

The controls are never completely watertight. Chips can be resold via third countries, which is called smuggling. And anyone who is not allowed to buy the hardware can instead rent computing power from a cloud provider abroad, that is, use computers over the internet. Whether such rental models fall under the rules is one of the open points of contention.

Export controls in the news

In business news, export controls almost always appear in three contexts. First, in chipmakers' quarterly results, when revenue losses in China are explained. Second, in political announcements from Washington, Brussels, or Beijing. Third, in countermeasures: China itself restricted the export of rare earths, that is, certain metals needed for electronics.

A common misconception is to equate export controls with tariffs or sanctions. Tariffs make goods more expensive but do not prohibit them. Sanctions are comprehensive punitive measures against a country or a person. Export controls, by contrast, are targeted: they concern precisely defined technologies, regardless of whether a punitive action is intended.

Private users, too, feel the effects indirectly. If an AI model becomes available more slowly in Europe or a provider raises prices, scarce computing power may be behind it. And whenever Chinese companies release powerful models despite the restrictions, it is discussed anew each time whether export controls work at all.