Data Center CMBS Spreads Widen as Investor Risk Pricing Framework Comes Under Pressure
nashnova research
Data-center CMBS issuance has hit roughly $17 billion this year — more than triple the prior two years combined; AAA spreads now sit at 1.65 percentage points, far above office and retail, as legacy real-estate credit frameworks buckle under a fundamentally different asset class.
$17 billion and counting — what is happening in this market?
Data-center CMBS — bonds backed by loans on data facilities, not offices or malls — have reached roughly $17 billion in 2025 issuance, about 8% of all new commercial-property bonds.
This means → a once-marginal category is flooding into mainstream CMBS at triple speed, forcing investors to engage whether they are ready or not.
Citi expects issuance to jump another ~50% next year, to $18–20 billion. In plain terms = the supply wave has not peaked; this is the warm-up.
Why are spreads wider than for office buildings?
AAA-rated data-center CMBS now price at 1.65 percentage points over the floating-rate benchmark. Office sits at 0.93, retail at 1.05, industrial at 1.25.
This means → at the same credit rating, the market demands nearly twice the risk premium for data centers versus offices — rating agencies say "equally safe," but buyers disagree.
Last week, a $356 million bond backed by a 30-megawatt facility near Elk Grove Village, Illinois, priced its top-rated tranche well wide of initial guidance — the third such episode in recent months.
Why do traditional property-analysis tools break down here?
The metrics that matter for data centers — grid capacity, power cost, cooling design, compute density — have almost no overlap with conventional real-estate credit analysis.
CWCapital senior managing director Alex Killick put it plainly: real data centers are measured in "compute units and megawatts," which is "a completely different world for real-estate investors."
In plain terms = legacy analysts evaluated location, rent rolls, and vacancy rates. Now the due-diligence checklist starts with electricity bills and chip-upgrade cycles — an almost entirely new knowledge base.
Tenant concentration and opacity — where is the risk hiding?
Most data-center CMBS are structured as single-asset, single-borrower deals: one large loan against one facility or campus, leased to a handful of hyperscale cloud tenants.
This reflects a key fragility: tenant identities and lease details are typically confidential, leaving outside investors unable to independently verify demand stability.
Axonic Capital portfolio manager Steven Jury keeps his data-center allocation low and insists on diversification across tenants, uses, and geographies. His core question: "What are these assets worth in five, ten, or twenty years — and who still needs them then?"
How have site selection and re-leasing risk been upended?
Traditional commercial property prizes transit access and urban centrality. Data centers prize cheap power and available transmission capacity — the definition of "prime location" has been rewritten.
Re-leasing risk — finding a new tenant if the current one leaves — is equally unfamiliar: bespoke electrical and cooling systems may not suit another operator, meaning higher retrofit capex, longer vacancy, and lower recovery rates.
Trepp's head of applied research, Stephen Buschbom, said data centers "look more like infrastructure and complex technology projects than traditional real estate."
Chip turnover and regulatory pushback — what is still unpriced?
Rapid AI-chip iteration could drive sharp spikes in power and cooling demand. Killick noted that traditional property obsolescence is modeled in decades, "but a chip installed six months ago in a data center may soon be replaced" — that alone makes due diligence extraordinarily difficult.
Data centers have become a political issue from local to national level: concerns over utility costs and infrastructure strain have sparked calls to restrict new builds in some communities, making the future regulatory landscape unusually hard to forecast.
Citi's Jeffrey Berenbaum captured the worst-case scenario: if hyperscale tenants all walk away, investors could be left holding "the world's largest pickleball court" — and the market has yet to build a mature framework to price that tail risk.
市场有风险，内容仅供研究参考，不构成投资建议。