DataCentersX > Sites

Data Center Site Deployments

The Sites pillar covers named data center deployments and ranked deployment datasets. It differs from Types, which covers the kinds of facilities that exist, and from Workloads, which covers what runs inside them. Sites is the pillar where specific instances live: the named AI factories that host frontier training runs, the scientific HPC clusters at national laboratories, the sovereign hubs under national AI strategies, the ground infrastructure for orbital networks, and the ranked lists of the largest deployments in the US and worldwide. A reader working from this pillar is asking "what are the actual facilities, and what distinguishes each?"

The pillar organizes deployment coverage into two layers. Case Studies are editorial deep dives on individual named sites where the site's architecture, workloads, or strategic context is distinctive enough to warrant standalone treatment. The ranked deployment datasets (Top 20 US, Top 20 Worldwide, Top Planned) provide bulk reference coverage of the broader landscape with standardized attributes per site. Both layers reference each other: a site appearing in the Top 20 US ranking may also have a case study page if it is particularly distinctive, and case studies may reference rankings for comparative context.

Case studies

Case studies are named sites that earn editorial deep dives because of distinctive architecture, strategic importance, or representative value for a broader trend in the industry. The twelve case studies currently covered span frontier AI training factories, scientific HPC infrastructure, sovereign AI hubs, specialized compute for astronomy and physics, and the ground infrastructure for emerging orbital networks.

Ranked deployment datasets

The ranked deployment datasets provide bulk reference coverage of the largest deployments in the US and worldwide, plus the major planned deployments under development. Each list carries standardized attributes per site (operator, location, capacity, workload mix, operational status) and is updated as new public data becomes available. The datasets complement the case studies: a site may appear in both layers, with the ranking entry providing concise reference data and the case study providing editorial depth.

What the case study landscape reveals

Reading across the twelve case studies shows the shape of modern data center deployment. Four of the twelve (xAI Colossus, Stargate, Meta Hyperion, Tesla Dojo) are frontier AI training infrastructure at gigawatt scale or approaching it. One (Tesla Cortex) is the AI inference counterpart to Dojo, co-located with the training site and supported by the same silicon supply chain. Two (NEOM and Fermi Hypergrid) represent sovereign or quasi-sovereign AI strategies with distinctive power sourcing (behind-the-meter nuclear in one case, energy-surplus national strategy in the other). Four (Rubin Observatory, SKA Array, HL-LHC, Oak Ridge) represent scientific HPC, each with a specific mission that shapes its compute and storage architecture in ways that differ from general-purpose AI training. One (Starlink Ground Infrastructure) represents the ground-side of orbital network infrastructure, a category that will expand substantially when the SpaceX Orbital Data Center constellation (covered under Types:Orbital and Space Datacenters) reaches deployment.

The pattern across the twelve case studies is that AI has pulled site-level infrastructure attention toward gigawatt-class integrated builds while not displacing the separate traditions of scientific HPC, sovereign computing, and network infrastructure that continue to operate on their own architectural principles. Future case studies will likely add frontier orbital compute sites as the SpaceX and Blue Origin constellations deploy, additional sovereign AI hubs as national programs mature, and the Terafab-connected silicon-and-deployment integrated model that Tesla is pioneering may recur with other operators.

Where Sites sits in the DatacentersX structure

The Sites pillar answers "what are the actual facilities?" The Types pillar answers "what kinds of facilities exist?" The Workloads pillar answers "what runs inside them?" Sites cross-references all three extensively: each case study identifies the site's type (AI Factory, HPC Cluster, Hyperscaler, etc.) and the dominant workloads it hosts, and the ranked datasets provide the bulk reference for readers who want to understand the overall deployment landscape rather than any individual facility. The case studies also cross-reference into Stack for infrastructure architecture details, Energy for power sourcing and sustainability posture, and cross-network to SiliconPlans network peers where the site sits at a network boundary (Tesla sites bridging to SemiconductorX Terafab coverage, Starlink bridging to AustinIO and SpaceX coverage).

Related coverage

xAI Colossus | Stargate | Meta Hyperion | Tesla Dojo | Tesla Cortex | Fermi Hypergrid | NEOM AI Center | Rubin Observatory | SKA Array | HL-LHC | Oak Ridge Lab | Starlink Ground Infrastructure | Top 20 US | Top 20 Worldwide | Top Planned | Types | Orbital and Space DCs

Planned & Under Development

The next generation of data centers is being designed at unprecedented scales, often measured in gigawatts of power draw to support frontier AI workloads.