The US government is reportedly preparing to lift export restrictions on sales to China of Nvidia’s second most powerful chip, following a thaw in relations after the two leaders met in South Korea last month.
If the report is accurate, Nvidia will be permitted to sell its H200 chip to Chinese data centre operators and AI developers. The H200 is around 18 months behind the company’s most advanced chip, the Blackwell B200, which the US Commerce Department still plans to restrict from China.
The move is seen as a compromise between the two nations, according to Semafor, which first reported the news. The US remains firmly opposed to China gaining an upper hand in the AI race, but it has shown a willingness to work with both China and Nvidia to maintain trade. Nvidia has pushed hard for access to the Chinese market, arguing that a full ban would only strengthen domestic rivals such as Huawei.
This follows reporting from two weeks ago that President Donald Trump was considering allowing Nvidia to sell some of its chips to China. The Chinese market accounts for roughly 13 percent of Nvidia’s revenue, although this share is expected to decline as Chinese firms shift to homegrown solutions. Nvidia CEO Jensen Huang has admitted he is unsure whether China would purchase the “degraded chips” the company is permitted to offer.
Nvidia’s position as the top dog
The enthusiasm around Nvidia has cooled slightly over the past month, with the company’s stock price falling by about 11 percent since the start of November. Nvidia was the first company to reach a $5 trillion market cap, but that has since slipped to $4.5 trillion. Even with revenue and net income rising, the chipmaker has faced several waves of investor questions on issues ranging from its market dominance to the true size of the AI market to the scale of its major partnerships.
Although data centre operators focused on AI still have thousands of orders for Nvidia’s most powerful chips in their backlog, recent purchases of Google and Amazon AI accelerators have prompted some investors to wonder if Nvidia’s tight market control will weaken in the next few years. AI accelerators can outperform Nvidia GPUs on specific tasks, such as the large-scale training of transformer models.
Nvidia may also be feeling the effects of its close association with OpenAI across a number of major projects and partnerships. As momentum has shifted from OpenAI to Google in the perceived AI race, investors have cooled on several companies linked to OpenAI, while those seen as aligned with Google have seen their stock prices surge over the past month.
China’s homegrown talent
Growing competition in the West may have increased pressure on Huang to reopen commerce with China, which still lacks a homegrown chipmaker capable of matching Nvidia’s sophistication. Analysts estimate that full access to the Chinese market could add more than $15 billion to Nvidia’s revenue over the next two years.
The Chinese government has poured billions into the sector, but Huawei remains the only competitor with real scale, and major AI developers do not widely use its chips. Even so, performance has improved over the past year, and several smaller firms, including Cambricon Technologies, Moore Threads, and MetaX, are beginning to show promise.
Also read: Stargate attracts more funding for a Texas-sized AI factory, where Lancium’s CEO describes how a $500 billion AI campus in Texas is racing to build megacentres and upgrade the power grid.