Epoch AI finds China is more exposed than the US to a chip supply shock

A common assumption is that the US, dependent on Taiwanese fabs, has the most to lose if chip supply is disrupted. Epoch AI's Daniel Carey argues the opposite in a new analysis. By separating semiconductors out of the OECD's inter-country input-output tables, which track 50 industries across 80 countries, the work measures how much chip revenue each dollar of final demand pulls in. In 2022 every $1,000 of Chinese final demand generated $15.20 for chipmakers, against $5.70 for the same US spending, a ratio of 2.7 to 1. China leaned even harder on specific producers: its exposure ran 4.7x the US figure for Taiwan's chip industry and 4.0x for South Korea's.

To turn exposure into consequences, the authors calibrate a trade model to 2022 and simulate shocks over a two to three year adjustment. In a combined scenario of a large fall in Taiwanese productivity plus a full China-West decoupling, advanced processor prices in China rise 17-fold and real national expenditure falls 3%, while in the US prices rise about 20% and expenditure falls about 0.6%. The direction held across every robustness check, though the size swung widely: China's processor price effect ranged from a 6% rise to a 100-fold increase depending on assumptions.

The caveats are substantial and stated plainly. The tables end in 2022 and miss the AI server boom, the model covers only the trade channel and ignores inventories, sanctions, and recessions, and newer 2024 data narrows the exposure gap from 2.8x to 2.2x. Chinese self-sufficiency is also rising, from 12% of domestic chip demand in 2016 to 23% in 2022.

Why it matters

If you plan compute capacity or supply chains with an eye on US-China tension, this reframes who absorbs the pain: the data points to China's economy taking the larger hit, which is the reverse of the usual framing. The wide range in the price estimates is a reminder to treat any single number as a scenario, not a forecast.