Huawei, which is at the center of Beijing’s drive for tech self-sufficiency, is reportedly preparing to test a new AI processor that could match one of Nvidia’s most popular products.
The chip, dubbed Ascend 910D, would be Huawei’s most powerful AI chip, according to a report from the Wall Street Journal citing unnamed sources. Huawei expects to get the first batch of samples as early as late May.
The Chinese tech giant reportedly hopes that the new chip can beat Nvidia’s H100 processor. The chip, released in 2022, is widely used for AI applications. The U.S. has banned export of the H100 to Chinese companies since late 2022.
Huawei did not immediately respond to Fortune’s request for comment.
Semiconductor self-sufficiency
Huawei is one of a group of Chinese companies that are helping support China’s self-sufficiency in semiconductors amid rising tensions between Beijing and Washington.
Because of U.S. export controls, China’s chip industry is barred from buying the most advanced equipment used to make cutting-edge chips. Chinese companies can’t buy cutting-edge chips directly either.
Earlier this month, the Trump administration stopped Nvidia from selling its H20 chip to China. The U.S. chip company developed the H20 to comply with earlier U.S. rules on chip exports.
Washington has targeted Huawei since 2019, but the Chinese tech company has proved its ability to navigate U.S. restrictions. In 2023, Huawei released a smartphone with a domestically produced advanced processor, revealing the company’s ability to innovate in spite of Washington’s sanctions.
Huawei’s Ascend 910C is reportedly being used for inference by DeepSeek and other AI startups.
Beijing is also leaning on local companies to use domestic alternatives to Western-made technology.
Still, despite recent successes, Huawei’s AI chips remain less powerful than Nvidia’s cutting-edge products. U.S. rules also mean that Huawei needs to rely on mainland Chinese chipmakers like Semiconductor Manufacturing International Corp. (SMIC). The Chinese chipmaker reportedly has lower yields for its chipmaking process compared with market leader TSMC, thus increasing its production costs.
