Wholesale Colocation Data Center Feasibility Study: Ground-Up Development
Ground-up wholesale colocation build in a primary market: megawatt capacity and power-availability analysis, a multi-year lease-up, and base and downside development returns.
Engagement Snapshot
Representative engagement profile. Project parameters are illustrative of this market. Market figures reflect current third-party data and are sourced in the exhibit below.
The question put to the analysis
A developer proposed to build a ground-up wholesale colocation data center of roughly 48 megawatts of critical IT load in a primary market, financed through a construction loan and institutional and private-equity equity, leased to multiple tenants by the kilowatt. The lender and the equity needed an independent assessment of whether the power could be delivered on a financeable timeline, whether the market would absorb the capacity at the assumed rents, and whether the project would earn a development yield wide enough over the market capitalization rate to justify the construction risk.
A data center is not a conventional building, and the feasibility question reflects that. Its capacity is measured in megawatts of power, not square feet, and the binding constraint on a data center is almost never demand and almost always power: whether the utility can deliver the load, and when. The revenue is built from the IT load, a price per kilowatt per month, and a lease-up curve, and the economics are judged not on a stabilized cap rate but on the development spread, the gap between the yield on total project cost and the cap rate the stabilized asset would trade at. A study that treats a data center as ordinary industrial real estate, or that assumes power is available, misstates the asset and the risk. This engagement profile illustrates how a feasibility study resolves the power question first, derives revenue from megawatts and price per kilowatt, models the lease-up, and tests the development on yield, coverage, and return across a base and downside case.
Power and capacity: the binding constraint
The analysis began with power, because in a data center the utility interconnection timeline, not the market, sets whether the project is feasible at all.
The analysis sized the total utility demand from the IT load and the power usage effectiveness, because a data center draws more power than its IT load alone: at a PUE of roughly 1.4, the 48 megawatts of critical IT load require approximately 67.2 megawatts of total facility power, the figure the utility must actually deliver. The central diligence task was then the interconnection: confirming the substation and transmission capacity, the utility's commitment, and the energization schedule, because grid connection in primary markets commonly runs multiple years and is the single most important variable in the timeline. The analysis treated a documented utility commitment with a financeable energization date as a condition of feasibility rather than an assumption, and flagged that a project without one carries a fundamentally different risk profile regardless of how strong the market looks.
The market and absorption
With power addressed, the analysis turned to whether the market would absorb the capacity.
The market read was favorable but the analysis modeled the lease-up rather than assuming instant fill. Primary-market vacancy sat at record lows and the great majority of capacity under construction was already pre-committed, reflecting demand from AI and cloud that has outpaced supply, which supported the absorption thesis. But a speculative or partially pre-leased build still carries lease-up risk, so the analysis modeled a ramp from roughly 30 percent leased in the first year to stabilized occupancy above 90 percent by the third, and noted that lenders impose pre-leasing requirements precisely because the lease-up is where the multi-tenant model carries its risk. The diversification across multiple tenants was itself part of the credit, giving the asset mark-to-market rent upside on renewals that a single-tenant lease does not.
The revenue calculation
Revenue was built from the IT load, the price per kilowatt, and the lease-up, with the math shown.
The revenue build followed the data center logic directly: the critical IT load in kilowatts, times the price per kilowatt per month, times twelve, produced an annual stabilized revenue of approximately $106.56 million at the blended rent. Under a triple-net or modified-gross structure with power passed through to tenants, the analysis modeled a stabilized margin near 70 percent of base rent, producing a stabilized net operating income of roughly $74.6 million, while noting that the actual margin depends on the lease structure and the recoverable costs. The analysis flagged that asking rents had been rising at record rates as supply tightened, which supported the rent assumption but also introduced the risk that new supply could eventually compress it.
Development cost and the yield-on-cost test
The economics were judged on yield-on-cost against the market cap rate, the test that governs a development.
This is the test a data center development turns on. The yield-on-cost, the stabilized net operating income divided by total project cost, came to roughly 13.5 percent in the base case against a stabilized market cap rate near 6.5 percent, a development spread of 300 to 450 basis points that compensated for the construction and lease-up risk. The analysis was candid that the base case was aggressive and modeled a conservative case, with lower stabilized rent, a fuller operating cost load, and a slower lease-up, that brought the yield-on-cost down to the 9 to 11 percent range and the spread toward the 75 to 150 basis point minimum that developers require, below which the project would not justify the risk. The analysis also priced the powered-shell alternative, in which the tenant funds the fit-out, as a roughly $288 million lower-cost, lower-margin path.
Coverage and returns
The credit was tested on coverage and the development on return across a base and downside case.
The base-case coverage was robust: a stabilized debt yield near 17 percent against the 7 to 10 percent floor lenders require, and a debt service coverage ratio around 2.5 times, reflecting the wide development spread once stabilized. The construction was modeled at roughly 60 percent loan-to-cost with the balance in institutional and private-equity equity, consistent with how speculative colocation is financed. The downside case, modeling lower rents as supply catches up, a slower lease-up, and construction-cost or equipment overruns, compressed the coverage and the return materially. The real exposures are the power-delivery timing, the lease-up pace, and the equipment lead times, which is why the analysis resolved power first, modeled the lease-up explicitly, and stress-tested cost and rent together.
Where the risk concentrates
The risks in a ground-up data center are specific and the analysis addressed each. Power availability and interconnection delay are the dominant risk, because the project is not feasible on any timeline the utility cannot meet, addressed by treating a documented utility commitment and energization schedule as a condition rather than an assumption. Supply-chain and equipment lead times are the related construction risk, since transformers, switchgear, and generators carry lead times measured in years and can delay energization, modeled in the timeline and the downside. Lease-up and absorption risk is the multi-tenant model's central risk, addressed through a modeled ramp and the pre-leasing lenders require.
Construction cost and timeline overruns are a capital risk for a project where electrical and mechanical systems are most of the cost, stress-tested in the downside. Rent compression from new supply is a revenue risk in a market where rents have risen sharply. Obsolescence and density shifts are a technology risk, as rising rack densities and the move to liquid cooling can strand power and cooling capacity, which the analysis flagged for the design. Power cost, sustainability and grid-strain, water use for cooling, and interest-rate sensitivity round out the set, with environmental and site due diligence handled as separate scopes.
Conclusion
The engagement reached a conditional affirmative, with the conditions led by power. The project was feasible only on a documented utility commitment with a financeable energization date, which the analysis treated as the threshold question rather than an assumption, and the total facility power was sized from the IT load and the PUE so the utility requirement was explicit. The market supported the capacity at record-low vacancy, but the lease-up was modeled rather than assumed, and the economics were judged on the development spread, which held in the base case and compressed toward the minimum in a candid downside.
The value of the feasibility study was not in declaring a strong data center market a good place to build, but in resolving the power and interconnection question first, in deriving revenue from megawatts and price per kilowatt the way the asset actually earns, and in showing the lender and the equity both the wide base-case spread and the real exposure if power slips, the lease-up lags, or supply catches up.
Market Data Exhibit
Pro forma figures are representative of this market. Industry benchmarks are indicative; verify current power availability, vacancy, rents, development costs, cap rates, and financing terms before citation in a client deliverable.