Huawei open-sources models to push its Ascend AI chips: report

China-based Huawei has open-sourced two of its Pangu series artificial intelligence models, in an effort to drive adoption of its Ascend AI chips, which are considered China’s top competitor to Nvidia (NASDAQ:NVDA), according to CNBC.

Nvidia and other U.S. chipmakers are not allowed to export high-performing processors to China, prompting domestic companies to develop their own. Huawei open-sourcing its models is an attempt to strengthen its position across the AI value chain in the face of U.S. chip export restrictions, the report said.

The new open-sourced large language models include the Pangu Pro MoE 72 billion parameter hybrid expert model and ascending-based model reasoning technology.

“This move is another key measure for Huawei to implement the rising ecological strategy, promote the research and innovative development of large model technology, and accelerate the application and value creation of artificial intelligence in thousands of industries,” Huawei said.

Huawei is also planning to expand its AI data centers to new countries, the report said. Constellation Research analyst Ray Wang told CNBC that the chip-to-model strategy is similar to what Google (GOOG)(GOOGL) is doing with its open-sourced Gemma models and in-house chips.