Chips 2 min read

China's chip problem is everyone's chip problem

On October 7, the Bureau of Industry and Security at the US Department of Commerce issued sweeping export controls on advanced semiconductor technology to China. The rules restrict the sale of advanced AI chips, chip manufacturing equipment, and the know-how to make them.

NVIDIA’s A100 and H100 GPUs, which power most large-scale AI training, are now restricted. AMD’s high-end accelerators, same thing. ASML’s EUV lithography machines, already informally blocked, are now formally blocked.

The US is trying to freeze China’s semiconductor capabilities at roughly their current level. No access to advanced chips. No access to the tools to make advanced chips.

I’ve spent two weeks mapping the dependencies. The picture is more tangled than most people realize.

The dependency web

Semiconductors are the most globally interdependent technology in existence. No single country controls the full supply chain. Here’s a simplified map:

Design software (EDA tools): dominated by US companies (Synopsys, Cadence, Mentor). Without these tools, you can’t design modern chips.

Design: US companies lead (NVIDIA, AMD, Qualcomm, Apple). But Chinese companies like HiSilicon (Huawei’s chip division) and SMIC are growing.

Manufacturing (leading-edge): TSMC makes about 92% of the world’s most advanced chips. In Taiwan. Samsung makes most of the rest. In South Korea. Intel is third, distant, and rebuilding.

Manufacturing equipment: ASML (Netherlands) makes all EUV lithography machines. Tokyo Electron (Japan), Lam Research (US), Applied Materials (US), and KLA (US) make the other critical tools.

Materials: silicon wafers (Japan, Germany), photoresists (Japan), specialty gases (various).

The point: no country is self-sufficient. The US controls design tools and some equipment. The Netherlands controls lithography. Japan controls materials. Taiwan controls manufacturing. China controls a large share of downstream assembly and packaging.

What the export controls mean

The controls target three things: advanced chips (anything above a certain performance threshold), the equipment to make them, and US persons who might help China develop the capability domestically.

The “US persons” provision is the most aggressive. American citizens and permanent residents are now prohibited from supporting the development or production of advanced chips in China, even if they work for non-American companies.

This is the US using its position in the dependency web as a weapon. It works because the web is structured in a way that routes through American control points. Even companies outside the US use American EDA tools, American equipment components, or employ American engineers.

The risks

I see two.

The first is retaliation. China controls rare earth minerals, battery materials, and a huge share of global manufacturing. If semiconductor restrictions escalate into a broader trade conflict, the consequences ripple across every industry that depends on Chinese manufacturing. Which is most industries.

The second is motivation. The surest way to make a country invest in domestic capability is to cut off foreign supply. Japan’s semiconductor industry in the 1980s emerged partly because American companies didn’t take Japanese competition seriously. China now has every reason to pour unlimited resources into building its own chip industry. They may not reach leading-edge for years or decades. But “years or decades” isn’t the same as “never.”

Reuters reported that China’s semiconductor investment fund is already scaling up. SMIC is pushing the limits of its non-EUV manufacturing. Huawei is reportedly stockpiling chips.

What I keep thinking

The semiconductor supply chain is a marvel of human cooperation. Thousands of companies across dozens of countries, each contributing a specialized piece, producing something that no single country could make alone.

The export controls are the beginning of that cooperation fracturing along geopolitical lines. Maybe it’s necessary. Maybe the security risks of selling advanced AI chips to a strategic competitor outweigh the benefits of an integrated supply chain.

But the web was beautiful in its interdependence. And watching it start to tear is uncomfortable.

92% of the world’s most advanced chips come from a single island. That fact was already scary before export controls. Now it’s the central fact of global technology policy, and everyone is finally paying attention.


Related thinking:

a

astro

Thinking about AI, robots, space, and the future. Writing it down so I don't forget.