MBA NewsLink Roundtable: Discussing Data Centers, Demand and Deal Structures With KBRA

(Data Center Stock photo courtesy Christina Morillo via pexels)

The rapid expansion of artificial intelligence is driving unprecedented demand for the physical infrastructure that supports it. Most notably: data centers. Recent industry research estimates that meeting AI-driven demand could require trillions of dollars in global capital investment over the coming years, underscoring how data centers are increasingly positioned at the intersection of commercial real estate, infrastructure, and capital markets. As development pipelines accelerate, questions around scale, power and financing structure have moved to the forefront of market discussions.

Against this backdrop, MBA NewsLink convened a roundtable discussion with analysts from KBRA’s ABS, CMBS, and Project Finance teams to examine how data center issuance has evolved across these rated sectors in recent years. Drawing on transaction experience and sector-specific credit perspectives, the discussion explores how underwriting considerations, risk profiles, and investor expectations differ—depending on how and where data center assets are financed.

MBA NewsLink: From a market-wide perspective, what has driven the increase in data center issuance in recent years, and how has the asset class matured from a credit standpoint?

Fred Perreten, Managing Director, KBRA: At a high level, the growth in issuance reflects both strong underlying demand and greater capital markets participation. Issuance averaged $4 billion annually from 2018 to 2022, rose to roughly $10 billion in 2023 and 2024, and then jumped to $27 billion in 2025. This growth has largely been driven by newly constructed and recently stabilized assets that were built to support hyperscaler ai demand, along with more issuers tapping the capital markets for more attractive financing. At the core of it all is sustained growth of roughly 20% per year in global data creation.

From a credit standpoint, the asset class has matured since we completed our first retail colocation rating in 2021, when operating histories were still relatively limited. Since then, the five master trusts we rate have upsized several times, either through asset additions or increased cash flow supporting additional debt issuance, giving us repeated visibility into portfolio performance. Market data has also improved, particularly around rents and occupancy. We are also seeing more refined lease structures, especially for mega campuses, with stronger termination protections and more capital cost reimbursement borne by tenants, which better aligns risk.

That said, the asset class is still developing in certain respects. There is limited data on hyperscale lease renewals and no workout history to assess recoveries under stress.

MBA NewsLink: What characteristics tend to differentiate this asset class from more traditional commercial real estate collateral?

Fred Perreten: A key distinction is that data centers operate more like infrastructure assets than traditional real estate. Their value is driven less by physical space and more by power, redundancy, and fiber connectivity.

First, access to power is central. A data center is best described as a power contract with a building attached. Rent is quoted in dollars per kW per month rather than per square foot, and capacity constraints are measured by available power, not physical occupancy. Market participants now track vacancy based on power capacity rather than square footage.

Second, redundancy is critical. Facilities are built with on-site backup generation to ensure uninterrupted service if the utility grid fails. Given the operational and, in some cases, life-safety implications of downtime, resilience is essential. Backup systems are capital-intensive, and MEP costs represent a significantly larger share of total construction cost than in traditional CRE.

Finally, fiber connectivity differentiates the asset class. Location is evaluated in terms of latency [and proximity to major fiber routes], reflecting the need for fast, reliable data transmission rather than traditional real estate considerations.

MBA NewsLink: How do data centers fit within CMBS structures, and what characteristics tend to differentiate these transactions from transactions backed by more traditional CRE collateral?

Perreten: CMBS has proven to be a strong fit for data center financing. In 2025, CMBS accounted for 42% of structured finance issuance, compared to 25% from 2018-24. The credit analysis is broadly similar to other CRE asset classes, focusing on legal structure, sponsorship strength and operating expertise, tenant credit, lease term and renewal prospects, and in-place rents and vacancy relative to the market.

In the hyperscale segment, which comprises most CMBS transactions, properties are typically leased to one or a small number of tenants on long-term leases, often with investment grade credit characteristics. As a result, we tend to focus on the likelihood of renewal, how critical the asset is within the tenant’s broader data center network, and the level of tenant capital invested into the facility. Given that term default risk is often mitigated by tenant credit quality, maturity risk and takeout financing receive heightened scrutiny.

Structurally, these transactions resemble traditional CMBS SASB deals, including interest–only features, defined prepayment terms, tax and insurance reserves during trigger events, springing cash management, release provisions, and prohibitions on additional debt. Leverage at a given rating category tends to be higher than in traditional CMBS, reflecting both substantial tenant investment and the infrastructure-like characteristics of the assets, which may support renewal prospects.

MBA NewsLink: ABS structures have increasingly been used to finance digital infrastructure. What makes data centers a fit for ABS, and what are the key credit considerations?

Alan Greenblatt, Managing Director, KBRA: Data centers can be a fit for ABS when the collateral generates either long-dated, contracted revenues from creditworthy tenants or granular contracted cash flows across a diversified customer base. This can include single-tenant hyperscale assets or diversified colocation portfolios, particularly when sector demand tailwinds are strong enough to support refinance and extension outcomes.

From our perspective, the starting point is the durability of cash flow. In some transactions, revenues may be tied to a limited number of large tenants under longer-term arrangements, which can introduce event and renewal risk even where credit profiles are strong. In others, such as colocation platforms, cash flow may be generated from a broader, more diversified customer base with shorter contractual commitments. In either case, tenant concentration, lease profile, and renewal dynamics tend to shape the credit discussion.

Beyond that, market and asset fundamentals also matter—particularly access to power and supply conditions in key markets—as those factors can influence retention and re-leasing prospects over time.

Finally, transaction structure plays an important role in how asset performance ultimately affects bondholders. That includes leverage levels, reliance on appraisals, the design of cash flow triggers and sweep mechanics. For example, unlike CMBS, which are structured as REMIC transactions backed by a first-lien mortgage, ABS data center transactions use a master trust structure in which the issuer holds direct ownership interests and may issue additional debt over time, resulting in cross-collateralization and potentially higher leverage that can amplify the impact of asset performance on noteholders.

MBA NewsLink: Drawing on recent ABS and CMBS data center transactions rated by KBRA, what themes or structural features stand out in today’s market?

Alan Greenblatt: In ABS, where we have rated several colocation platforms, a few themes stand out. We are seeing larger, scaled master trust platforms with repeat issuance and incremental collateral additions, reflecting a more programmatic market. Structures are generally consistent, featuring DSCR-based cash traps, early amortization triggers, six-month senior interest reserves, and LTV sweep tests that force deleveraging if performance weakens. Some deals have also incorporated PIK features in subordinate classes.

Collateral composition continues to balance fee-owned and leasehold assets. Short weighted average remaining contract terms, typically in the two-to-three-year range for colocation portfolios, remain a defining characteristic, elevating rollover risk and reinforcing the importance of structural protections.

Fred Perreten: On the CMBS side, where hyperscale assets are more common, tenant termination options and the associated termination fees are a central focus. Terminate rights typically arise either from landlord performance breaches or for convenience. On the performance threshold side, leases generally permit several short duration outages or one or two longer duration outages; exceeding those limits can result in penalty payments to the tenant and, if sufficiently severe, might trigger terminate rights. For convenience terminations, we have seen fees that can represent up to 100% of all remaining lease payments.

MBA NewsLink: Project Finance has played a growing role in large-scale data center developments. What distinguishes these transactions from real estate–oriented financings?

Adeeti Amin, Managing Director, KBRA: Project finance for data centers is centered on cash flow durability rather than asset value. The analysis is anchored in the project’s ability to generate stable, contracted cash flows sufficient to service debt under a range of operating conditions. In contrast, real estate-oriented financings such as CMBS also consider cash flow, but asset value and loan-to-value ratios tend to play a more prominent role in the overall framework.

Data center project financings typically rely on long-term leases with creditworthy counterparties, often investment-grade hyperscalers, and strength of the tenant is central to the overall credit profile of the transaction. These structures are generally static and include traditional project finance features such as cash flow waterfalls, debt service reserve accounts, distribution tests, and covenant packages. Renewal dynamics also receive heightened scrutiny, with consideration given to in-place rents, market positioning, power access, and tenant capital investment—factors that are particularly important for project finance transactions.

Refinancing considerations focus on the sustainability of contracted cash flow over time, often supported by meaningful amortization through the lease term or expected extensions. In this way, data center project financings resemble infrastructure transactions more than traditional property financings, with repayment capacity driven primarily by contracted revenue and structural protections rather than by asset value.

MBA NewsLink: Power availability and long-term energy arrangements are central to the data center thesis. How do these factors affect the credit profile of project finance data centers?

Adeeti Amin: Power availability has become one of the most critical credit considerations in data center project financings. The key questions are whether sufficient megawatts are fully contracted, deliverable on schedule, and supported by executed electric service agreements, since delays in interconnection or substation development can directly affect rent commencement and overall project performance.

Reliability and redundancy are equally critical, including backup generation and N+1 or N+2 configurations that support continuous operations. Long-term energy arrangements are also important in the context of lease tenors. When power contracts extend beyond the initial lease term, they may enhance renewal practicality; conversely, in markets facing transmission or generation constraints, uncertainty around incremental capacity may dampen renewal prospects.

In markets with limited new supply, facilities with secured and energized capacity often benefit from lower relocation risk and stronger renewal prospects. As a result, power certainty has become an increasingly important differentiator and a key factor in assessing long-term stability and refinancing prospects.

MBA NewsLink: Based on recent project finance and structured finance transactions, how are sponsors and lenders addressing construction risk, scale, and long-term operating assumptions?

Adeeti Amin: In project finance transactions, construction risk mitigation centers on contractor credit quality, guaranteed maximum price (GMP) contracts, and liquidity buffers such as contingencies, retainage, delayed equity funding, and sponsor liquidity. The headroom between the substantial completion and ready-for-service (RFS) dates further mitigates rent commencement risk.

Many hyperscale developments now use modular powered shell designs, which simplify execution relative to more bespoke builds. Phased campus buildouts and staggered stabilization dates also allow lenders to monitor execution risk and funding cadence more closely.

At this time, there simply isn’t enough capacity to fulfill hyperscaler’s needs. Given the scale of recent developments—often exceeding 100MW per building and hundreds of MW per campus—transactions are structured around tenant credit strength, lease duration, and committed capital investment.

Sponsors are anchoring long-term operating assumptions in lease form, power security, and DSCR resilience. The unifying theme remains consistent: project finance structures are designed to demonstrate durable, contracted cash flows capable of servicing debt through construction, stabilization, renewal, and refinancing phases.

Fred Perreten: In structured finance, construction risk is generally addressed before securitization. In terms of scale, deal sizes have increased meaningfully over the past 18 months, with a CMBS deal reaching $4 billion and many in the $1-3 billion range. ABS platforms often grow over time through repeat issuance rather than issuing all debt upfront.

For CMBS, value and leverage remain central considerations, with attention to sustainable cash flow after factoring in operating expenses, leasing costs, and capital needs. Data center capital expenditures differ from traditional commercial real estate (CRE), often expressed on a power-capacity basis rather than per square foot and materially higher in absolute terms.

ABS structures are more dynamic, typically featuring interest-only periods followed by cash sweeps that accelerate amortization. Performance sensitivity to tenant rollover, rent adjustments, downtime, and operating costs is an important consideration, and some transactions embed structural protections that address these risks directly.

MBA NewsLink: Looking ahead, what should lenders, investors, and sponsors be watching most closely as data center development and financing activity continues to expand?

Fred Perreten: As with any CRE asset, demand for capacity and tenant credit quality is of paramount importance, particularly given the high tenant concentration in hyperscale transactions. We are paying close attention to lease renewal trends, given the limited historical rollover in this relatively young asset class. While near-term demand remains robust, an important watchpoint is whether hyperscalers begin to consolidate capacity to their mega campuses, which would reduce demand for capacity from third party providers.

(Views expressed in this article do not necessarily reflect policies of the Mortgage Bankers Association, nor do they connote an MBA endorsement of a specific company, product or service. MBA NewsLink welcomes submissions from member firms. Inquiries can be sent to Editor Michael Tucker or Editorial Manager Anneliese Mahoney.)