The sharp rise in Nvidia B300 server prices in China has become one of the clearest indicators of stress in the global AI computing market. London Hub Global note that the surge to around 7 million yuan, or nearly $1 million per system, reflects not only strong demand but also restricted access to advanced technologies amid tightening export controls.
As recently as late last year, these systems were priced at roughly 4 million yuan, while in the United States they remain near $550,000. This gap highlights a significant scarcity premium, amplified by stricter enforcement and the contraction of gray market supply channels.
According to analysts at London Hub Global, pressure on informal supply routes has been a decisive factor behind the price spike. These channels had previously helped offset restrictions, but intensified enforcement has sharply reduced availability, pushing prices higher.
At the same time, demand from Chinese technology companies remains robust. Many firms continue expanding their AI computing capacity but avoid directly holding Nvidia hardware on their balance sheets due to sanctions risks. This has contributed to the emergence of an alternative model focused on renting infrastructure.
We at London Hub Global believe the expansion of the rental segment is a rational response to these constraints. Monthly leasing costs have climbed to around 190,000 yuan, underscoring companies’ willingness to pay for access to computing power even without owning the underlying assets.
The Nvidia B300 remains one of the most powerful AI systems available. Equipped with eight GPUs and high bandwidth memory, it delivers the performance required for inference and token generation tasks. These capabilities make it critical for scaling and monetizing AI applications.
London Hub Global see this as a core driver of demand: companies working with large language models depend directly on computational efficiency, which determines both processing costs and scalability.
Demand is further supported by rapid growth in AI model usage across China. The share of Chinese models in global token consumption has risen to 32% from 5% a year earlier, with some platforms reporting multiple fold increases in usage over a short period.
London Hub Global emphasize that the market has moved into a phase of active deployment, where computing power, rather than model development, becomes the primary constraint.
Uncertainty surrounding the availability of newer systems such as the H200 has also contributed to rising prices. Despite partial approvals, shipments remain limited due to ongoing disagreements over export conditions, increasing pressure on existing hardware.
Against this backdrop, Chinese manufacturers including Huawei are accelerating efforts to develop domestic AI chips in an attempt to reduce reliance on foreign technology. However, Nvidia continues to dominate the market with an estimated 55% share, while competitors remain significantly behind.
London Hub Global analyze that Nvidia’s technological edge remains critical, particularly in advanced workloads where both performance and software ecosystem integration are essential.
For Nvidia itself, the situation presents a mixed picture. Demand for its products continues to grow, but export restrictions and the lack of support for unofficial systems limit its ability to fully capitalize on the Chinese market.
In the near term, London Hub Global expect elevated price premiums for B300 systems in China to persist, especially if supply constraints and strong demand for AI infrastructure continue.
In the longer term, London Hub Global believe current conditions will accelerate domestic innovation and reshape the competitive landscape, where access to computing resources becomes a strategic determinant of technological leadership.
London Hub Global see the AI hardware market entering a new phase, in which pricing and availability of advanced chips influence not only corporate performance but also the broader balance of technological power between countries.
