Nvidia tightens China sales terms as H200 demand collides with geopolitical risk

Nvidia

• Nvidia now requires full upfront payment for H200 AI chips from Chinese buyers
• New sales terms reflect regulatory uncertainty and US-China tech tensions
• Chinese firms have ordered over two million H200 chips for 2026 delivery

Did our AI summary help?

Nvidia is reportedly tightening its sales terms for Chinese customers by requiring full upfront payment for its H200 artificial intelligence chips, reflecting growing uncertainty around regulatory approvals on both sides of the Pacific. According to a report from Reuters, the US chipmaker is asking buyers in China to pay the entire amount in advance, with no scope for refunds or order changes once a deal is signed.

The shift marks a significant departure from Nvidia’s earlier commercial practices in the region. In the past, some customers were allowed to place partial deposits when ordering high-end processors. Under the new terms, those options have largely disappeared. While a small number of buyers may still be permitted to rely on commercial insurance or asset-backed collateral, the overall framework is far more rigid than before. Nvidia has declined to comment publicly on the reported changes.

The tougher stance appears to be closely linked to political and regulatory uncertainty. While China is expected to allow Nvidia to sell the H200 domestically, approval is not unconditional. As previously reported by Bloomberg, Beijing wants to ensure the chips are not used by the military, state-owned enterprises, or within sensitive infrastructure projects. These restrictions complicate compliance and increase the risk profile for suppliers and buyers alike.

At the same time, Nvidia must navigate US export controls that have become increasingly unpredictable. The company has already felt the financial impact of policy shifts in Washington. During the Trump administration, new licensing requirements for exporting the H20 chip to China forced Nvidia to write down inventory worth $5.5 billion. That episode underscored how quickly regulatory decisions can translate into material losses, particularly when advanced chips are caught in the crossfire of geopolitical rivalry.

Against this backdrop, asking for full upfront payment serves as a form of risk management. By securing cash before shipment, Nvidia reduces its exposure if approvals are delayed, denied, or retroactively constrained. It also shifts a greater share of uncertainty onto customers, who must now commit capital without guarantees that orders can be modified or cancelled.

Despite these headwinds, demand for Nvidia’s H200 chips in China remains strong. Reuters reports that Chinese companies have placed orders for more than two million units for delivery in 2026, prompting Nvidia to increase production. The appetite reflects the central role of high-performance GPUs in training and deploying large AI models, an area where Nvidia continues to dominate despite growing competition.

The situation highlights the delicate balance Nvidia is trying to maintain. On one side is surging global demand for its AI hardware, particularly from markets eager to keep pace with advances in machine learning and data centres. On the other is the political reality of operating between two governments with increasingly divergent priorities and escalating concerns around technology transfer.

For now, Nvidia appears to be prioritising financial protection over commercial flexibility. Requiring full payment upfront may slow some deals, but it limits downside risk at a time when regulatory outcomes remain uncertain. As US China tech tensions show little sign of easing, such cautious approaches are likely to become the norm rather than the exception for companies operating at the cutting edge of semiconductor technology.