The US government will impose additional restrictions on exports of Nvidia’s H20 chips to China.
Despite reports earlier in the week claiming the Trump Administration had opted not to place additional restrictions on the chips following CEO Jensen Huang’s attendance at a Mar-a-Lago dinner, Nvidia will now require a license to export the chips to China.
In a filing with the SEC dated April 9, Nvidia said that as a result of restrictions, its first quarter results are expected to include up to approximately $5.5 billion of charges associated with H20 products for inventory, purchase commitments, and related reserves.
However, according to a report from Reuters, despite being told about the incoming restrictions in early April, the company did not warn some of its biggest customers about the changes.
First announced in 2023, Nvidia’s H20 GPUs are a less sophisticated version of its H100 processors and were designed specifically for the Chinese market in compliance with US export controls.
Orders for the H20 chips “surged” following demand for DeepSeek’s AI models, Reuters reported, and Tencent, Alibaba, and ByteDance all “significantly increased” their orders of the chip. A report from April puts the combined total of H20 exports at around $16bn for the first three months of the year.
Speaking to DCD in 2018, Huang said China represented about a third of the company’s business, but this figure has now dropped to around 13 percent of the company’s total revenue.
On Thursday (April 17), Huang visited Beijing at the invitation of a trade organization, with The Guardian reporting that China Daily, the Communist Party’s official English-language outlet, publishing a photo on Huang with the caption: “Three months after pledging to continue cooperation with China during his last visit.”
Elsewhere this week, Nvidia announced plans to produce AI supercomputers 'entirely' in the US, saying it would work with its manufacturing partners to commission more than 1 million square feet of manufacturing space to build and test Blackwell GPUs in Arizona and AI supercomputers in Texas.
