Amid the remote deserts of Xinjiang, in regions like Yuli County, China is quietly executing one of its most ambitious tech initiatives. According to Bloomberg, the government and private tech players are planning to install over 115,000 Nvidia AI chips across 39 data centers, with the goal of building a powerful national AI infrastructure—even under the constraints of U.S. export bans.

The project’s desert location is no coincidence: it offers geographic isolation, renewable energy potential, and strong state incentives. But above all, it raises a critical question: Is the U.S. truly stopping China’s AI momentum, or inadvertently enabling it through regulatory blind spots?

Dependence on Banned Chips

Two Biden administration officials estimate that China already possesses around 25,000 restricted Nvidia H100 and H200 chips, despite clear export restrictions. While Nvidia publicly denies any involvement in illicit sales, U.S. Commerce officials warn that large-scale smuggling operations are real and ongoing, often routed through intermediary nations and third-party shell companies.

Firms like DeepSeek are striving to compete with the likes of OpenAI and Google, but without access to cutting-edge hardware, the road ahead remains steep. Still, China appears committed to closing the gap—by any means necessary.

Is Nvidia Playing Both Sides?

Although Nvidia claims compliance with all U.S. export laws, some analysts suggest the company has an incentive to quietly tolerate regulatory loopholes—especially when indirect sales through global resellers and subsidiaries continue to generate significant revenues.

This “two-faced” strategy allows Nvidia to publicly align with U.S. policy while benefiting from China’s insatiable AI demand, even without officially doing business there.

China’s AI Vision vs. Washington’s Controls

While China’s desert-based data centers are still dwarfed by U.S. hyperscalers like AWS or Azure, they reflect a strategic pivot: gradual development of self-reliant AI infrastructure aimed at reducing dependence on Western tech.

It’s all part of the broader “Made in China 2025” agenda—an ambitious effort to place AI and deep tech at the center of national economic and military competitiveness.

Third-Country Workarounds

Direct chip transfers into China are blocked, but Beijing has increasingly turned to countries like the UAE, Malaysia, and Eastern European states to circumvent those restrictions. With cleverly registered subsidiaries and masked “civilian research,” advanced U.S. semiconductors continue to find their way into Chinese hands.

This creates a semi-legitimate global supply chain where tracing end-use becomes practically impossible. In essence, China opens new backdoors faster than Washington can close the front ones.

An AI Ecosystem Beyond Washington’s Reach

In response to tightening U.S. sanctions, China is deepening partnerships with countries such as the UAE, Malaysia, and Kazakhstan. These relationships allow advanced technology transfers under the guise of civilian or academic use. This strategy not only undermines U.S. export controls, but also helps form a new global AI ecosystem—one that operates independently of Washington. The outcome: China’s tech stack continues to advance rapidly—even under pressure.

Conclusion: China Is Winning in the Shadows

Ultimately, the AI conflict between China and the U.S. is not just about chips—it’s about global control over the next wave of innovation. While Washington builds firewalls, Beijing is building bridges, networks, and alternative routes.

Even if the U.S. doesn’t intend to help China, it may already be doing so—by failing to plug every leak.