Nvidia is racing to increase production of its H200 artificial intelligence chips after receiving a surge of orders from Chinese technology companies, and has asked Taiwan Semiconductor Manufacturing Co (TSMC) to ramp up manufacturing capacity, according to people familiar with the matter.
Chinese technology firms have placed orders for more than 2 million H200 chips for delivery in 2026, far exceeding Nvidia’s current inventory of about 700,000 units, two of the sources said.
Nvidia has approached TSMC, the world’s largest contract chipmaker, to begin producing additional H200 processors, with manufacturing expected to start as early as the second quarter of 2026, one of the sources said. The precise number of extra chips Nvidia plans to order has not yet been finalised.
The discussions and the scale of Chinese demand have not been reported previously.
The move underscores Nvidia’s challenge in balancing surging global demand for AI accelerators with constrained manufacturing capacity, while also navigating geopolitical restrictions on technology exports.
The H200 is Nvidia’s latest high-performance chip designed for training and running large artificial intelligence models and is widely used in data centres powering generative AI services.
However, shipments of the H200 to China remain politically sensitive. Although the administration of U.S. President Donald Trump has recently allowed exports of the chip to China, Beijing has not yet formally approved the deliveries, raising uncertainty over whether the orders can be fulfilled.
“The situation creates both an opportunity and a risk for Nvidia,” one source said. “On the one hand, Chinese demand is extremely strong. On the other, the company has to be careful about regulatory approvals and the risk of further export controls.”
The sources said Nvidia has decided which H200 configurations it will offer to Chinese customers and has set prices at around $27,000 per chip. Pricing details had not been reported previously.
If fulfilled, the Chinese orders would account for a substantial portion of Nvidia’s global output of advanced AI processors, potentially tightening supplies for customers elsewhere in the world, including U.S. and European cloud computing firms that are also racing to expand their AI infrastructure.
Nvidia has not publicly disclosed how it allocates chip supply across regions, and both Nvidia and TSMC declined to comment on the matter.
Industry analysts said any further strain on supply could push up prices and lengthen delivery times for AI hardware globally.
“Demand for AI accelerators continues to outstrip supply, and every large order has ripple effects across the market,” said Lin Wei, a semiconductor analyst at a Taipei-based research firm. “If Chinese demand absorbs a bigger share of production, others will feel it."
