Colocation Lease
Definition
A colocation lease is a data-center arrangement where a customer places its own servers, networking equipment, or GPU infrastructure inside a third-party facility and pays for capacity, power, cooling, security, and connectivity. In AI infrastructure, colocation terms increasingly revolve around power availability, rack density, cooling design, uptime, and expansion rights.
Why it matters
Colocation leases turn physical infrastructure into recurring revenue, but the economic quality depends on contract structure. A lease priced by power, rack, square foot, or reserved capacity can produce different margins and risk allocation. For AI workloads, investors should examine whether the lease supports high-density GPUs, liquid cooling, power pass-throughs, and customer commitments long enough to justify capex.
Common misconceptions
- •Colocation is not the same as cloud computing; the customer may own or control much of the hardware.
- •A signed lease does not guarantee AI-ready economics if power delivery or cooling upgrades are not complete.
- •Data-center revenue can be constrained by power and cooling before square footage is fully used.
Technical details
Common pricing structures
Contracts may price by committed power capacity, racks, cages, cabinets, cross-connects, managed services, or metered usage. AI colocation often centers on power density and cooling capability rather than generic floor space.
Power costs can be fixed, passed through, indexed, or subject to separate power purchase arrangements. The allocation of power-price volatility matters for margin stability.
Key contract terms
Important terms include committed capacity, term length, renewal options, expansion rights, service-level agreements, uptime credits, power delivery, cooling specifications, security, remote hands, cross-connect pricing, and termination rights.
AI tenants may require specialized networking, liquid cooling support, and rapid deployment schedules that exceed standard enterprise colocation requirements.
Diligence questions
Is revenue tied to take-or-pay capacity, actual usage, or month-to-month occupancy?
Who pays for power volatility, cooling retrofits, and hardware refresh-related changes?
Can the facility support the tenant's expected rack density over the full lease term?