Summary
- S&P Global Ratings has warned that hyperscale data centre growth is creating insurance capacity constraints.
- Construction-phase total insured values for large campuses can reach the $20bn to $50bn range.
- Partial insurance, retained risk, and alternative capital could become more prominent in AI data centre financing.
S&P Global Ratings has warned that hyperscale data centre growth is creating an insurance gap that could influence how large AI infrastructure projects are financed.
The ratings agency’s report, The insurance gap is reshaping hyperscale data center finance, examines how the scale of new campuses is pushing beyond conventional single-risk insurance limits. Specialist insurance-market summaries of the report indicate that construction-phase total insured values for hyperscale campuses can move into the $20bn to $50bn range.
S&P’s earlier sector work has also identified data centres as a growing pool of insurable risk, with rising demand for cover linked to large, specialised, power-intensive campuses. The issue is not basic availability of insurance, but whether full conventional coverage can be secured for the largest projects at a price and structure acceptable to lenders, sponsors, and customers.
The warning adds a financial constraint to the AI infrastructure buildout. Power, land, grid connections, cooling equipment, transformers, switchgear, and construction capacity are already visible bottlenecks. Insurance capacity is less visible, but it can still shape financing, lender protections, project covenants, and the amount of risk retained by developers.
Asset scale is changing the risk model
Modern AI campuses are not ordinary commercial buildings with a server-room risk attached. They combine high-value buildings, extensive electrical systems, cooling plant, backup generation, networking equipment, customer IT load, and, in some cases, adjacent energy infrastructure. During construction, multiple buildings and systems may be exposed before the asset is operational and revenue-generating.
That concentration creates a difficult underwriting problem. A single campus can carry construction risk, natural catastrophe exposure, fire risk, equipment delay exposure, business interruption, and technology dependency. If the site is built for one or a small number of hyperscale customers, outage and delay risks can carry wider contractual consequences.
Traditional insurance markets are used to large infrastructure projects, but hyperscale data centres combine asset value with pace and replication. Developers want to build quickly, often in regions where grid infrastructure is itself under pressure. Insurers must assess not only each site, but the aggregation of similar risks across geographies, operators, suppliers, and natural hazard zones.
More layered insurance structures are likely to follow. Projects may be covered up to a probable maximum loss or maximum foreseeable loss rather than full site value. Sponsors and customers may retain more exposure. Lenders may require additional reserves, contractual protections, or alternative risk-transfer arrangements. Captive structures and capital markets instruments could become more common where conventional insurance is insufficient or uneconomic.
European developers will feel the financing effect
The issue has global roots, but it has direct European relevance. Europe is moving towards larger AI campuses while also dealing with slower planning systems, tighter environmental scrutiny, complex grid connection queues, and higher construction costs. Insurance complexity adds another due-diligence layer to projects that are already difficult to finance.
Insurance risk could also influence site selection. Locations with higher flood, wildfire, seismic, heat, or grid reliability risk may become more expensive to cover. Sites with stronger resilience design, better separation between buildings, robust fire suppression, tested emergency power, clear water and cooling strategies, and stronger operational controls may be more attractive to underwriters and lenders.
That creates an engineering consequence. Resilience decisions made at design stage can affect capital availability later. Fire compartmentation, battery safety, generator arrangements, fuel storage, water supply, drainage, security, and redundancy influence insurability, lender confidence, and customer risk allocation.
The insurance gap will not stop the AI buildout, but it will make the economics more complicated. As data centre campuses reach infrastructure-scale values, underwriting capacity joins power, planning, and procurement as part of the project permission structure. Schemes with resilient design and transparent risk allocation are likely to move more cleanly through financing than schemes built only around headline megawatts.
