Tencent Leases 100,000 AI Chips from Oracle in Five-Year Contract Valued at Approximately $7 Billion
nashnova research
Tencent signed a five-year lease with Oracle for roughly 100,000 advanced AI chips housed in Southeast Asian data centers — the largest offshore compute deal by a Chinese tech firm to date, valued at around $7 billion, and the direct cause of Tencent's first negative quarterly free-cash-flow in over a decade.
What exactly did Tencent buy?
Tencent signed a five-year lease covering multiple Oracle data centers in Southeast Asia, securing access to roughly 100,000 advanced AI chips unavailable inside China.
The chips are Nvidia's top-tier processors, barred from sale or deployment in China under U.S. export controls.
This means → Tencent chose "rent the machines abroad" over "build alternatives at home," trading upfront cash for a time window.
How is the $7 billion being paid — and what did it do to the books?
The deal is valued at roughly $7 billion, with an upfront payment of about 30% — around $2.1 billion on signing.
That prepayment dragged Tencent's Q2 free cash flow to negative RMB 13.8 billion (~$2 billion) — the first negative quarter in over a decade.
Capital expenditure in the same quarter jumped 176% year-on-year to RMB 53 billion (~$7.9 billion), mostly earmarked for AI infrastructure and compute prepayments.
In plain terms = Tencent spent more in one quarter than it earned, pushing its cash-flow balance into the red — all because of this AI "arms" purchase.
Why are Chinese tech firms leasing chips offshore?
Advanced-chip supply inside China is short, and U.S. export controls keep tightening. Offshore leasing has become the main path to compute power.
These leases are permitted under current U.S. rules, letting companies legally use restricted chips in Southeast Asian data centers.
ByteDance and Alibaba remain the largest clients of Southeast Asian data centers; as controls intensify, lease prices, contract lengths, and prepayment ratios are all rising.
This reflects a hardening pattern: Chinese tech firms' AI compute depends more and more on "rent abroad," not "build at home."
What is Tencent doing with the compute?
Its latest Hunyuan large language model has improved significantly, narrowing the gap with leading Chinese models such as Alibaba's Tongyi Qianwen.
An embedded AI agent called Xiaowei is now live inside WeChat, letting users order food, book appointments, and run tasks across millions of WeChat mini-programs via natural language. WeChat has over 1.4 billion monthly active users.
Its enterprise AI office agent, WorkBuddy, is growing rapidly and reportedly leads the Chinese PC office-agent market.
What to watch next?
Neither Tencent nor Oracle has commented; deal details await official confirmation.
This means → the metric the market cares about most is whether Tencent's free cash flow turns positive in the coming quarters — that will determine how investors judge its financial resilience.
Put simply = the money is spent, the AI capabilities are coming online, but the ledger is still open — the next few quarters of cash flow are the scorecard.
市场有风险，内容仅供研究参考，不构成投资建议。