The named campuses behind the global build-out — who is building them, where, how much power they are being sized for, and what status they are at today. Open any row for the build detail, then jump straight into the equipment lines that project needs.
The flagship Stargate site, built on the Lancium Clean Campus. The first two buildings went live on Oracle Cloud Infrastructure in September 2025 and the expansion takes the campus to roughly 1.2 GW.
Around 2 GW with an on-site gas microgrid. The first buildings are expected late 2026 with full build-out around Q4 2028.
Foundations are underway for a campus of roughly 2.2 GW, powered mainly by on-site generation rather than grid supply.
A roughly 1.3 GW site designed around 70% renewables, supplemented by storage and grid supply — one of the more grid-friendly Stargate builds.
About 1.2 GW with dedicated on-site generation and storage, the pattern that lets Texas sites bypass long interconnection queues.
A roughly 1.4 GW site reported for Michigan. Local opposition and grid constraints make its timing the least certain of the US Stargate sites.
More than $50bn planned on 2,250 acres: 2 GW has been approved with a path to 5 GW. Entergy Louisiana is building dedicated gas generation for the site.
Roughly 555,000 GPUs across two sites running on on-site turbines, with reported live capacity near 1.1 GW heading toward 2 GW. Built at a speed that reset expectations for the sector.
One of the largest single compute clusters in operation, built on Amazon's own Trainium silicon rather than GPUs. Capacity is still ramping.
315 acres and about 1.2m sq ft. More than 90% of the facility runs on closed-loop liquid cooling, with 72 Blackwell-class GPUs per rack in the live phase.
About 2 GW delivered over five to seven years, launched behind the meter on up to 2.67 GW of gas-fired capacity under an agreement with Chevron.
A large multi-phase campus ramping through 2026 as part of Microsoft's Fairwater generation of AI sites.
Hundreds of megawatts scaling toward GW class in one of the most established US cloud corridors.
The largest data center market in the world absorbed 1,102 MW in 2025 alone. QTS keeps expanding across the Ashburn and Richmond corridors.
A multi-campus programme inside established Google regions, where grid interconnection rather than land is the pacing item.
A roughly 4.75 GW campus plan with grid authorisation in place — the largest data centre development proposed in Latin America to date.
Oracle is delivering the first buildings for OpenAI on the former Ford Saline plant site, targeting energisation through 2026 with the campus designed to scale toward 1.4 GW.
A $1.5bn campus reported for El Paso as part of Meta's multi-gigawatt Texas build-out, with construction expected to run into the late 2020s.
Amazon's Trainium-based Rainier campus continues to expand alongside the Claude workloads it was built for, making Indiana one of the densest AI compute clusters anywhere.
Google's long-running Texas and Oklahoma campuses keep adding AI capacity under its $85bn 2026 capex programme, largely on ERCOT and SPP grid power.
Reported as one of AWS's next-generation AI clusters alongside Rainier, part of the more than $100bn AWS is committing to US AI infrastructure.
The next phase of the Fairwater campus adds further closed-loop liquid-cooled halls as the first phase ramps into production.
A roughly 1 GW AI campus converted from a former steel site, financed through an $8bn lease structure with power secured from Entergy Louisiana.
Nscale's US expansion supplies contracted AI capacity to Microsoft, part of the wave of neocloud operators building for hyperscale tenants.
Vantage's side of the Texas Stargate Frontier programme, delivered on the fast-track schedule that has become standard for frontier AI campuses.
A long-running renewable-powered cloud region that Microsoft keeps expanding for AI workloads.
A hydro-powered campus inside the Arctic Circle — one of the earliest large-scale Nordic data centres and still expanding.
300 MW being added on a 3 km² site next to the largest transformer station in Europe, scalable toward 1 GW of 100% renewable power.
Google's first Norwegian data centre, built in phases through 2030 on a 200-hectare site about 50 miles south-west of Oslo.
230 MW now with another 290 MW planned, 100% hydropower and a target of 100,000 NVIDIA GPUs. One of Europe's first AI gigafactories.
A timber-framed AI mega campus in a hydro-rich region, built around heat reuse.
A converted paper mill cooled with seawater — still the reference point for low-water data centre design.
A 550 MW campus with more than €7.5bn of planned investment; the first 110 MW phase is already fully leased.
About €1bn starting at 50 MW with room to reach 128 MW, part of TikTok's European data localisation programme.
A second campus near Berlin with potential for 186 MW of IT power, secured ahead of demand.
The second building has opened on the Warsaw campus; the full site can reach 60 MW and 50,000 m² of gross space.
New AI-focused campuses expanding AWS's Spanish footprint in one of Europe's cheapest renewable-power regions.
Planning approval secured for a hyperscale campus of three large buildings on a former industrial site near Leeds.
Land secured for a 350 MW campus coming online from 2028, starting with 120 MW.
An 800,000 m² site with up to 500 MW planned; the first 60 MW phase is scheduled to start in early 2027.
Google's newest UK campus, roughly $1bn, built to absorb growing UK AI demand while its London region scales.
Microsoft's expanded UK programme, part of a $30bn commitment to UK AI infrastructure spanning sites in Essex and south Wales.
A Netherlands AI campus developed with Nvidia silicon as the anchor reference design, on one of Europe's fastest-growing FLAP-D markets.
The UAE portion of Stargate UAE, delivered with Microsoft technology and G42 as the local partner — the anchor of the Gulf's sovereign AI build-out.
Pure DC's Iberian expansion adds AI-ready capacity to Spain's growing Mediterranean digital corridor.
Meta's Spanish campus programme pairs a Madrid metro site with a larger inland plot, built around waterless cooling in one of Europe's drier regions.
A second Norwegian Nordic campus for Bulk alongside N01, built for export-grade AI capacity on cheap hydro power.
The first phase of the xAI–HUMAIN joint venture, intended to pair sovereign Saudi power and land with xAI's training stack.
The next tranche of Stargate UAE, extending the initial 1 GW commitment toward multi-gigawatt scale.
$15bn for a 1 GW campus across 600 acres; ground was broken in April 2026 and three new subsea cables will land at Visakhapatnam to serve it.
About ₹1.6 lakh crore (~$17bn) for a 1.5 GW AI cluster with captive solar and battery storage; the first 500 MW is targeted for October 2028.
Roughly $1.8bn for a 500 MW facility that anchors Meta's Indian AI capacity.
$11bn AI-native development targeting 1 GW across 400 acres.
A new cloud region due to go live in mid-2026, sized for AI and cloud workloads with a 600 MW programme behind it.
An AI supercluster targeting go-live by August 2026 — the smaller, faster end of the Indian build-out.
An MoU-backed 400 MW AI data centre cluster in Telangana.
A 1 GW AI cluster inside a planned 5 GW campus. The first 200 MW — roughly 100,000 GB300-class systems — is due live in 2026, built by Khazna Data Centers.
The first two sites of a national AI platform, each starting at up to 100 MW, on a roadmap to 1.9 GW by 2030 and 6 GW by 2034.
Phase one of a campus planned at 1.5 GW inside NEOM's industrial city.
An AI infrastructure joint venture of up to $20bn. Capacity has not been disclosed, so it shows as n/d rather than a guess.
Google's $2bn Malaysian data centre and cloud region — one of the largest foreign digital investments in the country.
Part of Microsoft's $2.2bn Malaysian investment, with AI capacity in the Kulim high-tech corridor.
Johor has become the densest data centre corridor in Southeast Asia, supplying capacity to Singapore across the causeway.
AirTrunk's Johor campus built to hyperscale specification for Microsoft AI capacity in the region.
AirTrunk's Japanese hyperscale platform, built for the AI demand flowing into Japan's cloud regions.
Part of Microsoft's $10bn Japanese commitment, adding AI-dedicated capacity to the existing Japan East and Japan West regions.
Google's Australian build-out pairs new cloud regions with AI capacity, supported by renewable PPAs.
Australia's largest hyperscale platform keeps expanding to serve AI workloads across ANZ.
STT GDC's Indian expansion adds AI-ready capacity in the two markets absorbing most of the country's hyperscale demand.
Indian operators building AI-dedicated halls alongside their colocation footprints, largely for domestic inference and sovereign AI programmes.
Africa's largest independent colocation campus. JB5 added twelve data halls at Isando in June 2025 and the JB4 Bredell expansion completed in August 2025, taking the combined critical IT load to roughly 120 MW.
East Africa's first hyperscale-grade campus, launched in May 2025 with 4.5 MW and 780 racks across three halls in phase one. Later phases (NBOX1.2 and NBOX2) extend the site.
Digital Realty's third Lagos facility, opened in August 2025 with close to 2 MW of IT capacity. Lagos carriers run heavy on-site generation because grid availability sits far below installed capacity.
Announced in 2024 as a $1bn geothermal-powered campus with a 100 MW first phase and a stated ambition of 1 GW. Commercial terms were still being negotiated with the Kenyan government through 2026 and the first phase may be reduced.
A $1bn green hyperscale campus announced in March 2026, built in two phases from 10,000 m2 to 500,000 m2 and targeting high-density workloads for global cloud providers. No IT capacity has been published yet.
Vantage's first African campus, backed by a reported $2bn commitment to South Africa. The initial 16 MW building at Isando in Ekurhuleni is being followed by further halls on a site zoned for roughly 80 MW.
An AI-focused facility in the Western Cape being built with NVIDIA-accelerated compute in mind, part of Cassava's plan to bring AI capacity into Africa rather than serving it from Europe.
The most geographically spread African platform: carrier-neutral facilities already live in Uganda, Ethiopia, Mozambique, Tanzania, Côte d'Ivoire and the DRC, with further builds in Angola. Individual halls are small, so capacity is aggregated here.
Open-access hyperscale-grade capacity on both the west and east African coasts, connected to WIOCC's subsea cable landings so carriers and cloud providers can interconnect directly in-region.
Announced in 2024 as Google's first African cloud region, paired with an Africa Connect subsea programme. Capacity has not been broken out publicly, so the figure here is a planning estimate for the region rather than a published number.
One of the continent's largest colocation footprints, expanding in Lagos, Nairobi, Accra and Cape Town with standardised halls that can be replicated quickly across markets where skilled build labour is scarce.
A second Lagos facility announced to follow the LOS1 acquisition, extending Equinix's west African interconnection footprint for enterprises and network operators.
A geothermal-powered AI campus in Kenya's Rift Valley, one of the first African sites designed for frontier training workloads.
Teraco's Gauteng campuses remain Africa's largest interconnected data centre platform, adding capacity for cloud and AI inference.
Lagos is emerging as West Africa's primary data centre market, with capacity built for cloud on-ramps and financial services.
Africa's first Nvidia-powered AI factory, intended to give African enterprises and researchers sovereign AI compute.
Submarine-cable-anchored capacity on both African coasts, built to serve hyperscale and content traffic.
No projects match that search.
Sources: company announcements, operator press releases, government and state filings, and reporting from Data Center Dynamics, Data Center Frontier, Reuters, BloombergNEF, SemiAnalysis and Epoch AI, as of September 2026. Capacities mix operational, under-construction and announced figures and are frequently self-reported; treat them as programme targets rather than energised capacity.
How to read this list. Operational means energised and serving; Under construction means ground is broken; Planned or Announced means capacity is committed but not yet built. Where an operator has not published a figure, capacity shows as n/d rather than an estimate.