As the United States contemplates implementing stricter trade restrictions to prevent advanced chip technology from being exported to China, Nvidia, a leading U.S.-based semiconductor company, is reportedly developing a modified version of its artificial intelligence chips to adhere to these regulations.
According to sources familiar with the matter, Nvidia is creating a variant of its Blackwell AI chips specifically for the Chinese market. The company plans to collaborate with a local distribution partner, Inspur, to market and distribute this chip, which is tentatively named the “B20,” in China.
The B20 is anticipated to begin shipping by the second quarter of 2025. Nvidia has not provided any comments regarding these developments.
The company already has three chip models designed to comply with U.S. export restrictions, including the H20. Nvidia recently lowered the prices of the H20 in response to weak sales, aiming to compete with similar products from the Chinese firm Huawei. Despite initial challenges, reports indicate that sales of the H20 are now on the rise. According to data from SemiAnalysis, Nvidia is expected to sell over one million H20 chips in China this year, amounting to approximately $12 billion, even with the current U.S. trade limitations. This forecast is nearly twice that of Huawei’s sales expectations for its Ascend 910B chip.
However, analysts from Jefferies warn that Nvidia’s H20 chips may face challenges under further U.S. trade regulations. They indicated that during the upcoming annual review of semiconductor export controls in October, it is very likely that the H20 could be prohibited from being sold to China. Potential methods for enforcement include a product-specific ban, a reduction in the computing power limit, or restrictions on memory capacity.
Additionally, the U.S. may broaden its export controls to include chips sold to neighboring countries like Malaysia, Indonesia, and Thailand, or even extend these limitations to overseas Chinese firms, though the latter would be more complex to implement, analysts noted.