Nvidia has informed major Chinese clients, including top cloud service providers, about its plan to release the modified H20 chip within the next two months, two sources said.
The move reflects the US chipmaker’s efforts to maintain access to a vital market amid Washington’s escalating curbs on advanced semiconductor exports to China.
The original H20, Nvidia’s most powerful AI chip authorized for sale in China, was effectively removed from the market after US authorities told the company last month that the product would now require an export license.
In response, Nvidia has adjusted its design strategy, setting new technical thresholds for revised chip models, one source said.
The new version will feature notable downgrades, including significantly reduced memory capacity, according to the same source.
Another source added that downstream users might be able to reconfigure the module to alter performance within permitted limits.
Nvidia declined to comment.
The US Commerce Department did not immediately reply to a request for comment.
China represented \$17 billion in revenue, or 13% of Nvidia’s total sales, in the fiscal year ending January 26.
Underscoring the importance of the Chinese market, Nvidia CEO Jensen Huang traveled to Beijing last month, shortly after the export license requirement for the H20 was imposed.
During talks with Chinese officials, Huang reiterated the country’s critical role in Nvidia’s global business.
Washington has restricted sales of Nvidia’s top-tier chips to China since 2022, citing fears over their potential military use.
The H20 was developed following an expansion of US export controls in October 2023.
Chinese tech firms including Tencent, Alibaba and ByteDance increased orders of the H20 amid rising demand for affordable AI solutions, particularly from startups like DeepSeek, Reuters reported earlier this year.
Reuters also reported last month that Nvidia had received \$18 billion in H20 orders since January.