Oracle has terminated a $7 billion, five-year agreement to lease approximately 100,000 advanced AI chips to Chinese technology giant Tencent. The decision follows direct pressure from the Trump administration seeking to restrict China's access to critical AI hardware.
The cancelled deal represents a substantial revenue loss for Oracle. Over five years, the contract would have generated roughly $1.4 billion annually—a meaningful portion of Oracle Cloud Infrastructure's revenue base at a time when the division is racing to compete with AWS and Google Cloud for AI compute workloads.
For Tencent, the loss threatens its AI infrastructure expansion plans. The company depends on high-performance chips to train and deploy large language models and power its cloud services. The 100,000 units were central to closing the gap with U.S. competitors in AI capability.
The Trump administration's move aligns with broader U.S. export controls targeting advanced semiconductors and AI technology. The strategy aims to degrade Chinese firms' technological capabilities by choking off access to high-end components.
Tencent now faces constrained options. Global supply of advanced AI chips remains tight, with lead times extending well into 2027. The company can pursue costly domestic chip development—a multiyear undertaking with uncertain returns—or accept export-controlled alternatives inferior to Nvidia's offerings. Both paths will slow its AI research and product deployment.
The deal's collapse exposes the business risk facing U.S. technology companies serving China. Oracle had positioned OCI as an alternative to Nvidia for customers seeking GPU capacity, a pitch that suddenly carries geopolitical liability. For U.S. cloud and chip companies, Chinese revenue and strategic accounts are now subject to national security review and government veto.
Oracle must now reallocate 100,000 units of advanced silicon. While AI compute demand remains robust, finding a replacement customer or customer set to absorb this capacity and match the deal's scale will prove difficult. The company's sales and cloud divisions face the challenge of filling a $7 billion hole in a market where most large buyers—AWS, Meta, Microsoft—build their own chips or have long-term supply agreements locked in.
