How Tencent Bypasses US Export Bans to Access Nvidia’s Latest GPUs via Datasection’s Osaka and Sydney Data Centers

• Why is Tencent Turning to Offshore GPU Solutions?
• The Financial Mechanics Behind the Deal
• How Does This Impact the Global AI Race?
• What Are the Risks and Contingencies?
• The Bigger Picture: A Fragmented Tech Ecosystem

In a strategic MOVE to circumvent US export restrictions, Tencent has secured access to Nvidia’s cutting-edge Blackwell and B300 GPUs through Datasection’s data centers in Osaka and Sydney. This $1.2 billion deal highlights the growing demand for AI infrastructure and the creative solutions Chinese tech giants are employing to stay competitive. Here’s a deep dive into the implications, financials, and future of this high-stakes arrangement.

Why is Tencent Turning to Offshore GPU Solutions?

With US export bans blocking China’s access to Nvidia’s most advanced chips, Tencent and other Chinese tech giants have been forced to look abroad. Datasection, a Japanese firm that pivoted from marketing solutions to AI infrastructure last year, has emerged as a key player in this space. Their Osaka facility houses 15,000 Blackwell B200 GPUs, while the upcoming Sydney cluster will feature 10,000 next-gen B300 chips—both unavailable in China due to US restrictions. "Six months ago, 5,000 B200s could handle our AI models," said Datasection CEO Norihiko Ishihara. "Now we need 10,000." This explosive demand explains why Tencent locked in three-year contracts (with two-year extension options) for these offshore resources.

The Financial Mechanics Behind the Deal

Datasection’s $406 million Osaka project (including 5,000 B200s for $272 million) and the $800 million Sydney B300 cluster represent a massive bet on the AI boom. Interestingly, Tencent routes payments through Tokyo-based NowNaw to obscure the arrangement—a common practice among Chinese firms navigating geopolitical tensions. Datasection’s stock has soared 185% this year, though it’s retreated from summer highs as investors worry about overcapacity. The company recently raised ¥50 billion via warrants issued to Singapore’s First Plus Financial Holdings, potentially diluting shareholders by 200%. "Our contracts account for regulatory risks," Ishihara noted, adding that US Commerce Department and Nvidia approvals were secured for all GPU deployments.

How Does This Impact the Global AI Race?

While the US recently approved export of a downgraded Nvidia chip to China, analysts like Bernstein’s Lin Qingyuan believe offshore leasing will remain popular. "It’s the safest option for Chinese tech firms," Lin observed. Tencent, Alibaba, and ByteDance all train models overseas and import the results—a workaround that’s spawned a new "neocloud" sector featuring players like CoreWeave (US) and Nebius (Europe). Datasection plans to deploy over 100,000 Nvidia GPUs across future sites, with Europe already in its sights through appointments like Spanish politician Pablo Casado Blanco as chairman.

What Are the Risks and Contingencies?

Short-sellers have questioned Datasection’s ties to Tencent and First Plus, though the company maintains full compliance with all regulations. Ishihara admits worst-case scenarios could force week-long operational pauses if US-China tensions escalate, but calls GPUs "exceptionally liquid assets" due to insatiable global demand. The Sydney B300 cluster—marketed as the world’s first hyperscale AI installation using these chips—illustrates how quickly the goalposts are moving. Where 15,000 GPUs seemed revolutionary in Q2 2025, by Q4 that’s barely enough for baseline operations.

The Bigger Picture: A Fragmented Tech Ecosystem

This saga underscores how geopolitics is reshaping tech infrastructure. As Tencent’s spokesperson stated: "Our cloud computing usage is transparent and legal"—a necessary disclaimer when operating in regulatory gray areas. With Datasection eyeing European expansion and Nvidia preparing its next-gen chips, the cat-and-mouse game between innovation and export controls shows no signs of slowing. One thing’s certain: in the AI arms race, necessity breeds inventive solutions.