The ACA Subsidy Cliff Is Back for 2026

For the years between retiring and turning 65, health coverage usually comes from the individual marketplace, and its cost is decided by a subsidy — the premium tax credit at 26 U.S.C. §36B. How much of that credit a household gets, or whether it gets any, turns on one measured number.

That number is not spending. It is not net worth. It is modified adjusted gross income, and a household can spend a great deal while reporting relatively little of it, or spend modestly and report all of it, depending entirely on which accounts the money comes out of.

There is a second thing to know before the mechanics, and it is the reason this page exists now rather than a year ago. The hard cap on eligibility was switched off for the 2021 through 2025 tax years, and it is switched back on for 2026. A great deal of accurate writing from those years describes the credit tapering smoothly away. That description was correct when it was written and it is not correct for this plan year.

What is the ACA subsidy cliff?

It is an eligibility test with a hard edge rather than a slope.

Section 36B gives the credit to an “applicable taxpayer,” and 26 U.S.C. §36B(c)(1)(A) defines one as a taxpayer “whose household income for the taxable year equals or exceeds 100 percent but does not exceed 400 percent of an amount equal to the poverty line for a family of the size involved.”

Read that as a definition of who is eligible at all, because that is what it is. Above the ceiling, the household is not an applicable taxpayer. It is not a taxpayer receiving a smaller credit. The credit is not reduced; the category does not apply.

That is what the word cliff is doing. A dollar of income that crosses the line does not cost a dollar of subsidy. It costs the entire subsidy, which is why the last dollar under the ceiling and the first dollar over it can be separated by an amount of money wildly out of proportion to the difference between them.

Did the cap really come back for 2026?

Yes, and by expiry rather than by any new legislation.

The suspension lived at §36B(c)(1)(E), and it was written with its own end date built in. It reaches a “taxable year beginning after December 31, 2020, and before January 1, 2026.” A tax year beginning in 2026 is outside that window, so the provision no longer reaches it and the ceiling in 26 U.S.C. §36B(c)(1)(A) applies again on its own terms.

This matters for how you read anything else on the subject. During the suspension the credit really did phase out gradually — a household above the old ceiling paid a capped share of its income toward a benchmark plan rather than losing the credit outright. Articles, calculators and videos produced in that period describe a real rule that was really in force. It has simply run out.

So the practical instruction is narrow and worth stating plainly: check the date on anything you read about this, including anything on this site, and treat guidance written between 2021 and 2025 as describing a rule that has since lapsed.

What counts as modified adjusted gross income here?

Adjusted gross income, plus three add-backs, and it is not the same MAGI used elsewhere in retirement.

Start from adjusted gross income — the bottom line of the front of the return. For this credit, §36B(d)(2)(B) then adds back excluded foreign earned income, tax-exempt interest, and the portion of Social Security benefits that is not otherwise included in gross income.

That last add-back deserves emphasis, because it is the one that surprises people who have learned the rules for other income tests. The whole Social Security benefit counts here, not just the taxable part. A household that has arranged its affairs so that most of its benefit escapes income tax has changed nothing about this measure.

What reaches the measure, then, includes withdrawals from traditional IRAs and 401(k)s, realized capital gains, dividends, taxable interest, tax-exempt interest, pension income, and the full Social Security benefit.

What does not reach it includes qualified withdrawals from a Roth IRA, the return of your own principal when you sell from a taxable account, and cash moved out of savings. Those are not income. The gain on the sale is income; the basis is your own money coming back.

This is the entire lever, and it is why the same lifestyle can produce very different measured incomes. It is also why the lever only exists if the accounts do. A household holding everything pre-tax has no combination to choose from.

If it helps to have the numbers in one place, I keep them all in the Ultimate Retirement Guide.

Why is a dollar threshold the wrong thing to memorize?

Because the line is defined as a percentage, and the thing it is a percentage of changes every year and with every household size.

The statute fixes the multiple. It does not fix a dollar amount. The poverty guideline it multiplies is republished by the Department of Health and Human Services each January, and it is a different figure for a one-person household than for a two-person one.

So a dollar threshold quoted anywhere — including in a video, including one of ours — is a snapshot of one household size in one year. It goes stale on a schedule and it was never right for a household of a different size. That is why no dollar figure appears on this page. The multiple is the durable part, and the arithmetic for your own household is a lookup rather than a memory.

Why does this land hardest between retiring and 65?

Because those are the years with both the most control over income and the most temptation to use it.

A household in that window usually has no wages, may not have claimed Social Security, and is not yet subject to required withdrawals. Nearly every dollar of measured income in those years is there because someone chose to realize it. That is unusual, and it is genuinely valuable.

It is also exactly when the standard advice says to do the things that create measured income. Converting to a Roth while the brackets are low, realizing gains while there is room beneath a threshold, filling out a bracket that will not be available later — all of these are reasonable, and all of them raise the number this credit is tested against.

So the cliff does not argue against any of that. It prices it. In the years before Medicare, a conversion is competing not only against a future tax rate but against a subsidy that disappears at a fixed edge, and the comparison has to be made with both on the table. That is a household-specific calculation, and which way it comes out depends on the size of the subsidy at stake, the size of the conversion, and how many such years remain.

Two further timing consequences follow from the same fact, and both are worth knowing before they are decided by accident:

Claiming Social Security early adds to this measure and does not come back off. The whole benefit counts, and the decision is durable in a way a single year’s withdrawal is not. A claim made for good reasons at the earliest age available can also be a permanent addition to the number this credit is measured against for every remaining pre-Medicare year.

A single sale can decide the year. Realized gains are income here, so a rebalance, a concentrated position finally sold, or a property disposed of can move a household across the line on its own — and the crossing is discovered at filing, not at the moment of the trade.

How is this different from IRMAA?

The shape is the same and almost everything else is different, which is what makes conflating them expensive.

Both are cliffs. Both are decided by a modified adjusted gross income rather than by taxable income, and in both a small crossing costs the full step. Beyond that the resemblance ends.

IRMAA applies to people enrolled in Medicare, reads a return from two years earlier, and reduces a subsidy by a defined step. It is reassessed every year, so a single year over a line produces a single year of surcharge. The mechanics, the appeal routes and the lookback are set out in full on the IRMAA page.

The premium tax credit applies before Medicare, is reconciled against the current year on the return you file for it, and is not stepped at all above the ceiling — it ends. And because it is reconciled, a household that estimated its income low and received the credit in advance can face repayment of what it already received.

The practical difference is timing. IRMAA is decided by a year you can no longer change. This is decided by the year you are living in, which means it can still be managed while the year is open, and it means an unnoticed crossing surfaces as a balance due rather than as a premium adjustment.

What should you actually do with this?

Three things, none of which require a decision today.

Find out where your household’s line is, using your household size and the current year’s guideline rather than any figure you have read. It is one lookup, and it is the only number here that is specific to you.

Estimate the measure, not your spending. Add up what you expect to realize this year — withdrawals, gains, dividends, interest including tax-exempt interest, pension, and the full Social Security benefit if you are claiming. Spending is not in that list.

Do it while the year is still open. This is the rare threshold that can still be influenced after you know roughly where you stand, because most of what feeds it is discretionary and the reconciliation happens at filing. A December estimate is useful in a way a December IRMAA estimate is not.

If the answer is that you sit comfortably on one side of the line, then this is a constraint you can stop thinking about. If it is that you are near it, the question of what a conversion is worth against a subsidy is a real one with a real answer, and it depends on figures that are yours rather than anyone’s average.