Startup Taxes, Stage by Stage

Startup tax planning has a strange shape: the highest-stakes decisions come earliest, when the company is worth the least and founders are paying the least attention. These are the five.

1. Entity choice. Delaware C corp for the venture path, LLC for many bootstrapped paths. This one decision controls investor eligibility, how losses flow, and whether QSBS is even possible.
2. 83(b) elections. Thirty days from each restricted stock grant, no extensions, no do-overs. The cheapest filing you will ever make relative to what it protects.
3. Accounting method. Cash or accrual, decided with your first return. Most early startups qualify for the simpler cash method; investors will eventually want accrual-basis books regardless.
4. The R&D credit election. Loss-making startups can convert research spending into a payroll tax offset worth up to $500,000 per year, but only by electing on a timely filed return.
5. State registrations. Remote hires quietly create payroll registration duties in every state where an employee works, and growing sales can create state tax filing duties where customers sit.

The entity question sorts by path, not by preference. If you intend to raise from institutional venture investors, the Delaware C corporation is the convention: it is the structure funds are built to invest in, and C corp stock is the only stock that can qualify for the QSBS exclusion at exit. The trade-off: a C corp's early losses stay locked inside the corporation as carryforwards instead of reducing your personal tax bill today.

If you are bootstrapping toward profit instead of a Series A, the calculus flips. An LLC can pass early losses through to the owners' personal returns (subject to basis and at-risk limits), and once profits arrive, an S corporation election can trim self-employment taxes. Our partnership vs S corp comparison walks through that fork, and how to set up an S corp covers the mechanics when the time comes. If you want to pressure-test the S corp math against your own projected profit, run the S corp savings calculator.

Two formation-stage tax items founders routinely miss. First, the costs of getting started (legal fees, filing fees, pre-launch expenses) are not simply deductible: you can generally deduct up to $5,000 of startup costs and $5,000 of organizational costs in year one (phased out once each bucket tops $50,000), with the rest amortized over 180 months. Our startup costs deduction guide covers the details. Second, forming in Delaware while operating from another state means registering as a foreign entity where you actually work, so budget for two states' worth of annual fees, not one.

When founders take restricted stock that vests over time, the default rule taxes each slice as it vests, at whatever the stock is worth by then. An 83(b) election flips that: you are taxed once, at grant, when the stock is typically worth close to nothing. The election must be filed within 30 days of the stock transfer. The IRS does not grant extensions, and since late 2024 there is finally an official form (Form 15620), with electronic filing available since mid-2025.

The same grant date quietly starts a second clock: the QSBS holding period. Under the One Big Beautiful Bill Act, stock acquired after July 4, 2025 earns a 50% gain exclusion at a three-year hold, 75% at four years, and 100% at five years, replacing the old all-or-nothing five-year cliff for new stock. The per-issuer exclusion cap also rose from $10 million to $15 million for that new stock, and the company gross asset ceiling rose from $50 million to $75 million, so more startups qualify for longer. Stock issued before July 5, 2025 keeps the old rules: $10 million cap, five-year cliff, $50 million asset test.

The operating stage is where startup taxes become an accounting problem before a tax problem. Your first return locks in an accounting method: most early-stage companies pass the small business gross receipts test and can use the cash method, though venture-backed companies usually run accrual books anyway because that is what diligence expects. Clean monthly books are not a tax nicety; they are what makes the R&D credit defensible, the 1120 painless, and the eventual due-diligence data room boring in the good way.

The R&D credit is the single most underused tax asset in early-stage companies. A qualified small business (gross receipts under $5 million in the credit year and none before the five-year window) can elect to apply up to $500,000 of credit per year against employer payroll taxes, which means a company with zero income tax still gets paid for its research spending.

The related Section 174 story matters too. From 2022 through 2024, companies were forced to capitalize research costs and deduct them slowly, which manufactured phantom taxable income at loss-making startups. The One Big Beautiful Bill Act fixed it: new Section 174A restores full expensing of domestic research costs for tax years beginning after December 31, 2024, and companies with capitalized 2022-2024 research costs can accelerate the remaining deductions into 2025 or spread them over 2025 and 2026. Foreign research still gets capitalized over 15 years.

The first hire converts your startup from a filing-once-a-year entity into a filing-every-quarter one. Payroll means registering in each state where an employee works, depositing withholding on schedule, filing Form 941 quarterly, and issuing W-2s by January 31. Contractors get 1099-NECs by the same January 31 deadline once payments cross $2,000 for the year, the new OBBBA threshold effective for 2026 payments. And the W-2 vs 1099 line is not a preference: misclassifying an employee as a contractor is one of the most expensive routine mistakes a startup makes. Our W-2 vs 1099 guide draws the line.

Franchise taxes are the tax on existing. A Delaware corporation owes its annual report and franchise tax by March 1 each year: a $50 report fee plus tax that bottoms out at $175 under the authorized shares method or $400 under the assumed par value method. A Delaware LLC skips the report and pays a flat $300 by June 1.

Federal income tax compliance stays simple in concept: a C corporation files Form 1120 by April 15 for a calendar year (extendable six months to October 15 with Form 7004) and pays a flat 21% on taxable income, which for most startups in the burn years is zero. Estimated tax installments only enter the picture once you actually expect to owe, generally $500 or more for the year.

First, the reassurance: money raised by selling stock is a capital contribution, not taxable income. A $3 million priced round does not create a $3 million tax bill, and SAFEs generally do not create income for the company either. What fundraising does change is your QSBS math: each round pushes the company toward the $75 million gross asset ceiling, after which newly issued stock can no longer qualify. Early employees and founders keep their qualification; it is the late arrivals who miss it.

At exit, QSBS is the whole ballgame for shareholders who set it up correctly. The per-issuer exclusion is the greater of $15 million (for post-July 4, 2025 stock) or 10 times your basis in the stock, which is how large-basis late investments can exclude far more than the headline number. And because the cap applies per taxpayer, advanced planning ("Section 1202 stacking") uses gifts to family members or non-grantor trusts to multiply the exclusion across several taxpayers before a sale. That is planning to do years before the exit, not the month of. The full playbook lives in our QSBS guide.

The compact version, assuming a calendar tax year and a Delaware C corporation (the most common venture-path setup). Dates that land on a weekend or holiday roll to the next business day.

State deadlines stack on top of this: annual reports and franchise taxes in your home state if you foreign-qualified there, state payroll filings everywhere you have employees, and state income or gross receipts returns where you have filing obligations. The federal calendar is the floor, not the whole list.

Notice what drove the outcome: not clever strategies, just doing the boring filings correctly and catching the one election (the R&D payroll offset) that pays a pre-revenue company. That is startup tax in miniature.

Overbought: complexity before revenue. Multi-entity structures with a holding company and an IP subsidiary for a company with no IP revenue to shift. Aggressive tax strategies pitched to founders who have no income to shelter yet. Premature S corporation elections that trade a modest loss deduction today for the QSBS exclusion tomorrow. If a structure costs more per year than the tax it could plausibly save before your Series B, skip it.

Underbought: the cheap, boring things with enormous downside protection. 83(b) elections, which cost postage or a free e-filing and protect against being taxed on paper gains you cannot sell. Clean books before diligence, because every financing and acquisition process reprices sloppy accounting as risk. State payroll registrations for remote hires, which cost little to do on time and multiples more to unwind. And contemporaneous R&D credit documentation, because a credit you cannot substantiate is a credit you do not have.

If the map above tells you the boring things are behind and the decisions are ahead, that is the point where a startup-fluent CPA earns their fee. Here is how Taxstra works with startups, from formation-stage elections through the exit file.