Nvidia is reported to have halted production of artificial intelligence (AI) chips for export to China. Nvidia is said to plan to focus on producing the next-generation "Vera Rubin" chip instead of the H200 chip.
The Financial Times (FT) reported on the 5th (local time) that "TSMC has shifted its production facilities from H200 chip production to next-generation 'Vera Rubin' chip production." The move is seen as driven by delays in U.S. export approval to China for the H200 chip and the rise of potential Chinese regulatory actions.
U.S. President Donald Trump said in Dec. last year that he would allow exports of Nvidia's H200 chip to China. However, approval is said to be delayed as Nvidia has differences with the Commerce Department over procedures such as know your customer (KYC).
The Chinese government is also reported to be encouraging the use of Chinese-made AI chips, conveying guidance to its corporations to buy the H200 only "when necessary." Deputy Minister David Peters said at a House Foreign Affairs Committee hearing on the 24th of last month that the H200 had not been exported to China.
Colette Kress, Nvidia's chief financial officer (CFO), said on an earnings conference call on the 25th of last month, "We received U.S. government approval for a small number of H200 products for Chinese customers but have not yet generated revenue," adding, "We do not even know whether imports into China will be allowed."
Nvidia appears to have pivoted to next-generation chip production, where demand is clearly assured, to minimize uncertainty.
Nvidia currently holds an inventory of 250,000 H200 chips and is reported to plan to meet demand by drawing down existing stock if it becomes able to sell semiconductors to China.