Nvidia has stopped making chips intended for the Chinese market, gambling that regulators in Washington and Beijing will continue to muck it around.
The US chipmaker has shifted manufacturing capacity at Taiwan Semiconductor Manufacturing Company away from H200 chips and towards its next-generation Vera Rubin hardware, according to two people familiar with the move.
That switch shows Nvidia is no longer banking on meaningful H200 sales in China anytime soon, after months of chaos over US export approvals and the threat of Chinese restrictions.
The H200 is one of Nvidia’s older AI processors and was pitched as compliant with US export controls on advanced chips. Vera Rubin, unveiled earlier this year, is the latest chip architecture, built for more complex AI systems and in hot demand from top US tech groups, including OpenAI and Google.
One person with knowledge of the plans said, “Instead of waiting in a limbo, Nvidia has to move on to what it can achieve with certainty, especially when there’s a shortage of supply for its advanced stuff. This could, in a way, accelerate the Vera Rubin delivery and roll-out.”
Nvidia had been leaning hard on both Washington and Beijing to let it sell H200 chips in China. After US President Donald Trump signalled in December that he would allow sales, Nvidia started ramping up production.
It had been expecting orders of more than 1mn units from Chinese customers, the FT previously reported. Suppliers were working around the clock to meet deliveries pencilled in as early as March.
Nvidia chief Jensen Huang said in early January: “Demand was ‘very high, and we’ve fired up our supply chain and H200s are flowing through the line’”.
Then the approvals jammed up, as the state department pushed for tighter restrictions to make it harder for China to use H200 chips in ways that would dent US national security.
China is reportedly lining up its own restrictions on buying H200s to protect domestic chipmakers and push local AI outfits towards homegrown silicon.
People familiar with the situation said China’s customs still listed H200s as banned unless the importer has an approval letter from Beijing, even though no public details have been shared.
Nvidia chief financial officer Colette Kress said: “While small amounts of H200 products for China-based customers were approved by the US government, we have yet to generate any revenue. And we do not know whether any imports will be allowed into China.”
Nvidia has already built about 250,000 H200 chips, according to a person with knowledge of the matter. If both the Americans and the Chinese permit only limited orders, that pile should cover demand, they said.
China’s President Xi Jinping and Trump are expected to meet at the end of March, prompting analysts to guess whether a deal could loosen chip export controls.
Even if that happens, it would take up to three months for Nvidia to shift or expand supply chain capacity to crank out more H200s, the people said, with the current stock doing the job in the meantime.
