A data center outside Osaka now runs thousands of Nvidia’s newest chips for Tencent, giving the company a path around Washington’s hardware limits while keepingA data center outside Osaka now runs thousands of Nvidia’s newest chips for Tencent, giving the company a path around Washington’s hardware limits while keeping

Tencent uses Datasection’s Osaka and Sydney data centers to access Nvidia’s newest GPUs despite U.S. export limits

A data center outside Osaka now runs thousands of Nvidia’s newest chips for Tencent, giving the company a path around Washington’s hardware limits while keeping everything within legal bounds.

The GPUs sit inside a site owned by Datasection, a Japanese group that once sold marketing solutions but flipped into AI infrastructure last year.

That pivot pulled in more than $1.2 billion in contracts tied to Tencent through a third-party partner, according to people who know the deal.

These contracts cover a big slice of Datasection’s first batch of 15,000 Blackwell processors, all parked in Japan at a moment when China cannot import the same chips.

Datasection’s move placed it among Asia’s rising neocloud players, a group that includes CoreWeave in the U.S. and Nebius in Europe. These operators rent out Nvidia hardware at a huge scale and serve global tech firms hungry for compute.

Norihiko Ishihara, Datasection’s chief executive, said the demand shift has been wild. He said “less than half a year ago…5,000 B200 chips were sufficient to support AI models,” but that the floor has doubled and “10,000 should be the minimum requirement.”

His comment sums up the pace of this market and the reason companies like Tencent are moving fast to secure GPU capacity offshore.

Expanding offshore GPU access and building long-term contracts

China’s largest platforms have been forced into overseas workarounds after U.S. export rules blocked Nvidia’s best chips from entering the country. Datasection grew as this pressure intensified, with Trump canceling a Biden-era plan to close the loophole in May.

The Osaka project was finalized soon after. Earlier this month, the White House approved a lower-tier Nvidia chip for China, which could help Tencent start building its own domestic sites again.

But analysts say the offshore approach may stay popular. Lin Qingyuan at Bernstein Research said renting compute abroad may be “the more attractive choice for Chinese tech groups.”

Tencent, Alibaba, and ByteDance are all training models outside China and selling the output, said people familiar with these operations.

Datasection expects to run more than 100,000 Nvidia processors across future facilities. Its first 15,000 chips are mostly locked into Tencent contracts for three years, with options to extend. Ishihara would not confirm Tencent by name, calling it only a “major customer,” citing confidentiality.

Datasection’s stock price is up nearly 185% this year, though it has fallen from a summer high above ¥4,000 because traders fear over-investment and react to a short seller’s attack on the business.

Building Sydney’s hyperscale B300 cluster and facing scrutiny

In July, Datasection agreed to buy 5,000 B200 chips for $272 million for the Osaka site, supported by a $406 million three-year contract with one of the world’s largest cloud firms. Server crates filled with new GPUs arrived in August.

Soon after, Datasection and its partner signed another three-year, $800 million agreement to build a second AI data center in Sydney. This new cluster will use tens of thousands of B300 chips, which outperform the versions Nvidia is allowed to sell into China.

Datasection said the first 10,000 B300s will cost $521 million. Ishihara said the site will be “the world’s first hyperscale AI cluster by using B300 chips.” People with knowledge of the project said Tencent is expected to be the main user.

Tencent said it follows all laws and that its use of “cloud computing services is both transparent and legal.”

Ishihara said GPU purchases are his biggest expense. He spreads that cost over five years, while customer deals usually run three years with a two-year extension option.

Contracts run through a partner to protect client data; for Tencent, that partner is Tokyo-based NowNaw. Datasection can cancel agreements if U.S.–China rules tighten again.

The company has also come under attack. A short seller questioned links to Tencent and to First Plus Financial Holdings, a Singapore-based investor.

Datasection said its projects are “in full compliance with all applicable laws and regulations.” Ishihara later said U.S. Commerce Department and Nvidia approvals were obtained for GPU use.

Datasection is raising ¥50 billion through warrants issued to First Plus, which could dilute shareholders by up to 200%. The investor, owned by a Chinese national, wants to keep its stake below one-third.

Ishihara said First Plus does not want to consolidate Datasection into its accounts and surrendered voting rights to avoid Japan’s foreign exchange control scrutiny. First Plus did not comment.

The company is now pushing into cloud services and targeting Europe. It hired Spanish politician Pablo Casado Blanco as chair and added John Ellis Bush Jr to its board.

Ishihara said high demand for GPU power makes it easy to replace customers if export rules change again. In a worst-case scenario, he said “we may have to stop the operation for, let’s say, one week,” calling the asset “very sexy.”

Sharpen your strategy with mentorship + daily ideas - 30 days free access to our trading program

Market Opportunity
Union Logo
Union Price(U)
$0.002958
$0.002958$0.002958
-4.05%
USD
Union (U) Live Price Chart
Disclaimer: The articles reposted on this site are sourced from public platforms and are provided for informational purposes only. They do not necessarily reflect the views of MEXC. All rights remain with the original authors. If you believe any content infringes on third-party rights, please contact service@support.mexc.com for removal. MEXC makes no guarantees regarding the accuracy, completeness, or timeliness of the content and is not responsible for any actions taken based on the information provided. The content does not constitute financial, legal, or other professional advice, nor should it be considered a recommendation or endorsement by MEXC.

You May Also Like

Huang Licheng Holds Controversial 25x ETH Long Position

Huang Licheng Holds Controversial 25x ETH Long Position

The post Huang Licheng Holds Controversial 25x ETH Long Position appeared on BitcoinEthereumNews.com. Key Points: Huang Licheng, known as “Machi,” holds a 25x leveraged
Share
BitcoinEthereumNews2025/12/22 03:49
UK and US Seal $42 Billion Tech Pact Driving AI and Energy Future

UK and US Seal $42 Billion Tech Pact Driving AI and Energy Future

The post UK and US Seal $42 Billion Tech Pact Driving AI and Energy Future appeared on BitcoinEthereumNews.com. Key Highlights Microsoft and Google pledge billions as part of UK US tech partnership Nvidia to deploy 120,000 GPUs with British firm Nscale in Project Stargate Deal positions UK as an innovation hub rivaling global tech powers UK and US Seal $42 Billion Tech Pact Driving AI and Energy Future The UK and the US have signed a “Technological Prosperity Agreement” that paves the way for joint projects in artificial intelligence, quantum computing, and nuclear energy, according to Reuters. Donald Trump and King Charles review the guard of honour at Windsor Castle, 17 September 2025. Image: Kirsty Wigglesworth/Reuters The agreement was unveiled ahead of U.S. President Donald Trump’s second state visit to the UK, marking a historic moment in transatlantic technology cooperation. Billions Flow Into the UK Tech Sector As part of the deal, major American corporations pledged to invest $42 billion in the UK. Microsoft leads with a $30 billion investment to expand cloud and AI infrastructure, including the construction of a new supercomputer in Loughton. Nvidia will deploy 120,000 GPUs, including up to 60,000 Grace Blackwell Ultra chips—in partnership with the British company Nscale as part of Project Stargate. Google is contributing $6.8 billion to build a data center in Waltham Cross and expand DeepMind research. Other companies are joining as well. CoreWeave announced a $3.4 billion investment in data centers, while Salesforce, Scale AI, BlackRock, Oracle, and AWS confirmed additional investments ranging from hundreds of millions to several billion dollars. UK Positions Itself as a Global Innovation Hub British Prime Minister Keir Starmer said the deal could impact millions of lives across the Atlantic. He stressed that the UK aims to position itself as an investment hub with lighter regulations than the European Union. Nvidia spokesman David Hogan noted the significance of the agreement, saying it would…
Share
BitcoinEthereumNews2025/09/18 02:22
Aave DAO to Shut Down 50% of L2s While Doubling Down on GHO

Aave DAO to Shut Down 50% of L2s While Doubling Down on GHO

The post Aave DAO to Shut Down 50% of L2s While Doubling Down on GHO appeared on BitcoinEthereumNews.com. Aave DAO is gearing up for a significant overhaul by shutting down over 50% of underperforming L2 instances. It is also restructuring its governance framework and deploying over $100 million to boost GHO. This could be a pivotal moment that propels Aave back to the forefront of on-chain lending or sparks unprecedented controversy within the DeFi community. Sponsored Sponsored ACI Proposes Shutting Down 50% of L2s The “State of the Union” report by the Aave Chan Initiative (ACI) paints a candid picture. After a turbulent period in the DeFi market and internal challenges, Aave (AAVE) now leads in key metrics: TVL, revenue, market share, and borrowing volume. Aave’s annual revenue of $130 million surpasses the combined cash reserves of its competitors. Tokenomics improvements and the AAVE token buyback program have also contributed to the ecosystem’s growth. Aave global metrics. Source: Aave However, the ACI’s report also highlights several pain points. First, regarding the Layer-2 (L2) strategy. While Aave’s L2 strategy was once a key driver of success, it is no longer fit for purpose. Over half of Aave’s instances on L2s and alt-L1s are not economically viable. Based on year-to-date data, over 86.6% of Aave’s revenue comes from the mainnet, indicating that everything else is a side quest. On this basis, ACI proposes closing underperforming networks. The DAO should invest in key networks with significant differentiators. Second, ACI is pushing for a complete overhaul of the “friendly fork” framework, as most have been unimpressive regarding TVL and revenue. In some cases, attackers have exploited them to Aave’s detriment, as seen with Spark. Sponsored Sponsored “The friendly fork model had a good intention but bad execution where the DAO was too friendly towards these forks, allowing the DAO only little upside,” the report states. Third, the instance model, once a smart…
Share
BitcoinEthereumNews2025/09/18 02:28