Directory
Browse AI data centers by operator and location, with reported power capacity, GPU infrastructure and project status.
| Facility | Operator | Location | Power capacity | GPU infrastructure | Status | Opened |
|---|---|---|---|---|---|---|
| Equinix AM Amsterdam Campus | Equinix | Amsterdam, Netherlands | — | — | Operational | — |
| Google Eemshaven | Eemshaven, Netherlands | — | — | Operational | — | |
| AWS Northern Indiana AI Campuses | Amazon Web Services | Northern Indiana, United States | 2.4 GW | AI and cloud data-center capacity | Planned / developing | — |
| Digital Realty Ashburn Campus | Digital Realty | Ashburn, Virginia, United States | — | — | Operational | — |
| Meta Hyperion Richland Parish | Meta / Blue Owl Capital | Richland Parish, Louisiana, United States | 5 GW compute capacity | Large multi-gigawatt AI training cluster | Under construction | — |
| Meta Prineville Campus | Meta | Prineville, Oregon, United States | — | — | Operational | — |
| Microsoft Boydton Campus | Microsoft | Boydton, Virginia, United States | — | — | Operational | — |
| Microsoft Fairwater Wisconsin | Microsoft | Mount Pleasant, Wisconsin, United States | — | AI supercomputer infrastructure | Operational | 2026 |
| Stargate Abilene | OpenAI / Stargate partners | Abilene, Texas, United States | — | Frontier AI training and serving infrastructure | Operational | — |
Capacity and status reflect publicly announced figures. Map view is planned.
Sources attributed Data updated regularly Methodology
Proposals have been lodged for a one-gigawatt data centre campus situated at a retired nuclear power station in Scotland. Developer CX Tech plans to supply the substantial facility by establishing dedicated direct connections to neighbouring renewable generation assets.
Google has put an experimental data centre satellite into orbit alongside research evaluating the technical feasibility of extraterrestrial server deployments. The company's findings suggest that operational bit barns in space would require massive launch capacity from providers such as SpaceX, as well as solutions to intricate orbital networking and formation-flying hurdles.
A Stanford University academic is promoting Homa, a modern transport protocol positioned to replace standard TCP across networks. The technology proposes a redesigned networking foundation tailored to handle the heavy demands of modern artificial intelligence workloads.
NetApp has unveiled Novus, a next-generation storage architecture created to tackle zettabyte-scale capacity and clear metadata processing hurdles. The platform is designed to provide 100 terabytes per second of aggregate throughput via NFS and parallel NFS protocols, aimed at supporting hundreds of thousands of GPUs concurrently.
Google has launched an advanced processor into orbit as part of preparatory efforts to evaluate the feasibility of extraterrestrial computing platforms. The company estimates that SpaceX's Starship launch vehicle must complete approximately 1,600 missions before data centres in space can realistically operate.
Nebius has entered into a commercial agreement with AIB Data Centers to take 50 megawatts of capacity. The leased space is located within a facility in the south-eastern United States.
Traysar has entered into a partnership with Armada to deploy computing infrastructure beneath the Earth's surface for AI workloads. The venture, branded as Terminus, aims to house mission-critical digital systems in subterranean environments.
A data centre facility in Kyiv has ceased operations following severe structural damage sustained during a Russian bombing strike. The strike reflects a continuing campaign to disrupt telecommunications and digital infrastructure across Ukraine. The incident highlights the acute vulnerability of operational compute assets in conflict zones.
Data platform provider VDURA has unveiled its V12 architecture, designed to bring hyperscale-tier storage performance to AI factory environments. The platform seeks to eliminate data delivery bottlenecks that impede GPU utilisation across fast-growing specialist cloud operators. By accelerating data ingestion and retrieval cycles, the system aims to keep intensive machine-learning workloads consistently fed.
Nvidia has highlighted the investment economics behind gigawatt-scale AI factories, noting that infrastructure deployment averages around $60 million per megawatt. Operators evaluate financial returns across earning potential, operational lifespan, and the ability to repurpose computing assets over time. The analysis underscores how hyperscale developers calculate capital allocation when deploying massive AI clusters.
SpaceX is scheduled to transport Google Tensor Processing Units into low Earth orbit aboard satellites operated by Planet Labs. The mission represents an early step towards deploying orbital data centres capable of handling remote computing workloads. Placing enterprise-grade AI accelerators into space aims to reduce latency for satellite data analysis and test alternative infrastructure models.
Virginia has prohibited the use of non-disclosure agreements for data centre projects as part of an updated digital infrastructure framework. Introduced by Governor Spanberger, the package of policies also includes the establishment of an artificial intelligence task force.
Reports indicate that the Russian military has begun targeting data centre facilities in Ukraine using missile strikes. The physical destruction of this server infrastructure has disrupted operations and knocked several major piracy websites offline.
Cloud provider Civo has announced plans to build an initial 8MW edge data centre dedicated to artificial intelligence workloads in Hertfordshire. The project is designed to scale up to 38MW and represents the inaugural development in a planned 40-facility deployment strategy.
Castrol has launched Castrol CORE, a comprehensive thermal management service aimed at data centre operators. The offering transitions the company from supplying standalone cooling fluids to delivering an integrated package that includes hardware cooling technology, design capabilities, and ongoing lifecycle testing.
Amazon Web Services has been granted planning approval to build a new data centre development in Iver, Buckinghamshire. The green light clears the cloud provider to replace the existing Ridgeway Distribution Centre site with a purpose-built computing campus.
HPCwire has released its regular monthly overview tracking personnel movements across the high-performance computing ecosystem. The collection outlines recent corporate appointments, promotions and professional recognitions within the supercomputing sector.
Storage provider VDURA has made version 12 of its data platform generally available, targeting high-throughput environments in AI data centres and neoclouds. The release incorporates multi-tenancy, context-driven tiering, and improved energy efficiency, whilst also gaining formal validation on Supermicro hardware.
Researchers at Oak Ridge National Laboratory have deployed the Frontier supercomputer to construct an advanced quantum algorithm tailored for fluid dynamics. The project, facilitated by the Oak Ridge Leadership Computing Facility, demonstrates how world-leading open-science compute capacity can expand algorithmic modelling across complex physical sciences.
NxN has expanded into Portugal through the purchase of a 1MW facility located in Évora. The company intends to commit €100 million in further capital expenditure to build out its presence in the country.
AI data centers are facilities designed for dense clusters of GPUs and other accelerators. A single rack of current-generation systems can draw well over 100 kilowatts, several times a traditional enterprise rack, which is why liquid cooling, grid connections and on-site power have become central to new projects.
This directory tracks significant AI and hyperscale campuses with their operator, location, stated capacity and status, drawn from public announcements. For daily developments, follow our latest news and energy coverage.
Estimate GPU memory for a model
H100, H200, B200, GB200 & AMD specs
Compare where to rent GPUs
AI infrastructure basics
Today's sourced briefing