The 2023 US Basel III endgame proposal would have raised CET1 requirements for affected holding companies by about 16%. The re-proposal that replaced it on 19 March 2026 would cut them: by 4.8% for Categories I and II, 5.2% for Categories III and IV and 7.8% for banks below $100 billion, on Fed staff's cumulative estimates. Comments closed on 18 June 2026. As of mid-October no final rule has been issued, and the agencies are reported to be aiming for a vote in December.
The re-proposal is not a softened version of the original. The expanded risk-based approach (ERBA) would be mandatory only for Categories I and II, it would run as a single stack rather than a dual one, and the estimated effect runs in the opposite direction. Almost every design choice in the new package is a response to what happened to the old one, so that is where to start.
The July 2023 notice of proposed rulemaking (NPR) was maximalist on every dimension at once.
Scope. ERBA would have applied to all banks in Categories I to IV of the US tailoring framework, that is, every banking organization with $100 billion or more in assets.
Method. Standardized credit risk with no internal credit models, FRTB-based market risk, a standardized operational risk charge with an internal loss multiplier, and CVA.
Architecture. A dual stack: a bank would compute both ERBA and the existing standardized approach and be bound by the lower ratio, which preserved the Collins Amendment floor.
Definition of capital. It would also have required Categories III and IV to recognize AOCI in CET1, closing their opt-out.
The agencies' own estimate was an aggregate CET1 increase of roughly 16% for affected holding companies, and about 19% for the G-SIBs.
It met the most effective industry campaign in modern US financial regulation: public advertising, an unprecedented comment volume and dissents inside the agencies themselves.
In September 2024 the Fed's then vice chair for supervision previewed a re-proposal that would have roughly halved the impact. That re-proposal never appeared. The 2025 change of administration and agency leadership buried the original, and the March 2026 package formally rescinded it.
Proposed, contested, previewed as revised, never revised, then rescinded: 32 months from July 2023 to March 2026.
Three coordinated proposals: the Basel III (ERBA) NPR and the standardized approach NPR, issued jointly by the Fed, FDIC and OCC, and a G-SIB surcharge NPR issued by the Fed alone.
Three changes from 2023, and the first two are structural.
Mandatory scope narrows to Categories I and II. Categories III and IV fall outside mandatory ERBA, and any other banking organization may opt in, giving four quarters' notice and accepting the same definition of capital, including AOCI recognition. Separately, the market risk framework would apply to Category I and II holding companies and to any firm with at least $5 billion of trading assets and liabilities or trading activity of at least 10% of total assets.
The dual stack becomes a single stack. For Category I and II firms the standardized approach would no longer apply and the advanced approaches would be removed from the rule altogether. Each firm would compute one set of risk-based ratios.
That is the most interesting design move in the package. A dual stack satisfies the Collins Amendment by computing both numbers. A single stack has to satisfy it another way, and the route appears to run through the opt-in: if any banking organization can elect ERBA, ERBA becomes part of the "generally applicable" requirements that Collins uses as the floor, alongside the revised standardized approach. That is a legal argument rather than a settled point, and worth reading in the NPR itself.
Calibration falls for lending, rises for trading. Credit risk weights become more risk-sensitive (loan-to-value for most real estate, repayment history for retail, assessed creditworthiness for corporates). The operational risk charge is set by business volume, with a reduction for investment management, custody and similar fee businesses, and the internal loss multiplier is dropped, which is equivalent to fixing it at one. That lands the US where the EU and UK already chose to be: loss experience informs supervision rather than capital.
No internal credit models under ERBA. Market risk keeps a models-based approach, now built on expected shortfall with liquidity horizons, with a standardized approach as a credible fallback. Client-facing derivatives would be exempt from the CVA charge.
This is the proposal with the widest reach and the least attention. It would be the first broad recalibration of the generally applicable rule since 2013: loan-to-value risk weights for residential mortgages, corporate exposures cut from 100% to 95%, unassigned assets from 100% to 90%, and mortgage servicing assets moved from a threshold deduction to a flat 250% risk weight for every bank, community banks included.
Note who that touches. The generally applicable rule is what most of the roughly 4,200 US insured institutions compute, all the way down below Category IV. So the package's relief extends well beyond the largest banks, and through a different instrument from the one that gets discussed.
Method 2 coefficients would be updated to reflect changes in the financial system and economy, with an automatic annual adjustment for economic growth and inflation from then on. Certain systemic indicators would be calculated as annual averages rather than year-end values, surcharges would move in 10 basis point rather than 50 basis point increments to soften cliff effects, and the short-term wholesale funding indicator would lose its RWA denominator. It replaces the 2023 surcharge proposal.
The averaging point addresses a specific behavior. A surcharge keyed to year-end indicators rewards managing the balance sheet down into the measurement date, and averaging removes that incentive without changing the underlying measure.
Fed staff's estimates, from the board memo of 19 March 2026, run negative for every population. They are cumulative and include the effect of the Fed's stress test proposals, which is why the totals differ from the instruments alone.
| Change in aggregate CET1 requirements | Categories I-II | Categories III-IV | Below $100bn |
|---|---|---|---|
| Basel III (ERBA) NPR | +1.4% | - | - |
| G-SIB surcharge NPR | -3.8% | - | - |
| Standardized approach NPR: risk weights | - | -6.1% | -7.8% |
| Standardized approach NPR: AOCI recognition | - | +3.1% | - |
| Stress test changes | -4.3% (market shock and op risk), +1.9% (other) | -2.2% | - |
| Total | -4.8% | -5.2% | -7.8% |
| 2023 proposal, for comparison | +16% aggregate (G-SIBs about +19%) | outside scope | |
Read the last column against the 2023 row. The 2023 package raised requirements and was aimed at the tailoring categories. The 2026 estimates fall furthest for banks that were never inside ERBA's scope at all, and that relief comes entirely from the standardized approach NPR.
Two caveats. First, for Categories I and II the Basel III NPR on its own raises requirements, mostly through trading, and the net relief comes from the surcharge and the stress test, against which staff explicitly calibrated the higher operational and market risk charges. Second, the Fed finalized its stress test changes on 30 September 2026, largely as proposed, saying they are not expected to materially affect aggregate capital requirements. Read the stress-test slice of the March figures as illustrative.
Less than the headline suggests, but not nothing. The ERBA NPR adds capital for trading books, and the standardized approach NPR keeps the 2023 idea of requiring Categories III and IV to recognize most of AOCI in CET1, phased in over five years, at an estimated +3.1% of CET1 requirements for those firms. It is the clearest regulatory legacy of SVB's unrealized losses. The scope is easy to overstate: it reaches Categories III and IV only, and the opt-out survives below Category IV.
Because two other jurisdictions have calibrated around it.
The EU postponed FRTB to 1 January 2027 and then adopted Delegated Regulation (EU) 2026/1221 on 4 June 2026, published in the Official Journal on 11 September 2026. From 1 January 2027 it applies temporary multipliers and operational relief that offset the FRTB increase in market risk capital until the end of 2029, citing the expected delays in other major jurisdictions. That sits on top of the CRR3 changes that took effect in January 2025. The UK delayed Basel 3.1 to 1 January 2027, citing the need for clarity about US implementation, and set its internal models approach for market risk at 1 January 2028, with a further consultation on IMA adjustments (CP9/26) in June 2026.
So what the EU does when its relief expires after 2029, and how the UK finalizes its IMA settings, both depend in practice on this rulemaking. A proposal with no final rule is the anchor for two other jurisdictions' calibration decisions.
That is the divergence era: not that the three frameworks differ, but that each conditions its moves on the others, and the one furthest behind has the most influence.
The rulebook is in two states: the 2013 rule is binding law, and its Basel III endgame replacement exists only as the March 2026 proposals.
In force since 2013 or thereabouts: the Basel II-vintage standardized approach with flat 100% corporates, advanced approaches for Categories I and II, the AOCI opt-out, the Collins floor, the 4% tier 1 leverage ratio and the SLR, the Method 2 G-SIB surcharge, and a countercyclical capital buffer that has never been switched on.
In force or final, delivered recently and outside the endgame:
Proposed, no final rules: the three March 2026 NPRs, now reportedly heading for a vote in December, with the agencies aiming to finalize by year-end. One older proposal is easy to lose: the August 2023 long-term debt proposal for large non-G-SIBs, which has not been finalized.
The second list is the one to watch. The endgame is unfinished, but US capital relief has been arriving through instruments that did not need it to move.
For the complete capital stack, from CET1, buffers and the output floor to Pillar 2 and MREL, and how Basel, the EU, the UK and the US implement them, see Mirai RiskTech's guide capital regulation across Basel, the EU, the UK and the US: a complete reference.