Carbon's subprime moment

Every credit crisis begins the same way. An asset is carried on the books at a number nobody has independently tested, the number is supplied by the party with an interest in it being high, and a structural reason exists for everyone in the chain to leave the assumption unexamined until an external event forces the question. By the time the question is forced, the gap between the carrying value and the recoverable value is no longer a rounding error. It is the loss.

Corporate carbon books are now sitting in exactly that position, and an accounting standard is about to force the question. This is not a sustainability essay. It is a balance-sheet problem with a documented analogue, and the analogue is the structured-credit failure of 2008.

The structure of the analogy

01 · The asset is carried at a number nobody independently tested

A voluntary carbon credit enters a corporate book at cost. It then sits there. Under the accounting regime that prevailed until very recently, there was no requirement to test it for impairment, no mark, no periodic revaluation against anything. A credit bought at twelve dollars in 2021 was a twelve-dollar line whether or not the underlying project still stood, still stored carbon, or had ever represented a real additional tonne in the first place. The carrying value was an artefact of the purchase invoice, not of the asset's condition.

The dispersion underneath that single number is enormous and it is now measurable. By late 2025, afforestation, reforestation and revegetation credits rated BBB+ were trading above twenty-six dollars while lower-rated tonnes of the same nominal type sat around fourteen, and the weakest tier changed hands near three dollars (BeZero, MSCI Carbon Markets). High-quality removal credits rose roughly fourteen dollars to twenty-six across 2025. That is a more than three-times spread between the best and worst credits that a balance sheet records, in many cases, at the same flat historical cost. A book of "carbon credits" is not a homogeneous holding. It is a portfolio of wildly varying recovery, presented as a single asset class, marked at acquisition price.

02 · The grade comes from the party selling the risk

The 2008 failure had a precise mechanical cause that is too often described in moral terms. Structured-credit ratings were issuer-paid. The bank assembling and selling the instrument paid the agency that graded it. The agency competed for that fee against other agencies. The buyer of the paper relied on a grade that the seller had commissioned and could shop. The incentive did not have to be corrupt to be fatal. It only had to tilt the marginal judgement, repeatedly, in one direction, across millions of instruments, for years.

Carbon repeats the structure. A large share of the assurance that sits behind a credit is, in effect, seller-commissioned. Standards bodies that register and issue credits are funded by issuance. Project developers select and pay the validation and verification bodies that sign off their projects. The history is on the record. Under the Clean Development Mechanism, the predecessor regime, more than eight thousand projects were registered across more than a hundred countries between 2004 and 2012, and the vast majority of the claimed reductions were later judged unlikely to be additional. That is the original integrity failure, and the voluntary market inherited the architecture that produced it. The grade and the seller share a paymaster.

The incentive did not have to be corrupt to be fatal. It only had to tilt the marginal judgement, repeatedly, in one direction, for years.

03 · Everyone in the chain is paid to leave it unexamined

The deeper parallel is in the silence. In 2008 the originator, the arranger, the agency, the distributor and the holder were each individually better off not pressing on the assumption, right up to the moment the assumption broke. The carbon chain has the same property. The developer wants issuance. The broker wants the transaction. The corporate sustainability team that approved the purchases is the last party that wants those purchases re-examined, because a writedown is also an admission. The auditor, historically, had no standard that compelled the test. Nobody in the chain was the natural owner of the inconvenient question, so the inconvenient question went unasked.

The external event that forces the question

Credit assumptions do not get re-marked because someone has a change of heart. They get re-marked because an external force compels it: a default, a downgrade, a regulator, an auditor who can no longer sign. Carbon now has that force, and it arrived in May 2026.

The Financial Accounting Standards Board issued ASU 2026-02, creating ASC Topic 818, on 19 May 2026. The mechanics matter and they are unforgiving. Noncompliance credits, which is to say the voluntary market, are to be carried at cost less impairment at each reporting date, with a fair-value election available that runs changes through earnings. Impairment is irreversible. Presentation is gross: the credit asset is not netted against the related obligation, so a degraded book cannot be quietly offset out of view. The environmental credit obligation is measured on the carrying amount of the owned credits plus fair value for the unfunded excess. FASB has, in short, imported the discipline of impairment testing into an asset class that was specifically built to avoid it.

Read that as a finance professional rather than a sustainability one. Impairment testing at each reporting date means the controller now needs a defensible view of recoverable value. Irreversibility means there is no waiting it out: once written down, the loss is booked. Gross presentation means the audit committee will see the obligation and the asset side by side. A fair-value election means some companies will be running these holdings through the P&L every quarter. This is not disclosure hygiene. This is a recurring, board-level, auditor-tested question about an asset that, for most holders, has never once been independently marked.

04 · Why this is the 2005 stage, not the 2008 stage

The most useful thing the analogy gives us is a clock reading. Carbon is not at 2008. It is at 2005. At the 2005 stage of the structured-credit cycle the conflict was already documented, the evidence of poor underlying quality was already in, the academic and journalistic record already existed, and the market carried on pricing the paper as though the grade were independent and the assumption sound. The repricing had not happened yet, but everything required to predict it was visible to anyone who chose to look.

Carbon is there now. The conflict is documented. The quality dispersion is measured and public. The accounting trigger is signed and dated, with first application landing on reporting calendars. And the market is still, broadly, carrying credits at cost and trusting grades commissioned by the side that wants them high. The gap between the 2005 stage and the 2008 stage was roughly three years of everyone agreeing not to look. The instructive feature of that interval is that the people who marked their books early did not suffer the repricing. They caused other people to.

What gets repriced, and by whom

Three forces are converging on the same books at the same time, and each one independently lowers recoverable value. The accounting force is FASB, just described. The claims force is regulatory: the EU Empowering Consumers for the Green Transition Directive applies from 27 September 2026 and blacklists offset-based product claims such as carbon neutral and climate positive under all circumstances, with fines up to four per cent of annual turnover. The withdrawal of the softer Green Claims Directive proposal in June 2025 removed the safe harbour and raised, rather than lowered, the enforcement risk. The target force is standard-setting: the SBTi Corporate Net-Zero Standard V2.0, finalised 11 June 2026, relegates credits to contribution claims and bars them from counting toward Scope 1, 2 and 3 target implementation. Litigation has already begun to test the claims (the KLM ruling in Amsterdam in March 2024 found offset-based neutralisation statements misleading; the Delta class action survived a motion to dismiss in December 2024).

A credit that cannot support a claim, cannot count toward a target, and must be impairment-tested for fair value is a credit whose recoverable value has fallen on three axes at once. The reclassification machinery is already running. ICVCM Core Carbon Principles and CORSIA eligibility operate as a sorting mechanism that reprices entire vintages with no watchlist and no grace period. A whole vintage can be reclassified between one reporting date and the next, and under irreversible impairment that reclassification is not a warning. It is a loss event.

The people who marked their books early did not suffer the repricing. They caused other people to.

The missing independent layer

Return to the mechanical lesson of 2008, because it is the one that actually generalises. The fix was never to abolish ratings. It was to break the conflict: to insist that the party relying on a number should be able to get that number from someone who is not paid by the party selling the risk. The structural defect was payer identity, and the durable repair was payer independence.

Carbon has no such layer yet. The grading and assurance that sit behind the carrying value are still, in large part, commissioned by the sell side. The buyer holding the book, carrying the impairment risk, facing the auditor and signing the accounts, has no independent, finance-grade, continuously updated view of what the book is actually worth. Every incumbent that could supply one has a structural reason it cannot cleanly do so: it sells credits, or it rates for both sides, or it transacts, or it runs an investment arm. The independence that 2008 taught us to demand is precisely the thing the current market cannot offer, because the market is organised around the seller paying.