Meta
Meta builds for its own apps and models: Prometheus in Ohio is a 1 GW AI cluster and Hyperion in Louisiana scales to 5 GW.
Footprint, AI and capex
- Type Hyperscale cloud, internal hyperscaler, AI cloud or regional alternative.
- Internal hyperscaler 2
- Regions Live cloud regions, from the operator's own region page or region list.
- Not published
- Zones Availability zones (availability domains at Oracle) the operator states across its regions.
- Not published
- Countries Countries and territories with at least one live region, counted from the operator's region list.
- Not published
- Named locations Deployment locations the operator names where it publishes no region count.
- Not published
- Data centers Data center count in the operator's own words.
- Not published
- Owned vs leased What the operator discloses about owning or leasing its facilities.
- Not published
- AI capacity Announced AI data center capacity in GW or MW, or the scale of a named AI cluster.
- Prometheus (Ohio) 1 GW from 2026; Hyperion (Louisiana) up to 5 GW 1
- AI compute The operator's AI accelerators, instances or internal AI platform.
- Prometheus supercluster: hundreds of thousands of GPUs deployed 2
- Latest capex Latest reported capital expenditure with its period, plus guidance where given. Group-wide figures for parent companies.
- USD 34.9bn incl. finance-lease principal in Q3 2026; USD 135bn to 145bn guided for 2026 3
- Investment plan Multi-year AI and cloud investment programs, separate from annual capex.
- Not published
- Sovereign cloud Named sovereign, government or in-country cloud offerings.
- Not published
Quarter-by-quarter capex since 2023: Meta on Hyperscaler Capex.
Sources
- TechCrunch: Meta building 5 GW AI data center accessed 2026-11-11
- Meta engineering: inside Prometheus accessed 2026-11-11
- Meta Q3 2026 results accessed 2026-11-11
Compare with
All hyperscalersSpotted a newer number? Send the source to hello@hyperscalerlist.com and we update the record. How we collect the data.