US Pressure Kills Oracle's $7 Billion AI Chip Leasing Deal with Tencent

Oracle has canceled a $7 billion agreement to lease advanced AI chips to Tencent after U.S. officials raised concerns about the deal directly with the company. The Financial Times reported the cancellation on Nov. 4, citing two sources, and the collapse removes what was Tencent's largest overseas AI chip leasing arrangement.

The contract, signed earlier this year, committed Oracle to provide Tencent with access to up to 100,000 advanced AI chips across multiple data centers in Southeast Asia over five years. Tencent had already paid roughly 30% of the total contract value upfront. The company planned to use the arrangement to reach high-performance chips that are difficult to buy inside China, supporting its work on AI models and autonomous AI agents.

The mechanism matters here. U.S. export controls restrict the sale of advanced chips to China, but leasing computing power from data centers outside the country has remained legal under current regulations. Chinese companies have relied on this workaround for years. The Tencent case shows a different lever: Washington did not change the rules, it signaled concern about one specific contract, and the contract disappeared. That precedent suggests the leasing channel can be narrowed without any formal regulatory change.

Why did the deal collapse now?

According to the FT's sources, U.S. officials communicated their concerns about the arrangement to Oracle, and Oracle then canceled the contract. The report does not describe a formal enforcement action; the pressure appears to have been conveyed and acted upon within the existing regulatory environment.

The timing compounds the pressure on Chinese buyers. Just one day before the cancellation surfaced, reports emerged that Nvidia had asked its cloud customers to scale back their China-related business. Nvidia has told cloud customers that at least 51% of their services must be sold to non-Chinese companies, a condition that applies to new contracts. That threshold makes it structurally harder for Chinese firms to buy computing capacity from overseas providers, because every new deal must now be weighted toward other customers.

Tencent is being squeezed from the opposite direction as well. Senior officials from Guangdong Province visited Tencent's headquarters and urged the company to use and support domestically produced semiconductors. Beijing is pushing its technology firms toward domestic chipmakers as part of a broader drive for technological self-sufficiency, even though domestically produced chips still have clear limitations in training large-scale AI models. Chinese companies therefore remain heavily dependent on overseas computing resources, and that dependency is exactly what the canceled Oracle deal and Nvidia's new contract terms now constrain.

What is happening with chip supply?

The squeeze extends beyond the most advanced AI chips into Nvidia's mainstream professional products. Alibaba, Tencent and ByteDance have together ordered approximately 6 million of Nvidia's RTX Pro 5500 chips. Deliveries over the next four months are expected to total just 1% of that order volume, according to the reports.

The shortfall reflects a production decision, not a new export rule. Nvidia has concentrated its manufacturing capacity on next-generation products, which has limited its ability to supply existing models. For Chinese buyers, the practical effect is a shortage on top of the access restrictions: even chips they can order are not arriving in meaningful volume.

How did markets react?

Investors marked the news on Nov. 4. Oracle shares fell 2.0% in U.S. trading, while Nvidia declined 3.2%. Oracle's drop also reflected broader weakness among data center operators following the U.S. midterm election results, so the share price move cannot be attributed to the Tencent cancellation alone.

Taken together, the developments mark a shift in how U.S. technology restrictions reach Chinese customers: through individual contract interventions and commercial conditions set by suppliers, rather than only through published export rules. Unless the leasing channel is clarified or domestic Chinese chips close the training gap, Chinese AI developers will face rising costs and shrinking options for large-scale compute.