Data centers are priced differently from every other commercial property type. Rent is quoted per kilowatt, not per square foot. Leases run 10-20 years. Tenants are Fortune 50 companies. And the infrastructure investment — power, cooling, generators — means a 200,000 SF shell is worth ten times what the same shell would trade for as a warehouse.

This lesson covers how cap rates are set in data center real estate, how hyperscale NNN leases are structured, how to value a data center asset using the income approach, and what the development economics look like for a powered shell.

The cap rate landscape

Data centers trade in a different cap rate universe than most CRE. The combination of long lease terms, investment-grade tenants, and high replacement cost pushes pricing to near-bond levels.

| Asset Type | Cap Rate Range | |---|---| | Hyperscale Campus NNN | 5.00 – 6.00% | | Powered Shell (stabilized) | 5.50 – 6.50% | | Powered Shell (development yield-on-cost) | 7.00 – 8.50% | | Colocation (wholesale) | 6.00 – 7.50% | | Edge / Micro Data Center | 6.50 – 8.00% |

For comparison, other major CRE types:

| Asset Type | Cap Rate Range | |---|---| | NNN Retail (investment grade) | 5.00 – 6.50% | | Industrial / Warehouse | 5.00 – 6.50% | | Multifamily | 5.00 – 7.00% |

Hyperscale data centers compress to the low end of these ranges — and sometimes below them — because they combine three things that institutional capital prizes above all else: investment-grade credit, long lease terms, and genuine mission-critical necessity. Amazon, Microsoft, or Google cannot simply relocate a hyperscale campus on 90 days' notice the way a retail tenant can close a store. The switching cost is enormous.

Why do hyperscale leases compress cap rates close to bond levels? The answer is duration and certainty. A 20-year Amazon lease with 3% annual rent bumps is underwritten like a long-duration corporate bond with a real asset as collateral. When institutional investors model that income stream, they compare it to investment-grade corporate debt — and at today's treasury spreads, a 5.25% all-cash return on mission-critical real estate with inflation escalations looks very attractive.

Colocation and edge assets trade at wider caps because lease terms are shorter, tenant rollover risk is higher, and the income stream is less predictable.

The hyperscale NNN lease

A hyperscale lease bears almost no resemblance to a retail NNN lease. It is a complex, heavily-negotiated document, but it follows a recognizable anatomy once you know what to look for.

Base term

Hyperscale tenants sign long. Base terms of 10-20 years are standard, with 15-20 years typical for a purpose-built campus. The tenant needs this certainty because they will invest hundreds of millions of dollars in their own IT equipment inside your building. They cannot move that equipment cheaply.

Rent structure

Rent is structured as absolute NNN or flat NNN. The landlord collects rent and nothing else — taxes, insurance, maintenance, utilities, and capital expenditures all flow to the tenant. Some leases require the landlord to maintain the shell structure and roof; others push that too. Read carefully.

Rent is denominated in dollars per kilowatt per month ($/kW/month), not dollars per square foot. A 50MW facility leased at $120/kW/month generates very different rent per square foot than a retail building. The kW figure is what the lease is priced on.

Escalations

Most hyperscale leases include one of two escalation structures:

Fixed annual step-ups: 2-3% per year, compounding over the term
CPI-tied escalations: rent increases tied to the Consumer Price Index, sometimes with a floor and ceiling (e.g., CPI ± 1%, with a 1.5% floor and 4% ceiling)

Fixed escalations are preferred by landlords for simplicity. Over a 20-year term, a 3% annual bump means rent in year 20 is approximately 80% higher than year 1 — a significant inflation hedge.

Renewal options

Tenants typically negotiate multiple renewal options at each renewal, options are 5-10 years long, and there may be 3-5 of them. If the tenant exercises all options, the total lease term can reach 40-50 years. This is intentional — the tenant wants to retain the optionality to stay as long as the facility serves their needs.

Renewal rents are typically set at:

Fixed percentage bumps over the expiring rent (most landlord-friendly)
Fair market value (least landlord-friendly — creates uncertainty)
The greater of FMV or current rent (common compromise)

Corporate guarantee

This is critical. A hyperscale lease must be guaranteed by the investment-grade parent corporation, not a subsidiary. An Amazon Web Services LLC guarantee is worthless if the LLC has no assets. The guarantee should come from Amazon.com, Inc. — the publicly rated parent with the full balance sheet.

Always verify:

Which entity is the guarantor
Whether that entity is publicly rated
The current credit rating

A lease guaranteed by a special-purpose entity with no external credit rating is not a hyperscale NNN — it is an unsecured lease to an LLC.

Power commitment

Many hyperscale leases include a minimum power draw commitment: the tenant guarantees they will consume at least a defined minimum amount of power (e.g., 80% of contracted capacity). This protects the landlord against a scenario where the tenant has a lease but is not actually using the facility — which could signal impending vacancy.

ROFO / ROFR provisions

Tenants frequently negotiate:

Right of First Offer (ROFO): before the landlord markets the property for sale, they must offer it to the tenant first at a price the landlord sets
Right of First Refusal (ROFR): if the landlord receives a third-party offer, the tenant has the right to match it

These provisions reduce the landlord's exit flexibility but are often required by hyperscale tenants to protect their operational security. Understand them before pricing the deal — a ROFR can chill third-party bids at sale.

Expansion rights

Large tenants often negotiate the right to expand onto adjacent parcels or within the same campus at predetermined economics. These rights can be valuable or burdensome depending on your land position.

Why a 20-year Amazon lease beats almost any retail NNN

A 20-year absolute NNN lease with Amazon.com, Inc. as the corporate guarantor, 3% annual bumps, and four 10-year renewal options is one of the highest-quality income instruments in all of commercial real estate. The credit is rated AA (Amazon's current S&P rating), the term eliminates re-leasing risk for an entire investor career, the bumps outpace most retail NNN escalations (which are often flat), and the replacement cost of the asset makes any future sale resistant to distress pricing. Retail NNN can match none of those characteristics simultaneously.

Valuation: the income approach

The income approach dominates data center valuation. The steps are identical to any cap rate-based valuation:

NOI = Annual rent (NNN = rent only, no expense deductions)
Value = NOI / Cap rate

Worked example: 50MW hyperscale powered shell

Asset: Powered shell, 50MW critical load capacity
Tenant: Investment-grade hyperscale cloud provider
Lease: 20-year absolute NNN, 2.5% annual escalations
Rent: $120/kW/month

Step 1: Calculate annual base rent

50MW = 50,000 kW
Annual rent = 50,000 kW × $120/kW/month × 12 months
Annual rent = $72,000,000

Step 2: Calculate NOI

Because the lease is absolute NNN, there are no landlord expenses. NOI equals gross rent.

NOI = $72,000,000

Step 3: Apply cap rate

At a 5.5% cap rate (stabilized powered shell, investment-grade tenant):

Value = $72,000,000 / 0.055 = $1,309,090,909
Value ≈ $1.31 billion

This is why institutions chase hyperscale data center assets. A single stabilized campus can exceed a billion dollars in value — while delivering bond-like income with inflation protection.

Rent per kW is the primary pricing metric. Square footage matters for construction cost estimation, but the income and valuation of a data center are almost entirely driven by critical power capacity (kW or MW). When underwriting, always start with the kW figure.

Powered shell development economics

Development of powered shells has become one of the more active capital strategies in commercial real estate as hyperscale demand outpaces existing supply. The economics are driven by two figures: total development cost per MW and stabilized rent per kW.

Worked example: 5MW edge/powered shell

Land: $5,000,000
Shell construction: $80/SF × 200,000 SF = $16,000,000
Power infrastructure (substation, switchgear, generators, UPS, cooling): $30,000,000
Soft costs and contingency (architecture, engineering, permits, financing, 10%): $9,000,000
Total development cost: $60,000,000

Development cost per MW: $60M / 5MW = $12M/MW

This is within the typical development cost range of $8-15M/MW for smaller facilities. Larger hyperscale campuses benefit from economies of scale and often come in at $8-10M/MW.

Stabilized NOI at $85/kW/month (appropriate for an edge facility or smaller regional asset):

5,000 kW × $85/kW/month × 12 months = $5,100,000 NOI

Yield on cost:

Yield on cost = $5,100,000 / $60,000,000 = 8.5%

Exit value at 7.0% stabilized cap rate:

$5,100,000 / 0.070 = $72,900,000

Development profit: $72.9M exit value vs $60M cost = $12.9M development gain (21.5% profit on cost).

This is the development trade: build at an 8.5% yield on cost, exit at a 7.0% cap rate, and capture the spread as value creation. The spread between yield-on-cost and exit cap rate is the developer's margin. When that spread compresses — as it has in some primary markets as cap rates tighten — developers shift to secondary and emerging markets where construction costs are lower.

Colocation pricing: a different model

Colocation facilities operate under a fundamentally different pricing and leasing model than powered shells.

How colocation is priced

Colocation rent is charged by the unit of space occupied:

| Unit | Range | |---|---| | Retail cabinet (1/4 rack to full rack) | $1,000 – $3,000/month | | Wholesale cabinet | $400 – $800/month | | Cage (dedicated, 10-50 cabinets) | $8,000 – $40,000/month | | Wholesale ($/kW/month) | $80 – $130/kW/month | | Hyperscale wholesale ($/kW/month) | $100 – $150/kW/month |

Retail colocation — individual cabinets leased to small businesses, law firms, hospitals, and mid-market enterprises — commands the highest per-kW rent but with the shortest lease terms (often 1-3 years).

Wholesale colocation — large blocks of space leased to enterprise customers or smaller cloud providers — trades at pricing closer to powered shell economics, with 3-7 year terms typical.

Why colo trades at wider cap rates

Colocation NOI is higher on a per-kW basis than a flat-rent NNN powered shell, but colocation assets trade at wider cap rates (6.0-7.5%) because:

Shorter lease terms create more frequent re-leasing risk
Higher tenant churn — small retail colo customers turn over at higher rates than hyperscale tenants
Operational complexity — the landlord (or an operating partner) manages ongoing relationships with dozens or hundreds of tenants
Mark-to-market risk — rent is reset more frequently, which can be positive or negative depending on market direction

Colocation assets are operationally intensive. Unless you have an operating partner, the income stream is not truly passive.

Due diligence for data center pricing

Several data center-specific diligence items must be resolved before you can underwrite a deal with confidence.

Power committed vs. power available

Do not pay for power capacity that is not contracted. A facility with 50MW of available utility power but only 30MW of committed tenant demand should be priced on 30MW of contracted NOI, with the remaining 20MW treated as upside — not income. Sellers frequently present "available capacity" in marketing materials. Insist on seeing the executed leases and power service agreements.

Lease expiry profile

A facility with a single 20-year hyperscale lease expiring in year 3 is a very different risk profile than the same facility with 17 years remaining. Model the exact lease expiration dates. What is your re-leasing assumption if the tenant does not renew?

Renewal probability

Hyperscale tenants renew at extremely high rates — moving a data center is operationally catastrophic and capital-intensive. But the renewal probability is not 100%. Factors that reduce it include:

Tenant's own real estate strategy shift (moving to owned campuses)
Power capacity constraints at the site
Fiber connectivity limitations
Market rent significantly above or below the in-place rent

Tenant credit analysis

Run full credit analysis on the guarantor entity — not the operating subsidiary. Pull the most recent rating actions, read the rating agency commentary, and monitor for any outlook changes. An Amazon or Microsoft is not a credit concern, but some colocation operators and smaller cloud providers have weaker balance sheets. Treat them accordingly.

Replacement cost as a pricing floor

One advantage of data center real estate is that the power infrastructure creates a significant replacement cost premium. A facility that would cost $60M to replicate is unlikely to trade below $45-50M regardless of income, because buyers can compare the acquisition cost to a greenfield development. Always calculate replacement cost per MW as a sanity check against the income-approach value.

What to take away

Data center cap rates range from 5.0-8.0%, with hyperscale NNN compressing toward bond-like yields due to credit quality and lease length
The primary pricing metric is dollars per kilowatt per month, not dollars per square foot
A hyperscale NNN lease runs 10-20 years with 2-3% annual escalations, absolute NNN structure, and must carry a corporate guarantee from the investment-grade parent
Income approach: NOI / cap rate — for a 50MW hyperscale asset at $120/kW/month and a 5.5% cap, value approaches $1.3B
Powered shell development targets 8.0-8.5% yield on cost and exits at a 6.5-7.0% cap — capturing the spread as development profit
Colocation trades at wider caps (6.0-7.5%) because of shorter terms, higher churn, and operational complexity
Never pay for uncommitted power capacity — price only what is contracted
Corporate guarantee identity matters: the guarantor must be the rated parent entity, not a subsidiary LLC

Next lesson: site selection and location fundamentals — what makes one market command hyperscale rents while another struggles to attract any tenant demand.

Cap Rates, Lease Structures, and Pricing Data Centers