← Back to stories Detailed view of server racks with glowing lights in a data center environment.
Photo by panumas nikhomkhai on Pexels
凤凰科技 2026-05-25

Compute-power leasing firms say no new rules; overseas GPU buys “come with invoices,” industry sources say

Industry sources push back on rumors

Several sources in China’s compute-power leasing sector say they have not received any “new requirements” from regulators, pushing back against social-media reports of abrupt policy changes. It has been reported that major firms’ purchases of overseas accelerator cards are supported by commercial invoices and import paperwork, which industry players cite as routine compliance when sourcing GPUs abroad.

What this means for China’s AI compute market

The compute-power leasing business — which rents out GPU clusters for AI model training to startups and enterprises — has been under close watch as demand for large-scale accelerators soars. Some market participants pointed to purchases by major cloud and leasing firms, including well-known providers such as Baidu (百度), Alibaba Cloud (阿里云) and Tencent Cloud (腾讯云), as examples where overseas cards reportedly arrived with full invoicing and customs documentation. But those reports remain unverified and should be treated cautiously.

Geopolitics and export controls in the background

Why does documentation matter? Because advanced AI accelerators sit at the intersection of commercial demand and geopolitics. Since 2022 the U.S. and allied partners have tightened export controls on high-end chips to China and placed certain firms on entity lists. That context has increased scrutiny from customs and compliance teams, and it has been reported that companies are being more careful to show standard invoices to demonstrate legitimate commercial imports. Will regulators demand tighter rules anyway? For now, industry sources say no new directives have been issued publicly.

Stakes for startups and regulators

If true, the status quo gives breathing room to cloud tenants and compute renters who rely on overseas hardware to train models. But uncertainty remains. Market participants told reporters they are monitoring customs procedures and bilateral trade policy closely, and that any formal regulatory shift would quickly ripple through China’s fast-growing AI ecosystem. Reportedly, firms are keeping invoices and paperwork ready — just in case.

AISpace
View original source →