Data Centre Investment Calculator

A data centre investment calculator measures cap rate for colocation and hyperscale facilities. A 100M facility with 10MW of capacity at 150 per kW per month, 80% occupancy and 8M of annual operating expenses produces 14.4M of effective gross income, 6.4M of net operating income, and a 6.40% cap rate. It also reports the cost per kW, which on those figures is 10,000, or 10M per MW of capacity. Demand for compute capacity has grown alongside AI and cloud workloads, and lead times for new hyperscale capacity are commonly measured in years rather than months.

Working through the default figures: 10,000 kW at 150 a month gives 18M of gross potential revenue across a year. At 80% occupancy that becomes 14.4M of effective gross income. Operating expenses of 8M, covering cooling, security, staffing and power conditioning, leave 6.4M of net operating income, and dividing that by the 100M facility cost gives the 6.40% cap rate. That sits within the range commonly quoted for stabilised facilities; ground-up hyperscale developments are often underwritten to higher unlevered targets to compensate for construction and lease-up risk.

Several things shape the sector, and only some of them appear in the inputs. Compute demand growth is tied to AI and cloud workloads, which is where the occupancy and rate assumptions come from. Power availability has become a limiting factor in established hubs such as Northern Virginia, Dublin and Singapore, where grid capacity limits or connection moratoria constrain new build, and international energy analysis now tracks data centre electricity consumption as a category in its own right. Cooling efficiency is measured as Power Usage Effectiveness, which feeds the operating expense line. Customer concentration is a real risk, since a small number of hyperscale tenants account for much of the leased capacity. Lease terms commonly run ten to fifteen years, and capital intensity is high: the cost per kW this calculator reports is the direct measure of it. Access for most investors is through listed data-centre trusts and specialist infrastructure funds, since direct ownership is generally institutional.

A worked example

With the defaults, a facility cost of 100,000,000, capacity of 10,000 kW, a monthly rate of 150 per kW and occupancy of 80%, against 8,000,000 of annual operating expenses, the tool returns 6.40%. The breakdown shows effective gross income of 14,400,000, net operating income of 6,400,000, a cost per kW of 10,000 and the capacity entered.

What moves the number most

Occupancy is the sharpest lever, because every point of it multiplies against the full rate and capacity. On the defaults each ten percentage points of occupancy is worth 1.8M of revenue, which flows straight to net operating income: 90% occupancy gives a cap rate of 8.20%, while 70% gives 4.60%. The Monthly Rate per kW works the same way but is usually harder to move: raising it from 150 to 160 lifts the cap rate to 7.36%. Annual Operating Expenses pass straight through in the other direction, so an extra 1M of opex takes the result from 6.40% to 5.40%. Facility Cost sits in the denominator alone, scaling the cap rate without touching the income lines at all.

The formula behind this

Annual revenue is capacity in kW multiplied by the monthly rate per kW, multiplied by twelve months, multiplied by occupancy as a decimal. Net operating income is that revenue less annual operating expenses. The cap rate is net operating income divided by facility cost, expressed as a percentage. Cost per kW is facility cost divided by capacity, which gives the capital intensity figure the sector is usually compared on. Nothing here is financed, depreciated or taxed.

Where this fits in planning

This is a what-if tool rather than a forecast. It helps to test ideas: what happens if one of the inputs comes in higher or lower than first assumed. Running several sets of figures shows how sensitive the result is to each input, where a single set does not. Occupancy and the rate per kW are the two most worth stress-testing, since both depend on demand conditions that no calculator can see.