Nvidia is requiring Chinese customers to make full upfront payments for its H200 artificial intelligence chips, as the company seeks to protect itself against uncertainty surrounding regulatory approval from Beijing, according to two sources familiar with the matter.
Under the revised terms, Nvidia is enforcing unusually strict conditions that require customers to pay the full order value in advance. Orders cannot be canceled, refunded, or modified once placed, the sources said. In limited cases, customers may be allowed to use commercial insurance or asset-backed collateral instead of cash payments.
While Nvidia has historically required advance payments from Chinese clients, customers were previously permitted to submit partial deposits rather than full prepayment. For the H200 chip, however, Nvidia has adopted a far more rigid stance due to the lack of clarity over whether Chinese authorities will approve the imports.
The sources requested anonymity because the information has not been made public. Nvidia and China’s industry ministry had not responded to requests for comment at the time of publication.
Chinese technology firms have reportedly placed orders for more than 2 million H200 chips, each priced at approximately $27,000, according to Reuters. This demand significantly exceeds Nvidia’s available inventory of roughly 700,000 units.
Although Chinese chipmakers such as Huawei have developed domestic AI processors, including the Ascend 910C, their performance still falls short of Nvidia’s H200 when it comes to large-scale training of advanced artificial intelligence models.
According to one source, Chinese regulators recently asked some technology companies to temporarily pause their H200 orders while authorities determine how many domestically produced chips customers must purchase alongside each imported H200 unit. The Information first reported the pause earlier this week.
Nvidia CEO Jensen Huang said on Tuesday that demand for the H200 remains strong, adding that the company has accelerated its supply chain to increase output. Huang noted that he does not expect a formal approval announcement from Beijing, stating that the presence of purchase orders would signal regulatory clearance.
Balancing demand and regulatory risk
The tightened payment requirements highlight Nvidia’s effort to balance surging demand from China with regulatory uncertainty in both the U.S. and China.
The U.S. government had previously banned exports of advanced AI chips to China, but President Donald Trump reversed the policy last month, allowing H200 sales to proceed subject to a 25% fee paid to the U.S. government.
Nvidia has previously absorbed heavy losses tied to policy shifts. Last year, the company wrote down $5.5 billion in inventory after being barred from selling its H20 chip to China. Although the U.S. later reversed that restriction, China has since prohibited H20 imports.
By requiring full prepayment for H200 orders, Nvidia effectively shifts financial risk to customers, who must commit significant capital without assurance that Chinese regulators will approve the shipments or that the chips can be deployed as planned.
Chinese internet companies, including ByteDance, see the H200 as a major upgrade over existing alternatives. Nvidia’s second-most powerful chip delivers roughly six times the performance of the now-blocked H20 model, which was originally designed for the Chinese market.
Nvidia plans to fulfill early H200 orders from existing inventory, with initial shipments expected to arrive before the Lunar New Year holiday in mid-February. The company has also approached Taiwan Semiconductor Manufacturing Co about expanding H200 production, with additional manufacturing capacity potentially coming online in the second quarter of 2026.
Expanding output remains challenging, as Nvidia is simultaneously transitioning from its current flagship Blackwell chip to the next-generation Rubin architecture. The company is also competing with firms such as Google for limited advanced chipmaking capacity at TSMC.
