Data 2026-03

Scenarios

Five stress cases over the AI-compute supply chain, each dated by when it was last reviewed and by the age of its newest input.

Abqaiq 2.0: Saudi Grid and Compute Co-Location

Reviewed 2026-07-18Newest input 2026-Q1Hand-reasoned

Reviewed at the A5 keep-with-currency ruling (overhaul v2, 2026-07-18), which made the review commitment a condition of the page surviving. Next review due 2026-10-16 (a 90-day cadence). The figures on this page are human-authored and sourced individually; they are not produced by the cascade engine. Newest input dated from Newest data_vintage across the module's sourced claims, '2026-Q1' (IISS March 2026 hardening assessment) — src/lib/data/abqaiq-data.ts. Review state for all five analyses is on the index.

Saudi Arabia’s AI buildout draws power from the same Eastern Province grid the 2019 Abqaiq drone attack took offline. The exposure is collateral: compute next to petroleum, on a grid that is already a military objective. Two data centres share a named substation with a petroleum facility; the rest of the pairs below are proximity, and the table keeps the two apart.

Infrastructure Co-Location Risk

8 data centers, 7 petroleum facilities, 8 grid-coupled pairs (2 on a named substation). Indirect disruption model: grid co-location with targeted petroleum infrastructure. Active: Abqaiq 2019 ReplaySaudi data centers share power grid infrastructure with the world's most targeted petroleum facilities. The 2019 Abqaiq attack (0% intercept rate, 18 drones, no interceptions) is the proof case. Active: Abqaiq 2019 Replay

Choose attack scenario

Saudi data centers in the Eastern Province share grid infrastructure with the world's most-targeted petroleum facilities. The September 2019 Abqaiq–Khurais attack achieved 0% intercept across 18 drones, removing 5.7 mb/d of crude processing capacity for two weeks; co-located data centers would lose power as collaterally as the refineries did. 2 DC–petroleum pairs share a named substation and 6 more share a grid zone; a further 4 sit within 200 km across different zones, which is proximity rather than coupling.

Methodology →SEC annual report · Aramco 2024 · CSIS / UN Panel of Experts S/2020/326 · IISS March 2026

Active scenarioAbqaiq 2019 Replay

Co-Location AnalysisEvery DC–petroleum pair inside the screening radius, grouped by what backs it: a named shared substation, a shared grid zone, or proximity across different zones — which is not a coupling claim.2 on a shared substation · 6 shared zone · 4 proximity only

Grid & SubstationsEastern / Central / Western 380 kV zones, named substations, and which facilities draw from each — the operational basis of co-location risk.3 zones · 8 substations

Attack HistoryStructured ledger of 2019–2022 Aramco strikes — projectile counts, intercept rates, damage class, restoration days. The empirical floor.6 events

Attack Scenario DetailsReference cards for each modeled attack — precedent, payload, primary targets, collateral DCs, expected outage duration.4 scenarios

MethodologyIndirect-disruption model (grid → DC), severity formula, air-defense gap analysis, and source authority tiering for SEC + Aramco data.

Rerun this disruption in the engine

A caveat, not an equivalent: the cascade engine propagates through supplier–customer edges, and this scenario’s mechanism is grid co-location — data centres going dark because they share an electrical grid with a struck petroleum facility, which is a geographic adjacency the supply-chain graph does not encode. Opening the engine runs the different question of a direct supply-chain strike on the co-located operator (Alphabet, whose Dammam region anchors the analysis above); it does not reproduce the grid-sharing exposure that is this page’s actual subject. The grid model here stands on its own.