Nvidia has halted production of its H200 artificial intelligence chips designed for the Chinese market, abandoning what was once seen as a compromise solution to handle U.S. export controls.
The chipmaker stopped manufacturing its second-most advanced AI processors intended for Chinese customers, according to a Financial Times report citing people familiar with the situation. Nvidia has reallocated manufacturing capacity at TSMC away from H200 production toward its next-generation Vera Rubin hardware instead.
This shift comes despite Nvidia receiving licenses from the U.S. government last week to ship "small amounts" of H200 chips to customers in China. The move signals that company executives do not expect meaningful H200 sales in China in the near term, even with regulatory approval.
U.S. officials are simultaneously considering caps that would limit Chinese firms to buying 75,000 H200 chips each, according to people familiar with discussions within the Trump administration.
Shipments of Advanced Micro Devices' MI325 chips with similar capabilities would also count toward a customer's cap under the proposed restrictions.
Total shipments to China could still reach as many as a million units based on an upper bound set earlier in the regulatory process, but per-customer limits would constrain major Chinese tech giants like Alibaba Group Holding and ByteDance. Those companies privately told Nvidia they wanted more than double the 75,000-chip limit for their purchases.
Nvidia said last week that it still isn't getting any data center revenue from China and doesn't know whether Beijing will allow imports even if Washington gives permission. The company's shares fell almost 1% to $181 following news of the potential export caps, while AMD dropped to $197.07.
The H200 represents Nvidia's most powerful chip from its previous generation of products, serving as the industrial standard for training and operating AI software like ChatGPT until the company debuted its current-generation Blackwell line last year. It has six times the computational capabilities of what Trump's team had previously authorized for sale to China.
China hawks within and outside the Trump administration worry that H200 exports will only help China develop better AI models without benefiting U.S. interests.
However, Nvidia CEO Jensen Huang convinced President Trump that such exports would contribute to "a positive economic relationship with China" during discussions last year.
Nvidia's internal logic maintains that keeping Chinese AI companies dependent on American technology prevents Huawei from building the revenue and developer ecosystem necessary to compete globally. Commerce Secretary Howard Lutnick acknowledged in January that "there are plenty of people who disagree" with this reasoning but noted it was Huang's argument and "this is the president's deal."
Implementation details have proven challenging since Trump's December announcement authorizing H200 exports. U.S. officials added constraints designed to limit potential negative impacts while following presidential directives, including requirements that chip sales won't cut into availability for American companies and that Chinese customers perform rigorous due diligence checks against military use.
Nvidia has grown increasingly frustrated with what it views as bureaucratic obstacles undermining Trump's vision, complaining at times that U.S. export conditions are so onerous they dissuade Chinese firms from purchasing H200 chips altogether.
