Nvidia is Working on New Flagship AI Chip Model For China

The news agency Reuters revealed the American technology company Nvidia is working on a new flagship AI chip model for the Chinese market. Three anonymous sources explained that the new chip version will consider the strict export controls in the United States to make it compatible.

Back in 2023, the U.S. tightened export controls to avoid providing China with tools that could lead to empowering its military capabilities; forcing Nvidia to create specific products for the Chinese market and decreasing its sales in the region.

“Nvidia will work with Inspur, one of its major distributor partners in China, on the launch and distribution of the chip which is tentatively named the `B20’,” explained Reuters. Nvidia hasn’t made an official announcement on this new development.

A few months ago, in March, Nvidia announced that its new AI chips would be more affordable. The new generation series was named Blackwell and will be mass-produced by the end of this year to supply major AI and cloud computing companies.

According to Reuters, the new processors will be 30 times faster than previous models and will be better at tasks like processing chatbot answers. Nvidia’s sales have rapidly increased this year, with competitive prices to gain leverage against the company’s main competitor, the Chinese giant Huawei.

Nvidia’s top and most advanced product in the Chinese market at the moment is the H20 chip, and it’s about to hit over 1 million sales in China. The new advanced products for the Asian region could increase friction with the U.S. government as Biden’s administration has been restricting chipmaking equipment to China and reaching out to allies, like the Netherlands and Japan, to join the mission.

Official statements and more details on the new AI chip models should be announced soon.

Nvidia has also been recently accused of using YouTube videos to train AI models without permission, along with other major tech companies like Apple, Salesforce, and Anthropic.