On 17 January 2025, three days before the Trump inauguration, the Federal Reserve announced its withdrawal from the Network of Central Banks and Supervisors for Greening the Financial System. The stated reason was that the NGFS had broadened its scope beyond the Board's statutory mandate. The FDIC followed days later. The OCC left in February. The Treasury Department exited the same window. By the end of the first quarter of 2025, every US federal financial regulator that had been engaged with international climate-supervisory work was gone. The Fed, FDIC, and OCC subsequently withdrew their joint guidance on climate-related financial risk management for banks with more than $100 billion in assets.

The political-cycle reader interpreted this as the end of central-bank climate work. The narrative was tidy: Trump returns, the climate-finance scaffolding built across the Obama and Biden administrations gets dismantled, the working-level supervisory infrastructure unwinds as the political winds shift. The Davos commentariat moved on. The compliance officers and risk managers inside major banks waited for the supervisory expectations to weaken.

The supervisory expectations did not weaken. They compounded.

Eighteen months later, in May 2026, Frank Elderson, Executive Board member of the European Central Bank and Vice-Chair of its Supervisory Board, gave a lecture at the University of Oxford that brought inflation, banking supervision, litigation, and environmental stress into a single financial-risk story. The ECB's verdict on European bank climate-risk management, published in a parallel supervisory blog two weeks earlier, was that climate and nature risks are "very likely being underestimated" by banks under ECB supervision. The ECB cited EUR 822 billion in direct EU losses from climate-related events between 1980 and 2024 as the floor of what the supervisory framework now treats as a real, present, balance-sheet-affecting risk.

This piece is a case study of the gap between what the political cycle was saying about central-bank climate work in 2025-2026 and what the supervisors were actually doing. Five moves. The political cycle says X. The supervisors' spreadsheet says Y.

Move 1. The Fed exit was real. The reset of central-bank climate work was not.

The Fed's January 2025 withdrawal from the NGFS was a discrete, dateable event with real institutional consequences. The Fed stopped sending representatives to NGFS working groups. The shared US data flow into NGFS scenario calibration thinned. The transatlantic alignment that had characterised the 2020-2024 period weakened materially. None of that is contested.

What did not happen is the broader unwinding the political-cycle reader expected. The NGFS continued operating with its remaining 134 central-bank and supervisor members. Phase V of the long-term scenarios shipped on schedule. Short-term scenarios for the 2025-2030 horizon launched on 7 May 2025, less than four months after the Fed exit, providing the first dedicated framework for analysing near-term climate impacts on financial stability. The ECB, BoE, BIS, Bank of Japan, Monetary Authority of Singapore, and most major non-US central banks continued contributing data, methodology, and political support.

The casualty count of the Fed exit, at the working level, was the loss of US-specific physical-risk calibration in the NGFS scenarios. The rest of the supervisory machinery kept running.

The political-cycle reader saw the Fed withdrawal and projected an end-of-era reading onto it. The supervisors' spreadsheet shows the era continuing without the United States as a contributor.

Move 2. ECB collateral framework: climate moved into the price of refinancing.

The ECB's central institutional move in 2025-2026 happened in a domain the political cycle barely reads: the framework that governs how the ECB accepts collateral from banks against monetary policy operations. Collateral framework is not headline material. It is also where central-bank decisions become real money for commercial banks.

The ECB announced in late 2025 the introduction of a climate factor for refinancing operations, taking effect in the second half of 2026. The mechanism: when a commercial bank posts corporate bonds as collateral to borrow from the ECB, the haircut applied to those bonds will now reflect climate-related transition risk. The factor takes into account sectoral risks, the issuer's climate score, and the bond's remaining maturity.

Translated into the language of bank treasurers: a fossil-fuel corporate bond posted as collateral will fetch less euro liquidity from the ECB than an equivalent-rated bond from a transition-aligned issuer. The price differential is small initially, but the mechanism is now built into the daily plumbing of the Eurosystem. Every bank that funds itself through ECB operations will, from H2 2026 onwards, find that the composition of its corporate-bond holdings affects the cost of its central-bank funding.

This is not disclosure. It is not stress testing. It is not voluntary guidance. It is direct price formation in a wholesale funding market. The ECB has used the same tool, with the same effect, for credit ratings and remaining maturity since the early 2000s. Climate risk now joins the list.

The political-cycle reader does not track collateral haircut adjustments. The supervisors' spreadsheet records that climate risk is now an input to commercial-bank funding costs in the euro area.

Move 3. EU-wide stress test 2025: climate showed up as basis points of CET1 capital.

The 2025 EU-wide stress test, run jointly by the European Banking Authority and the ECB, tested 64 European banks against an adverse macroeconomic scenario through 2027. Climate risk was integrated into the macroeconomic scenario for the first time at full methodological maturity.

The aggregate results, published 1 August 2025:

The climate-attributable share of the CET1 depletion is 151 basis points of the total 370. Two-fifths of the system-level capital hit, in a single stress-test cycle, came from climate physical-risk channels.

This is the number that did not exist three years ago at this scale of attribution. The 2022 ECB climate stress test reported climate impacts in narrative terms with order-of-magnitude estimates. The 2025 test reports them in basis points alongside the other macroeconomic risks. The shift from narrative to numeric is a one-time threshold the supervisory framework has now crossed. Climate physical risk is now a quantified, comparable, capital-affecting input.

The 2026 thematic test moves to a reverse stress test on geopolitical risk, with climate physical-risk channels carried forward as a baseline assumption in the macro scenario. Climate is no longer the test's main subject. It is now the baseline.

The political-cycle reader cannot find the words "climate stress test" in the daily news flow of 2026. The supervisors' spreadsheet shows climate risk integrated into the baseline of the next test.

Move 4. Frank Elderson's May 2026 Oxford lecture: the language tells the story.

Frank Elderson has been the most consistent voice on climate at the ECB through the political shifts of 2024-2026. His 21 May 2026 lecture at Oxford is the cleanest single articulation of how the ECB now frames climate and nature risk in supervisory language.

Three observations from the text are worth pulling out, because they show how far the framing has matured.

First, the framing is not future-tense. Elderson does not say climate risks "may" or "could" affect bank balance sheets. He says they "already pose material risks". The hedged conditional that characterised central-bank climate speech in 2020-2022 has been replaced with declarative present-tense statements about current risk.

Second, nature risk has been bundled into the supervisory frame at parity with climate. Elderson treats biodiversity loss and ecosystem degradation as financial risks transmitted through the same channels as climate risk: collateral value, loan quality, sovereign risk, litigation exposure. The 2024-2025 distinction between climate (well-developed supervisory language) and nature (early-stage supervisory language) has compressed. By mid-2026 they share a single framework.

Third, Elderson is explicit about the boundary: "Central banks are policy takers, not climate and nature policymakers, and primary responsibility rests with elected governments." This is not a retreat. It is the supervisory framework's correct positioning: the ECB does not make climate policy, but it will price climate risk into the operations it does control (collateral, liquidity, capital requirements, supervisory review).

The political-cycle reader looks at speeches like this for political signals: is the ECB pulling back, leaning in, hedging? The supervisors' spreadsheet shows that the language has become more precise, the scope has widened to include nature, and the boundary has been clarified to remove political ambiguity.

Move 5. Implications for banks, allocators, and the climate-capital stack.

Five practical implications for the readers this publication serves.

For European banks, the supervisory expectations are now embedded in three concurrent channels: capital requirements via stress-test calibration, funding costs via the collateral framework, and supervisory dialogue via the ECB's 2024-2025 thematic review. Banks that read the 2025 Fed exit as a signal to deprioritise climate-risk management are misreading the channels they actually have to comply with. ECB-supervised banks face a tightening climate-risk regime regardless of what US regulators do.

For US banks operating in Europe, the asymmetry is real and growing. A US bank with material euro-area operations now faces ECB collateral haircuts on its European corporate-bond holdings, ECB stress-test inclusion if it crosses the SSM thresholds, and ECB supervisory dialogue on its climate-risk practices. The Fed's withdrawal from the NGFS does not insulate US bank European subsidiaries from European supervision. The compliance load has not decreased; it has only become asymmetric across jurisdictions.

For asset managers holding euro-area bank debt or equity, the climate-supervisory regime is now an input to the credit story. Banks more exposed to transition-risk-loaded corporate-bond holdings face a higher cost of ECB liquidity from H2 2026. Banks with weaker climate-risk management practices face supervisory review action under the 2024-2025 thematic findings. Both feed through to credit spreads and equity valuations on a multi-year horizon.

For ESG analysts, the central-bank channel is now a load-bearing source of forward-looking risk signal that does not show up in voluntary disclosure regimes. Bank-level supervisory expectations precede public disclosure by 12-24 months in many cases. An analyst who tracks only the disclosure regime is reading the climate-risk story with a lag.

For the climate-capital stack as a whole, the central-bank layer is now a durable backstop that survives political reversals. Equity markets reprice on political signal; bond markets and bank funding costs reprice on supervisory signal. The 2025-2026 episode demonstrated that supervisory work compounds across political cycles in a way disclosure regimes and voluntary frameworks do not. The climate-finance scaffolding that survives a Trump administration is the supervisory scaffolding, not the disclosure scaffolding.

The political-cycle reader keeps watching the front page. The supervisors' spreadsheet keeps compounding underneath it.

The Fed exit was real. The end of central-bank climate work was not. The quiet layer survived because it was never primarily about politics. It was about the daily plumbing of how banks fund themselves, how supervisors test capital adequacy, and how central banks price the risks they accept onto their balance sheets. That plumbing kept being upgraded in 2025-2026, while everyone with a microphone was talking about the political reversal.

The political cycle says US central-bank climate work is over. The supervisors' spreadsheet says European central-bank climate work just became load-bearing.