Report says TSMC capacity redirected to next-generation hardware
Nvidia has stopped production of its second-most advanced artificial intelligence chips intended for the Chinese market, the Financial Times reported on Thursday, citing two people with knowledge of the matter. The chips, known as H200, were designed for China under U.S. export controls, but the report said Nvidia has redirected manufacturing capacity at contract chipmaker TSMC toward its next-generation Vera Rubin hardware.
Reuters said it could not immediately verify the Financial Times report. TSMC declined to comment, and Nvidia did not immediately respond to a request for comment.
Licenses allow only small shipments as sales outlook remains limited
The reported production decision follows mixed signals on U.S.-approved shipments to China. Last week, Nvidia said it had received U.S. government licenses to ship “small amounts” of H200 chips to customers in China. However, the shift in production capacity suggests the company does not expect meaningful H200 sales into China in the near term.
A U.S. Commerce Department official said last month that none of Nvidia’s H200 chips had been sold to Chinese customers. While licensing approval has been in place, practical barriers have remained.
Export guardrails have delayed deliveries despite January clearance
In January, the Trump administration provided formal clearance for China-bound sales of Nvidia’s H200 chips, according to the information provided. Shipments, however, remained stalled because of guardrails built into the approval process, limiting the pace and scale of deliveries.
If the Financial Times report is accurate, reallocating capacity from H200 toward Vera Rubin would reflect a strategic choice to prioritize next-generation product ramps over a constrained market where regulatory friction has reduced near-term demand visibility.
Reallocation underscores how controls reshape AI chip roadmaps
The reported move highlights how export restrictions and licensing uncertainty can influence manufacturing plans, especially for advanced AI accelerators that require scarce foundry capacity. For Nvidia, shifting production at TSMC toward Vera Rubin could also reflect a broader effort to position supply for markets where demand is less constrained and revenue contribution is more predictable.
The outcome for China customers remains unclear, but the combination of limited license volumes, reported lack of sales to date, and a potential production halt would further narrow options for acquiring high-end Nvidia hardware through official channels.