Scenarios
Five stress cases over the AI-compute supply chain.
Abqaiq 2.0: Saudi Grid and Compute Co-Location
A disruption to electricity infrastructure could affect both Saudi data centres and petroleum facilities. This assessment assigns 2 data centre–petroleum pairs to a named shared substation (some connections are inferred from corridor proximity) and 6 more to the same regional grid zone. A further 4 pairs are within 200 km and in different zones. Establishing a common point of power failure would require the sites’ actual supply and backup arrangements.
Aramco reported that the September 2019 Abqaiq–Khurais attack initially suspended 5.7 million barrels per day of production; partial restoration was reported three days later. The compute scenario adds an assumed interruption to shared electricity infrastructure. 2 data centre–petroleum pairs are assigned to a named substation and 6 more to a grid zone; 4 pairs fall within the 200 km screen across different zones.
Every 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. Proximity records distance only.
Co-Location Risk Model
The scenario assumes that damage near a petroleum facility interrupts electricity supplied to a data centre. Shared substations and regional grid zones identify connections to investigate. Site-level supply diagrams and backup arrangements would be needed to establish the effect of a particular outage.
HUMAIN projects
HUMAIN announced a Saudi data-centre partnership with xAI on 19 November 2025. NEOM subsequently reported HUMAIN and DataVolt starting construction at Oxagon on 31 August 2026. Neither project has an assigned electricity connection in this map, so their exposure to the petroleum facilities shown here is unassessed.
Saudi Arabia’s AI programme →| Evidence | Data Center | Petroleum Facility | Distance | Same Grid | Shared Substation | Exposure Analysis |
|---|---|---|---|---|---|---|
| Substation | Google Cloud Dammam | Ras Tanura Refinery & Terminal | ~24 km | ● | Faras 380 kV | Both facilities are assigned to Faras 380 kV in this assessment. Some connections are inferred from corridor proximity. Their actual supply arrangements and backup would determine whether a disruption affects both. |
| Substation | Aramco Digital DC (Dhahran) | Abqaiq Processing Facility | ~60 km | ● | Abqaiq 380 kV | Both facilities are assigned to Abqaiq 380 kV in this assessment. Some connections are inferred from corridor proximity. Their actual supply arrangements and backup would determine whether a disruption affects both. |
| Shared zone | Aramco Digital DC (Dhahran) | Ras Tanura Refinery & Terminal | ~39 km | ● | — | Both facilities are assigned to the same regional grid zone. No shared substation is identified for this pair, so no common point of power failure is identified. |
| Shared zone | Google Cloud Dammam | Abqaiq Processing Facility | ~69 km | ● | — | Both facilities are assigned to the same regional grid zone. No shared substation is identified for this pair, so no common point of power failure is identified. |
| Shared zone | Google Cloud Dammam | Jubail Industrial Complex (SATORP) | ~78 km | ● | — | Both facilities are assigned to the same regional grid zone. No shared substation is identified for this pair, so no common point of power failure is identified. |
| Shared zone | Google Cloud Dammam | Jubail Desalination Plant | ~81 km | ● | — | Both facilities are assigned to the same regional grid zone. No shared substation is identified for this pair, so no common point of power failure is identified. |
| Shared zone | Aramco Digital DC (Dhahran) | Jubail Industrial Complex (SATORP) | ~93 km | ● | — | Both facilities are assigned to the same regional grid zone. No shared substation is identified for this pair, so no common point of power failure is identified. |
| Shared zone | Aramco Digital DC (Dhahran) | Jubail Desalination Plant | ~96 km | ● | — | Both facilities are assigned to the same regional grid zone. No shared substation is identified for this pair, so no common point of power failure is identified. |
| Proximity | AWS Saudi Arabia (announced region) | Khurais Oil Field | ~150 km | — | — | The facilities fall within the distance screen and are assigned to different grid zones. The pairing records their geographic proximity. |
| Proximity | DataVolt Hyperscale (Riyadh) | Khurais Oil Field | ~150 km | — | — | The facilities fall within the distance screen and are assigned to different grid zones. The pairing records their geographic proximity. |
| Proximity | Oracle Cloud Riyadh | Khurais Oil Field | ~154 km | — | — | The facilities fall within the distance screen and are assigned to different grid zones. The pairing records their geographic proximity. |
| Proximity | STC Data Center (Riyadh) | Khurais Oil Field | ~155 km | — | — | The facilities fall within the distance screen and are assigned to different grid zones. The pairing records their geographic proximity. |
Rows are ordered by their assigned connection class, then distance. Shared substation means a named substation entry lists both facilities; some of these connections are inferred from corridor proximity and need verification against supply diagrams. Shared zone means the records have the same regional grid assignment without a common named substation. Proximity means the two sit within the 200 km screening radius and have different zone assignments. The radius determines which pairs enter the table; it does not describe how far an electrical failure could propagate. A ~ on a distance means at least one endpoint’s coordinates are analyst-estimated.
Regional grid assignments, named substations and the facilities associated with each entry; some connections are inferred.
Dated accounts of attacks on petroleum facilities, with sources for the reported damage and recovery.
Assumed disruptions and outage durations, the facilities selected for each case, and the historical events cited as context.
How the distance screen and electricity assignments select pairs, and the evidence needed to assess an outage.