Key takeaways
- Nvidia reportedly requires majority non-Chinese AI cloud customers
- Overseas data centers face growing restrictions on Chinese demand
- China exposure adds pressure to Nvidia's struggling stock
Nvidia is reportedly directing its global data center partners to reduce their reliance on Chinese customers, potentially closing another route through which China's technology companies access advanced American artificial intelligence chips.
According to the Financial Times, Nvidia has instructed cloud computing providers to ensure at least 51% of their services go to non-Chinese companies. The newspaper cited five people familiar with the matter, describing a policy communicated verbally through sales teams rather than formal documentation.
The development comes as Nvidia stock has fallen more than 13% over the past month, with investors questioning AI infrastructure spending despite strong demand for the company's chips.
Nvidia's 51% Rule Could Reshape Overseas AI Data Centers
The reported restrictions apply globally to new cloud computing contracts rather than existing customer agreements.
Chinese companies have increasingly rented overseas computing capacity to train AI models without directly importing Nvidia's most advanced processors, which remain subject to US export controls.
Southeast Asian data centers have become particularly important because Chinese technology companies represent a substantial share of demand in several regional markets.
The new requirements could force cloud providers to find additional American, European or other non-Chinese customers before expanding their computing infrastructure.
However, separate reporting by The Information characterized Nvidia's instructions as nonmandatory guidance that has not been enforced as a strict customer-allocation rule.
That distinction matters because the financial consequences depend on whether operators must actually reject Chinese contracts.
The move also raises questions about previously reported overseas partnerships, including Tencent's $7 billion Oracle arrangement involving approximately 100,000 advanced AI chips.
Neither company has publicly confirmed that agreement, and there is no evidence Nvidia's reported instructions have changed its terms.
China's AI Chip Shortage Creates a $2.8 Million Problem
Nvidia's latest restrictions would arrive after years of escalating US controls on advanced semiconductor exports.
Its B300 chips cannot legally be sold directly into China under existing restrictions, creating substantial pricing disparities between international and Chinese markets.
According to the Financial Times, an Nvidia B300 server rack selling internationally for approximately $600,000 recently fetched $2.8 million in China.
That represents a premium exceeding 360%, highlighting the scarcity of advanced computing equipment.
Meanwhile, Nvidia has reportedly reduced its list of approved Asian AI chip customers by more than half through stricter compliance screening.
The company previously disclosed that H200 shipments to China represented less than 1% of quarterly data center revenue despite limited US export licenses.
Earlier discussions about reopening chip purchases for ByteDance and Alibaba had suggested some restrictions might ease.
However, tougher cloud access requirements could offset those potential gains.
What the China Restrictions Mean for Nvidia Stock
The immediate financial risk may fall on cloud operators that depend heavily on Chinese customers rather than Nvidia itself.
Some providers could struggle to replace that demand, potentially delaying future data center expansion and GPU purchases.
For Nvidia, the restrictions introduce another complication as the company faces a broader semiconductor stock selloff.
The company also risks encouraging Chinese developers to accelerate investment in domestic alternatives, including Huawei and Cambricon.
That trend is already visible in China's AI chip market, where domestic suppliers have gained opportunities as American export restrictions tighten.
