Valuing a Data Center: A Hyperscale-Lease Walkthrough

Introduction

Valuing a data center looks familiar at first, an income-producing property capitalized at a market rate, but the inputs are unlike anything in traditional real estate. The unit of value is power, not floor area; rent is quoted per kilowatt per month; a large share of revenue is a pass-through of electricity rather than rent; and the asset may not be fully energized on day one, so revenue phases in as power load is absorbed. Get those mechanics right and the valuation is tractable. Treat the building like a warehouse and the answer will be wrong by a wide margin. This walkthrough builds the value of a stabilized hyperscale facility from the ground up, using a single worked example to keep the numbers concrete.

The setup: a 50-megawatt facility leased entirely to a single investment-grade hyperscaler on a long-term, triple-net basis, with a base rent of $90 per kilowatt per month and fixed annual escalators. We will turn that into a stabilized net operating income, capitalize it, cross-check it with a discounted cash flow, and bound it with a sensitivity table.

Before any numbers, the analyst has to internalize three structural differences from office or industrial valuation. First, everything is denominated in power: capacity, rent, capital cost, and NOI are all quoted per kilowatt or per megawatt, because the scarce, value-driving input is deliverable electricity, not square footage. Second, electricity is usually passed through to the tenant at cost, so power reimbursement can represent 30% to 50% of gross revenue while contributing little to NOI, which means a naive revenue multiple badly overstates value. Third, value depends on energization: a facility leased but not yet powered earns nothing until load is absorbed, so the model must phase revenue rather than assume day-one stabilization.

Stabilized NOI (Data Center)

The normalized annual net operating income a facility produces once it is fully energized and leased, calculated as base rent plus any rent-like recoveries, less non-recoverable operating expenses and a capital reserve. Power passed through at cost is excluded from NOI because it is revenue and expense in equal measure.

These features sit on the lease, so the valuation is only as good as the analyst's read of the data center lease structure, particularly the base rent, the escalator, the term, and how power and PUE risk are allocated. A turnkey or hyperscale lease to a strong tenant produces bond-like cash flow; a powered-shell or short-term colocation profile does not, and the cap rate has to reflect that difference.

From Pure Real Estate to Operating Business

Data centers span a spectrum from pure real estate to operating business, and where an asset sits changes both the method and the cap rate. A powered shell leased triple-net to a single tenant is almost pure real estate: the landlord collects rent and the tenant runs everything, so direct capitalization of a bond-like NOI is the right tool. A multi-tenant retail colocation facility is closer to an operating business, with customer churn, interconnection revenue, and active management, which pushes the analysis toward a going-concern valuation and a wider cap rate for the operational risk. Our single-tenant hyperscale example sits near the real-estate end of that spectrum, which is exactly why its NOI is so clean and its cap rate so tight. Misjudging where an asset sits, valuing an operating colocation business as if it were a net-leased box, is one of the more common ways to misprice the sector.

The cleanest way to value the asset is direct capitalization of stabilized NOI, cross-checked with a DCF. The sequence is the same discipline used in the broader direct capitalization method, adapted to power-based inputs.

Building Stabilized NOI

Start with base rent. A 50-megawatt facility is 50,000 kilowatts of critical load. At $90 per kilowatt per month, annual base rent is 50,000 multiplied by $90 multiplied by 12, or roughly $54 million. Because the lease is triple-net, the tenant pays power, taxes, insurance, and most operating costs, so the bridge from base rent to NOI is short: subtract a modest non-recoverable management and structural reserve, on the order of $2 million, to reach a stabilized NOI of roughly $52 million. The discipline of separating base rent from pass-through power is the same NOI logic explained in property cash-flow mechanics, with power as the dominant pass-through line.

The bridge looks like this, and the line that matters most is the one that nets to zero:

What this bridge reveals is how clean the cash flow is. Under a triple-net hyperscale lease, stabilized NOI lands within a few percent of base rent, with no meaningful leasing costs, tenant improvements, or operating drag for the landlord to absorb. That is what gives a well-leased data center its bond-like quality: nearly every dollar of contracted rent falls to NOI, year after year, escalating on a fixed schedule. The valuation challenge is therefore less about forecasting expenses than about the durability of that rent and the rate at which to capitalize it.

Applying the Cap Rate

With stabilized NOI in hand, direct capitalization is mechanical:

Stabilized institutional-quality data centers trade in a cap-rate range of roughly 4.25% to 6.25%, with the tightest pricing reserved for long leases to the strongest tenants. For our facility, a 5.75% cap rate on $52 million of NOI implies a value of about $904 million, or roughly $18 million per megawatt. The per-megawatt cross-check matters: if the implied figure were far above or below the $10 million to $12 million typical build cost per megawatt, the analyst should understand why, since the gap between value and cost is the development spread the sector is built on.

Direct capitalization assumes a steady-state. A DCF is the better tool when a facility is still leasing up or when the escalators meaningfully change the cash-flow trajectory. The income approach for a stabilized asset and the property-level DCF should triangulate to a similar number when both are built carefully.

The DCF projects the lease cash flows year by year. With a 15-year lease and 2.5% annual escalators, the $52 million of year-one NOI grows steadily, and each year is discounted at a market rate, here around 7.5%, consistent with the 7.0% to 8.5% unleveraged IRRs institutional buyers underwrite. The residual is the larger driver of value: at the end of the hold, the analyst applies an exit cap rate to the following year's NOI to estimate a sale price, then discounts that back.

Exit Cap Rate

The capitalization rate assumed at the end of a hold period, applied to the next year's forecast NOI to estimate the asset's terminal sale value. Buyers typically set the exit cap modestly above the entry cap to reflect the asset's aging and the uncertainty of forecasting a sale years out.

Phasing the Lease-Up

The lease-up problem deserves its own attention because it is where data center DCFs most often go wrong. A facility rarely energizes all at once; power load is absorbed in tranches as utility capacity comes online and the tenant fills the building. Suppose our 50-megawatt facility delivers in three tranches: 17 megawatts energized at the start, another 17 megawatts nine months later, and the final 16 megawatts at month eighteen. In that case, year-one cash revenue is well below the $54 million stabilized base rent, because only a fraction of the capacity is paying for much of the year, and the asset does not reach full run-rate until roughly year two. A model that assumes full revenue from year one will overstate value, sometimes badly, because it ignores the months during which capital is sunk and revenue is partial. This is the valuation echo of the power constraint: the same time-to-power that gates development also delays the cash flow a buyer is paying for.

A Worked Residual

For a stabilized asset already at full load, the DCF is cleaner, and the residual does most of the work. Holding ten years, the NOI grows from $52 million at a 2.5% escalator:

Applying a 6.0% exit cap rate to the year-eleven NOI of $66.6 million gives a terminal value of about $1.11 billion. Discounting the ten years of growing NOI plus that terminal value at a 7.5% rate produces a present value of roughly $930 million, within a few percent of the $904 million direct-capitalization figure. When the two methods triangulate like this, the analyst can present the value with confidence; when they diverge, it usually means the cap rate, the escalator, or the lease-up assumption needs a second look.

The cap rate is the single most consequential input, and it is not a market average plucked from a survey. It is built up from the specific risk of this asset. Four drivers dominate, and they follow the same logic as the broader cap-rate framework applied to data centers.

As a rough calibration, each driver moves the cap rate by a quantifiable amount. A step down in tenant quality, from an investment-grade hyperscaler to a weaker neocloud, can add 100 to 200 basis points. A site with secured, energized power versus one stuck in an interconnection queue can be worth another 50 to 150 basis points of compression. Lease duration and product type each contribute on top. Stacking those adjustments onto a market baseline is how an analyst defends a specific cap rate rather than asserting one, and it is precisely the build-up logic an interviewer wants to hear.

A long lease to a top-rated tenant on secured power sits at the low end of the range, near the 4.25% floor; a speculative powered shell with rollover and uncertain power sits well above it. As a market reference, data center assets financed through CMBS have priced near a 6.5% cap rate, and Class A facilities have generally traded 100 to 150 basis points above the ten-year Treasury yield. Picking a point in the range is an argument about credit, duration, and power, not a lookup.

A model produces a value; the market confirms or contradicts it, and transaction evidence is the reality check. Deal activity surged through 2025: S&P Global Market Intelligence reported that data center M&A hit a record of more than $69 billion for the year, with US private equity investment in the sector reaching a five-year high, led by the roughly $40 billion acquisition of Aligned Data Centers. Within that flow, investor appetite skewed heavily toward the most contracted product: turnkey hyperscale was the preferred segment for about 42% of investors and powered shell for 28%, while hyperscale build-to-suit jumped to 49% from 31% a year earlier as the top opportunity. That preference ranking maps directly onto the cap-rate stack, with the most bankable product clearing at the tightest yields.

The buyer universe explains the resilience of pricing. Institutional investors, infrastructure funds, sovereign wealth vehicles, and REITs increasingly treat data centers as a long-duration core holding, and that capital gravitates toward assets with secured power, renewable integration, and proximity to cloud and AI clusters. The same deep pools that fund the private data center platforms and that the in-house teams at sovereigns and pensions deploy are the marginal buyers setting cap rates, which is part of why pricing has stayed firm even as rates rose.

Reinvestment and Obsolescence Risk

The terminal value in the DCF is not free money. A data center's shell, land, and core power infrastructure are long-lived, but the cooling and electrical fit-out tuned to one generation of chip density can require expensive reinvestment as densities climb. A disciplined valuation separates the durable elements from the refreshable ones and reflects, in the exit cap or a capital reserve, the cost of keeping the facility competitive across the hold. Ignoring that reinvestment drag is how buyers overpay for assets that look stabilized but face a costly retrofit.

Entitlement and Power Risk

Power and entitlement risk can erase value before a building ever earns rent. Between May 2024 and March 2025, more than $64 billion of data center projects were delayed or blocked amid local opposition driven by environmental, energy, and land-use concerns. For a speculative or pre-stabilized asset, that risk is a real haircut: a site without secured power and entitlements is worth a fraction of an identical site that has both. The premium for de-risked, energized capacity is not a market quirk; it is the value of certainty in a sector where certainty is scarce.

Financing and Levered Returns

Most buyers do not pay all cash, and the financing assumption shapes what they can pay. Data center assets are financed with mortgage debt, increasingly through CMBS that has priced near a 6.5% cap-equivalent yield, and leverage amplifies returns when the spread between the asset yield and the cost of debt is positive. A stabilized facility bought at a 5.75% cap and financed below that rate produces a levered cash-on-cash return above the unlevered yield, and the full picture is measured with the IRR and equity-multiple math common to all real estate, set out in IRR, equity multiple, and the waterfall. For a banker, this is not a footnote: the buyer's cost and amount of debt often determine whether it can clear the seller's price and still hit its return target, which is why the same asset can be worth more to a low-cost-of-capital institution than to a leveraged sponsor.

No data center value should be presented as a single number. The honest output is a range, driven mostly by the cap rate:

A 100-basis-point move in the cap rate swings the value by more than $150 million on this single facility, which is why defending the cap rate matters more than refining the NOI to the last dollar. Beyond the cap rate, the recurring pitfalls are worth stating plainly, because each one has produced real mispricings:

Each of these is avoidable with discipline, and each separates an analyst who understands the asset from one who has dropped data center numbers into a generic real estate template. The list is also a useful self-check: before presenting a value, run back through it and confirm none of the five is quietly baked into the model.

A data center is valued as a long-duration, power-based income stream whose worth turns on the lease, the cap rate, and the timing of energization. The mechanics rhyme with traditional real estate, capitalize NOI, build a DCF, defend a cap rate, but every input is translated into the language of power, and the analyst who keeps that translation straight will value the asset correctly while others double-count the electricity bill. For our example facility, that discipline produced a defensible range of roughly $830 million to $990 million, centered near $900 million, or about $18 million per megawatt, a figure that holds up under both direct capitalization and a discounted cash flow. The number is only as good as the cap rate behind it, which is why the entire exercise ultimately rests on a clear-eyed read of tenant credit, lease duration, and the security of power.