Quarterly taxes are not a quarter of your year
US estimated tax has unequal periods, a safe harbour instead of a quarter, and a base that isn't revenue. The arithmetic, worked, and what to pay each period.
Ask a US freelancer what to send the IRS on 15 June and most will describe a division problem: last year's tax bill, divided by four. That is the common mental model of estimated tax, and its premise fails in three separate places. The periods are not quarters. The amount is not a quarter of anything. The base is not your revenue.
Scope first. This is US federal income tax and self-employment tax; every state that levies an income tax runs its own estimated-payment calendar and safe harbour, none of which is covered here. None of it is tax advice either. It is how the arithmetic on the form works, so you know what to ask.
The periods run three months, then two, then three, then four
The 2026 payment dates are 15 April, 15 June, 15 September and 15 January 2027, per Form 1040-ES: four dates, spaced two, three and four months apart. The periods behind them match that spacing exactly, which is where "quarterly" stops being accurate.
The IRS states the periods as 1 January to 31 March (due 15 April), 1 April to 31 May (due 15 June), 1 June to 31 August (due 15 September), and 1 September to 31 December (due 15 January of the following year). Three months, two months, three months, four months. Schedule AI of Form 2210 states the same boundaries cumulatively, through 31 March, 31 May, 31 August and 31 December.
So the payment due on 15 June covers eight weeks of earning and the one due on 15 January covers seventeen. A project invoiced in early September has four months before its deadline; invoiced in April, six weeks. Which is why reserving tax monthly beats reaching for cash four times a year. Worklyn's tax set-aside calculator turns expected revenue, expenses and a country preset into a monthly figure; the payment dates are only when you hand some of it over.
Form 1040-ES says you can pay "all of your estimated tax by April 15, 2026, or in four equal amounts by the dates shown below." Equal amounts across unequal periods is not a contradiction, because the regular method never asks when you earned the money. It asks whether your running total clears a line by each date. That line is the whole subject.
The amount is whatever clears a safe harbour
Form 1040-ES sets the test in two parts. You must pay estimated tax for 2026 if you "expect to owe at least $1,000 in tax for 2026, after subtracting your withholding and refundable credits," and expect withholding and credits to be "less than the smaller of: a. 90% of the tax to be shown on your 2026 tax return, or b. 100% of the tax shown on your 2025 tax return." If your 2025 adjusted gross income exceeded $150,000 ($75,000 married filing separately), you substitute 110% for that 100%.
The word doing the work is smaller. You are not required to be accurate about 2026, only to clear whichever of the two is lower, and one of them is not an estimate at all: the tax shown on your 2025 return is already printed on a document you own, at Form 1040 line 24 with a few adjustments the form lists.
Take a 2025 return showing $14,200 of total tax on an AGI of $96,000, as an illustration. The harbour is $14,200, so $14,200 ÷ 4 = $3,550 on each of the four dates, and you are done thinking about it. Suppose 2026 goes much better and the eventual bill is $23,000. The harbour still holds: the remaining $8,800 is due on 15 April 2027 with no underpayment penalty, because the penalty is charged on the shortfall against the harbour rather than against the final bill. Had that 2025 AGI been $162,000, the multiplier becomes 110%: $14,200 × 1.10 = $15,620, or $3,905 a period.
Safe harbour therefore buys certainty about the penalty, not about the bill. Someone who grows 60% and pays the prior-year harbour all year is choosing to owe a large balance in April, which is a cash-flow decision worth putting through a runway calculation in February rather than discovering in April.
The base is not your revenue, and it isn't your profit either
Self-employment tax is 15.3%, being 12.4% for Social Security and 2.9% for Medicare, and it applies to 92.35% of your net earnings rather than all of them. Multiply the two: 0.153 × 0.9235 = 0.1413. That is 14.13 cents on every dollar of Schedule C profit, before a cent of income tax is calculated. On $70,000 of net profit, net earnings are $70,000 × 0.9235 = $64,645, and the self-employment tax is $64,645 × 15.3% = $9,891.
Two features of that number surprise people. The first is where it starts. Self-employment tax bites once net earnings reach $400, while income tax waits until you clear the standard deduction, $16,100 for a single filer in 2026. A freelancer with $12,000 of profit can owe roughly $1,700 in self-employment tax and nothing in income tax, which is where "I'm under the threshold" turns into a bill.
The second is where it stops. The 12.4% Social Security portion applies only up to $184,500 of combined wages and self-employment earnings for 2026, while the 2.9% Medicare portion has no ceiling. Wages fill that cap first, so a freelancer with $150,000 of W-2 salary has $34,500 of headroom before the marginal rate on profit drops from 15.3% to 2.9%. Half the self-employment tax then comes back as an above-the-line deduction, $4,945 in the $70,000 example, lowering income tax without touching self-employment tax. The qualified business income deduction, made permanent by the July 2025 law, behaves the same way: it sits further down Form 1040 and never reaches the self-employment tax worksheet.
Everything upstream of that 14.13 cents is your expense record, which is why deductions that survive scrutiny beat any planning trick; the four with the biggest 2026 numbers are in the deductions post. It also explains why the 90% route is harder than it looks: 90% of the current year's tax needs a defensible profit figure at four points in the year, which is a bookkeeping problem before it is a tax one. A monthly bookkeeping service closes your books and sends a report; the Bench alternative approach keeps the ledger beside the invoices and time logs, so the profit figure is current on the day you need it.
The three ways to satisfy the requirement
Route | What it needs from you | Effort | Who it suits |
|---|---|---|---|
90% of 2026 tax | A live estimate of full-year profit, self-employment tax and income tax, before each date | High, four times over | Income clearly falling, or predictable revenue |
100% of 2025 tax (110% if 2025 AGI > $150,000) | One number off your 2025 return, divided by four. That return must cover 12 months and show a liability | Lowest: one lookup, four transfers | Almost everyone whose income is flat or rising; any year you can't forecast |
Annualised income instalment (Form 2210, Schedule AI) | Books closed to 31 March, 31 May and 31 August; Schedule AI filed even when no penalty is due | Highest | Lumpy years: a mid-year start, or one big Q4 project |
You started freelancing in March, so what is due in June
This is the case general advice handles worst, and it has a clean answer. The prior-year harbour is computed from your prior-year total tax, not from your prior-year self-employment income. Someone employed for all of 2025 whose Form 1040 line 24 read $9,400 has a $9,400 harbour, or $2,350 per period, despite no freelance income last year at all. The harbour does not care where the tax came from.
Two conditions attach. Form 1040-ES requires that "your 2025 tax return must cover all 12 months," and tells you not to use the prior-year figure if you didn't file for 2025. A graduate in their first filing year has no prior-year harbour at all, and is thrown back on the 90% test or on annualising.
The timing has a trap in it too. Under the regular method a quarter of the harbour fell due on 15 April even though you began in March and earned almost nothing in the first period. That method ignores when income arrived, which is generous if you earn late and punitive if you earn nothing early. Form 1040-ES anticipates it, directing anyone with "a large change in income, deductions, additional taxes, or credits" after 31 March to the annualised income instalment method.
That method deserves one honest paragraph. It works, and it is the right tool for income that arrives in lumps: each instalment is recomputed from the profit actually earned by 31 March, 31 May and 31 August, so a year with nothing until July owes nothing until September. The cost is closing your books to three dates no other calendar uses, then filing Form 2210 including Schedule AI even if no penalty is owed. Most people who consider it are better served by the flat harbour and a worse cash position in April. Choose it when the alternative has you funding payments from money you have not been paid, which is what a year-round set-aside system exists to prevent.
Withholding is treated as paid evenly, and that changes the answer
One fact reorganises the whole subject for anyone with a job, a part-time contract or a salaried spouse. Estimated payments are credited on the date you make them. Withholding is not. The Form 2210 instructions state that "for withheld federal income tax and excess social security or tier 1 RRTA, you are considered to have paid one-fourth of these amounts on each payment due date unless you can show otherwise."
Read that literally. If $12,000 of federal income tax is withheld from wages in November, one-fourth of it, $3,000, is treated as paid on 15 April, another $3,000 on 15 June, and so on. Withholding is retroactive in a way estimated payments never are. A period you underpaid in April can be cured in October by changing a form.
The IRS says so plainly on its own estimated tax page: "If you receive salaries and wages, you can avoid having to pay estimated tax by asking your employer to withhold more tax from your earnings. To do this, file a new Form W-4 with your employer."
Work it through, as an illustration. A married couple, one salaried and one freelancing, face a 2025 total tax of $16,400 on an AGI above $150,000, so their harbour is $16,400 × 1.10 = $18,040. The salaried spouse's withholding will come to $13,000, leaving $18,040 − $13,000 = $5,040: either $1,260 on each of four dates, or one W-4 change adding about $5,040 across the remaining pay periods. The second needs no vouchers, no calendar and no memory in January, and it works even if you file the W-4 in September. The "unless you can show otherwise" clause cuts both ways, since you can also elect to have actual withholding dates counted, but the even split is the default.
None of which helps if there is no wage income in the household. Wholly self-employed, the vouchers are the only mechanism, and holding the money between dates is the actual skill. It is also where being paid with no form attached, now that the 1099 threshold has risen, makes your own records the only record. Worklyn's workspace for freelancers keeps invoices, bank feed and profit in one place for that reason.
The rule, given your two numbers
You need two facts: your 2025 AGI, and whether any federal income tax will be withheld anywhere in your household this year.
If your 2025 return covered twelve months and showed a liability, you have a prior-year harbour; if not, go straight to the 90% test or annualise. Given that, set your target H as the total tax on your 2025 Form 1040 line 24, multiplied by 1.10 if your 2025 AGI exceeded $150,000 ($75,000 married filing separately) and by 1.00 otherwise. Subtract expected 2026 withholding W. If W is at or above H, you owe no estimated tax and can stop. Otherwise pay (H − W) ÷ 4 on 15 April, 15 June, 15 September and 15 January; where a date has passed, catch up on the next one and cure the rest with a W-4 change if wages exist.
Leave that rule in two cases. If you confidently expect 2026 tax below H, the 90% test is the smaller number and you can pay 90% of your own estimate, accepting the forecasting work. And if income is so back-loaded that even the smaller target has you paying in April out of money arriving in August, annualise. Everyone else is paying a quarter of a number they already have, on four dates that are not quarters, against a bill that will differ from it. That gap is the method working, not failing.
Worklyn's CFO Mode tracks profit as it accrues and shows what is safe to spend against upcoming obligations, so the money for 15 September is identifiable in June rather than hopeful.