Data Center Development Model

As the built environment evolves to meet the demands of an increasingly digital world, data centers have emerged as one of the most sophisticated and capital-intensive asset classes in commercial real estate. What was once considered a niche subset of infrastructure has become a foundational layer of the global economy, driven by AI, cloud computing, SaaS, 5G, blockchain, and machine learning.

The shift is unmistakable. Institutional capital is rapidly moving into digital infrastructure assets. Hyperscale platforms, sovereign wealth funds, REITs, private equity, and infrastructure funds are deploying billions into the development of large-scale, mission-critical facilities. With that growth comes both a challenge and an opportunity for real estate professionals: to model, structure, and evaluate these investments with the same precision and discipline expected in any core asset class.

That need led to creating the Data Center Development Model — a powerful tool designed for real estate professionals underwriting digital infrastructure from development through stabilization. Whether you’re underwriting a single-tenant build-to-suit or a multi-phase hyperscale campus, this model equips you to forecast capital deployment, model phased ramp-up schedules, structure layered capital stacks, calculate OpEx recoveries, and evaluate stabilized exits with confidence.

More than just a financial model, this tool is a framework built for professionals working at the intersection of real estate, infrastructure, and technology. In what follows, I’ll explore the key drivers of value in data center development and walk through a modeling approach specifically designed for this asset class.

Data Center Development Financial Modeling

In traditional real estate, we’re taught to model value around square footage. Whether it’s office, multifamily, industrial, or retail, the core assumptions revolve around rentable space — how many square feet, at what rate, and over what absorption period.

But that framework doesn’t hold in data center development. Here, the monetized product isn’t space — it’s power. Specifically: IT Load, measured in kilowatts (kW). It’s the core unit of design, the basis for tenant leases, and the foundation for how investors and lenders size capital.

Power is what drives revenue. It dictates how infrastructure is built, what redundancy is required, and how the facility performs. In these deals, space is secondary. A data hall could be physically complete, but without an energized IT Load, it’s economically offline.

That’s why the financial model has to evolve. That said, a data center model must:

• Follow how power is ramped over time, not just delivered upfront
• Reflect the way rent is billed — by the kilowatt, not by the square foot
• Capture pro rata expense recovery based on power, not area
• Allow for real vacancy and ramp lag, even on fully executed leases
• Tie capital cost and revenue to a common metric: $/kW delivered

Whether you’re underwriting your first hyperscale lease or evaluating a three-phase GPU campus, your model needs to align with how capital flows into these deals. That means thinking in terms of delivered power, leased capacity, and stabilized NOI per kW.

The Data Center Development Model was built to meet that standard, combining real estate fundamentals with the technical precision required for digital infrastructure. For example, it captures the development draw schedule, tenant ramp timing, OpEx recovery, general vacancy & credit loss, and exit valuation — all centered around IT Load.

Because when it comes to data centers, modeling power isn’t just helpful — it’s essential.

The Core Drivers of the Data Center Development Model

The Data Center Development Model was built from the ground up to serve one purpose: to help real estate professionals underwrite institutional-scale digital infrastructure projects with the same rigor they bring to traditional asset classes.

This is a developer-focused, fully dynamic tool designed to handle scenarios where the business plan includes developing, leasing, stabilizing, and ultimately disposing of hyperscale data center assets. The model follows the full investment lifecycle — draw schedule, construction phasing, tenant absorption, operating cash flow generation, and valuation at stabilization — all tied to the economic engine of IT Load.

The Data Center Development Model template is purpose-built for real estate professionals, capital partners, and digital infrastructure developers — anyone tasked with structuring, underwriting, or managing high-voltage, capital-intensive real estate. The current version was built for clarity and adaptability. It features a primary underwriting tab for inputting assumptions, calculating monthly draws and lease-up, and outputting detailed return metrics for both levered and unlevered scenarios. It also includes a version control tab to track enhancements as we continue to expand the model’s capabilities.

The Data Center Development Model was built to meet the rigorous standards of institutional underwriting in today’s digital CRE environment. As we continue improving the model, we welcome your feedback on how it can better support your data center investment process

IT Load (KW): The Leasing Currency

At the core of the data center model — and this model — is IT Load, measured in kilowatts (kW). This is the product. It’s what tenants lease, what developers deliver, and what investors monetize.

Unlike traditional asset classes, where square footage defines the rentable area, in data centers, power defines revenue capacity. That’s why tenant ramp-up in this model isn’t tied to SF, but to how much IT Load each tenant absorbs over time.

Take a data center tenant, for example. The lease may commence at Month 26, but actual occupancy and rent start ramping at Month 29, growing by 417 kW per month until they reach their full 6 MW commitment. Each tenant could follow this structure, and the model tracks it precisely. This reflects how power ramps work in practice. Hyperscale tenants sign for large blocks of capacity, but they bring load online gradually as space is commissioned, hardware is deployed, and redundancy tests are completed.

The result: the model doesn’t just simulate space absorption — it simulates load absorption, which is how revenue actually materializes in a data center.

Recovery Income That Tracks Actual Ramp

In this model, recovery income scales with delivered IT Load, not square footage. Tenants don’t pay their share of OpEx — including utilities, maintenance, staffing, and security — until they begin drawing power. That nuance creates a much more accurate cash flow timeline.

Recovery is allocated on a pro rata basis: each tenant pays their share of total OpEx, based on how much of the IT Load they’ve absorbed. And because each tenant ramps on their own timeline, recovery income phases in gradually, month by month, not all at once.

This gives you a cleaner view into net operating income stabilization.

General Vacancy & Credit Loss: A Realistic Buffer

Even in fully leased facilities, real-world frictions exist — and institutional models need to account for them.

This model includes a general vacancy and credit loss factor, applied as a percentage haircut to gross revenue. It accounts for:

• Delays in energizing tenant load
• Temporary drops in usage (known as deramps)
• Billing disputes or operational misalignments
• Equipment shipping or installation lag

These aren’t theoretical risks. In real-world deals, power delivery can be delayed by substation schedules, network latency, or even weather. A ramp may slow temporarily while a tenant commissions hardware. Recovery may lag if metering or allocation data is disputed.

Power-Driven Operating Expenses

A core distinction between traditional asset classes and data centers lies in how operating expenses behave. In multifamily or office, OpEx is typically tied to occupancy or leased square footage. But in a data center, expenses — particularly utilities — scale with power consumption.

The model ties utility costs directly to ramped IT Load, using a Power Usage Effectiveness (PUE) input to estimate Total Facility Power. This includes not just tenant draw, but also the power required to operate backup systems, cooling, lighting, and infrastructure.

As each tenant ramps their load, utility expenses — by far the largest line item — scale accordingly. But the model doesn’t stop there. It breaks out every major OpEx category to reflect real operational behavior:

• Maintenance and repair of critical systems (UPS, chillers, fire suppression)
• Staffing and 24/7 security, often required in high-tier builds
• Insurance aligned with equipment density and critical uptime
• General & Administrative (G&A) for back-office and operating overhead
• Management fee, tied to Effective Gross Revenue (EGR)
• Flat property taxes, phased in as a dollar amount consistent with institutional practice
• Capital reserves, allocated on a $/kW/month basis to support long-term reliability

This approach gives users a clearer path to understanding how OpEx evolves alongside tenant ramp-up — and ultimately, how NOI stabilizes in a dynamic, load-based environment.

Hyperscale Classifications And The Role Of Uptime Certification

To build and model institutional-scale data centers, it’s not enough to understand how rent is calculated. You also need to understand what type of facility you’re building and what technical standards you’re meeting. These aren’t academic distinctions; they directly shape leasing assumptions, development scope, capital deployment, and risk profile.

Two frameworks are essential:

• Hyperscale classification, which speaks to the scale and tenant type.
• Uptime Tier Certification, which defines system resiliency and operational guarantees.

Hyperscale Classification: Understanding Scale

The term refers to the deployment of massive computing power, but the range of what qualifies — and the type of tenant involved — can vary widely:

Small-Scale Hyperscale (5–10 MW): Often regional SaaS platforms or enterprise cloud deployments. These builds may serve a single tenant but rely on more standardized infrastructure.

Core Hyperscale (10–40 MW): This is the institutional norm — large-scale deployments by cloud platforms like AWS, Google, or Microsoft. These facilities typically involve long-term leases, dedicated infrastructure, and layered redundancy.

Mega Hyperscale (50–200+ MW): These are full campus builds, often developed in phases. They support high-density AI workloads, GPU clusters, and sovereign cloud infrastructure. CapEx per kW is high, but so is long-term value and exit demand.

Uptime Certification: Quantifying Reliability

Alongside scale, investors and tenants evaluate data centers based on their Uptime Institute Tier Certification — a widely adopted standard for assessing resiliency and redundancy. It tells you, in practical terms, how well a facility can maintain uptime during maintenance, failure, or unexpected outages.

Tier I: Basic infrastructure, with a single path for power and cooling. Roughly 28.8 hours of downtime per year — not suitable for mission-critical workloads.

Tier II: Some redundancy added — usually N+1 components — but still single-path delivery. Around 22 hours of expected downtime annually.

Tier III: The institutional standard. Full concurrent maintainability — the facility can perform maintenance without shutting down. Redundant distribution paths and less than 2 hours of downtime per year.

Tier IV: Fully fault tolerant. 2N+1 redundancy ensures that even during failure events, operations continue uninterrupted. Designed for workloads where downtime is simply not an option — think financial exchanges, defense, or mission-critical AI.

Together, hyperscale classification and Tier Certification shape the financial reality of your project. They influence:

• CapEx assumptions
• Ramp timing and lease structure
• OpEx scaling
• Exit valuation and buyer pool

In short: these are not technical details. They’re underwriting fundamentals.

What’s Under The Hood – Data Center Development Model

Version Tab (Default View)

The version tab serves as the model’s homepage. It includes:

• A detailed change log outlining updates in the latest release.
• Helpful links to model tutorials, guides, and additional support materials.

Underwriting Tab (Primary Inputs – Default View)

The Underwriting tab serves as the central hub for inputting and reviewing the assumptions that drive the model. It’s where all key variables are entered, from development budget and capital stack to lease-up timing, OpEx recovery, and reversion assumptions.

The layout is designed for clarity and speed. Inputs are structured into six core sections, arranged vertically from top to bottom. You can scroll through each section manually, or use the ‘ONE-CLICK SECTION SHORTCUTS’ at the top of the tab to jump directly to the area you need.

Here’s how the tab is organized:

Summary – A high-level snapshot of the deal, including project name, location, deal type, IT Load, Total Facility Power, timing, and key metrics like. This section also includes shortcuts to each modeling section.

Description – Basic asset-level information..

Investment – The full capital stack: equity, senior debt, mezzanine financing, interest rates, funding timing, and draw structure.

Operating – Assumptions related to revenue, recovery income, ramp-up schedules, OpEx, and CapEx.

Reversion (Sale) – Exit cap rate, selling costs, and valuation logic based on stabilized NOI.

Returns – Output section showing both levered and unlevered return metrics, including IRR, equity multiple, and yield-on-cost.

To keep things consistent and user-friendly, all inputs are color-coded in blue font, while key outputs appear in black. This allows you to easily distinguish editable fields from calculated results — a convention we follow across all A.CRE models.

S- Curve Reference Tab (Default View)

Some basic backend settings related to the s-curve development cash flow, specifically the forecasting module, are housed in this tab.

---

Download the Data Center Development Model

To make this model accessible to everyone, it is offered on a “Pay What You’re Able” basis with no minimum (enter $0 if you’d like) or maximum (your support helps keep the content coming – typical real estate development models sell for $100 – $300+ per license). Just enter a price together with an email address to send the download link to, and then click ‘Continue’. If you have any questions about our “Pay What You’re Able” program or why we offer our models on this basis, you can contact our support team here.

We regularly update the model. Paid contributors to the model receive lifetime updates and unlimited downloads.

---