The European Central Bank's examination of private credit risk, published as a special feature of its May 2026 Financial Stability Review, has been received by the insurance industry as a clean bill of health. Insurance Europe found in it support for the assertion that European insurers have "limited direct exposure" to the asset class. Investors interviewed in these pages emphasised investment-grade portfolios, long liabilities, and prudent manager selection. The consensus: scrutiny is welcome, additional regulation unnecessary, and private credit here to stay.

The special feature deserves a closer reading than that. Taken in the original rather than in summary, it undermines both pillars of the industry's comfort — the exposure numbers and the ratings — and then does something the industry commentary has passed over in silence: it places the comparison with pre-crisis subprime on the table itself, and rests its dismissal of that comparison on grounds its own evidence contradicts.

The wrong number, measuring the wrong thing

Start with "limited direct exposure." The figure doing the work is the ECB's estimate that euro area insurers hold €211bn of private credit, some 2.3% of total assets. What the summaries omit is that 2.3% is already the most charitable count available. EIOPA's own broader measure puts insurers' private credit at 5.1% of total assets; the ECB arrived at its narrower figure principally by excluding mortgages and non-euro-area exposures. "Limited" is, in part, an artefact of definition — and the exposure it describes is concentrated, both in a small number of large institutions and geographically in Germany, France and the Netherlands.

But suppose we grant the narrow figure. The ECB's own simulation then demonstrates that it is the wrong measure of risk. The exercise runs in three stages: defaults on private credit itself (10% of exposures at 50% loss-given-default); a spillover into leveraged loans and high-yield bonds, with a severe shock to software-sector borrowers (30% default at 80% LGD); and a broad market repricing — equity markets down 30%, high-yield valuations down 25% — as sentiment reverses.

For insurers, the first two stages — the credit losses themselves — sum to less than €50bn. Add the third and total losses exceed €350bn, around 4% of total assets. The ECB is explicit that by far the largest impact comes from the third stage. The multiplier between those two numbers is the finding. The danger was never that insurers hold €211bn of private loans; it is that a private credit shock capable of triggering a general repricing detonates against insurers' far larger equity portfolios. The €211bn is the detonator, not the bomb — and a trade body pointing to the size of the detonator is answering a question nobody should be asking.

The sharpest edge of that transmission channel is one the ECB describes rather than models: the equity linkage between insurers and the alternative-manager complex that originates, levers and warehouses the credit — already prevalent in the United States, where private equity firms have spent a decade acquiring life insurers, and growing in Europe. And in a footnote, the special feature flags a channel it leaves out of scope entirely: private credit funds acting as protection sellers in banks' synthetic risk transfers. Risk "transferred" to entities whose assets are the correlated exposure is a sentence that should be read twice.

The circular defence

The second pillar is the ratings. UK annuity writers point out, correctly, that matching adjustment eligibility confines their private credit to investment grade. But privately placed, unlisted credit does not carry public agency opinions formed in a market. It carries private ratings, internal models mapped to agency scales, or bespoke assessments — procured, unpublished, and uncontestable. The defence is circular: the assets are safe because they are investment grade; they are investment grade because a rating was procured to make them eligible. When a chief credit officer explains that sub-investment-grade paper is "not very efficient" from a capital perspective, the candour is admirable and the implication clear: the binding constraint is capital treatment, not credit opinion. A system whose constraint is the label will optimise the label.

Pre-2008, at least, a sceptic could trade against the ratings: the ABX index existed precisely so that someone who believed the labels were wrong could say so at scale, and eventually the marks had to follow. Private credit has no equivalent. The asset class is marked by its own originators, quarterly or annually, with smoothness marketed as a feature. Flat watchlists and portfolios "doing well" are, in an unmarked asset class, unfalsifiable claims. The one corner of the market with genuine price discovery is meanwhile dissenting: the ECB notes sizeable redemption requests at semi-liquid vehicles since the start of 2026, withdrawal caps at some funds, and falling share prices at listed private-market firms. Where the asset class is market-priced, the market disagrees with the marks.

The comparison the ECB raised

Which brings us to the passage the industry commentary has not quoted. The special feature itself observes that US private credit, at roughly $1.4trn, is approximately the nominal size of the subprime mortgage segment in 2006 — about $1.5trn — before distinguishing the two on grounds of relative scale (subprime was 10.9% of US GDP; private credit is 4.7%), lower leverage, and funding that is long-term and not subject to run risk.

Take those grounds seriously, because the sector's safety case lives or dies on them. On leverage, the ECB's own document supplies the rebuttal: fund-level leverage, it notes, comes on top of leverage at the borrower and investor level, accumulating along the intermediation chain. NAV financing, back-leverage on fund stakes and rated feeder structures are rebuilding, earlier in the cycle, exactly the layered gearing whose sudden visibility defined 2008.

On run risk, the point should be conceded — openly, because it is real. Annuity liabilities cannot redeem. There is no overnight repo funding a bulk annuity book, no ABCP conduit to break. A mark-to-market shock does not become a funding run, and the system can carry unrealised losses for years without a forced-sale cascade. This is the industry's one genuinely sound defence, and it deserves better than to be deployed as if it settled the matter — because what it actually does is change the failure mode, not remove it.

Consider what made subprime a catastrophe: five conditions coinciding. Ratings manufactured by structure rather than earned by the underlying. Issuer-pays conflicts in the rating process. Concentration hidden behind apparent dispersion. Marks that could not fall until an external instrument forced them. And run-prone funding that converted a repricing into a collapse. The insurance private-credit complex has convincingly rebuilt the first four — with the conflicts arguably worse, since private ratings are invisible to everyone but the buyer and the regulator, and no tradeable instrument exists through which a sceptic can force the issue. What it lacks is the fifth. The sector has reconstructed four-fifths of 2008 and calls the missing fifth prudence.

The absence of run risk means the failure, if it comes, will not look like 2008. It will not look like anything. There will be no Lehman weekend, no index collapsing on a screen. An asset shortfall inside annuity books would surface slowly — through PRA reviews, through matching adjustment attestations that grow harder to sign, through buyout pricing that quietly stops improving — discovered years after the fees were paid, and borne by the one constituency with no ability to run: the pensioner. A crisis without a crash is not a smaller problem. It is a quieter one.

A floor, not a ceiling

Two forward-looking details in the ECB's analysis suggest the window for asking these questions is now. The growth vector of the asset class points toward defined contribution savings: vehicles outside the matching adjustment rules are explicitly freed to invest in the sub-investment-grade universe the annuity book cannot touch — so the insurer's balance sheet takes the manufactured-IG tranche, and the auto-enrolled DC saver takes the unwrapped residual. That is the channel through which 2008's losses became political, being rebuilt with official encouragement. And the ECB notes market intelligence suggesting up to 30% of the roughly $3trn required for AI data centre construction over the coming years could be financed by private credit — in which case today's exposure figures, limited or otherwise, are a floor rather than a ceiling, and increasingly a concentrated bet on a single technology cycle.

None of which is a prediction of failure. Private credit is credit, as its defenders rightly say, and much of it will perform. But "private credit is credit" cuts both ways: it is credit with the price discovery, the mark-to-market discipline and the contestable ratings removed — held against the retirement income of people who cannot run, will not be consulted, and will find out last. The ECB's special feature, read whole, is not the reassurance the industry took from it. It is a description, in a central bank's careful prose, of a familiar machine — missing one part, and being connected to the pension system.