Tencent’s Japan Workaround: How a Cross-Border Cloud Deal Unlocks Advanced Nvidia Power
- Induqin
- 22 hours ago
- 2 min read

Tencent is accessing advanced Nvidia chips by renting overseas computing power through Japan‑based Datasection, bypassing U.S. export restrictions. The $12 billion arrangement relies on Datasection’s large AI data centers using Blackwell processors. The deal highlights policy loopholes, the rise of “neocloud” providers, and how Chinese tech firms are shifting AI workloads abroad.
A novel partnership is allowing one of China’s largest technology companies to tap cutting‑edge Nvidia computing power without breaching U.S. export restrictions. Tencent has arranged access to advanced semiconductors through Datasection, a Japan‑based firm that has shifted its focus from marketing services to operating large‑scale AI data centers, according to reporting by the Financial Times.
At a facility near Osaka, Datasection controls roughly 15,000 of Nvidia’s high‑end Blackwell B200 processors. Rather than exporting the chips to China, the Japanese company is renting substantial portions of that computing capacity to Tencent via third‑party agreements. The arrangement, valued at around $12 billion, allows Tencent to use processing power that cannot be shipped directly to China under current U.S. rules—exposing a regulatory gap in Washington’s export controls.
Datasection’s rapid transformation has made it a prominent player in the fast‑growing “neocloud” market, where companies rent GPU capacity to global clients. Comparable operators include CoreWeave in the United States and Nebius in Europe. Beyond Japan, Datasection is constructing an AI data center in Sydney outfitted with Nvidia’s next‑generation B300 chips, hardware that exceeds what Chinese buyers are permitted to purchase outright. The first tranche of 10,000 B300 processors reportedly cost $521 million, and the site is expected to primarily support Tencent’s AI workloads. Datasection chief executive Norihiko Ishihara has described the project as the world’s first hyperscale AI cluster built on B300 technology.
The Tencent‑Datasection collaboration takes advantage of changes in U.S. policy. Export control measures introduced during the Biden administration to close such overseas‑use loopholes were reversed in May under President Donald Trump, clearing the way for Datasection’s Osaka deal. While the U.S. recently approved sales of lower‑performance chips to China, leading Chinese tech firms—including Tencent, Alibaba, and ByteDance—continue to rely on foreign data centers to meet their most demanding AI needs. Industry analysts say that leasing GPU capacity abroad can be more appealing than investing in downgraded hardware.
For Datasection, these dynamics have fueled rapid growth. The company’s shares climbed about 18% earlier this year, though they have since fallen by roughly half from a peak of ¥4,000 amid concerns about aggressive spending and pressure from short sellers. Datasection has also raised ¥50 billion through equity warrants sold to Singapore‑based First Plus Financial Holdings, which is owned by a Chinese national.
Despite heightened scrutiny, the company is moving ahead with international expansion plans, including a push into Europe. It has added prominent figures to its board, such as Spanish politician Pablo Casado Blanco and John Ellis Bush Jr., a member of the U.S. Bush family. Ishihara has argued that global demand for GPU capacity is strong enough to absorb any future regulatory shifts.
For Tencent, the Japanese arrangement highlights how China’s largest technology groups are adapting to geopolitical constraints by shifting AI training and computing overseas. The deal underscores lingering questions about the effectiveness of U.S. export controls—and signals the rising influence of neocloud providers in the intensifying global competition over artificial intelligence infrastructure.



