BEIJING — U.S. chipmaking giant Nvidia has reportedly found a way to sell high-end chips to Chinese companies — while remaining compliant with U.S. rules aimed at curbing China's access to the tech.
China accounts for 20% to 25% of Nvidia's revenue in its data center business, its biggest unit.
Nvidia is set to deliver three new chips to domestic manufacturers in the coming days, Chinese financial media Cailian Press said Thursday, citing sources.
The chips — called HGX H20, L20 PCle and L2 PCle — are based on Nvidia's H100 chip, the report said.
The H100 and A100 artificial intelligence chips were the first to be hit by new U.S. restrictions last year that aimed to curb sales to China. Nvidia said in a September 2022 filing the U.S. government would still allow it to develop the H100 in China.
Companies in China had then switched to Nvidia's H800 and A800 chips, but the U.S. subsequently clamped down on those sales last month with new restrictions.
The H20's computing power is only about 50% of that of the A100, said Bo Du, managing director at WestSummit Capital Management and a former engineer in the chip industry.
That's «basically saying goodbye to physical simulation,» he said in Mandarin, translated by CNBC. While it's possible to use clusters of lower-power chips to support large model calculations, he said there's no ideal solution given the costs.
«In the near term, Chinese manufacturers have no better option and they will continue to buy Nvidia's chips, while searching for replacements,» Du said, noting that some large internet companies have started to buy domestically-made AI chips at scale.
Demand for artificial intelligence computing power has only gone up as companies in China rush to develop local
Read more on cnbc.com