Nvidia: ‘We are working closely with the US to comply with China export rules’
Nvidia is ‘working closely’ with the US government to comply with strict export restrictions on AI chips to China. That’s what CEO Jensen Huang says. The US Commerce Secretary warned the company on Monday for circumventing those restrictions.
“Nvidia has worked very closely with the U.S. government to create products that comply with the [exportregels]Huang said according to Reuters during a press conference in Singapore. “Our plan now is to continue working with the government on a new range of products that meet the new regulations with certain limits.”
The CEO seems to indicate that he wants to continue introducing customized AI chips that are specifically for the Chinese market. The company was recently criticized for this by US Secretary of Commerce Gina Raimondo. “If you redesign a chip around a certain cutting line that [China] makes it possible to do AI, then I will regulate it the next day,” Raimondo said recently. She specifically targeted Nvidia.
Since last year, Nvidia has been releasing redesigned versions of its existing GPUs, which are less powerful and therefore do not fall under the export restrictions. The company did this for the first time last year with its A800, a modified version of the Nvidia A100 data center GPU with lower NVLink bandwidths. That A800 appeared a month after the US introduced its first export restrictions. Earlier this year, the American chip company also released an H800 GPU, which is based on the Hopper H100.
These chips have now also been banned due to new, stricter export rules that were introduced in October this year. Nvidia then introduced a number of new chips for the Chinese market. According to rumors, the company is also working on a GeForce RTX 4090D for Chinese consumers, as the regular RTX 4090 is no longer allowed to be delivered to China. The company was also expected to release a less powerful H20 GPU based on the H100, but its release was recently postponed until early next year.