Data Center Investment Returns, and Why the Headline Is Not Yours

What a projected data center yield is built from, the six places it leaks before it reaches you, and how to turn an advertised figure into the number you would actually receive.

A private data center offering will show you a number: 11%, 15%, 17.5%. It is not a return. It is the output of a model, and between that output and your bank account sit six deductions and two assumptions. This page takes the number apart.

How the headline is built

Almost every projection in this sector is the same arithmetic:

Contracted megawatts × 1,000 × rate per kW per month × 12 = gross revenue Gross revenue − operating expenses = net operating income (NOI) NOI ÷ total equity invested = the advertised yield

For a 10 MW facility at $200 per kW per month: 10,000 kW × $200 × 12 = $24 million of annual contracted revenue. Subtract operating costs and you have NOI; divide by the equity and you have the headline.

Two assumptions are doing most of the work, and they are usually in a footnote rather than the headline:

The rate. For context, CBRE put the global weighted average asking rate at $217.30 per kW per month in Q1 2025 (CBRE), with a North America outlook of $200+ per kW per month (CBRE). Those are asking rates, averaged across markets and deployment sizes. Large wholesale leases commonly price below small retail requirements, so published averages are a sanity check, not a benchmark. A model well above them needs a reason.

The occupancy. Most projections assume the building is full. On a project with no signed lease, year-one full occupancy is the single most aggressive assumption in the document, and the one to interrogate first. See how data center leases work.

The six leaks

Work through these in order and the headline usually loses a third or more.

1. Occupancy shortfall. Lease at 80% of the assumed capacity and you lose 20% of gross revenue — but operating costs do not fall by 20%, so NOI falls by more. Operating leverage cuts both ways, and this is the downside direction.

2. Timing. Rent starts when the tenant takes occupancy, not when you fund. If energisation slips a year, you have a year of capital earning nothing. On a 15% projection, a one-year delay roughly halves the return over a two-year holding period, and the capital is still exposed.

3. Asset management fees. An annual percentage of invested capital or assets. Charged whether the project performs or not.

4. Sponsor promote or carried interest. A share of profit above a threshold. This is the fee most aligned with you, because it only pays if you do well — but it does take a slice of the upside, and the advertised yield is often struck before it.

5. Acquisition, sourcing and financing fees. Usually a percentage of project cost, frequently paid at closing. These are earned on transacting, not on succeeding, which is why the balance between closing fees and promote tells you something about where the sponsor's incentives point.

6. Sponsor units. Units the sponsor holds that share in distributions. Your per-unit economics are diluted by however many there are. A per-unit NOI figure described as "net of Sponsor Units" has accounted for this; one that is not, has not.

The two questions that convert a projection into a return

"What is the net-of-all-fees cash yield, assuming the base case?" If the answer is not a single number, the fee structure is not clear enough to invest against.

"What is that figure if the building leases at 85% of the assumed rate, or six months late?" A sponsor who has modelled downside cases will have an answer. One who has not is showing you a single optimistic line and calling it a forecast.

Why higher advertised yields are not better deals

In a sensibly priced market, a higher projected return means more risk has been left in. The clearest expression of this is a tiered offering where the price rises and the target yield falls as milestones are cleared — land closed, power contracted, lease signed. The investor in the earliest tier is not getting a bargain; they are being paid for accepting risks the later tiers will not have to take.

So the useful question is never "which yield is highest" but "which risks are still live, and am I paid enough for the ones that are?" A project with a signed lease to a strong tenant and a signed electric service agreement has retired the two biggest risks in this asset class, and its return should look more like a REIT's than a development deal's. A project that has retired neither and offers only a modest premium is mispriced against you.

Comparing with the liquid route

Listed REITs distribute most of their taxable income because the dividends-paid deduction must equal or exceed 90% of taxable income excluding that deduction and net capital gain (IRS). The resulting yields are lower than private projections and they are actual distributions from leased buildings, not projections about unleased ones, marked to market every day and audited quarterly.

That is the comparison worth making: not 15% against 4%, but a projection about a building with no tenant against a distribution from hundreds of buildings that have them. See REITs vs private deals.

Nothing here is investment advice. See how this site is funded.