Sources: a Chinese fab making Huawei's AI chips is set to start production as soon as the end of 2025, with two more due in 2026; the fabs' ownership is unclear
Fabrication plants serving Huawei push to increase production as US cuts off access to Nvidia's top processors
Financial TimesZijing Wu
Context & Ripple Effects
Huawei’s AI-chip push has been constrained less by product plans than by manufacturing throughput: it previously struggled to ramp its Ascend server chip, while a reported increase in AI-chip yields to nearly 40% improved the economics of production.
The reported fabs extend a longer domestic-manufacturing effort that included plans for a dedicated Shanghai chip plant. They matter because restricted access to Nvidia’s top processors makes Huawei’s own supply chain a central constraint on AI hardware availability.
First-order effects
A fab serving Huawei is reportedly poised to add production capacity by the end of 2025, with two further facilities expected in 2026; that would expand the manufacturing base available for Huawei AI chips.
Huawei gains a potential route to reduce its production bottleneck, though the fabs’ unclear ownership leaves accountability and operational control opaque.
Second-order effects
More available fab capacity could help Huawei translate improving yields into greater chip supply, easing a constraint that had limited the ramp of its Ascend AI server chips.
Nvidia’s restricted access to the top end of the Chinese market gives Huawei a stronger incentive to prioritize domestic capacity; Chinese AI-system buyers would have a larger potential non-Nvidia supply source if the ramp succeeds.
Third-order effects
If these facilities begin production on schedule, China’s AI-hardware market would become more dependent on locally coordinated chip design and fabrication rather than access to leading foreign accelerators.
Unclear fab ownership may complicate assessment of the supply chain and of the practical reach of export controls, while capacity build-outs remain subject to yield and equipment constraints.
The trend: This is another step in the industrialization of China’s domestic AI-chip supply chain as access to leading foreign accelerators narrows.
FT on SMIC ramp up in AI chip production. Qwen AI: Based on the retrieved search results, here's the available information about SMIC's Ascend AI chip production projections: For 2025: Huawei is projected to ship approximately 700,000 Ascend AI chips in 2025, though yield [image]
“Necessity begets innovation,” said an investor in multiple Chinese semiconductor and large language model start-up. FT US attempts to block China are back firing. https://www.ft.com/...
by the way it would be very impressive if they avoided repeating the Nvidia/CUDA monopoly situation with Huawei, making all DL hardware stacks commoditized from the start. At this point, would require everyone adopting CANN probably? [image]
Scoop: China seeks to triple output of AI chips in boost to DeepSeek ambitions *New capacity: 3 plants for Huawei + SMIC doubling (now ~30K WPM) *Cambricon 690 & Huawei 910D to natively support UE8M0 FP8 *Unprecedented alignment to build own AI ecosystem https://enterprise-sharin…
This is absolutely insane... Not only will one of Huawei's dedicated AI chip factories start production as early as the end of this year, but SMIC is also going to double its 7nm production capacity next year. “Domestic output won't be an issue for long, especially with all the […
SMIC's advanced node are in its SMSC subsidiary, announced in 2016, w/ plan for 35k wpm in phase 1 (SN1) & another 35k in phase2 (SN2) Last yr, they already imported necessary ASML scanners for full 70k. Ramping up capacity required validation of US-free supply chain for 5-7nm. […
Huawei AI accelerators getting a dedicated plant opening next year with two more on the way (but nobody knows who actually owns the fabs lol). Output expected to be equivalent to what SMIC can currently do, and they're also doubling 7nm capacity.
“China's chipmakers are seeking to triple the country's total output of artificial intelligence processors next year, as Beijing races the US to develop the most advanced AI.”
FT: “smaller Chinese chip designers such as Cambricon, MetaX and Biren will be able to get much larger allocations of SMIC's capacity, spurring competition for a fast-growing China market left by Nvidia following US export bans.” [image]
1) Chinese No.2 AI chipmaker Cambricon's revenue reached RMB 2.881 billion, a YoY increase of 43 times. Net profit reached 1.038 billion RMB, compared with a net loss of RMB 530 million in the same period last year, marking a clear turnaround to profitability. [image]