Nvidia says it never told Chinese companies to pay the full amount before receiving its H200 chips. The company responded after an article earlier claimed it had set harsh new payment terms.
In a written statement, a spokesperson said the company “would never require customers to pay for products they do not receive.”
A source allegedly told Reuters that while advance payment was already part of Nvidia’s terms in China, buyers usually got the option to place a deposit instead of paying in full. That changed for the H200 chip.
Nvidia got stricter this time. The same person said the company started demanding full payment from Chinese clients because it wasn’t clear whether Beijing would approve the shipments. So if buyers agreed, they’d be risking their money without knowing if they’d even get the product.
Chinese officials are now planning to allow some imports of the H200 within this quarter, according to people with knowledge of the situation.
They said the chip would be approved only for select commercial use, not for the military, sensitive government offices, state-run firms, or critical infrastructure. If any of those organizations want the chip, their applications will be reviewed one by one.
The rule is similar to China’s other restrictions. Apple devices and Micron chips have faced the same treatment. No official has made a public announcement yet. But people involved in the talks said the internal decisions are already moving forward.
The H200 is not a top-tier chip. It came out in 2023 and began shipping in 2024. It belongs to Nvidia’s Hopper generation, and it’s behind both the Blackwell and Rubin chips.
That made it acceptable under U.S. policy. In early December, President Donald Trump reversed a previous ban and approved the export, but slapped on a 25% surcharge. That gave Nvidia a shot at getting back into the world’s largest chip market.
Last year, Alibaba and ByteDance told Nvidia they want to order more than 200,000 units each, according to a person close to the talks.
Other companies, like DeepSeek, are also interested as they’re all trying to build faster models to compete with OpenAI and other U.S. tech companies.
But there’s still a problem. Beijing hasn’t said which buyers count as part of “critical infrastructure.” That term isn’t defined clearly, and companies like Alibaba or Baidu often work with state clients, just like Amazon or Microsoft do with U.S. federal agencies. So even if they’re private, they might still get blocked depending on how China sees it.
Nvidia hasn’t spoken directly to Chinese regulators. Executives at the CES tech show in Las Vegas said they’re waiting for answers.
They confirmed that license requests have already been filed in Washington, and they’re just waiting on final U.S. approval. They also said demand from China is strong, but no shipments will happen until both governments give the green light.
Back in 2025, China’s government told companies to stay away from Nvidia’s H20, a weaker AI chip that the U.S. had allowed. China’s cyberspace agency also told Alibaba to stop buying Nvidia’s RTX Pro 6000D, a workstation chip that could be used for AI systems. At the same time, Beijing started pushing for local chip production and offered $70 billion in new subsidies to boost its industry.
Huang, who runs Nvidia, said the rules set by U.S. policymakers cut the company’s market share in China from 95% to zero. But he added that the company still expects to grow overall. Back in October, Nvidia said it would make $500 billion from its data center chips by the end of 2026. This week, the company said it’s now likely to beat that estimate.
Nvidia is the top supplier of AI accelerators, the chips that train and run large AI systems. The H200 is still being used, even though it’s not the latest. Local Chinese chipmakers like Huawei and Cambricon grew fast during Nvidia’s absence. Both companies now say they plan to expand production in 2026.
The smartest crypto minds already read our newsletter. Want in? Join them.


