Nvidia to resume H20 GPU chip sales to China, launches mainland-compliant model
By Liam Mo and Anne Marie Roantree
BEIJING/HONG KONG (Reuters) -Nvidia said on Monday it will resume sales of its H20 artificial intelligence chip to China and has introduced a new model tailored to meet regulatory requirements in the Chinese market.
Nvidia, the world’s most valuable company, is filing applications with the U.S. government to resume sales to China of the H20 graphics processing unit (GPU), and expects to get the licences soon, the company said in a statement. Deliveries are expected to begin shortly thereafter, it added.
“The U.S. government has assured NVIDIA that licenses will be granted, and NVIDIA hopes to start deliveries soon,” Nvidia said in a statement. The White House did not immediately respond to a request for comment.
CEO Jensen Huang is scheduled to hold a media briefing in Beijing on Wednesday when he attends a supply chain expo, his second visit to China after a trip in April where he stressed the importance of the Chinese market.
The move to resume sales of the H20 chips comes amid easing tensions between Washington and Beijing, with China relaxing controls on rare earth exports and the United States allowing chip design software services to resume in China.
The H20 chip was developed specifically for the Chinese market after U.S. export restrictions were imposed on national security grounds in late 2023. The AI chip was Nvidia’s most powerful legally available product in China until it was effectively banned by Washington in April.
The H20 ban forced Nvidia to write off $5.5 billion in inventories, and Huang told the Stratechery podcast earlier this year that the company also had to walk away from $15 billion in sales.
Nvidia’s AI chips have been a key focus of U.S. export controls designed to keep the most advanced chips out of Chinese hands, amid intense competition between the superpowers to dominate the AI race.
The company also announced the development of a new AI chip designed specifically for China, called the RTX Pro GPU. Nvidia described the model as “fully compliant” with U.S. export controls and suitable for digital twin AI applications in sectors such as smart factories and logistics.
In May, Reuters reported Nvidia was preparing to launch a new AI chip, based on the RTX Pro 6000D, in China at a significantly lower price point than the H20.
The graphics processing unit would be part of Nvidia’s latest generation Blackwell-architecture AI processors and was expected to be priced well below the the H20 due to its weaker specifications and simpler manufacturing requirements, sources said.
Huang has met with U.S. President Donald Trump and policymakers in Washington and later with officials in Beijing, as part of efforts to promote AI cooperation and highlight Nvidia’s support for open-source research and global AI development, the company said.
(Reporting by Liam Mo in Beijing, Anne Marie Roantree in Hong Kong and Surbhi Misra in Bengaluru; Editing by Sumana Nandy and Stephen Coates)