Data centers are not one size fits all. This guide sets out the spectrum of facility sizes and a framework for aligning physical footprint with real-world power, cooling and workload demands.
At one end are edge and micro deployments: small-scale sites or modular units handling workloads close to users and devices. In the middle sit enterprise and colocation facilities: private office-sized buildings or urban facilities that trade square footage for high connectivity.
At the far end are hyperscale and AI campuses: massive multi-building sites drawing hundreds of megawatts for cloud services or AI training.
3
Size tiers: edge, enterprise / colocation, hyperscale / AI
100s of MW
Drawn by hyperscale and AI campuses
The right-sizing framework
- Translate workload to IT load: map expected daily requests to the compute power required to process them.
- Optimize density and cooling: use high-density racks to keep footprints smaller without sacrificing total capacity.
- Plan phased expansion: scale in modular blocks or server generations to avoid stranding power and capital.