Three of the world’s largest banks have recently explored or bought credit insurance to hedge their exposure to AI infrastructure loans, according to the Financial Times - a sign that the surge in data center financing is beginning to test the limits of what bank balance sheets can absorb on their own.
Barclays has been sounding out insurers about buying default protection on more than $10 billion of AI infrastructure loans. HSBC has also explored using credit insurance on data center debt. JPMorgan Chase considered the approach but ultimately declined against it.
The common thread is the same pressure: concentration limits that cap how much a bank can lend to a single company, project, or sector are constraining appetite at the same moment that hyperscalers are demanding ever-larger loan packages.
Credit insurance resolves that constraint by transferring a portion of the default risk to insurers. In a typical transaction, a bank cedes 20% to 50% of its exposure, freeing up balance-sheet capacity to approve larger facilities or additional deals. For US banks in particular, the product offers something credit default swaps increasingly do not: manageable cost.
Azi Larsen, managing director and Lenders Solutions Group leader for North America at Marsh Specialty, told the FT that the rising cost of CDS and the scale of AI debt issuance have pushed US lenders to look at insurance as an alternative route to exposure relief.
The number of credit insurance providers for project finance has grown from a handful a decade ago to more than 70, including AIG, Chubb, and Axa, according to the FT. That growth reflects how structurally important the product has become to large-scale infrastructure lending. Premiums for data center exposure are running at 50% to 70% of what banks earn on that portion of their loan portfolios, according to the FT report. That pricing reflects both the demand pressure and the risk concentration that insurers are being asked to absorb.
Several providers of non-trade credit risk insurance had indicated they were approaching sector limits or were already full on data center exposure, according to a Willis Towers Watson survey released in May 2026. The same survey found credit risk insurers had already taken on significant data center exposure as the AI build-out accelerated. The pipeline of bank demand described in the FT piece is arriving into a market already signalling supply constraints.
The London market has been moving to absorb more of the class. Direct credit insurance premium at Lloyd’s grew by approximately a quarter in 2025 to around £1.6 billion, following Lloyd’s removal of premium income limits on credit and financial guarantee business in July 2025, according to Willis Re. New capacity has entered the market through syndicates including Atradius, Coface, and Credeq.
With property catastrophe rates softening and capital looking for a home in specialty lines, credit has become a natural destination, though Willis Re has cautioned that reinsurers will scrutinize data center concentrations closely at the January 2027 renewals.
Banks seeking to offload risk on AI infrastructure loans need specialist brokers to identify available capacity, structure the transfer, and navigate insurer appetite that is increasingly selective on data center exposure. Charlie Radcliffe, chief commercial officer of BPL Global, a specialist credit and political risk insurance broker, told the FT that banks are willing to pay premium rates in order to secure majority lender status or demonstrate support for a particular client. That dynamic makes the placement viable but requires precise matching of risk to available appetite.
The specific insurers approached by Barclays and HSBC have not been publicly named, and several of the transactions described remain exploratory. What has become clear is that credit risk transfer on AI infrastructure has moved from an occasional trade to a recurring placement category, one where capacity management and insurer selectivity will shape outcomes as much as price.
