Roth conversion or ACA subsidy? Before 65, you usually can't have both

The Roth conversion ladder has one requirement: low-income years to convert in. If you retire before 65, you have another thing that wants those exact same years: the premium tax credit that makes health insurance affordable until Medicare starts. Both strategies consume the same scarce resource, which is room at the bottom of your tax return. Most people planning an early retirement discover this conflict late, usually after modeling one strategy in isolation and being very pleased with it.

This piece is the decision math for the bridge years: when the conversion wins, when the subsidy wins, and why 2026 made the collision much more expensive than it used to be.

Both strategies are MAGI strategies

A Roth conversion is taxed as ordinary income in the year you convert. That income lands in your modified adjusted gross income, and MAGI is exactly the number the ACA uses to size your premium tax credit. There is no separate ledger. Every dollar you convert is a dollar of MAGI, indistinguishable from wages as far as the subsidy formula cares.

The subsidy formula works backward from the benchmark plan: the second-cheapest silver plan on your exchange. The law says your share of that benchmark premium is capped at a percentage of your income, and the credit pays the rest. Lower MAGI, bigger credit. Below 250% of the federal poverty level, silver plans also carry cost-sharing reductions, so a low MAGI buys you lower deductibles on top of lower premiums.

What changed in 2026: the cliff is back

From 2021 through 2025, enhanced credits removed the income ceiling: above 400% of the poverty level, your benchmark premium was capped at 8.5% of income no matter how high your income went. Those enhancements expired at the end of 2025. In 2026 the original structure is back, and it does not taper. At 400% of the federal poverty level the credit simply ends.

For 2026 coverage the relevant poverty guideline is $15,650 for a single person plus $5,500 for each additional household member. So the cliff sits at $62,600 of MAGI for a single person and $84,600 for a couple. These are edges, not slopes. One dollar of MAGI past the line and the entire credit is gone for the year.

A $1 Roth conversion can cost $20,000. Not marginal tax on the dollar. The whole subsidy, forfeited.

Run the numbers for a couple, both 60, in a typical market where the benchmark silver plan for two people their age runs around $2,400 a month. At $84,600 of MAGI their share of the benchmark is capped just under 10% of income, roughly $8,000 a year, so the credit covers about $20,000. At $84,601 of MAGI the credit is zero and they pay the full $28,800 themselves. That is the steepest marginal tax rate most households will ever face, and it is triggered by the last dollar of an otherwise sensible Roth conversion.

The decision math, per converted dollar

A conversion is a bet that today’s tax rate on the converted dollar beats the future rate you would have paid withdrawing it. In the bridge years, the honest version of “today’s rate” has to include the subsidy you give up:

• Below the cliff, the subsidy phases out gradually. As MAGI rises toward 400% FPL, your capped share of the premium rises with it. Each converted dollar costs its bracket rate plus a few cents of shrinking credit. Converting in the 10-12% bracket often still wins against a future 22-24% withdrawal, even net of the phase-out.
• At the cliff, the cost is discontinuous. The dollar that crosses the line carries the entire remaining credit as its price. No future-bracket argument survives a five-figure cost on one dollar. If you convert at all, you convert to a ceiling set safely below the cliff, with room for the dividend or fund distribution you forgot about.
• Below 250% FPL, count the cost-sharing too. Conversions that push you from 240% to 260% of FPL don’t just trim the premium credit. They can move you to a plan variant with a deductible thousands of dollars higher. If someone in the household actually uses medical care, that is real money.

Three ways people thread the needle

Fill to the cliff, not the bracket.

The standard ladder advice is “fill the 12% bracket.” For a couple, the 12% bracket runs well past $84,600 of taxable income. In the bridge years the binding constraint is usually the subsidy cliff, not the bracket edge. The planning number to write down is 400% of FPL minus everything else that lands in MAGI, minus a safety margin.

Alternate years.

Take the subsidy in even years with near-zero conversions, then do one large conversion in each odd year and pay full freight for coverage that year. Two years of forfeited credit hurts less than five, and the conversion years can be sized into the 22% or 24% bracket where the per-dollar math still clears. Whether the alternating pattern beats steady under-the-cliff conversions depends on your balances, your state, and your premiums. It is a modeling question, not a slogan question.

Spend basis, convert the difference.

Living expenses drawn from taxable-account principal barely register in MAGI (only the gain portion does). Retirees with large taxable accounts and high basis can eat from basis, keep MAGI low, and use the freed-up room under the cliff for conversions instead of groceries. This is the quiet advantage of the taxable bucket that pure tax-deferred savers don’t have.

The window is narrower than it looks

One more date matters. Medicare premiums carry their own means-testing, IRMAA, and it looks back two years. Conversions at 63 and 64 set your Medicare premiums at 65 and 66. So the conversion-friendly years are not “everything before 65.” They are the years after work income stops and before the IRMAA lookback begins, with the ACA cliff patrolling all of them. For someone retiring at 58, that is a five-year window in which every strategy in this post is competing for the same MAGI.

Where Thermal fits

Thermal projects your bridge-years health coverage from your income, state, and household size, alongside a retirement engine that models conversions against RMDs, IRMAA, and the tax rules of 51 jurisdictions. Seeing the subsidy math and the conversion math on the same screen, from the same numbers, is the whole point. The strategies above stop being folklore once you can check them against your own plan.