Bottom line: Hyperscale leases are priced by the kilowatt, not the square foot. Modeling them on $/SF hides about $1.92M per $5/kW/mo swing on a typical 32MW deal. Model on power from day one.
The $1.92M calculation
For a typical 32MW hyperscale facility: $5/kW/mo × 32,000 kW × 12 mo = $1.92M/year in revenue missed if the model rounds to PSF. At portfolio scale, this can represent a valuation error, not just a rounding error.
Public CBRE/JLL 2024 benchmarks put hyperscale at $130–150/kW/mo versus retail colo at $250+.
What PSF models get wrong
- They ignore ramp schedules: MW contracted, MW billing, and MW committed.
- They flatten tenant mix: GPU high-density, hyperscale, and retail colo wear very differently.
- They disconnect power pricing from lease economics.
How to model hyperscale
- Lease on $/kW/mo, with term, escalations, and TCV.
- Schedule on MW—contracted, ramped, and churned by month.
- Use responsive opex tied to occupancy and tenant mix, not static $/SF.
- Phase construction—tie shell, MEP, and expansion to MW delivery.
- Report on power—IRR/EM and waterfall flow from the same power-aware model.
DIJASOFT says it does all five natively because it was built for digital infrastructure investments, not generic office.
Explore the topic guide
This post is part of the BOFU cluster for data center investment companies. Start with the pillar for the primary buying keyword: Data Center Investment Management Software →, with an evaluation checklist, comparison table, and FAQ designed for featured snippets.
*Companion: Data Center Pro Forma Software: Excel vs. Platform—when to graduate from 40-tab Excel.*