Data center financing is not commercial real estate financing with servers in the building. It is infrastructure financing with real estate as one component. Lenders who understand this asset class underwrite power capacity in megawatts, water rights and cooling capacity, fiber connectivity density, and tenant credit quality — not just the four walls and a roof. The building is the least interesting part of the capital stack. This guide breaks down how capital actually flows into hyperscale, colocation, and edge data center projects in 2026, what lenders look at, and where the financing market is headed as AI workloads push power and cooling demands to levels the industry has never seen.
1Why Data Centers Are Infrastructure, Not Real Estate
Walk into a typical commercial real estate lending meeting with a data center deal, and you will quickly realize the disconnect. Traditional CRE lenders evaluate location, comparable sales, replacement cost, and cap rates. Data center lenders evaluate power capacity, cooling efficiency, network connectivity, and the creditworthiness of the tenants who will occupy the facility for 10-20 years. These are fundamentally different underwriting exercises.
The shift happened gradually over the past decade, but it accelerated in 2024-2025 as AI training and inference workloads drove power density requirements from 5-10 kW per rack to 40-100+ kW per rack. At those densities, the mechanical and electrical infrastructure — power distribution, cooling systems, backup generation — represents 60-70% of total project cost. The shell building is 15-20%. The land is 5-10%. Lenders figured this out. The underwriting followed.
The Core Underwriting Shift
Traditional CRE lenders ask: "What is the building worth?" Data center infrastructure lenders ask: "What is the power capacity worth?" A 100,000 square foot shell with 50 MW of committed power is a fundamentally different asset than the same shell with 5 MW. The power makes the deal.
Here are the primary metrics that data center lenders now use to size loans and assess risk:
| Metric | Definition | Lender Target |
|---|---|---|
| PUE (Power Usage Effectiveness) | Total facility power / IT equipment power | 1.2-1.4 (good), below 1.2 (excellent) |
| WUE (Water Usage Effectiveness) | Annual water usage (liters) / IT equipment energy (kWh) | Below 1.0 L/kWh preferred |
| MW Capacity | Total IT load capacity in megawatts | Committed utility power with interconnection agreement |
| Tier Rating | Uptime Institute classification (Tier I-IV) | Tier III minimum for institutional capital |
| N+1 Redundancy | Backup capacity for critical systems | N+1 minimum for power and cooling |
| Fiber Connectivity | Number of carrier-neutral fiber paths | Minimum 2 diverse fiber paths; 4+ preferred |
PUE and WUE deserve special attention. PUE, defined by the Uptime Institute, measures how efficiently a data center uses energy. A PUE of 2.0 means for every watt delivered to IT equipment, another watt is consumed by cooling, lighting, and other overhead. Industry average PUE in 2025 was approximately 1.58 according to the Uptime Institute's annual survey. Best-in-class hyperscale facilities operate at 1.1-1.2. Lenders increasingly require PUE projections as part of the underwriting package because PUE directly impacts operating margins and, therefore, debt service coverage.
WUE (Water Usage Effectiveness), introduced by The Green Grid in 2011, measures how efficiently a data center uses water. The formula is simple: Annual Water Usage (liters) ÷ IT Equipment Energy (kWh). Lower is better — a WUE of 0 means the facility uses no water for cooling at all.
WUE Benchmarks (2024-2026)
- Industry average: ~1.8 L/kWh — most legacy facilities with evaporative cooling towers
- Hyperscale leaders: 0.15-0.24 L/kWh — achieved through advanced air cooling and closed-loop systems
- Lender expectation: Below 1.0 L/kWh — increasingly a baseline for new financing
- Gold standard: 0 L/kWh — zero-water cooling (air-cooled, immersion, or chip-level liquid cooling in closed loop)
- AI workload trend: High-density AI racks are projected to push average WUE to 0.45-0.48 L/kWh by 2025-2026 as liquid cooling becomes standard for GPU clusters
The critical trade-off: evaporative cooling reduces energy use (better PUE) but consumes massive amounts of water (worse WUE). A large data center can use 1-5 million gallons of water per day through cooling towers. In water-stressed regions — Phoenix, Las Vegas, Northern Texas — this creates both regulatory risk and financing risk. Lenders now view a facility with WUE of 1.8 L/kWh in a drought-prone market as a fundamentally different credit than one achieving 0.3 L/kWh through air-cooled or liquid-cooled systems.
Common strategies to lower WUE include: water-free cooling (air-cooled chillers or dry coolers — higher capex but eliminates water dependency), liquid immersion cooling (servers submerged in non-conductive dielectric fluid — eliminates evaporative cooling entirely), and sustainable water sourcing (reclaimed greywater or rainwater harvesting to reduce strain on potable water supplies).
2Three Types of Data Centers, Three Capital Structures
Not all data centers are created equal, and the financing options diverge dramatically based on scale, tenant profile, and use case. Here is how capital structures differ across the three primary data center types.
Hyperscale: $100M - $1B+ Per Facility
Hyperscale data centers are purpose-built for a single tenant — typically one of the major cloud providers (AWS, Microsoft Azure, Google Cloud), a large AI company, or a major enterprise. These facilities range from 20 MW to 200+ MW of IT capacity, with campuses sometimes reaching 1 GW across multiple buildings.
The financing model for hyperscale is closest to project finance or infrastructure finance. The credit story is simple: a single investment-grade tenant on a 15-20 year lease, with the lease structured as a triple-net (NNN) obligation. Lenders are underwriting the tenant's credit rating, not the building. A 100 MW hyperscale facility leased to Microsoft on a 20-year NNN lease is, from a lender's perspective, essentially a Microsoft corporate bond with a building attached.
At this scale, the developer typically has a pre-existing relationship with the hyperscale tenant. The tenant may provide a letter of intent (LOI) or an executed lease before the developer seeks construction financing. Many hyperscale developers are publicly traded REITs with access to corporate debt and equity markets in addition to project-level financing — which is why this segment has seen $10B+ individual capital raises in 2025 alone.
Colocation: $10M - $200M Per Facility
Colocation facilities serve multiple tenants, ranging from enterprise customers leasing full cabinets or cages to smaller businesses renting individual rack space. Power capacity typically ranges from 1 MW to 30 MW. The multi-tenant model creates both diversification and complexity.
The financing challenge with colocation is absorption risk. Unlike hyperscale, where you build for a specific tenant, colocation facilities must attract and retain multiple customers. Lenders want to see pre-leasing — typically 40-60% or more of planned capacity — before funding construction. The tenant credit profile is mixed: some enterprise clients with strong credit, some smaller companies with less financial transparency.
Colocation operators with established brands and existing customer relationships have a significant advantage in financing. A first-time developer building a speculative colocation facility in a secondary market will face much higher equity requirements (40%+) and may need bridge financing to get through the lease-up period before qualifying for permanent debt.
Edge: $2M - $20M Per Facility
Edge data centers are the fastest-growing segment of the market, driven by 5G deployment, IoT, autonomous vehicles, and the need to reduce latency for real-time applications. These are smaller facilities — often 500 kW to 5 MW — located closer to end users in secondary and tertiary markets. Some are purpose-built micro data centers; others are retrofitted commercial or industrial spaces.
Financing edge is different because the capital stack is equipment-heavy. The building (or modular unit) may represent only 20-30% of total project cost, with the remainder going to power distribution, cooling, IT hardware, and networking equipment. This creates an opportunity to layer multiple financing products.
The modular data center trend has been a game-changer for edge financing. Prefabricated, containerized data centers from manufacturers like Schneider Electric, Vertiv, and Compass can be deployed in 12-16 weeks and financed as equipment rather than real estate — simplifying the underwriting process and reducing construction risk to near zero.
| Factor | Hyperscale | Colocation | Edge |
|---|---|---|---|
| Project Size | $100M - $1B+ | $10M - $200M | $2M - $20M |
| Power Capacity | 20 - 200+ MW | 1 - 30 MW | 0.5 - 5 MW |
| Tenant Model | Single (investment-grade) | Multi-tenant | Owner-occupied or 1-5 tenants |
| Lease Terms | 15-20 years NNN | 3-7 years | 1-5 years or owner-operated |
| Equity Requirement | 20-35% | 25-40% | 10-30% |
| Primary Lender Type | Infrastructure PE / CMBS | Banks / Private credit | SBA / Equipment lenders |
3Capital Structure by Project Phase
Data center projects move through distinct phases, and the optimal capital source changes at each stage. Understanding this lifecycle is critical because using the wrong capital at the wrong time either costs too much (using equity where debt is available) or creates execution risk (using debt where equity should absorb the uncertainty).
Land Acquisition and Entitlements
This is the highest-risk phase and almost always requires equity or bridge capital. You are purchasing land, securing zoning approvals, environmental clearances, and — critically — applying for power interconnection with the local utility. The interconnection process alone can take 12-24 months in high-demand markets and represents the single biggest entitlement risk for data center development.
Capital source: Sponsor equity, joint venture equity, or short-term bridge loans (if the developer has existing assets as collateral). Lenders will not provide construction financing until power interconnection is committed and major entitlements are secured.
Typical cost: 15-25% of total project budget consumed in pre-development.
Shell Construction
Once entitlements and power commitments are secured, the project becomes financeable with construction debt. The shell — the physical building, site infrastructure, parking, and basic utilities — is the most straightforward portion to finance because it closely resembles traditional CRE construction lending.
Capital source: Construction loan, typically 12-24 months with extension options. Interest reserve is standard. For hyperscale projects with executed leases, construction loans can reach 65-70% of shell cost. For speculative colocation, expect 50-60% LTV with recourse to the sponsor.
Typical rate (2026): SOFR + 250-400 bps depending on sponsor, tenant commitments, and market.
Tenant Improvement and Fit-Out
This is where data center financing diverges most sharply from traditional CRE. The fit-out — power distribution (switchgear, PDUs, UPS), cooling systems (CRAH/CRAC units, chillers, cooling towers), backup generation, fire suppression, and security systems — can represent 40-60% of total project cost. A 10 MW facility might spend $80-150M on the shell but $120-200M on the mechanical and electrical infrastructure.
Capital source: Mezzanine debt, equipment financing, or private credit. Some developers finance the fit-out as part of the construction loan; others layer in equipment-specific financing to reduce overall cost of capital. Equipment lenders will finance generators, UPS systems, and cooling equipment on 5-10 year terms secured by the equipment itself.
Key consideration: Equipment financing rates are often lower than mezzanine because the lender has a hard asset to repossess. Smart operators finance the long-lived mechanical infrastructure (generators, chillers — 15-20 year useful life) separately from IT equipment (servers, switches — 3-5 year useful life).
Stabilization and Permanent Financing
Once the facility reaches 85%+ occupancy (or is fully occupied by a single hyperscale tenant), construction debt can be replaced with permanent financing at significantly lower rates and longer terms. This is where the infrastructure nature of data centers works in the borrower's favor — institutional lenders love the predictability of long-term data center leases.
Capital source: CMBS conduit (for larger, stabilized deals), insurance company portfolio lending (for long-term fixed rates), or bank permanent loans. Single-tenant hyperscale facilities with investment-grade tenants can access the lowest rates — comparable to corporate bond yields plus a spread.
Typical terms (2026): 55-70% LTV, 10-25 year terms, fixed rates from mid-5% to low-7% depending on tenant credit, facility quality, and market.
Expansion and Upgrades
Existing data centers frequently need to expand capacity or upgrade cooling systems to support higher-density workloads. This is where C-PACE (Commercial Property Assessed Clean Energy) financing has emerged as a powerful tool. C-PACE can finance energy efficiency improvements, cooling system upgrades, and renewable energy installations with 20-30 year terms, fixed rates, and no personal guarantee — repaid through a property tax assessment.
Capital source: C-PACE for qualifying energy and water efficiency improvements. Supplemental construction loans for building expansions. Equipment financing for new IT infrastructure. C-PACE is particularly attractive for cooling system retrofits (converting from air-cooled to liquid-cooled or immersion cooling) because the energy savings directly offset the assessment payments.
C-PACE advantage: Non-recourse, fixed-rate, 20-30 year amortization. Can be layered behind existing senior debt with lender consent. Typical C-PACE sizing for data center cooling upgrades: $5-50M+.
4Water Rights: The New Underwriting Factor
Water is the underwriting factor that most data center developers underestimate — and it is becoming a deal-killer in drought-prone regions. A large data center campus can consume 1-5 million gallons of water per day for evaporative cooling. To put that in perspective, a 50 MW hyperscale facility using traditional cooling towers may consume as much water annually as a 10,000-acre almond farm in California. Lenders in Arizona, Texas, Nevada, and other water-constrained states now require detailed water rights documentation as part of the loan package.
Water Risk Is Financial Risk
In January 2025, Chandler, Arizona imposed a moratorium on new data center development due to water supply concerns. Mesa and Goodyear followed with additional restrictions. In Texas, several municipal water districts have implemented tiered water pricing that increases the effective cost of cooling by 200-400% for high-consumption users. Lenders underwriting new data center construction in these markets now require either (a) documented water rights sufficient for the facility's projected 20-year consumption, or (b) a cooling technology that dramatically reduces water dependency.
The cooling technology choice directly affects both capital expenditure and lender appetite. Here is how the three primary cooling approaches compare from a financing perspective:
| Cooling Technology | Water Usage | Capex Impact | Lender Preference |
|---|---|---|---|
| Evaporative (Cooling Towers) | High: 1.5-2.0+ L/kWh WUE | Lowest capex | Acceptable in water-rich regions only |
| Air-Cooled (Dry Coolers) | Minimal: 0.1-0.3 L/kWh WUE | 15-25% higher capex | Preferred in drought-prone markets |
| Direct Liquid Cooling | Near-zero water consumption | 20-40% higher capex | Increasingly preferred for high-density AI workloads |
| Immersion Cooling | Zero water consumption | 30-50% higher capex | Emerging preference; limited track record for lender comfort |
The industry is clearly moving toward WUE below 1.0 L/kWh as a baseline lender expectation. Facilities that can demonstrate WUE below 0.5 L/kWh — achievable with air-cooled or liquid-cooled designs — have a measurable advantage in both the cost and availability of debt capital. Some institutional lenders now apply a "water risk premium" of 25-75 basis points to facilities in drought-prone regions that rely on evaporative cooling, directly impacting the project's returns.
Junior vs. Senior Water Rights: A Hidden Lending Risk
In the Western U.S., water is governed by the doctrine of "first in time, first in right." Data centers that hold junior water rights face higher interest rates or outright lending restrictions — because those rights can be curtailed during drought. Lenders increasingly require documentation of water right seniority as part of due diligence. (Source: JD Supra, Gravel2Gavel, Feb 2026)
Zero-Water Financing and the Green Premium
The link between water efficiency and capital access is no longer theoretical. A recent $2.1 billion ABS issuance required new-build data centers to use zero water for cooling to qualify for the funds (Source: Equinix reporting). Sustainability-linked loans now offer lower interest rates to operators that meet WUE targets — a measurable "green premium" in data center financing.
Community "social license" is also a financing factor. Public opposition to "mega-thirsty" data centers has led to permit denials in multiple markets. Lenders may now require developers to front local water infrastructure upgrades as a condition of financing — adding cost but reducing long-term regulatory and reputational risk.
How Water Risk Affects Your Financing Terms
For developers, the calculus is straightforward: the higher upfront capex of water-efficient cooling technologies is offset by (a) lower operating costs, (b) better financing terms, (c) reduced regulatory risk, and (d) broader access to capital. In water-constrained markets, air-cooled or liquid-cooled designs are not a premium choice — they are a financing prerequisite.
5Power Capacity and Interconnection: The Revenue Determinant
If water rights are the emerging underwriting factor, power capacity is the established one. Megawatt capacity directly determines revenue potential, and revenue potential determines loan sizing. The industry rule of thumb for underwriting is $8-15M per MW of IT capacity, which means a 50 MW facility represents $400-750M in potential enterprise value. Every lender in the data center space starts the underwriting conversation with one question: "How many megawatts, and are they committed?"
"Committed" is the operative word. Lenders distinguish sharply between (a) utility power that is available in the area, (b) power that has been requested via an interconnection application, and (c) power that is secured through an executed interconnection agreement with the utility. Only the third category counts for underwriting purposes. An interconnection agreement — sometimes called a power service agreement or electric service agreement — is the document that guarantees the utility will deliver a specific amount of power to the site by a specific date.
The Power Queue Problem
In Northern Virginia (the world's largest data center market), the power interconnection queue at Dominion Energy exceeded 30 GW of pending requests by late 2025. Average wait times for new utility connections stretched to 3-5 years. In parts of the Dallas-Fort Worth market, ERCOT interconnection timelines reached 24-36 months. This power bottleneck is the single biggest constraint on new data center development — and it is the single biggest risk factor lenders evaluate.
Developers who have already secured committed power capacity hold a scarce and valuable asset. Lenders recognize this, and committed power increasingly functions as de facto collateral — separate from and often more valuable than the real estate itself.
Renewable energy procurement adds another dimension to power underwriting. Many hyperscale tenants — particularly the major cloud providers — require 100% renewable energy matching as a lease condition. Developers who can demonstrate renewable energy procurement through power purchase agreements (PPAs), on-site generation, or renewable energy certificates (RECs) have an advantage in both tenant attraction and lender appetite. From a financing perspective, a PPA with a creditworthy renewable energy provider is an additional revenue-grade contract that lenders can underwrite.
6Who Lends on Data Centers in 2026
The data center lending landscape is more specialized than most borrowers expect. Traditional commercial banks rarely lead data center transactions — the asset class requires infrastructure expertise that most bank CRE teams do not have. Here is where the capital actually comes from, organized by deal type and phase.
Infrastructure Private Equity Funds
Players: KKR (Global Infrastructure Partners), Blackstone Infrastructure, Brookfield Infrastructure Partners, DigitalBridge, Stonepeak Infrastructure.
Sweet spot: $50M-$1B+ equity investments in development-stage and stabilized hyperscale and large colocation assets. These funds can provide the full capital stack — equity, preferred equity, and debt — or participate at a single level. They have deep data center operating expertise and can move quickly on large transactions.
What they want: Scale, committed power, creditworthy tenants (or clear path to them), experienced developer/operator. Minimum equity check is typically $25-50M.
CMBS Conduits
Sweet spot: Stabilized, single-tenant hyperscale facilities with investment-grade tenants on long-term leases. CMBS works well for these assets because the cash flow profile — predictable NNN rent from a rated credit — is exactly what CMBS bond investors want to see.
Typical terms: 55-70% LTV, 5-10 year terms with 25-30 year amortization, fixed rate. Non-recourse. CMBS lenders typically require 12+ months of stabilized operating history.
Limitation: CMBS does not work for construction or pre-stabilization. Multi-tenant colocation is harder to securitize due to tenant rollover risk.
Insurance Companies
Players: MetLife, Prudential, TIAA (Nuveen), New York Life, Principal — plus smaller life companies with infrastructure mandates.
Sweet spot: Long-term, fixed-rate permanent financing for stabilized data centers. Insurance companies love the asset class because the long lease terms (15-20 years) match their liability duration. They will provide 10-25 year fixed-rate loans, often at competitive rates due to the asset-liability match.
Typical terms: 50-65% LTV, 10-25 year fixed rates, 25-30 year amortization. They want Tier III or better, established operator, and ideally investment-grade anchor tenant.
Equipment Lenders and Lessors
Sweet spot: IT hardware (servers, storage, networking), UPS systems, generators, HVAC/cooling equipment. Equipment financing is often the most efficient way to fund the non-real-estate portions of a data center, particularly for edge and smaller colocation facilities.
Typical terms: 80-100% of equipment cost, 3-7 year terms (matched to equipment useful life), rates based on borrower credit and equipment type. Secured by the equipment — lender can repossess and resell.
Strategic use: Layer equipment financing behind real estate debt to reduce the equity required for the total project. A $15M edge data center might use $6M in SBA 504 (real estate), $5M in equipment financing (servers, cooling, power), and $4M in sponsor equity — achieving 73% total leverage.
C-PACE Providers
Sweet spot: Energy efficiency upgrades, cooling system replacements, on-site renewable generation, and new construction in states with C-PACE enabling legislation. C-PACE is uniquely valuable for data centers because so much of the capital spend qualifies — cooling systems, LED lighting, energy-efficient power distribution, and building envelope improvements.
Typical terms: 20-30 year fixed rate, non-recourse, no personal guarantee. Repaid through a property tax assessment. Requires lender consent for properties with existing senior debt.
Key advantage: C-PACE does not appear on the borrower's balance sheet as traditional debt. It is a tax assessment. This makes it attractive for operators who want to preserve borrowing capacity for other projects.
Private Credit Funds
Sweet spot: Bridge and mezzanine capital for data center development, particularly for projects that are too complex or too early-stage for traditional bank lending. Private credit fills the gap between sponsor equity and senior construction debt.
Typical terms: SOFR + 500-900 bps, 2-5 year terms, interest-only. Higher cost of capital, but faster execution and more flexible underwriting than traditional lenders.
When to use: Pre-stabilization colocation facilities that need capital to complete lease-up. Land banking for future development. Mezzanine to reduce equity in large construction deals.
Commercial Banks
Banks rarely lead data center financing, but they participate in syndications for larger deals and occasionally provide construction loans for well-sponsored colocation projects with strong pre-leasing. Regional banks in major data center markets (Northern Virginia, Dallas, Phoenix, Chicago) have developed some expertise, but their hold sizes ($20-50M) mean they typically participate rather than lead. Banks are most competitive on edge-scale deals where SBA programs or conventional commercial loans apply.
7How PeerSense Helps with Data Center Financing
Data center financing sits at the intersection of infrastructure finance, commercial real estate lending, equipment financing, and energy project finance. Most borrowers — even experienced CRE developers — do not have relationships across all of these capital sources. A colocation developer who knows the bank market may not have access to infrastructure PE. An edge operator familiar with SBA programs may not know that C-PACE can finance their cooling upgrade at a lower cost of capital.
PeerSense maps your data center project — regardless of type, phase, or size — to the right capital sources from our network of infrastructure lenders, equipment financiers, C-PACE providers, and private credit funds. We understand the metrics that matter (MW capacity, PUE, WUE, tenant credit, interconnection status) and present your deal in the language that data center lenders speak.
No retainers. No upfront fees. Referral fee is established at engagement and paid at closing.
Have a Data Center Deal?
Tell us about your project — type, MW capacity, location, and timeline. We will match you with the right capital sources.
Tell Us About Your DealThe Bottom Line
Data center financing in 2026 rewards developers who understand that they are building infrastructure, not just real estate. The capital markets have caught up to this reality — the most competitive financing goes to projects with committed power, efficient cooling, documented water rights, and creditworthy tenants. Whether you are developing a 200 MW hyperscale campus or a 2 MW edge facility, the fundamentals are the same: secure your power, solve your cooling, prove your demand, and match the right capital source to each phase of the project lifecycle. The developers who get this right are building the physical layer of the AI economy. The capital is there for projects that are structured correctly.