Cambricon Technologies Corp. is gearing up to more than triple its AI chip output in 2026, positioning itself as a leading domestic alternative to Nvidia in China’s rapidly evolving data center market. The move aligns with Beijing’s broader strategy to reduce reliance on US technology and promote homegrown semiconductor champions.
The Beijing-based chip designer is targeting deliveries of around 500,000 AI accelerators next year, according to people familiar with internal planning. Up to 300,000 of those units are expected to be its most advanced Siyuan 590 and 690 processors, designed for large-scale AI training and inference workloads in cloud and enterprise environments.
Cambricon plans to lean heavily on Semiconductor Manufacturing International Corp. for production, tapping SMIC’s “N+2” 7‑nanometer process for its cutting-edge chips. This partnership underscores Beijing’s push to build a full domestic supply chain, from chip design to fabrication, amid tightening US export controls on advanced manufacturing technologies.
The aggressive ramp-up comes as Chinese authorities discourage the use of Nvidia hardware in sensitive or state-linked data centers, reshaping procurement policies across the sector. New guidelines requiring state-funded projects to rely on local AI chips are opening doors for players such as Cambricon, Huawei and Moore Threads to capture share once dominated by US firms.
Huawei is simultaneously preparing to double output of its top AI accelerators, intensifying competition within China’s domestic chip ecosystem. Moore Threads, another up-and-coming GPU and AI chip supplier, is using its Shanghai listing to highlight ambitions in graphics and data center compute, signaling a crowded but fast-growing local landscape.
Market reaction has been swift, with Cambricon shares advancing in Shanghai as investors bet on stronger demand for local AI silicon. SMIC and rival foundry Hua Hong Semiconductor also saw gains in Hong Kong trading, reflecting optimism that foreign chip restrictions will translate into sustained orders for Chinese manufacturers.
Despite the enthusiasm, Cambricon has publicly cautioned investors about speculative reports around its production and customer pipeline. In a statement posted on WeChat, the company labeled recent media information about its products and forecasts as inaccurate, without naming specific outlets, and urged stakeholders to rely only on official disclosures.
Technical and manufacturing hurdles remain a key risk to Cambricon’s expansion plans, including yield challenges on advanced 7‑nanometer nodes that could constrain effective output. Analysts note that even with capacity secured at SMIC, the company must improve yields and software ecosystem support to fully capitalize on the domestic demand wave.
Nvidia CEO Jensen Huang has acknowledged that US export restrictions have effectively blocked his company from China, creating space for local rivals such as Cambricon and Huawei. While the Trump administration is weighing limited approvals for Nvidia’s newer H200 accelerators, uncertainty over Chinese acceptance means domestic AI chipmakers are likely to keep gaining ground
