Tencent Secures $7B Oracle Deal for 100,000 AI Chips in Southeast Asia

Tencent has finalized a $7 billion, five-year lease agreement with Oracle to access 100,000 advanced AI chips hosted in Southeast Asian data centers. This move allows the company to circumvent U.S. export restrictions on high-end silicon. The deal ensures the Chinese tech giant maintains its artificial intelligence development pipeline despite trade controls that limit direct access to state-of-the-art domestic hardware.

"We’re comfortable in making significant investments in AI because there is a substantial upside potential, and there is also clear downside protection," said Tencent president Martin Lau.

Financial Structure of the Cloud Lease

The agreement is structured as a multi-year cloud services contract instead of a direct purchase of hardware. According to a report on the deal, Tencent will provide an upfront payment of approximately 30% of the total $7 billion contract value to secure the capacity.

This capital-intensive strategy has impacted the company's balance sheet. In its second-quarter earnings, Tencent reported that free cash flow turned negative at RMB 13.8 billion. CFO John Lo attributed this shift to heavy AI infrastructure spending and the compute prepayments required to sustain the firm's model training operations.

Bypassing Export Controls

By hosting the chips in Oracle data centers located in Southeast Asia, Tencent is utilizing a lease-to-access model to bypass U.S. government restrictions that prohibit the sale of advanced AI processors, such as those produced by Nvidia, directly to Chinese entities. This approach treats the compute power as a service. It allows Tencent to train large-scale models without taking physical possession of the restricted hardware.

This development shows the lengths to which technology firms go to maintain competitive parity in the global AI race. As U.S. export controls restrict the flow of physical AI accelerators into China, cloud-based leasing via international partners has become a primary mechanism for large Chinese enterprises to access the compute resources required for machine learning development.