For three years, one of the most valuable incentives in the tax code quietly turned into a penalty, and most founders never noticed, because their return still got filed and the number at the bottom still looked normal. Then, in 2025, the rules changed again, this time in founders' favor. And a large number of the founders I work with are still operating as if they did not.
What broke, and what just got fixed
The research incentive comes in two parts founders constantly conflate: the R&D tax credit, a dollar-for-dollar reduction of tax owed, and the deduction for R&D expenses, how you write those costs off in the first place.
From 2022 through 2024, a change in Section 174 forced businesses to capitalize and amortize domestic R&D costs over five years instead of deducting them immediately. In plain terms, a company that spent a dollar developing its product could no longer write off that full dollar the year it was spent. For lean, product-driven founders, that created tax bills on money they had already spent building the thing.
The One Big Beautiful Bill Act reversed it. Under the new Section 174A, domestic research expenses are once again immediately deductible for tax years beginning after December 31, 2024. Foreign research costs still have to be amortized over fifteen years, a distinction that matters if any of your development happens offshore.
The part most founders are about to miss
Two time-sensitive openings, and both close. First, retroactive relief for smaller businesses. Eligible small businesses, generally those with average annual gross receipts under roughly $31 million, can elect to apply the new expensing rules retroactively to 2022, 2023, and 2024 by amending those returns. For a founder who overpaid during the capitalization years, that can mean a real refund, and the election window runs out in mid-2026.
Second, the transition election. Businesses can choose to deduct their remaining unamortized R&D costs fully in 2025, or spread them across 2025 and 2026. Which choice is better is not universal. It depends on your income in each year, which is exactly the kind of multi-year decision that should be designed, not defaulted.
The word research makes founders think lab coats. The tax definition is far broader. If your business develops or meaningfully improves a product, a software platform, a process, or a formulation, and there is genuine technical uncertainty in doing it, you may be conducting qualified research, whether or not anyone calls it R&D.
Why this sits in tax architecture, not tax filing
Your CPA files what happened. Claiming the credit correctly, and choosing between the retroactive amendment, the 2025 full deduction, and the two-year spread, is a design decision that depends on your income across multiple years and the rest of your structure. The documentation bar is rising too: the IRS now wants more detailed, business-component-level substantiation, with the expanded Section G of Form 6765 becoming mandatory for most businesses in 2026 and beyond.
If you spent real money building your product between 2022 and today, three questions are worth answering before the retroactive window closes: did we qualify in the capitalization years and overpay, should we amend, and for 2025 and 2026, which deduction path fits our income curve. None has a generic answer. All three have a deadline.
This is general information, not tax advice. Eligibility, elections, and deadlines depend on your specific facts and current IRS guidance, designed in partnership with Engineered Tax Services.