Nvidia is reportedly set to launch a new, lower-cost artificial intelligence (AI) chip designed specifically for the Chinese market, responding to restrictions on exporting its more expensive models. According to Reuters, mass production of these new AI chips is scheduled to begin in June. This initiative marks Nvidia's third attempt to create chips compliant with U.S. regulations for sale in China.

The new chip, part of Nvidia’s latest Blackwell-architecture AI processors, is expected to be priced between $6,500 and $8,000, significantly lower than the restricted H20 model, which sold for $10,000 to $12,000. This price reduction reflects the chip's weaker specifications and simpler manufacturing requirements. The chip will be based on the RTX Pro 6000D, a server-class graphics processor, and will utilize conventional GDDR7 memory instead of more advanced high bandwidth memory. It will also forego Taiwan Semiconductor Manufacturing Co’s advanced Chip-on-Wafer-on-Substrate packaging technology.

An Nvidia spokesperson stated that the company is still evaluating its options, emphasizing that until a new product design is approved by the U.S. government, Nvidia remains effectively excluded from China's $50 billion data center market. China is a critical market for Nvidia, accounting for 13% of its sales in the past financial year.

The U.S. government's restrictions, which led to the effective ban of the H20 in April, cited concerns over the technology potentially being used in, or diverted to, supercomputers in China. Nvidia CEO Jensen Huang noted that Nvidia’s market share in China has dropped from 95% before 2022 to 50% currently due to these export restrictions. Huang also warned that continued export curbs could drive more Chinese customers to purchase chips from competitors like Huawei, which produces the Ascend 910B chip.

The restrictions have already impacted Nvidia, forcing the company to write off $5.5 billion in inventory and forgo $15 billion in sales. The latest export restrictions impose new limits on GPU memory bandwidth, capping it at 1.7-1.8 terabytes per second, compared to the H20's capability of 4 terabytes per second. The new GPU is forecast to achieve approximately 1.7 terabytes per second using GDDR7 memory technology, complying with the export control limits.

In addition to the chip slated for June production, Nvidia is reportedly developing another Blackwell-architecture chip for China, with production potentially starting as early as September. Nvidia's main rival in this market is Shenzhen-based Huawei, which is testing and preparing to ship its Ascend 910D AI chip.