AI Data Centers Are Displacing Traditional Offices in the $4.5 Trillion CMBS Market

The commercial mortgage-backed securities (CMBS) market, the primary financing engine for America's office towers and shopping malls, is undergoing a structural realignment. Driven by the massive capital requirements of the artificial intelligence boom, data centers are rapidly displacing traditional real estate as a core component of institutional debt portfolios.[1]

This shift is fundamentally altering the risk profile of a $4.5 trillion securitized credit market. While traditional commercial real estate grapples with elevated vacancy rates and refinancing hurdles, capital is flowing aggressively toward digital infrastructure.[1]

The scale of the required investment is unprecedented. Morgan Stanley estimates that global investment in data centers—encompassing land, buildings, power infrastructure, and computing hardware—will demand approximately $2.9 trillion in capital expenditures through 2028.[3]

While large cloud providers like Amazon Web Services, Microsoft Azure, and Google Cloud are expected to fund roughly $1.4 trillion of that total through internal cash flows, an estimated $1.5 trillion will require external financing. Securitization markets are stepping in to bridge that gap.[3]

AI-generated

Data center operators rarely move directly into securitization. Development typically begins with construction loans, project finance, or private credit, where investors assume the risks of construction, equipment procurement, and power availability.[3]

Once a facility is complete, fully leased, and generating predictable cash flows, developers often refinance the asset to lock in lower-cost, long-term funding. This is where the ABS and CMBS markets come into play.[3]

Combined issuance of ABS and CMBS by data center operators surpassed $25 billion in 2025, exceeding the total of the previous three years combined. As of mid-2026, data center CMBS represents approximately 6% of the single-asset, single-borrower (SASB) market, up from negligible levels in 2020.[3][5]

Understanding the distinction between ABS and CMBS structures is critical for investors navigating this space. In an ABS structure, the securities are backed directly by the property interests and the corresponding lease cash flows generated by the data center customers.[4]

AI-generated

ABS deals often utilize a master trust structure, which allows the issuer to add new assets to the collateral pool over time as the portfolio grows. This makes ABS highly suitable for operators managing multiple facilities with diverse tenant rosters.[5]

Conversely, a data center CMBS transaction is typically a single-asset, single-borrower securitization backed by a traditional mortgage loan on the underlying real estate. The collateral pool is fixed at issuance, and the primary receivables are the principal and interest payments due under that specific mortgage.[4]

Credit analysis for a CMBS deal is therefore driven primarily by the underlying asset's value, the specific lease profile of the facility, and the enforceability of the mortgage security, rather than platform-level diversification. These structures are particularly well-suited for massive, stabilized hyperscale facilities leased to single, high-credit-quality tech tenants.[4]

This transition introduces a new set of risks that traditional real estate investors are unaccustomed to underwriting. The most prominent is technological obsolescence.[7]

A premium office building in Manhattan might maintain its core utility for half a century. A data center built today, however, could become obsolete within five to ten years if it cannot support the escalating power density and liquid cooling requirements of next-generation AI processors.[7]

This technological risk directly impacts the refinancing profile of CMBS deals. Unlike ABS structures, which often feature soft bullet maturities where the issuer is not strictly obligated to repay the principal on the scheduled date, CMBS deals are structured as fixed-maturity instruments.[4]

AI-generated

If the property value falls short at maturity because the facility's infrastructure is outdated, the borrower's ability to refinance the mortgage is severely compromised. Investors must evaluate whether the sponsor has the willingness and capital to recapitalize the asset if valuations decline.[4]

Furthermore, the AI leasing boom is highly concentrated. According to Trepp, AI companies accounted for 22.7% of office leasing in major tech markets in the first quarter of 2026, up from 15.3% a year earlier. This demand is heavily clustered in a small group of markets, led by San Francisco, leaving other regions exposed to broader commercial real estate weaknesses.[2]

In response to these dynamics, bond buyers are becoming more selective. While the broader CMBS market remains stable, spreads for certain data center-backed securities have widened as investors demand higher yields to compensate for potential regulatory hurdles and construction delays.[6]

"We're not talking about how to price in risk for the deal collapsing," said Sam Murphy of Bracewell, a law firm specializing in digital infrastructure. "We're talking about how to make sure it can go an extra six to 12 months to cover whatever additional regulatory scheme gets placed on them."[6]

The integration of digital infrastructure into the commercial mortgage market marks a permanent evolution in real estate finance. As the AI supercycle accelerates, the ability to accurately price technological lifespan will become just as critical as evaluating location and tenant credit.[7]