Nvidia has reportedly turned to Taiwan Semiconductor Manufacturing Company to increase production of its H200 artificial intelligence chips, as demand from Chinese technology firms continues to accelerate. According to a report by Reuters, discussions between the two companies are already underway.
Chinese technology companies are said to have placed advance orders exceeding 2 million H200 chips for delivery in 2026. At present, Nvidia holds an estimated inventory of around 700,000 units, according to sources familiar with the matter.
To meet future demand, Nvidia has asked TSMC to begin manufacturing additional H200 chips, with production expected to start in the second quarter of 2026. The precise volume of chips Nvidia plans to order has not been disclosed.
The talks with TSMC, along with the scale of Chinese demand and related pricing details, have not previously been reported. The potential expansion comes as Nvidia is simultaneously ramping up production of its newer Blackwell chips and preparing to launch its next-generation Rubin architecture.
The H200 belongs to Nvidia’s older Hopper platform and is manufactured using TSMC’s 4-nanometer process. Any increase in H200 output could further tighten global AI chip supplies, as Nvidia balances strong Chinese demand with limited availability in other regions.
Regulatory uncertainty remains a key risk. While the administration of U.S. President Donald Trump has recently allowed exports of the H200 to China, Beijing has not yet approved any shipments of the chip.
Nvidia has reportedly finalized which H200 variants it will offer to Chinese customers and plans to price the chips at approximately $27,000 each. Final pricing will vary depending on order size and customer agreements.
In a statement to Reuters, Nvidia said it continues to actively manage its supply chain. The company added that licensed H200 sales to approved Chinese customers will not affect its ability to supply clients in the United States. Nvidia also emphasized that China remains a highly competitive market, warning that broad export restrictions could ultimately benefit non-U.S. competitors.
Initial Chinese orders are expected to be fulfilled using Nvidia’s existing inventory. The first batch of H200 chips is anticipated to arrive before the Lunar New Year holiday in mid-February.
Of Nvidia’s current stock, roughly 100,000 units consist of GH200 Grace Hopper superchips, which combine Nvidia’s Grace CPU with its Hopper GPU. The remainder are standalone H200 chips. Both versions are expected to be offered to Chinese customers.
Demand has been driven primarily by large Chinese internet companies, which see the H200 as a significant upgrade over the chips currently available to them. An eight-chip H200 system is expected to cost around 1.5 million yuan, compared with approximately 1.2 million yuan for the discontinued H20 module.
The H200 delivers roughly six times the performance of the H20, a downgraded chip Nvidia previously designed for the Chinese market before shipments were blocked. Despite the higher price, Chinese buyers reportedly view the H200 as attractively priced, especially given that it represents an estimated 15% discount compared with grey-market alternatives selling for more than 1.75 million yuan.
