May 24, 2025 – BEIJING/TAIPEI
Nvidia is preparing to launch a new artificial intelligence (AI) chip for the Chinese market at a significantly reduced price compared to its recently restricted H20 model, according to sources familiar with the matter. The chip, based on Nvidia’s latest Blackwell architecture, is expected to enter mass production as early as June.
The new graphics processing unit (GPU) is reportedly designed with more basic specifications and manufacturing requirements, contributing to a lower price range of approximately $6,500 to $8,000. This contrasts with the H20 chip, which was priced between $10,000 and $12,000 prior to U.S. export restrictions that effectively banned its sale in China in April.
According to two of the sources, the upcoming chip will be based on Nvidia’s RTX Pro 6000D, a server-grade processor. It will use standard GDDR7 memory rather than the more advanced high-bandwidth memory (HBM) employed in higher-end models. Additionally, the chip will not utilize Taiwan Semiconductor Manufacturing Co.’s (TSMC) advanced Chip-on-Wafer-on-Substrate (CoWoS) packaging technology.
The product’s name, specifications, and pricing have not yet been officially disclosed. Chinese brokerage GF Securities speculated earlier this week that the GPU might be called the “6000D” or “B40,” though it did not cite sources or confirm the final designation.
An Nvidia spokesperson stated that the company is still assessing its options under the current export constraints. “Until we settle on a new product design and receive approval from the U.S. government, we are effectively foreclosed from China’s $50 billion data center market,” the spokesperson said. TSMC declined to comment on the matter.
China remains an important market for Nvidia, accounting for approximately 13% of the company’s revenue in the past fiscal year. The company has had to redesign its chips for the Chinese market multiple times in response to evolving U.S. export controls aimed at limiting China’s access to advanced semiconductor technologies.
Following the ban on the H20 chip, Nvidia initially explored a downgraded version of the product, but ultimately shifted to developing an entirely new design. Nvidia CEO Jensen Huang confirmed that the company’s older Hopper architecture, used in the H20, could no longer be modified under current U.S. regulations.
Sources also indicated that Nvidia is working on a second Blackwell-based chip for the Chinese market, with production potentially starting as early as September. However, specific details about that product have not been confirmed.
Nvidia’s share of China’s AI chip market has declined from approximately 95% in 2022 to around 50% currently, according to Huang. Huawei, a major Chinese competitor, has gained market share with its Ascend 910B chip. Huang cautioned that continued U.S. restrictions may accelerate the shift of Chinese customers toward domestic alternatives like Huawei.
The H20 export ban also led Nvidia to write off $5.5 billion in inventory. Additionally, Huang stated during an interview on the Stratechery podcast that the company forfeited an estimated $15 billion in potential sales.
The latest U.S. export restrictions include new limitations on GPU memory bandwidth—a critical factor for high-performance AI processing. Jefferies, a global investment bank, estimates the current bandwidth cap to be around 1.7 to 1.8 terabytes per second, compared to 4 terabytes per second for the H20. GF Securities estimates that Nvidia’s upcoming chip will meet these limits by delivering roughly 1.7 terabytes per second using GDDR7 memory.
