Marsh seeks US capital relief for loans backed by credit risk insurance as banks manage rising AI data center exposure and global investment approaches $6.7 tn by 2030.
Marsh is urging US regulators to recognize credit risk insurance more broadly when calculating bank capital requirements, as lenders seek to manage growing exposure to AI data center financing.
The insurance broker argues that existing rules discourage US banks from using insurance guarantees to reduce regulatory capital requirements, putting them at a disadvantage to European competitors.
- Marsh is lobbying US regulators to recognize qualifying credit insurance guarantees for bank capital relief.
- Barclays has purchased insurance on AI infrastructure loans, while HSBC has explored similar transactions involving data center financing.
- Rising demand is testing insurer capacity, with premiums potentially consuming 50-70% of banks’ earnings on insured loan portions.
Marsh wants investment-grade insurance subsidiaries to qualify as eligible guarantors when their parent companies, rather than the subsidiaries themselves, issue investment-grade debt.
The proposal comes as global data center investment is projected to reach a cumulative $6.7 tn by 2030, according to McKinsey, increasing demand for financing and credit risk protection.
Marsh seeks regulatory capital relief for US banks
Guy Carpenter, Marsh’s reinsurance arm, met Federal Reserve staff on June 30 to propose changes to the eligibility requirements for insurance guarantees.
The proposal addresses situations where an investment-grade insurance subsidiary does not issue unsecured debt because its parent company handles financing. Marsh wants the subsidiary to qualify as an eligible guarantor based on an appropriately rated parent’s investment-grade debt issuance.
The proposed change would not automatically reduce capital requirements for data center loans. Recognition would still depend on the insurer, policy terms and applicable regulatory standards.
Existing restrictions concerning correlated credit exposure, or wrong-way risk, and monoline credit insurers would remain.
Marsh argues that clarification would improve access to insurance-based credit risk transfer, particularly for smaller US banks without extensive capital markets operations.
Azi Larsen, managing director in Marsh’s credit specialties practice, said the company has been advocating changes to US regulations and hopes for a positive decision by year-end.
European banks already use qualifying credit insurance arrangements to obtain capital relief, while US lenders generally lack equivalent treatment for the structures under discussion.
Federal Reserve Vice Chair for Supervision Michelle Bowman said on November 10 that new bank capital proposals were expected in the coming weeks. These include changes to capital surcharges for the largest lenders and risk-based requirements for midsize and large banks.
Barclays and HSBC seek insurance for AI infrastructure loans
Major banks are increasingly considering credit insurance to manage concentrated exposures to AI infrastructure projects.
Barclays has purchased protection on AI infrastructure loans after previously exploring insurance covering more than $10 bn of data center debt.
Some proposed arrangements could transfer over half the credit risk on individual facilities, although there is no confirmation that the entire portfolio has been insured.
HSBC has explored similar protection for project finance loans, including data centers. The bank has also worked on a significant risk transfer involving more than $1 bn of data center-linked loans, but completion of an insurance transaction has not been confirmed.
JPMorgan reportedly considered credit insurance before deciding against it.
Credit risk insurance allows banks to transfer an agreed share of potential loan losses to insurers, helping control sector concentration and preserve lending capacity.
Banks pay premiums for protection against specified credit losses. Even without regulatory capital relief, insurance can help lenders manage internal borrower limits and maintain relationships with major clients.
Insurance capacity and costs constrain AI loan protection
The project finance credit insurance market has expanded from a handful of providers a decade ago to more than 70 insurers.
Demand is nevertheless testing available capacity. A May 2026 Willis survey found that some insurers had already reached or approached their data center sector exposure limits.
Insurance premiums can absorb 50%-70% of a bank’s earnings on the insured portion of a loan, making protection expensive even when it supports additional lending.
Banks must weigh those costs against the benefits of reducing exposure and maintaining lending relationships.
Insurers also face concentration risks. Data center borrowers share exposure to electricity costs, financing conditions, regulation and AI demand, raising the possibility of simultaneous losses across multiple projects.
The Bank of England has warned that technological changes in computing equipment and cooling systems could reduce the useful life of data center infrastructure, weakening collateral values and borrowers’ ability to service long-term debt.
The IMF has separately noted increasing reliance on external financing for capital-intensive investment.
What regulatory changes could mean for banks and insurers
Marsh’s proposal focuses on a specific technical obstacle to recognizing insurance guarantees, rather than a general reduction in capital requirements for AI infrastructure loans.
If regulators approve the clarification, more insurance arrangements could qualify for capital relief, potentially increasing demand for credit risk transfer among US banks.
Insurance can reduce a lender’s retained exposure, but the underlying credit risk moves to the insurer rather than disappearing from the financial system.
The distinction is particularly relevant when multiple insured loans depend on the same AI infrastructure demand and financing conditions.
