Nvidia cannot sell advanced AI chip to China: White House

The Trump administration said it does not plan to let Nvidia sell its most advanced AI chip, the Blackwell, to China at this time.

White House spokesperson Karoline Leavitt made the statement during a press conference on November 4.

This follows recent comments by President Donald Trump, who said these chips should be reserved for US companies, and not exported to China or other countries.

Speculation had grown about the possibility of selling a scaled-down version of the Blackwell chip to China after Trump suggested in August that such sales might be allowed.

Trump also indicated he might raise the chip issue with Chinese President Xi Jinping at a summit in South Korea last week, but said the topic was not discussed.

🔗 Source: Reuters

🧠 Food for thought

Implications, context, and why it matters.

White House stance on Blackwell exports clouds Nvidia’s China revenue picture

At Nvidia, China revenue was about $17 billion last fiscal year (~13% of total) and fell to $2.8 billion in the latest quarter as market share slid from ~95% to near zero under U.S. export controls alongside Chinese restrictions 1 2.
H20 drove $4.5 billion in charges in fiscal Q1. The lower-performance data center GPU was built to meet U.S. export rules for China, yet Q2 saw no H20 sales into China, with about $650 million in unrestricted units sold to a buyer outside China 3 4. Booked AI chip orders exceed $500 billion through the end of 2026, while guidance assumed no H20 shipments to China, which keeps China exposure unclear with Blackwell exports to China paused 2 3.

US cloud giants and allied markets likely to take Blackwell supply

With Blackwell barred from China for now, 6 million shipped over the past four quarters and 14 million projected over the next five look set to flow to U.S. cloud providers plus allied markets as global data center capex rose 53% year over year in Q1 2025, driven by hyperscalers (the largest cloud plus internet platforms) 2 5.
Nvidia is building over one million square feet of U.S. manufacturing in Arizona and Texas, planning up to half a trillion dollars of AI infrastructure over four years 6. This gives suppliers plus colocation operators (data center companies that rent power, space, and cooling) clearer timelines to align capacity with hyperscaler buildouts.