China's Military Is Hunting Nvidia's Top Chips for AI Weapons, Congressional Report Warns
![]() |
| Nvidia H100 Tensor Core GPU, the class of chip at the centre of the dispute. |
China's military is seeking Nvidia's most advanced AI chips for AI-enabled weapons, a congressional commission said in its annual report released on Wednesday, 18 November. The US-China Economic and Security Review Commission, created by Congress in 2000, said Beijing was evading American controls to get the semiconductors behind its push for "intelligentized warfare" — the PLA's term for weaving AI into weapons, command and control, kill chains and logistics.
![]() |
| The US Capitol in Washington; Congress created the commission in 2000. |
The reason is a gap China has not closed, the report says. Domestic AI chips, including Huawei's Ascend and Cambricon's Siyuan series, still trail Nvidia and Advanced Micro Devices. Nvidia's CUDA software remains a barrier Chinese firms have yet to overcome, alongside gaps in leading-edge logic fabrication, memory and chipmaking equipment. "The resulting gap in compute is substantial, largely due to U.S. export controls on advanced chips," the report says.
![]() |
| Close-up of an Nvidia H100 accelerator board, showing the processor die and memory modules. |
Beijing has found workarounds. Chinese companies have obtained remote access to Nvidia-powered compute through data centres in third countries, including Malaysia, without chips crossing into mainland China. Reuters found a concrete case: Northwestern Polytechnical University — a military-linked school blacklisted by the US since 2001 — bought two Nvidia H200 chips for 874,700 yuan, about $123,000, via Shaanxi Bosuan Network Technology, an October 21 procurement record shows. A wider Reuters review found more than $3 billion in Chinese computing projects using Nvidia's restricted chips, though Nvidia says export controls have effectively shut it out of China's AI data-centre market.
On Monday 16 November, the Commerce Department issued a 166-page interim final rule, "Addressing Remote Access to Advanced Semiconductors", requiring licences before data centres anywhere supply more than 12 million aggregate TPP — roughly 750 Nvidia H100 or H200 chips — to firms based in China, Macau or restricted destinations. The rule formally rescinds the Biden-era AI Diffusion Rule from January 2025. This article covers the commission's report and the PLA's pursuit of the chips — distinct from the already-queued article on the new licensing rule.
Nvidia itself is pulling back. Interim chief executive Colette Kress said on Tuesday's earnings call that Nvidia would be "no longer shipping licensed Hopper data centre products to China", after Hopper sales reached $2.4 billion in the third quarter of 2026. Jensen Huang resigned as chief executive on Thursday 12 November after admitting an affair with Alice Huang, a former Megaspeed executive whom US officials suspect of spying for China. Megaspeed is suspected of diverting Nvidia chips via Malaysia data centres and has been under Commerce investigation since 2024. Enforcement is now the test: whether the new rule closes the remote-access route, and whether Congress tightens controls further.
![]() |
| Nvidia's headquarters sign in Santa Clara, California. |
Sources: Reuters (18 Nov 2026); Reuters via The Hindu (12 Nov 2026); Al Jazeera (11 Nov 2026); Financial Express (11 Nov 2026); Fox News (19 Nov 2026); US-China Economic and Security Review Commission, 2026 Annual Report to Congress (18 Nov 2026).



