October 12, 2026

Archived daily issue.

AI Systemic Risk Index
56 / 100
Elevated→ Unchanged from October 11

AI risk holds at 56 as hidden lease exposure broadens beyond Oracle

The AI Systemic Risk Index remains 56 / 100 — Elevated. New scrutiny of hyperscaler financing shows that Oracle is not an isolated case: Meta disclosed about $279 billion of uncommenced lease obligations at June 30 and another $68 billion signed in July, while Oracle disclosed $288 billion of additional uncommenced lease commitments. CapEx Sustainability rises to 56 and Financing / Credit rises to 78, but the headline score does not change. Both one-point moves carry above-normal uncertainty because future lease commitments are not the same as current funded debt, and the companies retain substantial cash flow and financing capacity.

Demand / ROI
35
Unchanged from 35: the new evidence concerns financing structure and long-dated commitments, not a fresh deterioration in realized AI demand or customer ROI.
CapEx Sustainability
56
Up 1 from 55: Meta and Oracle have committed to hundreds of billions of dollars of future data-center leases, increasing fixed-cost exposure before all facilities are operating.
Financing / Credit
78
Up 1 from 77: large uncommenced leases, residual-value guarantees, debt and external financing extend AI infrastructure risk beyond conventional bonds and loans.
Market Concentration / Valuation
54
Unchanged from 54: creditors are scrutinizing hidden leverage, but today's evidence does not establish a new equity-valuation break or broader repricing.
Physical Infrastructure
66
Unchanged from 66: power and construction constraints remain serious, but no newly verified system-wide physical disruption occurred since the previous public edition.
Macro Spillover
48
Unchanged from 48: higher long-term obligations have not yet produced a confirmed new spillover into broad employment, inflation, defaults or economy-wide growth.

What Changed

Compared with the October 11 public edition, the overall index remains 56. Demand / ROI remains 35, Market Concentration / Valuation remains 54, Physical Infrastructure remains 66, and Macro Spillover remains 48. CapEx Sustainability rises to 56 from 55, and Financing / Credit rises to 78 from 77. Neither one-point change is unusually large. Both are flagged as uncertain because lease commitments must be distinguished from funded debt and because payment timing, subleasing, project completion, customer prepayments and operating cash flow can materially change the eventual burden.

Why It Mattered

AI infrastructure risk is increasingly distributed across legal and financing structures rather than concentrated in a single company's visible debt balance. A hyperscaler can retain a strong credit profile while developers and lenders fund facilities against long-term leases, guarantees and expected customer demand. That structure can efficiently move construction risk to specialized capital. It can also create correlated exposure if demand disappoints, power delays commissioning, asset values fall, or multiple tenants try to renegotiate at once. The relevant question is therefore not whether one company can pay today, but how much fixed exposure the ecosystem has accumulated before the associated capacity and revenue are fully operating.

Watch List

Meta's next update on its approximately $347 billion of disclosed uncommenced lease obligations, 2026 capital spending and cash flow; Oracle's $288 billion of uncommenced lease commitments, customer prepayments and Project Jupiter milestones; new disclosures of residual-value guarantees and special-purpose financing; AI-linked bond spreads and project-loan marks; power, pipeline and permitting schedules; hyperscaler earnings guidance; lease restructurings or cancellations; and evidence that financing stress reaches utilities, construction, suppliers, employment or non-AI borrowers.

Long-Term View

Leases and project vehicles are not inherently dangerous. They can match long-lived infrastructure with long-lived funding and let specialized investors bear construction and asset risk. The systemic concern is opacity and correlation: several large technology companies, model providers, developers, utilities and lenders may depend on the same assumptions about power delivery, chip economics, utilization and customer revenue. Clearer disclosure of uncommenced leases, guarantees, project timing and counterparty concentration would make the buildout more durable. If capacity comes online on time and produces measurable economic value, today's commitments can support productive infrastructure rather than financial instability.