Oracle Drops $7 Billion Tencent AI Chip Lease Deal Under US Pressure

Oracle has walked away from a proposed $7 billion deal that would have let Tencent lease up to 100,000 advanced AI chips over five years. The Financial Times, cited by Reuters on November 4, reported that pressure from the Trump administration killed the arrangement before it was signed.

The numbers were substantial. The five-year agreement would have given the Chinese technology giant access to processors that US export controls bar from direct shipment to China, routed instead through data centres in Southeast Asia. It was, in effect, a workaround worth seven billion dollars. Workarounds of that size attract attention.

Why did the deal collapse?

US officials raised concerns about the proposed arrangement, according to the report. The structure — chips hosted outside China, serving a Chinese customer — sat in a grey zone that Washington appears determined to close. The Financial Times account makes clear that the scrutiny was decisive rather than speculative.

Beijing added its own friction. Chinese officials had urged Tencent to support domestic chipmakers rather than rent American silicon abroad. A deal squeezed from both directions rarely survives.

The cancellation underscores a broader trend: US technology companies face growing scrutiny when they provide computing infrastructure to Chinese businesses, even when the services are delivered through facilities located entirely outside China. Geography, it turns out, is no longer the test. The customer's nationality is.

How are Chinese firms getting compute?

Chinese technology companies have increasingly looked outside the country for advanced computing capacity. The reason is simple. Domestic chip supplies remain limited and cannot yet match the most advanced processors used for AI model training.

Southeast Asia has become the destination of choice. ByteDance and Alibaba rank among the region's largest Chinese clients for computing infrastructure. Tencent, meanwhile, has increased its own infrastructure spending as it develops its Hunyuan AI models — the very effort the Oracle agreement was designed to support.

The proposed deal would have expanded Tencent's AI computing capacity and strengthened its position in the AI market. Its withdrawal removes that path, at least for now.

What does this mean for US cloud providers?

The restrictions have already reshaped how American cloud and technology providers serve Chinese customers. Nvidia has reportedly told cloud service providers to ensure that at least 51 percent of their services go to non-Chinese customers.

That figure is telling. A majority-American customer base is now a compliance requirement, not a commercial preference. For providers with large Chinese revenue streams, the instruction forces a rebalancing of their business.

The wider picture

The Oracle withdrawal reflects the expanding reach of US-China technology restrictions into international AI infrastructure deals — those involving advanced processors, cloud computing and data-centre services alike. A lease structured through third-country facilities, once a plausible route around export rules, now carries political risk that outweighs its commercial logic.

For Tencent, the immediate question is where it finds training capacity next. Domestic chips do not yet match the leading processors, and Washington has just demonstrated that offshore leasing is no safe harbour. Expect Chinese cloud and AI spending to shift further toward Southeast Asian capacity held under less exposed structures — and expect US officials to keep watching who, exactly, ends up using it.