Nvidia requires full upfront payment for H200 chips in China | REUTERS

Key Concepts

• Nvidia H200: Nvidia’s high-performance AI chip, crucial for large-scale AI model training.
• Upfront Payment: Requiring full payment before shipment, a deviation from standard industry practice.
• Export Controls: US regulations impacting the sale of advanced technology to China.
• AI Chip Demand: High demand for AI chips, particularly Nvidia’s, from Chinese tech companies.
• Security Concerns: Chinese government anxieties regarding the use of advanced chips by military and critical infrastructure.

Nvidia’s Payment Terms for H200 Chips to China

Nvidia is now demanding full upfront payment from Chinese customers ordering its H200 AI chips, according to reports from Reuters citing two unnamed sources. This represents a significant shift in standard business practices and is directly linked to ongoing uncertainty surrounding approvals from the Chinese government for these shipments. The new terms are described as “unusually strict,” explicitly prohibiting order cancellations, refunds, or any post-placement configuration changes. As of the time of reporting, neither Nvidia nor China’s Ministry of Industry and Information Technology have publicly responded to requests for comment regarding these changes.

Scale of Demand and Inventory Discrepancy

The demand for H200 chips from Chinese tech companies is substantial. Reuters previously reported that orders have already exceeded 2 million units. These chips are priced at approximately $27,000 each, resulting in a total order value exceeding $54 billion. This demand significantly outstrips Nvidia’s current inventory, which stands at approximately 700,000 H200 chips. This inventory shortfall is a key driver behind the stricter payment terms, suggesting Nvidia is prioritizing securing revenue given limited supply.

Performance Gap and Domestic Alternatives

While Chinese companies are developing their own AI processors, these alternatives currently lag behind the performance capabilities of Nvidia’s H200, particularly when it comes to the large-scale training of sophisticated AI models. This performance gap explains the continued strong demand for Nvidia’s products despite geopolitical tensions and potential restrictions. The reliance on Nvidia’s technology highlights a critical dependency within the Chinese AI ecosystem.

Potential Approvals and Restrictions

Recent reporting from Bloomberg suggests a potential compromise may be in the works. China is reportedly preparing to approve some H200 imports as early as this quarter, but with significant restrictions. These approvals will be limited to “select commercial uses.” Purchases by the Chinese military, sensitive government agencies, critical infrastructure operators, and state-owned enterprises will be explicitly prohibited due to national security concerns. This indicates a targeted approach by the Chinese government, aiming to balance the need for advanced AI capabilities with the imperative to protect sensitive sectors.

Implications of Upfront Payment

The requirement for full upfront payment is a risk mitigation strategy for Nvidia. It protects the company from potential financial losses if the Chinese government were to ultimately block the shipments after production has begun. It also provides Nvidia with immediate capital, which can be reinvested in further chip production or research and development. This move underscores the complex interplay between geopolitical factors, supply chain dynamics, and financial considerations in the high-stakes world of AI chip manufacturing.

Synthesis

Nvidia’s decision to demand full upfront payment for H200 chips to China reflects the heightened uncertainty surrounding US-China trade relations and export controls. The massive demand for these chips, coupled with limited supply, further exacerbates the situation. While China may approve some imports for commercial use, restrictions on sensitive sectors demonstrate ongoing security concerns. This situation highlights the strategic importance of AI chips and the lengths companies and governments are willing to go to secure access to this critical technology.