The AOCI opt-out is the election that lets US Category III and IV banking organizations exclude accumulated other comprehensive income (AOCI) from common equity tier 1 (CET1). In practice it keeps unrealized losses on available-for-sale (AFS) securities out of regulatory capital ratios. The Basel III endgame re-proposals issued on 19 March 2026 would remove the opt-out and require all Category I-IV firms to recognize most AOCI in CET1, phased in over five years. As of mid-October 2026 the rule has not been finalized.
The US has no capital charge for banking-book interest rate risk, so this plumbing decides whether rate losses reach capital ratios at all. Below: how the AOCI filter works, why it mattered at Silicon Valley Bank, and what the re-proposal would change.
When interest rates rise, the market value of fixed-rate securities falls. For securities classified as available-for-sale (AFS), those unrealized gains and losses accumulate in AOCI, a component of equity. If AOCI counts in regulatory capital, those losses show up in capital ratios as rates move. If not, they are visible in the accounts but invisible to the ratios supervisors and markets watch.
The AOCI opt-out (often called the AOCI filter) is the second case: an eligible bank makes a one-time election to neutralize most AOCI for capital purposes, so AFS losses are disclosed but do not reduce CET1.
Securities classified as held-to-maturity (HTM) sit outside this question entirely. They are carried at amortized cost, so their unrealized losses are recognized nowhere, neither in AOCI nor in capital.
Under the 2019 tailoring framework, the answer depends on the firm's tailoring category. Category III broadly covers firms with $250 billion or more in total assets (or elevated risk indicators), and Category IV covers firms with $100 billion to $250 billion that do not fall into a higher category.
| Category | AOCI in CET1 today | Under the March 2026 re-proposals |
|---|---|---|
| Categories I-II (the largest, most complex firms) | Must recognize AOCI in CET1 | Unchanged: recognition continues |
| Categories III-IV | May opt out of AOCI recognition | Must recognize most AOCI in CET1, with a five-year transition |
For Categories I and II, a rate rise that devalues the AFS portfolio flows straight into CET1. For Categories III and IV that have opted out, it does not.
Silicon Valley Bank was a Category IV firm that had elected the opt-out, and that election is a central piece of how its interest rate exposure stayed off the capital radar. SVB carried billions of unrealized losses with pristine reported ratios: its AFS losses were neutralized by the opt-out, and its much larger HTM losses, which the opt-out debate does not even reach, were recognized nowhere at all.
The capital blindness compounded the measurement failures (a short-run earnings focus, de-emphasized economic value, assumptions changed when limits were breached) that make SVB the defining case study of the US framework's failure mode, covered in detail in our SVB IRRBB case study. The Fed's Barr report (April 2023) documented the whole chain. Of the structural fixes discussed after SVB, the AOCI provision is the one still moving through rulemaking.
A natural objection: the annual supervisory stress test scenarios specify Treasury yields, term spreads and mortgage rates. Does that not capture what the AOCI opt-out hides?
It does not, for two reasons. First, the exercise is a capital test, not an interest rate risk test: its scenarios are macro narratives, typically a recession with rate cuts, not a grid of rate shocks designed to probe banking-book exposure. Second, and decisively here, unrealized losses flow into the stress results only to the extent capital rules recognize them. A Category IV bank that has opted out of AOCI recognition carries its AFS losses through the stress test the same way it carries them through reported ratios: invisibly.
The closest the US has come to a system-wide banking-book rate exercise sat outside the core test: the Fed's 2024 exploratory analysis (elements announced in February 2024, aggregate results published in June 2024) included funding stresses in which a large share of deposits repriced rapidly under different rate paths. It had no capital consequence, but it is the nearest US relative of the EU's supervisory IRRBB exercises.
The July 2023 Basel III endgame proposal would have extended AOCI recognition to Categories III and IV. It stalled. The provision survived: the re-proposals issued on 19 March 2026 by the Fed, FDIC and OCC carry it forward. In brief:
The persistence is telling. Other post-SVB measures, such as the 2023 long-term debt proposal for large regional banks, have not advanced since the 2025 change in supervisory leadership. The US answer to SVB has been to fix supervision's reflexes and the AOCI loophole, not to write an interest rate risk capital rule.
According to the FDIC Quarterly Banking Profile for the second quarter of 2026, unrealized losses on securities across insured institutions totaled $326.7 billion (5.5% of amortized cost), up $1.6 billion from the prior quarter but down $68.6 billion from a year earlier. Of that, $109.8 billion sat in AFS portfolios and $216.9 billion in HTM portfolios.
The split matters. Ending the opt-out brings AFS losses at Category III and IV firms into CET1; banks below $100 billion are outside the mandate unless they elect ERBA. The HTM piece, roughly two-thirds of the total, stays outside capital under any version of the rule.
Not formally, but if adopted it gives the US, in effect, a partial Pillar 1 treatment of banking-book rate risk. The comparison with the EU and UK shows the design choices.
It is narrower than an IRRBB capital charge. AOCI recognition captures AFS securities only: not loans, not deposits, not the structural rate position of the balance sheet that an economic value (EVE) measure captures. And HTM remains outside: a bank that classifies its bond portfolio as held-to-maturity keeps its unrealized losses out of both AOCI and capital, a gap supervisors watch but no rule addresses.
It is also, in one respect, harder than the European approach. The EU's supervisory outlier test flags banks whose economic value sensitivity breaches thresholds, with discretionary consequences. The UK capitalizes IRRBB through supervisor-set Pillar 2A add-ons, again via supervisory judgment. AOCI recognition is automatic: rates move, AFS values move, CET1 moves, with no supervisor in the loop. Within its narrow perimeter, it is the most mechanical rate-risk-to-capital link in any major jurisdiction.
The final rule, its effective date and therefore the start of the five-year clock are all still pending. For Category III-IV treasurers, the consequences are already visible in outline:
For the complete picture, from EVE and NII sensitivities and the six supervisory shock scenarios to outlier tests and how Basel, the EU, the UK and the US differ, see Mirai RiskTech's guide IRRBB regulation across Basel, the EU, the UK and the US: a complete reference.