- Nvidia’s new AI chip is priced at $6,500-$8,000 to comply with export rules.
- New Blackwell-based AI chip lacks high-bandwidth memory, reducing cost.
- Nvidia’s market share in China has fallen from 95% to 50% due to export curbs.
Nvidia is set to release a new, lower-cost artificial intelligence (AI) chip for China. This decision comes after the company faced export restrictions on its high-end H20 model. According to sources, the new chip will be priced between $6,500 and $8,000, making it more affordable than the $10,000 to $12,000 price range of the H20 model.
The chip is designed on Nvidia’s Blackwell architecture and is simpler to manufacture than other chips. It will use GDDR7 memory rather than high-bandwidth memory (HBM) and will exclude advanced Chip-on-Wafer-on-Substrate (CoWoS) packaging technology. The company will begin mass production of the new chip in June to meet China’s demand for AI technology.
AI Chip to Enhance Nvidia’s Market Position
The new AI chip will enhance Nvidia’s competitive advantage in China, a key market for the company’s revenue. China accounted for 13% of Nvidia’s sales in the last financial year. After the U.S. government put restrictions on exports, Nvidia has seen a decrease in its market share.
Nvidia’s market share in China has dropped from 95% to 50% since 2022. The company’s CEO, Jensen Huang, warned that continued restrictions could drive more Chinese customers to rival companies, particularly Huawei. Huawei’s Ascend 910B chip has become popular in the market as an alternative to Nvidia’s products.
While the new Nvidia AI chip is less powerful than the restricted H20 model, it provides better performance for AI tasks. Analysts predict that Chinese domestic technologies, like Huawei’s chips, will soon close the performance gap within one to two years. Nvidia continues to lead because of its connection of AI clusters with CUDA, which is important for developers.
Nvidia’s Diversification Efforts
Nvidia previously attempted to create a less powerful version of the H20 for China, but the U.S. rules made it more difficult. The company shifted to the Blackwell architecture, developing a chip with standard memory that complies with new export controls. The new AI chip is expected to reach a memory bandwidth of 1.7 terabytes per second, which falls within the export limits.
Moreover, sources indicate that Nvidia plans to work on another variant set for production in September. Although the specifications are still unknown, it could meet the needs of the Chinese market.
However, Nvidia’s earnings for the quarter, which will be released on May 28, are expected to be good. Analysts predict the company will report $43.4 billion in revenue in Q2, up 66% year-over-year.