Nvidia will launch a new, cost-effective artificial intelligence (AI) chipset for China, built on its cutting-edge Blackwell architecture. The company plans to begin mass production in June 2025.
Sources familiar with the development report that Nvidia will price the graphics processing unit (GPU) between $6,500 and $8,000, making it significantly more affordable than the recently restricted H20 chip, which sold for $10,000 to $12,000. The lower price results from simpler manufacturing and reduced specifications, ensuring compliance with stringent U.S. export restrictions.
The new chip, likely named the 6000D or B40, is based on Nvidia’s RTX Pro 6000D server-class processor and uses conventional GDDR7 memory instead of high-bandwidth memory (HBM). It avoids advanced Chip-on-Wafer-on-Substrate (CoWoS) packaging technology from Taiwan Semiconductor Manufacturing Co. (TSMC), aligning with U.S. regulations capping memory bandwidth at 1.7–1.8 terabytes per second, compared to the H20’s four terabytes per second. Chinese brokerage GF Securities forecasts the GPU will achieve approximately 1.7 terabytes per second, staying within export limits.
This marks Nvidia’s third attempt to tailor a GPU for China, the world’s second-largest economy, which accounted for 13% of its sales ($17 billion) in the fiscal year ending January 2025. U.S. export curbs, intensified in April 2025, banned the H20 chip, forcing Nvidia to write off $5.5 billion in inventory and forgo $15 billion in sales. Nvidia CEO Jensen Huang noted that the older Hopper architecture, used in the H20, can no longer be modified to meet current restrictions.
China’s AI market remains critical for Nvidia, though its share has dropped from 95% pre-2022 to 50% today, with Huawei’s Ascend 910B chip gaining ground. Huang warned that continued U.S. restrictions could push Chinese customers toward Huawei. Nvidia is also developing another Blackwell-based chip for China, with production expected by September 2025, though specifications remain unconfirmed.
