U.S. export controls currently prevent the direct sale of advanced American technologies, including Nvidia graphics processing units (GPUs), to China. As a result, the Chinese AI company Moonshot AI is reportedly exploring moving its data and compute workloads out of China to gain access to recent Nvidia hardware.
According to CNBC, Moonshot AI is negotiating with at least one European data center to rent compute capacity containing 8,000 or more Nvidia GPUs, mainly for inference tasks, while also expressing interest in acquiring capacity for model training. The company is targeting access to Nvidia’s latest Blackwell-series AI GPUs.
Legal loophole and legislative response
Tom's Hardware notes that a legal loophole allows Chinese firms to use American technology by operating in foreign data centers when direct import is barred. The U.S. House of Representatives has passed the Remote Access Security Act to try to close that loophole; the bill still needs to clear the Senate and receive the President’s signature before becoming law.
Meanwhile, several Chinese AI labs are taking advantage of the current situation by training and running models in overseas data centers that house Nvidia chips.
Nvidia's capacity limit for Chinese customers
Separately, Nvidia has instructed data center operators that going forward no more than 49 percent of a facility’s capacity may be used by Chinese companies. This global limitation is intended to reduce direct access to the technology by Chinese customers, even if they seek it through foreign infrastructure.
AI competition between the U.S. and China
The United States and China are engaged in intense competition to develop leading AI models and achieve global influence through their technologies. According to the reporting, Moonshot AI’s model Kimi K3 lags behind some U.S. competitors but can be simpler to run and potentially much cheaper, which helps explain the company’s interest in renting large-scale foreign GPU capacity.
Taken together, Chinese firms’ shift toward foreign data centers, U.S. legislative efforts, and Nvidia’s capacity restrictions are shaping near-term strategies for global access to advanced AI hardware.
