The ACA subsidy cliff is back for 2026

The enhanced premium tax credits expired 2025-12-31. One dollar over 400% of the poverty line now costs an early retiree the entire subsidy — and the repayment cap is gone too.

For five years there was no subsidy cliff. There is one again. If you are retiring before 65 and buying your own health insurance, this is the single highest-stakes number in your plan. This spoke sits under the Retirement hub.

For the full menu of coverage options — COBRA, a spouse's plan, Medicaid, part-time work — see health care before Medicare. This page is about the cliff itself.

What expired, and what replaced it

The American Rescue Plan Act (2021), extended by the Inflation Reduction Act, temporarily rewrote the premium tax credit. Those enhanced rules expired on December 31, 2025 and were not extended. Plan year 2026 runs on the original, pre-2021 statute.

Precisely two things changed:

1. The 400%-of-FPL income limit came back. From 2021 to 2025 there was no upper income limit — a household above 400% of FPL could still receive a credit. Under current law, above that line the credit is zero.
2. The applicable percentages went back up. The enhanced rules capped the benchmark silver premium at 0% of income at the bottom of the scale and 8.5% at the top. The reverted table runs from 2.10% to 9.96%.

What did not happen: the cap did not disappear. Below 400% of FPL you still have one. The Congressional Research Service puts it plainly — on expiration, "the maximum income limit of 400% of the FPL would be reinstated and the applicable percentages would revert to higher levels resulting in lower subsidy amounts." Reverted, not removed.

Source: IRS Rev. Proc. 2025-25.

If you take one thing from this page: a household at 250% of FPL still gets help. Anyone telling them "the subsidy is gone" is costing them money.

The cliff in dollars

Take a single 60-year-old. The cliff sits at $62,600 of ACA MAGI.

At $62,600 — exactly 400% of FPL. They are in the top band, so their benchmark silver plan is capped at 9.96% of income. They pay $6,235 for the year. The premium tax credit covers the rest.

At one dollar more. The credit is zero. They pay the full national-average benchmark premium: $15,914.

One dollar of income costs roughly $9,679. The effective marginal tax rate on that dollar is not 22% or 37%. It is somewhere around one million percent. There is nothing else like it in the tax code.

The practical consequence is a dead zone above the cliff. Cross it, and you are worse off than you were until your income climbs back by roughly the whole lost credit — about $9,679 of extra MAGI, and more once you pay income tax on those extra dollars too. Earning an extra $4,000 in that zone makes you poorer. Not metaphorically. Arithmetically.

A caution on precision: $15,914 is a national average for a 60-year-old. Benchmark premiums vary substantially by county, and they rise steeply with age — a 64-year-old faces a bigger number than a 60-year-old, so a bigger cliff. Use the bridge tool with your own figures, and get a real quote for your county. The shape of the problem is universal; the exact dollars are local.

The cliff by household size

Plan-year 2026 is measured against the 2025 HHS poverty guidelines (48 states + DC).

Each additional person in the household adds $5,500 to the poverty line, and therefore $22,000 of extra cliff headroom.

The version of this that catches couples: the year one spouse turns 65, they move to Medicare and drop off the marketplace plan. The household's tax family may not change, but the coverage family does — and the plan you are buying gets cheaper while the cliff arithmetic shifts. Model that year separately. It is rarely the same answer as the year before it.

The repayment cap is gone — and almost nobody has noticed

This is the part that is missing from essentially every article on the subject, including the ones that correctly explain the cliff.

Premium tax credits are usually paid in advance, monthly, straight to your insurer, based on the income you estimated at enrolment. At tax time you reconcile the estimate against reality on Form 8962. Estimate too low, and you took credit you were not entitled to.

Until 2025, a cap limited how much of that you had to pay back — a table of a few hundred to a few thousand dollars, depending on income and filing status, and it applied only below 400% of FPL. That cap no longer exists.

The One Big Beautiful Bill Act, section 71305, repealed IRC §36B(f)(2)(B) for tax years beginning after December 31, 2025. The IRS says so twice, in two separate primary documents:

Section 71305 of the OBBBA removes § 36B(f)(2)(B), which limited the tax increase from excess advance payments for certain households, effective for taxable years beginning after December 31, 2025. Accordingly, the inflation adjustment to § 36B(f)(2)(B) is removed from this revenue procedure.

— IRS Rev. Proc. 2025-32, §3.04

There is no repayment cap for tax years after 2025. For tax years after 2025, you must repay the full amount by which your advance credit payments exceed your Premium Tax Credit.

— IRS Fact Sheet FS-2025-10, Q31, updated December 23, 2025

Read those together with the cliff and you get the genuinely dangerous scenario:

You estimate $62,600 of income for 2026 and take the advance credit — roughly $9,679 across the year, paid to your insurer month by month. In November a mutual fund throws off an unexpected capital gains distribution. Your actual MAGI lands just over the cliff. You were never eligible for any credit. So at filing you repay all of it, in one lump, with no cap. A fund distribution you did not choose and could not decline just produced a five-figure tax bill.

Two defences:

• Update the marketplace when your income changes. Not at year end — when it happens. The advance payment adjusts and the reconciliation shock shrinks.
• Consider taking no advance credit at all. You may decline APTC and claim the whole credit on your return instead. You fund the premiums yourself during the year and get the credit as a refund, but you cannot be asked to repay something you never received. For anyone with lumpy or unpredictable income — a business sale, big taxable-account distributions, variable consulting work — this converts an uncapped clawback into a cash-flow problem. That is a trade many early retirees should take.

If a calculator shows you a repayment capped at a few hundred dollars for 2026, it is running on pre-2026 law. Check it against Q31.

The playbook for 60 to 65

The order matters, because the cheap levers should be exhausted before the expensive ones.

1. Know your two lines before you do anything. The floor (100% of FPL) and the cliff (400%). Everything below is about landing between them.

2. Spend from Roth and from basis first. Neither touches MAGI. Selling $50,000 of a taxable holding with a $40,000 cost basis creates $10,000 of MAGI, not $50,000 — only the gain counts. Roth withdrawals create none at all. This is the cheapest lever there is and most people underuse it.

3. Harvest losses against gains. You control realisation timing. Net capital losses offset gains fully and up to $3,000 of ordinary income per year, with the rest carried forward. Use specific-lot identification so you are choosing which gain you realise, not letting the broker choose for you.

4. Use the above-the-line deductions. HSA contributions reduce MAGI dollar for dollar — and from January 1, 2026, bronze and catastrophic exchange plans count as HSA-qualified. Deductible IRA contributions and a SEP-IRA or solo 401(k), if you have any self-employment income, do the same at a larger scale.

5. Only then think about Roth conversions — and think hard.

Roth conversions and the cliff

A Roth conversion is MAGI, in full, in the year you do it. Convert $30,000 while sitting $9,679 below the cliff and you have not made a tax-efficiency trade. You have bought a tax bill and forfeited the entire premium tax credit.

The honest way to think about it: the conversion's real cost is the income tax plus the credit you destroy plus, from 63, the IRMAA surcharge two years later. Only compare that total against the future tax you are avoiding.

Which produces a genuinely counter-intuitive rule. If you are already over the cliff, converting more is comparatively cheap. The credit is already gone; there is no second cliff to fall off. So the worst possible place to be is just barely over — you paid the full price of the cliff and got nothing for it. Either stay clearly under, or, if you are going over anyway, go over decisively and convert a meaningful amount while you are there.

The Roth conversion tool prices the cliff and IRMAA together, which is the only way the comparison means anything.

IRMAA starts biting at 63

Medicare's income-related surcharge looks back 2 years. Your 2026 income sets your 2028 premium. So from age 63, a conversion is simultaneously an ACA-cliff decision and a Medicare-premium decision.

The first IRMAA tier starts at $109,000 of MAGI for a single filer and $218,000 jointly, and crossing it costs about $1,148 per person per year. IRMAA is a cliff too — a smaller one, and it repeats at each tier.

Between 63 and 65 you are threading both needles at once. That is the hardest income-planning window most people will ever face, and it lasts exactly two years.

Do not fall off the other side

The cliff pushes people to cut income. Cut too far and you hit the floor.

Below 100% of FPL — $15,650 for one person, $21,150 for two — you generally get no marketplace premium tax credit at all. In the 40 states and DC that expanded Medicaid, you land on Medicaid instead, below 138% of FPL. In the ten states that have not expanded, you can be too poor for a subsidy and too well-off for Medicaid, and qualify for nothing at all.

An early retiree spending from Roth and basis can have a MAGI near zero while living comfortably. That is not a win. Full detail in health care before Medicare.

The target is a band, not a minimum.

This law is live — check it before you act

Everything above is current law as verified on 2026-07-11. It may not survive the year.

• The enhanced credits lapsed 2025-12-31.
• In January 2026 the House passed a three-year extension. The Senate did not take it up.
• A bipartisan Senate compromise — shorter extension, income caps, minimum premium payments — was under discussion but had not passed.
• Nothing has been enacted. The cliff is the law today.

If an extension does pass, it could apply retroactively to plan year 2026, which would change the arithmetic on this page after you had already made decisions on it. That cuts both ways: it is an argument for not doing anything drastic and irreversible right now purely to duck the cliff, and it is emphatically not an argument for assuming rescue. Congress has already let the deadline pass once.

We mark this figure volatile in our source file and re-check it at every publish. If you are reading this well after July 2026, verify against KFF and HealthCare.gov before relying on it.

Key takeaways

• The cliff is real and it is back. Above 400% of FPL — $62,600 single, $84,600 for two — the premium tax credit is zero. One dollar destroys roughly $9,679 for a single 60-year-old.
• But the credit did not vanish below the line. The benchmark premium is still capped, from 2.10% to 9.96% of income. Do not let anyone tell you there is no help.
• The repayment cap is gone for 2026 (OBBBA §71305). Underestimate your income and you repay every dollar of advance credit, uncapped. Declining the advance credit removes that risk entirely.
• A Roth conversion is an ACA decision before 65 and a Medicare decision from 63. Price all three costs, or do not price it at all.
• There is a floor as well as a cliff. Aim for the band between them.

Last verified 2026-07-11. An extension of the enhanced credits was live in Congress and could change this page retroactively.

Educational only — not financial, tax, or medical advice.