Colocation

Colocation (colo) is a data centre service model where the facility provider supplies power, cooling, physical security, and network connectivity, while the tenant provides and operates their own IT equipment. In GPU infrastructure, colocation has become the primary deployment model for neoclouds and hyperscalers building large GPU clusters — it is faster and less capital-intensive than constructing purpose-built facilities. Colocation contracts for GPU workloads typically cover 3-25MW of IT capacity, with lease terms of 5-15 years and pricing around $130-250/kW per month depending on location and density.

GPU colocation differs from traditional enterprise colo in several ways: higher power density (40-120kW per rack vs 5-10kW), liquid cooling requirements, heavier floor loading, and more demanding electrical distribution. "Wholesale" colo provides dedicated halls or buildings; "retail" colo provides individual cages or racks. GPU deployments typically require wholesale arrangements due to power density and cooling requirements that cannot be accommodated in shared retail environments.

We work with data centre operators to navigate the GPU colocation market — connecting them with neocloud and hyperscaler tenants, structuring referral and agent relationships, and advising on capacity pricing. Our engagement with European data centre platforms has involved matching 5-15MW capacity requirements with available facilities across Germany, the Netherlands, Luxembourg, and the Nordics.