Operational Hyperscale and Colocation Campuses Worldwide

The worldwide operational hyperscale and colocation data-center footprint is the infrastructure the immersive medium operates on. As of mid-2026, the operational capacity is approximately 100 gigawatts of IT load globally, with the Northern Virginia data-center cluster carrying approximately 25 percent of global capacity, Ireland and Frankfurt anchoring the European cluster, and Singapore and Tokyo anchoring the Asia-Pacific cluster. This article catalogs the operational campuses by region and operator, with notes on capacity, power consumption, and the hyperscaler-versus-colocation distinction.

The Hyperscaler Cohort

Six US-headquartered hyperscalers and three Chinese hyperscalers operate the largest single-vendor data-center portfolios.

Amazon Web Services (AWS). Operates approximately 35 Regions worldwide, each containing multiple Availability Zones, with each AZ comprising one or more data centers.1 Operational power capacity is estimated at approximately 18 gigawatts globally (rough composite estimate; AWS does not consistently disclose capacity).

Microsoft Azure. Operates approximately 60 Azure Regions worldwide.2 Operational power capacity is estimated at approximately 12 gigawatts globally.

Google Cloud. Operates approximately 35 Regions worldwide.3 Operational power capacity is estimated at approximately 7 gigawatts globally. Google's own product infrastructure (Search, YouTube, the broader Google product surface) operates additional substantial capacity outside the Google Cloud reportable footprint.

Meta. Operates approximately 25 data-center campuses worldwide as of 2024 to 2025.4 Operational power capacity is estimated at approximately 8 gigawatts. Meta's data-center investment has accelerated through 2024 to 2025 to support the AI buildout (Llama, Reality Labs computing, the broader Meta AI infrastructure).

Apple. Operates approximately 10 data-center campuses across the United States, Europe, and Asia.5 Capacity is not publicly disclosed but estimated at approximately 2 to 3 gigawatts.

Oracle Cloud Infrastructure. Operates approximately 50 Regions worldwide.6 Oracle's 2025 Stargate partnership with OpenAI and Microsoft has expanded the Oracle capacity-build commitments substantially.

Chinese hyperscalers include Alibaba Cloud, Tencent Cloud, and Baidu Cloud, with substantial capacity inside China and growing international footprints.7

The Colocation Cohort

Several colocation operators run substantial multi-tenant data-center capacity available to hyperscaler-and-enterprise lease.

Equinix. The largest single colocation operator, with approximately 250 data centers across 75+ metropolitan markets worldwide.8 Equinix specializes in the Internet Exchange Point co-location that supports peering and direct cloud-to-network connectivity.

Digital Realty. Approximately 300 data-center facilities across multiple continents.9 Digital Realty's portfolio includes the Interxion European footprint acquired in 2020.

NTT Communications. Substantial Asia-Pacific footprint and growing global presence.10

KIO Networks. The largest Latin American colocation operator.

Regional colocation providers. A long tail of regional and national colocation operators serves smaller markets and specialized use cases (financial-services co-location near exchanges, broadcast-and-media co-location for content delivery, government and defense co-location for classified workloads).

The Northern Virginia Cluster

The Northern Virginia data-center cluster (concentrated in Loudoun County, Prince William County, and Fairfax County) is the largest single data-center concentration in the world.11 Operational capacity in 2024 to 2025 is approximately 4.5 gigawatts; under-construction capacity exceeds an additional 5 gigawatts. Approximately 70 percent of global Internet traffic transits Northern Virginia at some point.

The cluster's institutional infrastructure includes the Ashburn Internet Exchange (one of the largest Internet Exchange Points in the world), substantial submarine-cable landing-station presence, and integrated peering arrangements across the major US carriers. The cluster is reaching power-grid constraints; the Virginia state utility (Dominion Energy) and the regional transmission organization (PJM Interconnection) have raised capacity-allocation concerns through 2024 to 2026.

Other Major Regional Clusters

Ireland (Dublin). The largest European data-center cluster, approximately 1.0 gigawatt operational capacity. Substantial AWS, Microsoft, Google, and Meta presence. Power-supply constraints have been a recurring policy concern for the Irish government.

Frankfurt (Germany). Approximately 0.9 gigawatts. Europe's largest financial-services co-location market and a substantial hyperscaler presence.

London-Slough corridor (UK). Approximately 0.8 gigawatts. Substantial financial-services and hyperscaler presence.

Amsterdam. Approximately 0.6 gigawatts.

Singapore. Approximately 0.7 gigawatts. The Singapore data-center sector has been under a government-imposed moratorium on new data-center construction since 2019, with selective approvals resumed under sustainability requirements.

Tokyo. Approximately 0.6 gigawatts.

Sydney. Approximately 0.4 gigawatts.

Mumbai-Hyderabad-Bengaluru (India). Combined approximately 0.9 gigawatts, growing rapidly through 2024 to 2026.

Beijing-Shanghai-Shenzhen (China). Combined substantially larger (composite estimate approximately 3 to 5 gigawatts) but exact figures are not consistently disclosed.

Power Consumption and Sustainability

Operational hyperscale data centers consume approximately 1.5 to 2.0 percent of global electricity as of 2024 to 2025, per International Energy Agency estimates.12 The projection through 2030 is substantial growth, driven primarily by AI-workload demand. The IEA's Electricity 2024 report and successor analyses project data-center electricity consumption could double or triple by 2030 under high-AI-demand scenarios.

The implications for sustainability commitments are substantial. Major hyperscalers have committed to renewable-energy power-purchase agreements and to operational-emissions reductions, but the absolute electricity-consumption growth substantially exceeds the renewable-energy-supply growth in most major markets. Carbon-emissions commitments from the major hyperscalers are increasingly difficult to achieve under the AI-workload demand.

How to Read This Article

The hyperscale-and-colocation infrastructure is the operating layer for the immersive medium. Proposed and Under-Construction AI Gigasites treats the proposed AI gigasites under construction. Country and Regional Totals Through 2030 traces the country-and-regional totals through 2030. Streaming Media's Use of Global Internet Infrastructure treats the streaming-media use of the infrastructure as the working benchmark.

See Also

Notes

  1. AWS Global Infrastructure. https://aws.amazon.com/about-aws/global-infrastructure ↩︎

  2. Microsoft Azure Global Infrastructure. https://azure.microsoft.com/en-us/explore/global-infrastructure ↩︎

  3. Google Cloud Locations. https://cloud.google.com/about/locations ↩︎

  4. Meta data-center campuses. https://datacenters.atmeta.com ↩︎

  5. Apple data-center infrastructure. https://www.apple.com/environment ↩︎

  6. Oracle Cloud Infrastructure Regions. ↩︎

  7. Chinese hyperscalers (Alibaba Cloud, Tencent Cloud, Baidu Cloud). and https://cloud.tencent.com and https://cloud.baidu.com ↩︎

  8. Equinix global data-center footprint. ↩︎

  9. Digital Realty global facilities. https://www.digitalrealty.com ↩︎

  10. NTT Communications data centers. https://services.global.ntt ↩︎

  11. Northern Virginia data-center cluster, industry analysis. https://www.srgresearch.com ↩︎

  12. International Energy Agency, Electricity 2024. https://www.iea.org/reports/electricity-2024 ↩︎