US chipmaker Nvidia is considering expanding production of its H200 artificial intelligence chips after receiving demand from Chinese customers that is higher than its current manufacturing capacity, according to people familiar with the discussions.
The development follows a recent announcement by US President Donald Trump, who said the US government would permit Nvidia to export its H200 processors to China while imposing a 25% fee on such sales. The H200 is Nvidia’s second most powerful AI chip currently available.
Sources say interest from Chinese technology companies has been exceptionally strong, prompting Nvidia to explore the possibility of increasing output. The discussions remain confidential, and those involved requested anonymity.
In a statement, an Nvidia spokesperson said the company is carefully managing its supply chain to ensure that approved sales of the H200 to Chinese customers do not affect deliveries to clients in the United States.
Several major Chinese firms, including Alibaba and ByteDance, have reportedly contacted Nvidia in recent days to inquire about purchasing large quantities of the H200 chip. However, final approvals are still pending, as Chinese authorities have not yet authorised imports of the processor.
According to sources, Chinese government officials held urgent meetings to assess whether the H200 should be allowed into the country. A decision is expected after evaluating potential risks to China’s domestic semiconductor industry.
At present, production volumes of the H200 remain limited, as Nvidia is prioritising manufacturing of its newer Blackwell chips and preparing for its upcoming Rubin architecture.
China Pushes Domestic AI Chips
Supply constraints have been a key concern for Chinese customers, who are seeking clarity from Nvidia on availability timelines. The company has shared general guidance on current supply levels, though exact figures were not disclosed.
The H200 entered large-scale deployment last year and represents the most advanced chip from Nvidia’s previous Hopper generation. It is manufactured by Taiwan Semiconductor Manufacturing Company (TSMC) using its 4-nanometre process.
TSMC declined to comment on specific capacity allocations, referring instead to its broader strategy for managing capacity amid surging global AI demand.
Demand from China remains strong because the H200 is currently the most powerful AI processor accessible to Chinese firms. Industry experts say it significantly outperforms the H20, a reduced-capability chip Nvidia released for the Chinese market in late 2023.
Trump’s decision to allow H200 exports comes as China intensifies efforts to strengthen its domestic AI chip ecosystem. Analysts warn that wider access to advanced foreign chips could slow progress among local manufacturers, whose products still lag behind Nvidia’s technology.
According to investment experts, the H200 offers computing performance several times higher than China’s most advanced locally produced accelerators. This gap has led cloud service providers and large enterprises to lobby for eased restrictions on imports.
Sources said one proposal discussed by Chinese officials would require buyers of H200 chips to also purchase a certain proportion of domestically produced chips.
For Nvidia, expanding production presents challenges, as the company competes with other global tech firms for limited advanced manufacturing capacity at TSMC while transitioning to its next-generation chip platforms.