Nvidia prepares custom AI chip for China amid US restrictions | Cpu hardware list and functions | Cpu vs gpu vs ram | Gpu vs cpu | Turtles AI

Nvidia will launch an AI chip specifically tailored for the Chinese market as early as September. Based on the Blackwell RTXPro6000, it will be stripped of HBM and NVLink to comply with new US regulations. CEO Huang is returning to China to meet with authorities and strengthen his commitment to the country.

Key points:

• Chip based on the Blackwell RTXPro6000, modified to meet US requirements.
• High-bandwidth memory and NVLink interconnect will be removed for compliance.
• Launch expected in China in September, following regulatory approval from Washington.
• Jens Huang will be in Beijing to strengthen relations with the government and local operators.

Nvidia, a global semiconductor leader, intends to resume sales of AI chips in China as early as September, unveiling a specially limited version of the Blackwell RTXPro6000. The project was born in response to US export restrictions, which prohibit shipments of higher-performance versions of chips to China. The processor intended for the Chinese market will lack HBM memory and the NVLink interface, key components of Nvidia’s most advanced solutions, in order to comply with US regulations.

This chip, known internally as the RTXPro6000D or B40, will use GDDR7 memory and will not benefit from TSMC’s CoWoS packaging, making it compliant but less performing than the H20, from which it is derived. The estimated price is between $6,500 and $8,000, significantly lower than the approximately $10,000–$12,000 price tag of the H20 model, but designed to maintain Nvidia’s competitive offering in the country. The company aims to ensure continuity of supply by ensuring adequate supplies despite regulatory uncertainty.

To secure US government approval, Nvidia has initiated consultations with the administration and will not launch the product until legal certainty is obtained. Meanwhile, some Chinese customers have already tested prototypes and expressed interest in large orders, although demand will not match that generated by H20. The main reason is the cost and complexity of a potential transition from Nvidia’s CUDA ecosystem to alternative solutions.

During the recent Computex conference, CEO Jensen Huang called US restrictions ineffective, noting that they favor domestic development in China and allow competitors like Huawei to grow. His scheduled appearance in Beijing for the Supply Chain Expo in September includes scheduled meetings with Premier Li Qiang and Vice Premier He Lifeng—events still subject to confirmation by China. These meetings aim to reaffirm the strategic importance of the Chinese market for Nvidia, which is worth approximately $17.1 billion, or 13% of global sales.

According to Reuters, production could begin as early as June, but the commercial launch will be postponed until September to await final approval. At the same time, Nvidia is strengthening its presence in China with a new research center in Shanghai, aimed at better understanding the needs of the local market while operating within US regulatory constraints.

Nvidia is carefully balancing US regulations with its strategic interest in China, adapting its technical and diplomatic portfolio.