Despite the U.S. government's increased semiconductor restrictions on China, American AI chip giant NVIDIA cannot give up the Chinese market.

According to reports, NVIDIA is jointly developing a new flagship AI chip based on the Blackwell architecture specifically for the Chinese market. This chip will comply with U.S. export control regulations and is named "B20." However, the report did not mention the performance or specifications of this AI chip.

In response, NVIDIA declined to comment when approached by the Titanium Media App on July 22.

It is reported that during the GTC conference held in March this year, NVIDIA CEO Jensen Huang officially launched the new generation AI graphics processor (GPU) based on the Blackwell architecture. This architecture platform can build and run real-time generative AI on trillion-parameter-level large language models (LLM). The world's most powerful AI chip based on this architecture, the GB200, uses TSMC's 4nm process and contains 208 billion transistors. The B200 is expected to be 30 times faster than its predecessor on certain tasks and is anticipated to enter mass production as early as September this year. AI chips based on the Blackwell architecture are expected to start delivery in the second quarter, achieve large-scale delivery in the third quarter, and data centers can run on the new products by the fourth quarter, potentially generating significant revenue this year. He emphasized that there is currently a shortage of AI computing power in the market, and the demand for the previous generation Hopper architecture products will continue to increase.

However, due to the impact of the U.S. Department of Commerce's Bureau of Industry and Security (BIS) export control regulations, the Chinese market has always been a concern for NVIDIA's sales.

On March 30 this year, the U.S. upgraded its chip export ban to China, stating that it would implement a "case-by-case review" policy for AI semiconductor products exported to China, fully restricting the sale of advanced AI chips and semiconductor equipment from NVIDIA, AMD, and others to China.

In the earnings report for the first quarter of the 2025 fiscal year released in May, which ended on April 28, the revenue from Chinese customers in NVIDIA's data center business dropped from 19% in the 2023 fiscal year to a mid-single-digit percentage (5%) in the 2024 fiscal year.

A previous market report indicated that due to higher memory capacity, the manufacturing cost of the H20 would be higher than that of the H100, making the price of the H20 about half that of the H100, thus NVIDIA made compromises on profit margins. Renowned investment bank Morgan Stanley also stated in a report that the demand in the Chinese market is strong, and NVIDIA cannot afford to give up the sales revenue from the H20 chip.