TechNewsReel
Live

Chinese AI Giants Use Cloud Rental Loophole to Access Banned Nvidia Blackwell Chips

Tencent and other firms are bypassing U.S. export bans by renting remote compute power from overseas neocloud providers.

TechNewsReel Newsroom · August 19, 2026

Chinese artificial intelligence giants are circumventing U.S. export restrictions on high-end Nvidia GPUs by renting compute power from overseas cloud providers. This strategy allows firms to access cutting-edge hardware remotely, bypassing the physical shipment bans designed to limit China's AI capabilities.

In a prominent example, Tencent has accessed Nvidia Blackwell B200 chips through a rental agreement with Datasection, a Japanese neocloud provider. According to reporting from the Financial Times via Wccftech, Datasection secured more than $1.2 billion in contracts with a single large customer—identified as Tencent—to provide access to a significant portion of its 15,000 Nvidia Blackwell processors. Datasection's infrastructure, which includes B200 and B300 AI chips, is deployed across facilities in Japan and Australia.

The Infrastructure Gap

The U.S. government has implemented strict export controls to prevent China from acquiring advanced AI chips, such as the H100 and Blackwell series, to curb the development of frontier AI and military applications. In response to these restrictions, Nvidia developed downgraded versions of its hardware specifically for the Chinese market, such as the H20. However, these compliant chips often lack the raw performance required for training the most advanced large-scale models.

By utilizing "neocloud" services in regions like Japan and Australia, Chinese firms can leverage the world's most powerful hardware without ever taking physical possession of the chips. The Financial Times described this approach as a "legal but geopolitically fraught strategy" that allows tech giants to maintain their competitive edge despite Washington's restrictions.

Strategic Implications

This rental loophole fundamentally undermines the strategic goals of U.S. export controls. While customs and shipping regulations can block the movement of physical crates of GPUs, they are less effective against the remote provision of compute services. This shift effectively moves the battleground from border security to cloud governance.

Industry analysts suggest that the ability to rent top-tier compute is far more compelling for Chinese firms than purchasing downgraded, China-specific hardware. The performance gap between the B200 and compliant alternatives is significant enough to justify the high cost of overseas rentals.

The Path Forward

As this loophole becomes more widely utilized, it is likely to prompt a policy shift in Washington. To close the gap, the U.S. government may implement "Know Your Customer" (KYC) requirements for cloud providers. Such regulations would force neocloud operators to verify the identity of their clients and block remote access for restricted entities.

For now, the reliance on third-party providers in neutral territories remains a viable path for Chinese AI development, though it leaves these firms dependent on the stability of overseas infrastructure and the evolving legal landscape of international cloud services.

Sources

Get a notification when a big story breaks. A few a day at most — no spam.