An Introduction to US Data Center Financing Structures for AI Infrastructure Development

Financing the AI build-out has forced hyperscalers to make a strategic pivot. Faced with a wave of planned projects and escalating construction costs, hyperscalers are increasingly supplementing historically self-funded expansion with debt financing. That shift has expanded the range of available structures and increased reliance on non-traditional credit providers, including private credit funds.

The financing analysis changes as a data center moves from development to stabilization. During construction, lenders focus on project execution, anchor-tenant credit, projected lease revenues, sponsor experience, access to reliable power and utility infrastructure, and local and state permitting complexity. Once an asset is operational and generates predictable lease income, sponsors typically seek permanent capital with longer tenor and more flexible terms.

This alert addresses the principal financing structures being used across the data center lifecycle, with particular attention to how each structure allocates credit, collateral, construction, tenant, power, and residual-value risk.

Construction and Development Financing

Project-Level Construction Debt 

The central design of data center capital formation is limited recourse construction financing at the individual project level. These facilities are typically shorter-term in nature, funding the build-out period through the onset of revenue-generating operations. Unlike conventional commercial real estate lending, which focuses on the appraised value of real property collateral, data center lenders conduct a layered credit assessment that includes anchor-tenant credit, projected lease streams, developer execution history, power and utility proximity, and permitting complexity.

Holding Company Financing 

Developers overseeing multiple projects may establish a holding company (HoldCo) credit facility in addition to individual project-level loans. The borrower is a parent entity above ring-fenced project subsidiaries, and the HoldCo lender typically takes collateral in the parent’s equity ownership in the project-level entities rather than a direct lien on the data center assets. Developers with diversified portfolios may cross-collateralize multiple projects under a single HoldCo facility to improve blended credit metrics and reduce aggregate borrowing costs relative to standalone financing. Debt service at the HoldCo level is funded through cash distributions upstreamed from operating subsidiaries. Typical lender protections include centralized cash management and reserve requirements, approval rights over material amendments to key project agreements, excess cash flow sweeps, restrictions on upstream distributions and leakage, periodic reporting and information covenants, and financial maintenance tests.

Private Credit Direct Lending and Club Facilities 

Private credit has become a cornerstone of large-scale data center development, particularly where borrowers need execution certainty, bespoke covenants, or leverage unavailable in the syndicated bank market. Direct lenders or private credit funds may provide senior secured or unitranche debt at the project, portfolio, or special purpose vehicle (SPV) level, underwriting principally to anchor-tenant credit, contracted lease revenue, secured power, construction milestones, and sponsor support.

Corporate Bonds With Strategic Credit Backstops 

Some data centers are financed through high-yield or investment-grade bonds issued at the developer or platform level and supported by strategic customer arrangements. A Google, Fluidstack, and TeraWulf transaction illustrates this approach: TeraWulf issued high-yield bonds to build a data center, Fluidstack agreed to lease it for graphics processing unit clusters, and Google provided a conditional backstop of Fluidstack’s lease obligations in exchange for equity warrants.

Takeout and Permanent Capital

Operational Data Center Refinancing 

Once construction is complete and the facility generates predictable lease income, the project is stabilized and its risk profile improves. A well-timed refinancing can lower the weighted average cost of debt, unlock equity value, and position the asset for a sale or recapitalization. Long-term providers underwrite chiefly on tenant credit quality and the durability of contracted cash flows, and commonly execute through syndicated term loans, institutional private placements, securitizations (asset-backed securitization (ABS) or CMBS), and other long-duration fixed income.

Data Center ABS and Master Trust Structures 

ABS has become a leading tool for monetizing stabilized cash flows and building repeatable financing capacity. In a typical data center ABS, the sponsor contributes fee or leasehold interests in operating facilities, related leases, and cash flows into a bankruptcy-remote structure, often a master trust that can issue multiple series of debt over time and add qualifying collateral as the platform grows.

Single-Asset and Single-Borrower CMBS 

Commercial mortgage-backed securities, particularly single-asset and single-borrower CMBS, offer another takeout path for stabilized facilities or portfolios. Unlike ABS, which often involves a broader pledge of operating-company equity, cash flows, mortgages, and leases, CMBS is generally secured by mortgage loans on identified real estate and related lease assignments. Single-borrower CMBS works best for stable, predictable assets with associated real estate and lease cash flows.

Sale-Leaseback Arrangements 

Sale-leasebacks allow operators and hyperscalers to monetize owned real estate or infrastructure while retaining operational use. The owner sells the facility, or a portion of the campus, to a financial investor, real estate investment trust, infrastructure fund, or SPV and simultaneously enters a long-term lease to continue operating the asset. These arrangements are increasingly relevant given the substantial real estate component of data centers and the pressure — felt even by the largest hyperscalers — to relieve balance-sheet strain.

The core financing analysis turns on matching capital structure to the asset’s stage of development and the risk profile of the underlying cash flows. Construction-stage financing underwrites execution risk, power access, permitting complexity, anchor-tenant strength, sponsor support, and construction milestones. Stabilized-asset financing underwrites tenant credit quality, lease tenor, renewal risk, operating performance, collateral durability, residual value, and the ability of the structure to continue producing cash available for debt service.

Private credit offers speed, confidentiality, structural flexibility, and a willingness to absorb construction- and ramp-stage risk. Those same attributes create investor-facing considerations because private credit terms are confidential and largely exempt from US Securities and Exchange Commission reporting, requiring institutional and retail investors to gauge tolerance for exposure to ventures in which the scale and recourse of private lenders remain undisclosed.

For bond financings supported by customer arrangements, investors look beyond the developer’s standalone credit to the economic substance of the tenant lease and any customer commitment, termination fee, guarantee, or backstop. The analysis turns on whether the backstop is unconditional or contingent, whether it survives construction delays or tenant default, whether the bonds are structurally subordinated to project-level creditors, and whether the deal creates securities-disclosure, rating, or litigation risk if the economic support proves less durable than investors expected.

ABS offers structural advantages, including access to long-duration institutional capital, tranche-based pricing, asset-level ring-fencing in a bankruptcy-remote vehicle, repeat-issuance capacity through a master trust, and the ability to finance a growing pool of facilities under a single platform. The risks are also structural: repayment depends on continued lease income, and diligence typically centers on servicer and operator performance, eligibility criteria for additional collateral, debt-service-coverage triggers, cash traps, early-amortization events, reserve sizing, technology obsolescence, power availability, and sponsor operational flexibility after closing. Because leverage often runs at roughly 65-70% loan-to-value, the cushion is thin; these notes are typically floated as private offerings in either the 144A or 4(a)(2) markets.

CMBS offers a well-developed investor base, more standardized documentation than ABS, a direct link to real estate collateral, and suitability for large hyperscale facilities leased to investment-grade tenants. The trade-offs flow from that real estate orientation: CMBS offers less flexibility to add or substitute assets than a master-trust ABS, and its value is more sensitive to appraised value and cap-rate assumptions than to enterprise-level cash flows. Lenders must also underwrite tenant termination rights, mortgage-level reserve and escrow requirements, leasehold or ground-lease complications, environmental and engineering diligence, and the potential need to upgrade or replace mechanical, electrical, and plumbing systems as AI workloads and cooling demands evolve.

Sale-leaseback analysis centers on the seller-lessee’s lease economics and accounting. The parties must confirm lease classification for accounting and tax treatment because that classification determines whether the transaction delivers the intended balance-sheet relief. Rent escalators, purchase options, rights of first refusal, renewal rights, restoration obligations, tenant-improvement ownership, end-of-term conditions, residual value risk, and downstream financing capacity all affect the purchaser-lessor’s return and lender analysis.

Data center financing is no longer limited to conventional construction debt or balance-sheet funding. As AI-related infrastructure demand increases, sponsors, lenders, and investors may use a wider menu of project-level, platform-level, securitized, mortgage-backed, sale-leaseback, and hybrid structures. Each structure can expand access to capital, but each also creates distinct questions about collateral, recourse, lease durability, power availability, technology obsolescence, construction timing, accounting treatment, disclosure, and residual value.

The “Beignet” financing of Meta’s Hyperion campus illustrates the direction of travel. The approximately $30 billion transaction, which closed in October 2025 to fund Meta’s Hyperion campus in Richland Parish, Louisiana, paired Meta with Blue Owl Capital in a joint venture SPV named Beignet Investor LLC, with Blue Owl-managed funds holding 80% and Meta retaining 20%. Morgan Stanley assembled more than $27 billion of debt, anchored by PIMCO with participation from other institutions including BlackRock, alongside roughly $2.5 billion of equity. The notes were issued in 144A format, carried an investment-grade A+ rating from S&P, mature in 2049, are fully amortizing, and priced at approximately 225 basis points over then-current US treasuries.

Several features make the Beignet structure a template for the industry.

• Hybrid Financing: The structure fuses project finance with long-term fixed income in a single package, spanning both the construction phase and roughly 20 years of post-stabilization operations.
• Balance-Sheet Efficiency: Because the SPV, not Meta, owns the campus and leases it back to Meta, Meta accesses capital without recording the associated debt as a liability on its balance sheet; it reports only its equity stake and corresponding lease obligations.
• Residual Value Guarantee: Meta provided investors a residual value guarantee to mitigate lease renewal and termination risk. If Meta elects not to renew or terminates a lease and certain conditions are met, its maximum payment equals the shortfall between the property’s fair value at the time and the residual value guarantee threshold.

On July 13, Meta announced that it expects the Hyperion campus will reach five gigawatts of compute capacity and is now estimated to cost more than $50 billion, nearly double its original estimate. Furthermore, on July 20, Meta announced a new partnership with BlackRock to build a $12 billion data center in El Paso, Texas. The partnership is expected to be structured similarly to the Hyperion project, with BlackRock owning an 80% stake and Meta owning the remaining 20%.

The data center asset class offers a distinctive opportunity for credit providers, but the sector’s speed and technical complexity demand a specialized analytical framework. Private markets are expected to play an ever-larger role in funding this build-out. Lenders and investors who develop sector-specific expertise, insist on thoughtful documentation, and maintain rigorous diligence — particularly around collateral quality, the risks posed by rapid innovation, and power availability — will be best positioned to participate in this dynamic and fast-evolving industry.