$9.3 billion. That's SMIC's 2025 revenue, up 16% from the year before and an all-time high for China's largest chipmaker. Analyst estimates from LSEG suggest the company could top $11 billion in 2026.
SMIC isn't alone. Hua Hong, another major Chinese foundry, reported record fourth-quarter revenue of $659.9 million. And Cambricon Technologies, China's leading AI chip startup, swung from a 452 million yuan loss in 2024 to a 2.1 billion yuan ($306 million) profit in 2025, with full-year revenue surging from 1.2 billion yuan to 6.5 billion yuan - a 5x jump.
The pattern is clear: US export restrictions designed to slow China's AI capabilities are doing the opposite for China's domestic semiconductor companies. Cut off from Nvidia's top-end chips, Chinese tech firms are buying local. Cambricon now plans to ship roughly 500,000 AI accelerators in 2026, including up to 300,000 units of its flagship Siyuan 590 and the next-generation Siyuan 690.
Two forces are driving this. Electric vehicles and their supporting infrastructure are propping up demand for mature-node chips (the less advanced kind used in cars and appliances). Meanwhile, AI demand for more advanced chips is, as one industry analyst put it, "through the roof." Beijing's push for 100% semiconductor self-sufficiency by 2027 adds government spending on top of market demand.
For anyone building AI products, the practical takeaway is that the global chip supply chain is splitting into two parallel tracks. Chinese AI models will increasingly run on Chinese hardware, which means performance characteristics, availability, and pricing for AI compute could diverge significantly between markets over the next few years.