Nvidia has informed Chinese customers that it aims to begin shipping its second-most powerful AI chips, the H200, to China before the Lunar New Year holiday in mid-February, three sources familiar with the matter told Reuters.
The U.S. chipmaker plans to fulfil initial demand using existing inventory, with expected shipments ranging between 5,000 and 10,000 chip modules, equivalent to 40,000 to 80,000 H200 AI chips, according to two of the sources.
Nvidia has also told clients that it intends to expand production capacity for the H200, with new capacity orders expected to open in the second quarter of 2026, a third source said.
However, significant uncertainty remains, as Beijing has not yet approved any H200 purchases, and shipment timelines could change depending on regulatory decisions.
Approval still pending
“The whole plan is contingent on government approval,” one source said. “Nothing is certain until we get the official go-ahead.”
The sources declined to be identified because the discussions are private.
In a statement to Reuters, Nvidia said:
“We continuously manage our supply chain. Licensed sales of the H200 to authorized customers in China will have no impact on our ability to supply customers in the United States.”
China’s Ministry of Industry and Information Technology did not immediately respond to a request for comment.
Major policy shift under Trump
If approved, the shipments would mark the first deliveries of H200 chips to China since U.S. President Donald Trump announced earlier this month that Washington would allow such sales, subject to a 25% fee.
Last week, Reuters reported that the Trump administration had launched an inter-agency review of export license applications for H200 sales to China, following the president’s pledge.
The move represents a major shift from the Biden administration’s policy, which had banned advanced AI chip exports to China over national security concerns.
Strategic implications
The H200, part of Nvidia’s Hopper architecture, remains widely used for artificial intelligence workloads despite being superseded by the company’s newer Blackwell chips. Nvidia has largely prioritised production of Blackwell and its upcoming Rubin platform, making H200 supplies relatively limited.
Trump’s decision comes as China accelerates efforts to develop domestic AI chip capabilities. While local firms have yet to match the performance of Nvidia’s H200, Chinese officials are concerned that renewed imports could slow domestic innovation.
Earlier this month, Chinese regulators reportedly held emergency meetings to assess the issue. One proposal under discussion would require H200 purchases to be bundled with a specific ratio of domestically produced chips, according to Reuters.
For Chinese technology giants such as Alibaba Group and ByteDance, which have expressed interest in the H200, the potential shipments would offer access to processors around six times more powerful than the H20, a downgraded AI chip Nvidia previously designed to comply with U.S. export restrictions.
