Oracle Cancels $7 Billion Tencent AI Chip-Leasing Deal Under US Pressure
- Dispatched: Nov 8, 2026 · 13:38 ET (18:38 UTC)
One-sentence summary: The Financial Times reported that Oracle cancelled a five-year, roughly $7 billion AI chip-leasing agreement with Tencent after US officials raised concerns, a deal that would have given Tencent access to as many as 100,000 advanced AI chips through Oracle data centers in Southeast Asia — Tencent's largest overseas compute lease to date.
Verification status: ⚠ Single-source report, not yet officially confirmed — first reported by the Financial Times, citing two people familiar with the matter; Reuters relayed the account but said it could not independently verify it; Oracle, Tencent and the White House have not publicly commented.
Core Facts & Consensus Delta:
The Financial Times published the report at 2026-11-04 00:00 ET / 05:00 UTC. Under the five-year agreement, Oracle would have leased as many as 100,000 advanced AI chips to Tencent across several Oracle data centers in Southeast Asia, at an estimated value of about $7 billion. According to the FT, as relayed by ZeroHedge, Tencent had paid roughly 30% upfront — about $2.1 billion — a payment the FT said had already weighed on Tencent's second-quarter free cash flow. Spread evenly over five years, the contract amounts to about $1.4 billion a year, roughly 1% of Oracle's expected fiscal 2028 revenue (Tikr estimate). The cancellation followed concerns raised by US officials with Oracle; officials in Guangdong, where Tencent is based, also pressed Tencent to back domestic Chinese chipmakers. Notably, overseas AI chip leasing remains permitted under existing rules — the deal was ended by official pressure, not by a new regulation. For context, Oracle had raised $20 billion ($10 billion in bonds and $10 billion in stock) two days earlier, using up the entire cap on new capital it had set for fiscal 2027 through 2030.
Supply-Chain & Market Read-Through: Cancelling the deal frees up Oracle's Southeast Asian compute capacity, but re-leasing it will be harder: Nvidia has told cloud customers that at least 51% of their services must go to non-Chinese companies, and is tightening chip-flow tracking among its Southeast Asian cloud partners. For Tencent, the overseas route to advanced chips narrows further, pushing it toward domestic chips and in-house compute. The direct impact on the memory and AI-chip supply chain is limited: the agreement covered compute leasing, not chip purchases, so orders and capacity schedules at HBM and memory makers (SK hynix KRX: 000660, Samsung Electronics KRX: 005930, Micron MU, SanDisk SNDK) are not directly affected. The episode does show, however, that Washington is using case-by-case pressure to restrict Chinese AI firms' access to advanced compute abroad, extending export controls from the rulebook into individual transactions.
Next Catalysts & Watchpoints:
- Oracle's fiscal Q2 results (mid-December): whether its $664 billion backlog shrinks because of the cancelled deal, and management's commentary on re-leasing the Southeast Asian capacity.
- Any public confirmation from Oracle or Tencent, including how the upfront payment is handled.
- Nvidia's fiscal Q3 earnings (after the close on 2026-11-17 ET): how the 51% non-Chinese-customer rule is being enforced.
Sources:
- PYMNTS (relaying Reuters, citing the Financial Times) (Published: 2026-11-04 13:40 ET / 18:40 UTC)
- TipRanks (Published: 2026-11-04 05:15 ET / 10:15 UTC)
- TIKR (Published: 2026-11-04 12:55 ET / 17:55 UTC)
Information only, not investment advice.
Read full 2026-11-07 daily market brief & cross-asset tape → All flash alerts
More flash alerts from 2026-11-07
- Nvidia Requires Southeast Asian Cloud Partners to Join Chip-Location Tracking, Tightening Controls on Chip Flows to China⚠ UNCONFIRMED
- Meta Indefinitely Pauses AI-Scraping Commercial Terms Update After Norway and Brazil Injunctions and UK Antitrust Probe✓ VERIFIED
- Meta Chief Strategy Officer Dave Wehner Departs to Become CFO of AI Inference Startup Fireworks⚠ UNCONFIRMED
- Nvidia Plans to Convert About 130 Brightspeed Telecom Buildings into Small AI Data Centers⚠ UNCONFIRMED
- Microsoft Tightens Oversight of Third-Party AI Compute Rentals from Nebius, IREN and Nscale✓ VERIFIED
- TSMC Denies Exploring 3D DRAM Production, Says It Stays Focused on Logic Foundry Business