Topics / Chips & Export Controls
Chips & Export Controls
The United States has spent three years trying to deny China the chips that train and run frontier AI. China’s answer has not been one thing: a domestic accelerator industry led by Huawei, a memory push led by CXMT and YMTC, workarounds at the manufacturing edge, and a hard bet on model efficiency. We track both sides of that contest, with dates and sources.
ReferenceOpen the export-control tracker→The state of play
- Nvidia has largely conceded China’s AI chip market: the H20 was halted in April 2025, and the H200, cleared in January 2026, has effectively not sold.
- Huawei leads the domestic field, shipping roughly 812,000 Ascend cards in 2025 (IDC), with Cambricon, Moore Threads, Biren and others behind it.
- The manufacturing ceiling still holds: SMIC’s advanced line targets about 2.6 million AI chips this year against domestic demand near 4.2 million, and HBM remains a dependency.
- China’s hedge is software and power: DeepSeek V4 runs on roughly 27 percent of the prior generation’s compute per token at long (1-million-token) context, and China generates more than twice the United States’ electricity.
Best issues on this topic
- Conceded
On Nvidia’s earnings call, Jensen Huang said the quiet part out loud: China revenue is essentially zero, and the company has "largely conceded" the market to Huawei.
- The Boarding
The full Nvidia-to-China chip sequence, from the A800 and H800 to the H20 halt and a 4.5 billion dollar write-off, one of the largest write-downs in semiconductor history.
- Blind Run
Washington cleared the H200 for ten Chinese firms with a 25%-of-revenue condition. Beijing told them not to buy, keeping the money pointed at domestic chips. Deliveries stayed at zero.
- 60,000 Chips, Zero From Nvidia
China’s national computing hub in Zhengzhou runs 60,000 accelerators made by Sugon, a firm tied to the Chinese Academy of Sciences, with zero chips from Nvidia, AMD or any Western supplier.
- The No-GPU Machine
A Chinese system topped the global TOP500 at 2.198 exaflops on 47,000 domestic CPUs and zero GPUs. On the AI-relevant benchmark, it came fourth. The asterisk is the whole export-control story.
- The Co-Design
A teardown found part of DeepSeek V4’s architecture was co-designed for Huawei’s Ascend 950DT inference chip. The model and the silicon are now being built for each other.
- The Harder Half
Cambricon became China’s first pure-play AI chip designer worth over a trillion yuan. The same report shows the distance left: a 2.9% share of the cards China bought in 2025, against Nvidia’s 55%.
- The Windfall
A Chinese memory middleman’s half-year profit jumped 743x. AI’s appetite for HBM is being paid for one layer down, by the budget phone in your pocket.
These link to the full issues on the newsletter. New pieces on this topic go out in the daily first.
Common questions
Can China build advanced AI chips without Nvidia?
For a growing share of inference and some training, yes. Huawei’s Ascend line leads and models like DeepSeek V4 have been trained or served on domestic clusters. But China still trails at the leading edge of manufacturing, and high-bandwidth memory remains a dependency.
What is the Huawei Ascend 950?
Huawei’s 2026 AI accelerator line, codenamed David. The 950PR, unveiled in March 2026, claims 2.87 times the single-card compute of Nvidia’s China-market H20 and uses Huawei’s own memory. The 950DT adds 144GB of in-house HiZQ memory for inference.
Did US export controls on China work?
Partly. Controls on chipmaking equipment have held, and SMIC still cannot build leading-edge chips at volume. But controls on finished chips pushed Chinese buyers onto domestic parts faster, and in May 2026 Nvidia’s Jensen Huang said the company had largely conceded the market to Huawei.