Geopolitics and Chips

On 31 August 2022 Nvidia told investors in a regulatory filing that the US government had informed it that exports of its A100 and forthcoming H100 data-centre GPUs to China would need a licence, and that about $400 million of potential sales in its current quarter could be affected (company, August 2022). China, including Hong Kong, had accounted for roughly a fifth to a quarter of Nvidia's revenue in fiscal 2022–23 (company, verify exact share). Five weeks later, on 7 October 2022, the Commerce Department's Bureau of Industry and Security (BIS) published the rules on which every later control was built: licences for advanced computing chips and the computers containing them, licences for the tools used to make advanced logic and memory chips in China, and a bar on US persons supporting such production at Chinese fabs.

Four years later the market Nvidia had lost was close to gone. Jensen Huang said in October 2025 that Nvidia's share of China's AI accelerator market had fallen from about 95% to close to zero (company claim). In August 2026 Nvidia's quarterly filing said it had shipped only a fraction of the H200 chips it was licensed to sell to China, under 1% of the quarter's data-centre revenue, and its guidance assumed no China data-centre revenue at all (company). Yet its total revenue had grown several-fold, to $215.9 billion in fiscal 2026 (company), because demand elsewhere absorbed every chip it could make.

This chapter explains how AI chips became an instrument of state power: the chokepoints, the US rules and their reversal after 2025, China's answer, and the Taiwan risk under all of it.