The reported $1bn-plus portfolio extends HSBC’s existing use of significant risk transfers into AI infrastructure. The loans have become large and segmentable enough for portfolio-level risk transfer; that does not mean the underlying credits themselves have become standardised.
HSBC is discussing a significant risk transfer tied to more than $1bn of loans financing data centres and other AI infrastructure, according to Bloomberg reporting. The talks are preliminary, HSBC has declined to comment and the final structure may change.
The headline amount is the reference portfolio. It is not the loss tranche investors would necessarily absorb, cash proceeds to HSBC or the amount of regulatory capital that would be released.
HSBC was also using this machinery well before the AI portfolio appeared.
In January, it sounded out investors on an SRT linked to roughly €2bn of investment-grade corporate loans. The contemplated risk-transfer slice was about 10% of that reference pool. In July, HSBC began preliminary work on another transaction tied to Asia-Pacific loans.
The data-centre proposal therefore extends an existing capital-management programme into a new concentration of lending.
Data-centre credit is becoming portfolio-manageable
There is already a sector precedent.
Toronto-Dominion Bank was reported in April to be considering an SRT against an initial data-centre reference portfolio of about $1bn. The proposed structure included a forward-flow mechanism that could bring subsequently originated loans into the protection pool.
The implication is narrower than saying data-centre loans have become standardised. Project structures, borrowers, collateral and contractual economics can remain highly bespoke.
What has changed is scale and segmentation. Banks have accumulated enough exposure with sufficiently identifiable risk characteristics to contemplate transferring a defined layer of credit risk to institutional investors.
For HSBC, that can preserve lending and borrower relationships while changing who absorbs losses within an agreed tranche. Capital relief follows only if the final structure transfers sufficient risk to satisfy regulators.
HSBC reported $127.7bn of CET1 capital, $906.4bn of risk-weighted assets and a 14.1% CET1 ratio at June 30. The proposed AI portfolio is small beside that balance sheet before considering that an SRT would normally protect only part of the reference amount.
The transaction is therefore a weak basis for arguing that HSBC has become unusually worried about AI lending. It is stronger evidence that the bank now regards the exposure as suitable for the same portfolio-management techniques it applies elsewhere.
AI risk is moving into institutional portfolios
The Bank of England has documented the widening role of non-bank capital in AI finance. It cited OECD estimates showing private credit’s share of financing for AI investment rising from 9% in 2024 to 34% in 2025 and has separately examined the expansion of SRT issuance.
An SRT creates another connection between those two markets. Banks originate and retain customer relationships; funds, insurers or other investors can be paid to absorb a specified portion of default losses.
The economics of HSBC’s proposed transaction remain unavailable. Attachment and detachment points, protection size, premium and collateral determine how much credit risk genuinely leaves HSBC and how much regulatory capital the bank can recycle.
Those missing terms prevent a judgment about the attractiveness of the individual trade. They do not obscure its strategic significance.
AI infrastructure credit has reached the point where HSBC is considering carving it into a dedicated institutional risk-transfer transaction. The build-out is creating a market for transferring its credit risk as well as demand for the loans themselves.