Morgan Stanley Flags Wider Spreads for Data-Center ABS Debt

Key Takeaways

• Morgan Stanley expects data-center ABS spreads to widen as issuance grows and investors scrutinize concentration, leverage, and residual values
• A 2026 transaction involving Meta and Microsoft illustrates the pricing premium investors may demand over comparable corporate debt
• Strong demand and low vacancy support data-center assets, but power constraints, construction growth, and refinancing exposure complicate underwriting

Morgan Stanley expects spreads on asset-backed securities funding data-center developments to widen further as developers and operators tap a broader mix of debt markets. The warning reflects a shift in investor attention: demand for computing capacity remains substantial, but financing structures are receiving more scrutiny as issuance accelerates.

According to Bloomberg, Morgan Stanley sees continued widening in data-center ABS spreads as companies race to fund new developments. Investors are reassessing tenant concentration, hyperscaler leverage, structural complexity, and the uncertain residual value of facilities built around rapidly evolving computing requirements.

The market has grown quickly. From 2018 through early 2025, more than $34 billion of data-center ABS debt was issued across 75 transactions, according to KBRA data cited by APSEC in 2026. That expansion has helped establish securitization as an important funding channel alongside corporate bonds, bank loans, commercial mortgage-backed securities, and private credit.

Exposure to a well-known technology tenant is not necessarily priced like exposure to that tenant's corporate balance sheet.

CLDHQ 2026-1A A22, backed by a data center where Meta and Microsoft are anchor tenants, traded at 148 basis points over the five-year Treasury. Comparable Meta corporate debt traded at 89 basis points over the same benchmark. The 59-basis-point gap shows how investors distinguish between direct corporate credit and a securitized claim dependent on leases, property performance, transaction protections, and the facility's longer-term competitiveness.

That distinction becomes more important when one or two hyperscalers account for much of a property's revenue. Meta and Microsoft may bring strong credit profiles, but ABS investors also need to evaluate lease expiration schedules, renewal options, tenant termination rights, equipment ownership, and the cost of adapting a building for another customer. A tenant can remain financially healthy while deciding that a particular site no longer fits its technical plans.

For now, operating fundamentals remain robust. CBRE reported that absorption across North American primary markets reached 2,497.6 MW in 2025, up 38% year over year. Vacancy fell to 1.4%, while supply expanded 36% to 9,432 MW. Those figures point to tight capacity even as developers bring substantial new infrastructure online.

Strong conditions can cut both ways, though. Low vacancy and rapid absorption support lease revenue and collateral valuations. At the same time, aggressive construction assumptions may leave transactions more exposed if demand growth slows, customers consolidate workloads, or newer facilities make existing sites less attractive. What happens when today's specialized AI infrastructure has to serve a different mix of workloads several years from now?

Power may be the harder constraint. Grid-interconnection timelines commonly average four years or longer in major regions, limiting how quickly planned campuses can become revenue-producing assets. Global data-center electricity demand rose 17% in 2025 and is expected to double by 2030, while consumption by AI-focused facilities is projected to triple. The financing risk is straightforward: delayed power can mean delayed tenant occupancy, higher carrying costs, and a longer wait for the cash flows supporting ABS payments.

Broader industry reporting also points to an unusually tight operating environment. FacilitiesNet has highlighted how roughly 1% vacancy is reshaping infrastructure planning as AI demand changes where and how capacity is developed. Yet scarcity alone does not settle questions about future value. Power density, cooling design, network connectivity, and access to additional electricity can influence whether an older facility remains competitive.

Transaction structure therefore matters. Credit enhancement, reserve accounts, debt-service coverage triggers, and master-trust arrangements can provide protection, but they also add complexity. Investors may demand additional spread when protections are difficult to compare across deals or when future borrowing could alter leverage within a financing structure.

Morgan Stanley's outlook does not suggest that data-center demand has weakened. Instead, it indicates that capital markets are becoming more selective about how that demand is packaged. Issuers with diversified tenants, durable power access, conservative leverage, and flexible facilities may face a different reception from projects dependent on a single hyperscaler or an uncertain grid connection. For borrowers, wider spreads translate into higher funding costs. For investors, they may represent compensation for risks that headline vacancy and absorption figures do not fully capture.