NVIDIA is reevaluating its manufacturing strategy with Taiwan Semiconductor Manufacturing Company (TSMC) as U.S. export controls create bottlenecks for H200 chip sales to China. Despite strong demand from Chinese tech firms, shipments remain stalled due to Beijing’s pending approval, forcing NVIDIA to reassess how it allocates limited TSMC production capacity. This pivot was followed by the Trump administration allowing exports of the H200 to China, with 25% sales tax, reversing the prior Biden-era restrictions.
Uncertainty Over Export Rules Dampens H200 Sales Outlook
Last week, Nvidia reported that it had received licenses from the U.S. government to resume ‘small amounts’ of shipments to customers in China. However, Nvidia’s recent action suggests that the firm is no longer expecting significant H200 sales in China in the near term, after months of uncertainty over the U.S. export approvals and potential Chinese restrictions.
Reportedly, the U.S. chipmaker has reallocated manufacturing capacity at chip contract maker TSMC away from making H200 chips to its next-gen Vera Rubin hardware, succeeding Blackwell and featuring the Rubin GPU. It is designed for more complex AI systems and has strong demand from top U.S. tech groups like OpenAI and Google.
Strong Demand Persists Despite $27,000 Price Tag
Washington tightened its restrictions on the sale of high-end semiconductors to China, while Beijing has signalled it might curb imports to support the domestic semiconductor champions. NVIDIA heavily urged Washington and Beijing to accelerate its H200 chips in China after Donald Trump provided a green signal in December, indicating that he would permit sales, prompting NVIDIA to begin stepping up production.
However, the approval process stalled soon after the State Department imposed tougher restrictions, making it harder for China to use the H200 chips in ways that would undermine U.S. national security. Furthermore, by restricting the purchase of H200s China was also planning to protect its domestic semiconductor industry, thereby encouraging local AI groups to use domestically produced chips.
Interestingly, despite the restrictions, Chinese tech giants have placed orders that exceeded about 2 million H200 units for delivery in 2026, which far surpasses the current inventory of ~700,000 chips. The H200 is seen as a sixfold performance leap over the restricted H20 chip previously available in China, making it tightly attractive despite a price of ~$27,000 per unit.
Industry Watches for Possible Easing of U.S. Export Controls
As the U.S. government allowed a small amount of H200 products for China-based customers, Colette Kress, the Chief Financial Officer of Nvidia, revealed on the earnings call last week that the firm has not generated any revenue yet and is uncertain if any imports will be allowed to China.
U.S. President Donald Trump and Xi Jinping, President of China, are due to meet at the end of this month, leading some industry analysts to speculate whether an agreement would be passed to lift the chip export controls. If so, it is reported that Nvidia might take up to three months for Nvidia to reallocate or add supply chain capacity to produce H200s.
