Nvidia‘s B300 AI servers are now selling for about $1 million in China, nearly double their roughly $550,000 price in the United States, after US export controls and a crackdown on chip smuggling sharply reduced supply.
The surge shows how AI hardware is turning into a scarcity trade, where access to chips now carries financial value beyond the technology itself.
If you are searching why Nvidia chips are so expensive in China, the answer is not just demand. It is that legal supply has been restricted while demand for AI computing has accelerated, forcing buyers into a tighter, riskier, and far more expensive market.
At first glance, this looks like a simple case of sanctions pushing up prices. The deeper financial reality is more disruptive. Restrictions have created a split market, where the same product carries two entirely different valuations depending on geography, compliance risk, and access routes. In China, the price of AI infrastructure is no longer set purely by performance or cost efficiency, but by scarcity and the difficulty of obtaining it.
The B300 server sits at the center of this shift. Designed for high-performance AI inference, it is one of the most powerful systems available, and demand for that computing capacity is rising quickly. According to Morgan StanleyChinese AI models have rapidly increased their share of global usage, meaning the need for processing power is expanding at the same time supply is being constrained. That imbalance is what drives the premium.
The financial mechanism is straightforward but powerful. When access is restricted, the price of the asset begins to reflect not just its function, but its availability and legal risk. Companies that still need advanced AI infrastructure are forced to pay significantly more through indirect channels, while others shift toward renting compute instead of owning it outright. In some cases, access to these systems is now being priced as an ongoing operating cost rather than a one-time capital investment.
This is where the market begins to change in ways that are less visible. Some Chinese firms are reportedly avoiding holding Nvidia hardware directly on their balance sheets to limit exposure to potential US sanctions. That shifts how AI infrastructure is financed and structured, pushing parts of the ecosystem into less transparent arrangements where ownership, control, and risk are separated.
What emerges is a different kind of AI economy. The value of hardware is no longer defined only by speed or capability, but by whether it can be legally acquired, maintained, and supported. Nvidia itself has made clear it does not support systems that fall outside compliance rules, which adds another layer of risk to buyers relying on restricted supply channels.
At the same time, this pressure is creating opportunity. Domestic competitors such as Huawei are positioning themselves to capture demand that can no longer be met easily by US suppliers. However, matching Nvidia’s performance at scale is not immediate, which keeps demand elevated and prices under pressure in the short term.
For investors and businesses, the implication is clear. AI demand is accelerating globallybut access to the most advanced infrastructure is becoming uneven and politically constrained. That raises costs in certain markets while strengthening pricing power for suppliers in others, even where direct sales are restricted.
The deeper insight is that export controls are no longer just policy tools. They are actively reshaping market dynamics, turning advanced semiconductors into controlled economic assets. What looks like a temporary price spike is actually part of a broader shift, where AI is evolving into a resource market defined by access, regulation, and strategic control.
In that environment, the companies that succeed will not only be those with the best technology, but those that can navigate supply constraints, manage regulatory exposure, and secure reliable access to compute. The price of Nvidia’s B300 in China is not just a reflection of scarcity. It is a signal that the economics of AI are changing.
