Key Moments
- Nvidia B300 servers in China have climbed to about 7 million yuan (approximately $1 million) each, nearly double late last year’s level, according to industry sources.
- The price spike has coincided with a U.S. crackdown on chip smuggling, constraining grey-market supply while Chinese tech firms continue to drive strong AI computing demand.
- A similar B300 server is priced at about $550,000 in the United States, highlighting the scarcity premium in China under tighter export curbs.
China AI Demand Pushes Nvidia B300 Prices to $1 Million
Strong appetite for artificial intelligence computing power in China has pushed the price of Nvidia’s B300 server to roughly 7 million yuan (about $1 million) per unit, nearly double its level late last year, according to four industry sources. The rise has accelerated as efforts to clamp down on chip smuggling have squeezed black-market and grey-market channels that had been critical for obtaining restricted hardware.
The B300 is described by the sources as Nvidia’s most advanced and powerful server currently used for AI applications in China. They said prices had been edging higher since the beginning of the year but surged more sharply once informal supply routes came under pressure. The sources requested anonymity due to the sensitivity of the topic and indicated that Reuters was the first to report that B300 servers had reached the million-dollar mark in China.
The sources added that Chinese technology companies are continuing to drive robust demand for compute capacity, even as many of them seek to avoid directly holding Nvidia equipment on their balance sheets to reduce the risk of running afoul of U.S. sanctions.
Nvidia Response and Export Compliance
In a statement to Reuters addressing questions about the situation, Nvidia said that the B300 is restricted from sale in China and emphasized that its partners must adhere to strict compliance requirements.
“As systems become increasingly large and complex, unlawful diversion is a recipe for failure,” it warned in a statement.
“Nvidia does not provide any service or support for such systems, and the enforcement mechanisms are rigorous and effective.”
Scarcity Premium Under Tighter U.S. Export Curbs
Two of the sources said that a B300 server configuration with eight B300 GPUs is priced at about $550,000 in the United States, up from around $500,000 late last year. In China, the same server has climbed from roughly 4 million yuan late last year to about 7 million yuan now, illustrating the scarcity premium that has emerged under tighter U.S. export restrictions.
| Region | Timing | Approximate B300 Server Price |
|---|---|---|
| United States | Late last year | $500,000 |
| United States | Current (as cited) | $550,000 |
| China | Late last year | About 4 million yuan |
| China | Current (as cited) | About 7 million yuan (about $1 million) |
The rising prices come as Chinese internet and AI firms seek the most cost-efficient infrastructure to generate tokens, the fundamental units of text processed by AI models, in an effort to commercialize both their models and their computing platforms.
The sources said the tightening of supply followed the March prosecution in the United States of Yih-Shyan “Wally” Liaw, a co-founder of Nvidia partner Supermicro. As supply has become constrained and prices escalated, some companies have found outright purchases economically unviable and have begun to consider renting capacity instead. According to the sources, rental rates have reached as high as 190,000 yuan per month on one-year short-term contracts.
Rising Token Usage and AI Workloads in China
Morgan Stanley said in a note that Chinese AI models nearly tripled their share of global token usage to 32% in March 2026, up from 5% a year earlier. The increase was attributed to advances in coding-related tasks and agentic capabilities.
The note also said that MiniMax, Zhipu and Alibaba’s Qwen each reported token volumes rising as much as six to seven times in February and March compared with December, underscoring the acceleration in AI workloads that is underpinning demand for high-end compute hardware.
B300 Capabilities and Market Positioning
Nvidia’s B300 is equipped with 288 GB of high-bandwidth memory and delivers 14 petaFLOPS of computing performance at FP4 precision. According to the description in the article, this places the chip among the most powerful options available for AI inference workloads.
Nvidia and its partners, including Supermicro, began shipping the B300 in September. Since then, ongoing uncertainty around exports of Nvidia’s H200 chips has added further support to B300 pricing. Although both governments have approved H200 exports, actual shipments to China have not taken place yet because of disagreements over the terms governing the sale.
Competitive Landscape in China’s AI Chip Market
The article states that Huawei and other Chinese AI chip designers have been moving to capitalize on the unresolved issues surrounding H200 exports. These firms are seeking to chip away at Nvidia’s reported 55% market share in China. Competing supplier AMD is cited as holding a 4% share in the same market.
TradingPedia.com is a financial media specialized in providing daily news and education covering Forex, equities and commodities. Our academies for traders cover 
