TL;DR:
- Major banks including JPMorgan Chase, Morgan Stanley and SMBC are looking to offload portions of debt tied to AI data-centre construction, as exposure approaches internal risk limits, the Financial Times reports.
- Lenders including JPMorgan and MUFG have spent more than six months distributing $38bn (£28bn) of construction debt for an Oracle-leased data-centre project in Texas and Wisconsin, with some banks seeking to sell at a discount to non-bank lenders.
- Resultsense view: when investment banks call the deal sizes “out of scale to anything we’ve thought about, ever”, that is a yellow card for the AI capex cycle. UK CFOs running long-dated AI infrastructure commitments should watch the secondary-debt market, not just the model launches.
The shift is being executed through a combination of significant risk transfers (SRTs) and direct loan sales, designed to free up balance-sheet capacity while keeping the bank in the deal. Companies including Oracle and CoreWeave have borrowed hundreds of billions of dollars to build data-centre capacity, predominantly in the US.
What the banks are doing
Banks are reportedly approaching investors about variants of significant risk transfers under which they shift the riskiest portions of data-centre loans off their books while continuing to hold a defined percentage. The structure resembles synthetic securitisation but carries the credit risk of single, very large counterparties — Oracle, CoreWeave, OpenAI’s various data-centre vehicles — rather than diversified pools.
Matthew Moniot, co-head of credit risk sharing at Man Group, told the FT the sizes involved are “out of scale to anything we’ve thought about, ever”. The UK and European credit markets have not seen private deal sizes of this kind since pre-2008 syndicated bank loans.
The risk concentration question
The pattern echoes warnings issued the same week by the Financial Stability Board, which separately flagged that AI firms accounted for more than a third of private credit deals in 2025, up from 17% over the previous five years. Both stories — banks distributing risk, private credit funds taking it on — point at the same phenomenon: AI infrastructure is now financed at a scale and concentration that traditional bank balance sheets cannot absorb alone.
Public opposition to large data-centre projects is adding to the risk. Maine recently passed a statewide moratorium on new data centres after public pressure, and similar push-back has emerged in Virginia, Ireland and parts of the UK. Project delays trigger draw-down covenants and refinancing triggers, both of which tighten the resale market for the underlying debt.
UK relevance
UK banks have so far had limited direct exposure to US AI data-centre construction debt — this is mostly a US-bank story. But the secondary effects matter: UK insurers and asset managers are likely buyers of the SRT structures and discounted loan tranches the US banks are now selling. UK pension funds with heavy private-credit allocations should be asking their managers explicitly about AI data-centre concentration.
Looking forward
Two leading indicators to track. First, whether SRT spreads on AI data-centre tranches widen materially against comparable infrastructure debt — that’s the cleanest market signal of distress. Second, whether US bank quarterly disclosures start breaking out AI data-centre exposure as a discrete line item. Both will move ahead of any actual project failures and give UK CFOs a window to recalibrate AI capex commitments before refinancing risk shows up in headlines.