Huawei’s plans to create more powerful chips for AI and smartphones have hit a major roadblock due to US sanctions.
Huawei is designing its next two Ascend processors, its answer to Nvidia’s dominant accelerators, around the same 7-nanometer architecture that’s been mainstream for years.
The decision comes as US-enforced restricts Huawei’s chipmaking partners from acquiring state-of-the-art extreme ultraviolet lithography systems from ASML Holding.
Huawei’s marquee chips will therefore be stuck with the aging technology until at least 2026.
China’s overall push to develop cutting-edge chips will lag further behind the US in 2025, when TSMC — chipmaker to Apple and Nvidia — begins to crank out 2nm chips, about three generations ahead, reported Bloomberg.
Further bad news for Huawei comes by way of reports that its main production partner, Semiconductor Manufacturing International, is struggling to churn out even these older 7nm chips at constant volumes. The Shanghai-based firm’s 7nm production lines have reportedly been plagued by poor yield and reliability issues.
The US government has restricted China from buying advanced equipment from American suppliers including Applied Materials and Lam Research, and Nvidia’s most powerful AI chips.
Huawei needs tens of millions of Kirin-branded processors for its own smartphones every year, and aims to make hundreds of thousands of Ascend AI chips.
This year, Beijing urged Chinese companies to avoid using Nvidia and instead adopt domestic alternatives such as Huawei.
In 2023, Huawei unveiled the Mate 60 Pro smartphone with a self-designed 7nm chip made by SMIC just as Commerce Secretary Gina Raimondo was touring Beijing.
This year, Huawei has been mostly silent on the processor powering its upcoming flagship smartphones. It is set to unveil its next-generation Mate 70 on November 26.
Chinese chipmakers have been attempting to push the limits of the older deep ultraviolet lithography machines from ASML since they are restricted from acquiring newer ones. However, they haven’t been successful yet in extracting enough out of those machines in order to advance their chipmaking manufacturing.
