Follow Beinsure
HSBC Holdings (NYSE: HSBC) is exploring credit risk insurance to reduce its exposure to data center loans as banks seek additional capacity to finance the rapid expansion of artificial intelligence infrastructure.
The London-based lender is holding preliminary discussions with insurers about a series of project finance transactions, including loans connected to AI infrastructure, according to Bloomberg, citing people familiar with the discussions.
- HSBC is in preliminary discussions with insurers about transferring credit risk on project finance loans, including exposures linked to AI infrastructure and data centers.
- Barclays is separately exploring insurance protection for more than $10 bn of AI-related debt, while Deutsche Bank has considered a risk transfer involving €2 bn in project finance loans.
- WTW reports that some credit insurers are approaching their exposure limits for data centers, as McKinsey estimates global investment requirements of $6.7 tn by 2030.
The talks follow an earlier assessment of insurers’ willingness to provide credit protection on a portfolio exclusively comprising data center loans. Details of the potential arrangements, including their size, pricing and participating insurers, have not been disclosed.
The discussions come as banks face increasing pressure to manage concentrated lending exposures while maintaining the capacity to finance new data center developments.
Credit risk insurance allows lenders to transfer an agreed portion of potential borrower default losses to insurers. Depending on the structure and applicable regulatory requirements, this protection can reduce banks’ capital requirements and make additional lending possible without selling the underlying loans.
The arrangements differ from conventional property and business interruption insurance, which protects data center operators against physical damage and operational losses rather than borrower defaults.
Banks turn to insurers to reduce AI infrastructure credit exposure
HSBC is among several international banks considering credit protection and significant risk transfer (SRT) transactions to manage exposure to AI infrastructure financing.
Barclays is exploring insurance coverage for more than $10 bn of debt linked to AI infrastructure, according to a separate Bloomberg report published on October 28.
The bank has been assessing insurer appetite for default protection in recent weeks, seeking to release lending capacity as large technology companies increase their infrastructure financing requirements.
Barclays is reportedly considering transferring more than 50% of the credit risk on certain AI-related facilities. The discussions remain exploratory, with no final decision on the size or timing of any transaction.
Deutsche Bank has also been working on a potential SRT involving approximately €2 bn ($2.3 bn) of project finance debt, including data center loans, according to Bloomberg reporting in September.
HSBC had previously been linked to another potential risk transfer. A September 18 report attributed to Bloomberg indicated that the bank was discussing an SRT referencing more than $1 bn of loans financing data centers and other AI infrastructure. The terms remained subject to negotiations.
These transactions can take different forms. Under credit risk insurance, insurers agree to cover specified losses in exchange for premiums. SRT structures typically transfer defined portions of a loan portfolio’s credit risk to investors, potentially through credit-linked notes or other protection arrangements.
Both approaches can help banks manage exposure limits and regulatory capital, although the amount of relief depends on how the transactions are structured and treated by supervisors.
Banks are also exploring loan sales and broader syndication to distribute the financial risks associated with large infrastructure projects.
Credit insurers approach exposure limits for data center financing
The expansion of AI infrastructure financing is increasing demand for insurance protection against borrower defaults, but insurers’ ability to absorb additional data center exposure is becoming a constraint.
Willis Towers Watson (WTW) reported in its 2026 Credit and Political Risk Insurance Capacity Survey that insurers had increased their participation in data center financing.
The survey, conducted in January among 76 credit and political risk insurers across Lloyd’s syndicates and company markets, found growing overall capacity and greater willingness to underwrite non-trade transactional credit risks.
Insurers also reported increasing interest in portfolio insurance, multi-obligor arrangements and transactions designed to help banks transfer credit exposure.
Demand from data center financing has nevertheless created concentration concerns. According to Bloomberg’s reporting on the WTW findings, several non-trade credit insurers indicated they were approaching their sector exposure limits or had already reached their available capacity for data center risks. This creates difficulties for banks seeking coverage on large portfolios where multiple borrowers may be exposed to similar economic, technological or operational conditions.
The challenge is particularly relevant for projects involving substantial construction commitments, long financing periods and dependence on a relatively small group of major technology customers.
Dan Riordan, head of political risk and credit at MSIG USA, told The Insurer in February that data centers and AI-related projects had generated the strongest growth in submissions for credit insurance.
He said banks were the principal buyers of this protection, even when projects involved hyperscalers with strong credit profiles.
According to Riordan, lenders are finding it difficult to accommodate the scale and duration of financing required for increasingly large data center projects.
AI data center investment could reach $6.7 tn by 2030
The scale of planned infrastructure expenditure explains why lenders are seeking additional sources of insurance and investment capital.
McKinsey estimates that global data center infrastructure will require approximately $6.7 tn in capital expenditure by 2030 to meet expected computing demand.
- $5.2 tn would be needed for facilities supporting AI workloads.
- A further $1.5 tn would support conventional IT and non-AI applications.
The consultancy also projects that global data center capacity demand could nearly triple by 2030, with AI workloads accounting for approximately 70% of that demand.
The financing requirements are already affecting credit markets.
In its September 2026 Financial Policy Committee record, the Bank of England reported that global AI-related debt issuance had reached approximately $450 bn by early September, based on Morgan Stanley estimates.
The central bank also cited JPMorgan projections that approximately $4.1 tn of AI-related capital expenditure could be financed through debt between 2026 and 2030.
Morgan Stanley separately estimated that private credit could finance around $700 bn of data center capital expenditure between 2026 and 2028.
The growing reliance on borrowing is expanding the involvement of commercial banks, private credit funds, institutional investors and insurers in AI infrastructure financing.
HSBC warns of AI credit risks as financing requirements increase
The expansion of AI-related borrowing is also raising questions about the ability of projects to generate sufficient cash flow to service their debt.
In its third-quarter 2026 earnings release, published on October 27, HSBC identified AI investment and technology valuations among the risks it continues to monitor.
The bank warned that expected productivity improvements and economic benefits from AI might take longer than anticipated or fail to materialize.
Such developments could contribute to a correction in technology-related asset valuations, affecting borrowers’ creditworthiness and the value of collateral supporting loans.
HSBC said these risks could affect its financial performance by weakening borrowers’ ability to meet their obligations and increasing financial pressure on customers.
The Bank of England has separately warned that the growing use of debt to finance AI infrastructure could transmit losses across financial markets if expectations for computing demand, technology adoption or future revenues prove incorrect.
Data center projects face additional uncertainties related to electricity availability, construction delays, refinancing requirements and the depreciation of computing equipment.
These exposures can become correlated when multiple projects depend on the same customers, technology suppliers or financing markets.
Banks can reduce their direct exposure through credit insurance and SRT transactions, but the underlying risk remains within the financial system and is transferred to insurers or other investors.
HSBC’s discussions with insurers have not yet resulted in a publicly confirmed transaction. The amount of credit protection sought and the insurers potentially involved remain undisclosed, leaving the outcome dependent on negotiations and the available capacity to underwrite additional AI infrastructure risk.
