Research and Development Tax Deduction vs. R&D Tax Credit: How They Stack Under Post-OBBBA Rules

Last filing season I watched three documents land in the same week, all describing “R&D stuff”: a credit study, an amortization schedule, and a note about capitalized labor. Same vocabulary, same two-letter acronym, completely different money math in each one. A deduction of some amount, a credit of some other amount, and a five-year amortization schedule for good measure. It looked like three people had described the same codebase using three different forks of the spec.

Here’s the thing I kept wanting to shout: they’re not competing options. The R&D deduction (Section 174, now Section 174A) and the R&D credit (Section 41) operate at two different points in the tax calculation, and the same research dollar can feed both. Section 280C is the rule that coordinates them. You don’t pick one. You run both, and then you make an election about how they interact.

Quick current-law status so we’re all on the same page: Section 174A, which restored immediate expensing of domestic research costs, is permanent, courtesy of the One Big Beautiful Bill Act enacted July 4, 2025. Section 41 got its permanence from the PATH Act of 2015. And the mechanical distinction that makes everything click: a deduction shrinks your taxable income before the tax bill is computed, so its value depends on your marginal rate. A credit comes off the bill after, dollar for dollar. Which is exactly why one dollar can do double duty.

(Quick aside so nobody feels lost: R&D and R&E, research and development versus research and experimentation, are pure vocabulary drift. Same thing. The distinction that actually matters is deduction versus credit.)

Section 174A and Section 41 at a glance

A Section 174/174A deduction reduces taxable income; a Section 41 credit, the IRC’s credit for increasing research activities, reduces the tax bill dollar-for-dollar. Same research dollar, two different levers, pulled at two different points in the calculation.

The Form 6765 bit matters in practice: the credit doesn’t exist as a line item until you compute it there. And here’s the wedge most coverage misses: these aren’t rival elections you pick between. They operate at separate stages of the tax computation, which is precisely why the same dollar can feed both. Section 41(d)(1)(A) even makes Section 174 eligibility the first requirement of the credit, so the statute links them, even though they regulate fundamentally different things. Hold that tension; we’ll come back to it.

Section 174 to Section 174A: the deduction’s full arc

Domestic R&E is immediately deductible under Section 174A for tax years beginning after Dec. 31, 2024. That sentence is the end of a genuinely weird story.

to 2021: the long stable default

Section 174 was enacted in 1954, and for nearly 70 years the default setting was simple: deduct your research expenses now. The scope was broad, almost suspiciously so. Direct and indirect costs, onshore and offshore: wages, supplies, rent, utilities, overhead. The original intent was a leveling move, not a loophole; small businesses without a dedicated research department could deduct development costs they couldn’t squeeze under Section 162.

There was even an opt-out worth remembering. The Section 174(b) election let a taxpayer choose capitalization instead, which mattered for loss-position, pre-revenue startups: electing capitalization deferred initiating a net operating loss, and NOL carryforwards run 20 years. Sometimes you didn’t want the deduction yet, because a 20-year clock was involved. Keep that in your back pocket; it comes back in the startup section.

The 2022 rug pull

For tax years beginning after Dec. 31, 2021, the Tax Cuts and Jobs Act flipped the default. Specified research expenditures had to be capitalized and amortized over five years domestic, fifteen years foreign. Higher taxable income, bigger upfront tax payments, messier planning. Your actual bill changed.

And the why is the good part: this wasn’t R&D policy. It was a revenue offset, included so the act could pass through budget reconciliation and dodge a filibuster. Procedural machinery, nothing more. Most observers, including people at the IRS, expected the change to get pushed out or reversed before it ever took effect.

Everyone assumed a patch was coming. The patch didn’t come, and four years of planning got shaped around a rule everyone thought was temporary.

The fix

OBBBA, enacted July 4, 2025, created Section 174A: immediate domestic expensing, reinstated and made permanent, for tax years beginning after Dec. 31, 2024. Made permanent is the load-bearing phrase. No more waiting for the next bill to decide your tax treatment.

The asymmetry is intentional, though: foreign R&E stays on 15-year amortization. The code is now opinionated about geography.

One pattern I’ve seen repeatedly since mid-2025, and I’ll describe it as a composite rather than any single engagement: a business pulls out an R&D study or amortization schedule prepared for 2022 through 2024, and it still spreads domestic costs over five years. The tell is the five-year spread in a document dated after July 2025. The law moved; the script didn’t.

Section 41: what the credit pays on and what it refuses

The credit side of the pair is actually the older mechanism with its own design goals, and it’s the one most people mean when they say “R&D tax credit.”

History and design

Section 41 was enacted in 1981, when Congress noticed research spending was declining and decided to counter-program it. Three words describe its mechanics, and each one matters: it’s nonrefundable (it can’t take your bill below zero), incremental (it pays on qualified spending above a base level, so it rewards growth, not existence), and a dollar-for-dollar general business credit (it comes straight off the computed bill). The PATH Act of 2015 made it permanent. Funny contrast: the credit has been the stable one in this story while the deduction got shuffled.

It’s also not the pharma-only club people assume. The credit was once associated mainly with biotech and pharmacology, but Treasury regulations substantially broadened eligibility. Your two-person startup can qualify. The trade-or-business connection needs a nontax profit motive plus substantial, regular involvement, held in good faith. Genuinely in the law: the expectation doesn’t have to be reasonable, just sincerely held.

The qualification gauntlet

The credit runs a four-part test. Four gates, not a legal exam:

1. The Section 174A test. First gate: are the expenditures even eligible for a 174A deduction, connected to your trade or business, R&D in the experimental or laboratory sense?
2. Technological information. You’re trying to discover information that develops a new or improved business component. And you don’t have to succeed. Failing to build the thing doesn’t disqualify the research, which is honestly generous.
3. Process of experimentation. An evaluative process that can weigh more than one alternative, aimed at function, performance, or reliability. Not style, not cosmetics. Did you actually iterate, or did you just pick a color?
4. Business component. Tests apply separately to each component, with a shrinking-back rule: you zoom in until you find the part that qualifies. A clever mechanism, and one of the more elegant things in this corner of the code.

Qualified research expenses are the sum of in-house research expenses (wages for qualified services, supplies, computers, including cloud and rental computers) plus contract research expenses, counted at 65%. Then the 80% rule: if at least 80% of an employee’s services during the period fit the qualified research criteria (“substantially all”), their entire annual wages count. Friendlier than it sounds.

One detail that trips everyone: eligibility keys off what the employee actually does during a specific period, not job titles. The IRS cares about your git log, not your LinkedIn.

Exclusions

The no-credit list: research after commercial production begins (once it’s shipping, the clock stops), adaptation of existing components (retrofitting isn’t researching), reproduction of an existing component from plans, blueprints, or publicly available information (copying isn’t discovery), efficiency surveys, routine data collection, ordinary quality-control testing, research outside the U.S., and funded research.

The funded-research exclusion is where the list gets interesting. It runs a two-pronged test: economic risk of research success, and retention of substantial rights to the results. Are you on the hook if it fails, and do you actually own what comes out? Here’s the insight the generic lists miss: the exclusions are about who bears risk and who owns output, not how innovative the work is. A technically brilliant project can fail the credit purely on contract structure.

Cost-by-cost scope map: deductible but not creditable

The credit’s qualified-expense list and Section 174’s cost set overlap, but they’re not the same list, and the differences are where the money hides.

The distinction below traces back to how software development costs are treated under Section 174:

A few of these deserve interpretation rather than a checkbox, the same deductions people miss on lists of the most overlooked tax deductions for tech professionals and small businesses. Same contractors, different percentages: contract research counts at 65% toward the credit but U.S. contractor costs count at 100% toward the deduction. The wage bases differ too: the credit reads W-2 Box 1, while the deduction’s base is gross wages plus benefits outside Box 1. Funded R&D gets shut out of the credit by the risk-and-rights test but still feeds a deduction analysis. And general and administrative costs, creditable never, deductible onshore.

Which lands the rule: Section 174 is the broader bucket. Failing the four-part test kills the credit, not the deduction. A cost can qualify under 174 and not 41, so you can’t assume one calculation covers both. The buckets are genuinely different sizes.

How they stack: Section 280C and the reduced-credit election

Yes, a business can claim both the R&D credit and the deduction. But Section 280C(c) coordinates them, and the coordination has a catch.

The default: your Section 174A deduction is reduced by any Section 41 credit you take. Unless you elect a reduced credit, in which case you keep the full deduction. Think of it as a trade-off dial, not a penalty. Where research costs are capitalized and the credit exceeds the allowable deduction, the same rule reaches into a different bucket: the capital account is reduced by the excess.

So the honest version of “you can get both” is this: the same dollar cannot generate a full deduction and a full credit. The tax code noticed you could claim both and wrote a rule about it. Which election wins depends entirely on the taxpayer’s own numbers. There’s no universal answer, and anyone selling you one hasn’t run your scenario.

Quick test: Run the numbers both ways — reduced credit with full deduction, or full credit with reduced deduction — and keep whichever nets more.

Practical sequencing, and it’s an efficiency win: compute Section 41 first, and do it concurrently with the 174/174A analysis. The credit work doubles as deduction homework, because the credit’s first requirement is 174 eligibility anyway.

And a warning from the enforcement file: an IRS directive, revised August 2005, flags invalid 280C(c)(3) elections on amended returns and refund claims. The IRS has seen people botch this election before. It’s worth getting right the first time.

Who owns the research? Substantial rights vs. right to exploit

Two ownership tests, same contracts. This is the corner of the topic where a clause you signed in year one decides what you can claim in year five.

Two standards, same paper

Section 41’s substantial-rights analysis reaches technological information and data created throughout the research process. The question: can you use the results without additional payment or approval? No permission slips required. Section 174’s right-to-exploit test, introduced through Notices 2023-63 and 2024-12, looks at the capitalizable work product itself.

The SRE product right means the right to use the resulting product in your trade or business, or exploit it through sale, lease, or license, without separate payment or approval. Can you ship, sell, or license this thing onward?

Notice 2023-63 drew its own boundary explicitly: it states it isn’t intended to change the rules for determining eligibility for or computation of the research credit under Section 41. The IRS said, in its own guidance, that this doesn’t touch the credit. And that’s why the divergence is real: 41 asks about everything you learned along the way; 174 asks about the artifact. The consequence is genuinely counterintuitive: you can retain substantial rights without a legally protectable right to exploit the final product. You can pass one test and fail the other.

The case trajectory

The courts have actually ruled on this, and the arc reads like a changelog:

• Lockheed Martin, 210 F.3d 1366 (Fed. Cir. 2000): the win case. Research for the U.S. government; substantial rights retained despite a non-exclusive right, because no payment was required to use the results. Non-exclusive didn’t sink them.
• Tangel, T.C. Memo. 2021-1: a contract styled as a “work made for hire” assigned all drawings, designs, equipment, and technical data exclusively to the buyer. The contract gave away everything, and that was that.
• Grigsby/Cajun Industries, 86 F.4th 602 (5th Cir. 2023): contracts conveyed “all right, title, and interest” in broad “work product” definitions. The taxpayer lost for failing to demonstrate retained rights. The contract language did the damage.
• Lewin, 335 F.3d 345 (4th Cir. 2003): an investment group bankrolled startups’ R&D in exchange for license-back rights but lacked the infrastructure to commercialize. The Section 174 deduction was denied. Rights on paper, no ability to use them.
• Harris, 16 F.3d 75 (5th Cir. 1994): supporting precedent establishing a similar requirement of intent and ability to exploit.

The drafting implication, and I’m flagging this as exactly what it is, speculation: had the Tangel and Grigsby providers reserved rights to underlying data or processes while granting exclusive rights to the final product, courts may have ruled differently. One clause might have flipped both cases. That’s a counterfactual, not settled law.

The pattern behind the cases, again as a composite rather than any real engagement: nobody re-reads the rights clause until a credit study or deduction position gets prepared, and the standard services agreement turns out to convey all rights in the work product. The tell is always the rights clause.

Startups and small businesses: the payroll offset and the 174(b) legacy

Yes, a pre-revenue startup can use the R&D credit against payroll taxes, if it meets the narrow $5 million / five-year tests. The alternative is electing capitalization under Section 174(b) to defer initiating a net operating loss, which starts a 20-year carryforward clock, a real strategic choice for loss-position startups, not a footnote.

• The mechanism: the credit is nonrefundable, so the payroll offset exists precisely for companies with no income tax bill to reduce. If the credit exceeds income tax liability, the remainder goes against the employer portion of FICA. Your payroll tax bill shrinks; your team’s paychecks don’t.
• The gate: less than $5 million in revenue (not profits) in a year following five years of no revenue. Revenue, not profit. Read that twice.
• The numbers changed for the better: pre-2023, the offset capped at $250,000 and covered Social Security taxes only. Starting 2023, the Inflation Reduction Act doubled the cap to $500,000 and extended it to Medicare. The cap doubled and the coverage widened. A genuine improvement worth naming.
• The odd lineage: the 2017 TCJA imposed amortization; the Biden-era IRA expanded the payroll offset. Both parties touched this system in different directions. That’s just factual political entropy.
• The 174(b) legacy: before all this, loss-position startups could elect capitalization under 174(b) to defer initiating a net operating loss, which starts a 20-year carryforward clock. Sometimes you didn’t want the deduction yet. It was a real strategic choice, not a footnote.

What changed in 2025, and why stale advice is the real risk

Early 2025 was a holding pattern: observers expected reinstatement of expensing under the incoming administration and Congress, with timing tied to TCJA renewal and Senate reconciliation. We knew it was coming; nobody knew when. Then OBBBA was enacted July 4, 2025, and the timing question resolved.

The live decisions now:

• Tax years 2022 through 2024 stay under pre-OBBBA capitalization. Anyone who filed in that window is still living with it.
• Catch-up menu: deduct remaining unamortized domestic costs entirely in 2025, or spread them over 2025 and 2026.
• Eligible small businesses may elect Section 174A retroactively by amending 2022 through 2024 returns. That’s the practitioner-reported path; outcomes on any specific amendment depend on the facts, so nobody should promise refunds.
• Rev. Proc. 2025-08 is the mechanism you’d use for changing your method of accounting for R&E expenditures.

Here’s the contrarian core, and I mean it: the most dangerous current error isn’t misunderstanding the law. It’s trusting R&D studies, amortization schedules, and estimated-payment assumptions still built under superseded rules. Advisers still recommending five-year domestic amortization spreads are reading an expired script. (That’s the same composite pattern from the history section, and it’s worth saying plainly: the document in your drawer dated before July 2025 may be describing a law that no longer applies to your 2025 return.)

And to answer the obvious follow-up: yes, research and development costs remain tax-deductible for 2025 and 2026. Domestic costs immediately under 174A; foreign costs on the 15-year amortization schedule. The deduction didn’t go anywhere. Its geography did.

The stakes and the debate: what each provision is worth at scale

Why did mandatory capitalization cause cash-flow problems for manufacturers? Short answer: because from 2022, a five-year domestic amortization schedule turned an immediate write-off into a slow drip, and the bill arrived before any offset did. The details, and their limits, are worth unpacking from two differently positioned sources.

The practitioner view

Practitioners report that middle-market manufacturers capitalized large portions of R&D from 2022 onward without corresponding Section 41 credit benefits, and that strain, practitioners say, may have squeezed their ability to innovate. Note the “potentially,” which is doing real work: this is a practitioner-attributed observation, not a dataset, and it’s distinct from any claim about aggregate research spending. Companies took the amortization hit without the credit offset, and that’s a cash-flow mechanics story.

The advocacy-attributed critique

Americans for Tax Fairness, an advocacy group rather than a neutral source, tells the scale story from the other side. Its findings: permanently restoring immediate expensing was estimated to cost roughly $200 billion over ten years; experts estimate corporations received as much as $35 billion in R&D credits in 2022 alone, a 550% increase since 2001 that outpaced both company-funded research and GDP growth. And from a corporate 10-K analysis, ATF found no evidence amortization deterred research investment: major corporations increased spending after 2022.

The effective-rate numbers it assembled: Northrop Grumman paid 10.5% on $12.4 billion in profits; Amazon paid 4.3% on $43.4 billion (2018 through 2020, against a 21% statutory rate); Boeing paid 5.4% from 2008 through 2015 against a 35% statutory rate, zeroing out its liability in five of those years. And the tension, stated factually rather than editorialized: R&D Coalition Leadership Committee members, including Northrop Grumman, Amazon, and Boeing, lobbied for expensing relief while their effective rates sat below statutory.

I’m not taking a side here, because the two claims answer different questions. The practitioner story is about cash-flow mechanics at the company level. The ATF story is about aggregate spending and who captures the benefit. Both sources are positioned, and neither is neutral.

One evidence set, three regimes: books, deduction, credit

This is my favorite part of the whole system, because it’s the moment the conflation stops being annoying and starts being elegant.

Take one engineer earning $200,000, doing genuinely new work. Those same salary dollars are, simultaneously:

• A capitalized asset under ASC 350-40 on the books. Engineering labor for new software gets capitalized as an asset under GAAP, a treatment made for investors and lenders, not the IRS. Note the scope: that’s the software development standard specifically, not a general rule for all R&D book treatment.
• A deductible expense under Section 174A on the tax return.
• Credit-base QRE material under Section 41 on Form 6765, if it passes the four-part test.

Three rulebooks, three audiences, all correct at once. Your CFO’s ledger and your tax return are reading different documents about the same person. (The IRS has even formalized the book-vs-tax overlap: directives from 2017 and revised in 2020 address the credit for taxpayers expensing R&D costs on their financial statements under ASC 730.)

Now the documentation gap, as a composite pattern: a typical credit file gets built from W-2 Box 1 summaries and job titles. Those inputs support neither the 174A gross-pay-plus-benefits base nor the 80% wage test. One project-level time-and-cost mapping built at the source serves all three regimes.

If you want to know what the IRS examiners actually read, it’s public. The Research Credit Claims Audit Techniques Guide dates to May 2008, with a pharmaceutical revision from April 2024, and there are software, aerospace, and research-credit guides alongside it. The reading list: payroll records, job descriptions, performance evaluations, calendars and appointment books. “Keep good records” becomes concrete when you know the examiners’ actual syllabus. And the guidance keeps moving, which is worth knowing: Rev. Proc. 2023-11 superseded Rev. Proc. 2023-08 within the same year on accounting methods for specific R&E expenditures, and the IRS recently released revised draft Form 6765 instructions after a public comment cycle, spelling out required information for valid refund claims.

One mapping, built once, used three times. For anyone who hates duplicate paperwork, this is the tiny wizardry hiding in the tax code.

Decision guide: how each provision moves your tax position

A few concrete moves, in order:

• Compute Section 41 first, concurrently with the 174/174A analysis. The credit work identifies potentially qualifying deduction costs, so you’re not doing the same mapping twice.
• Run your profitability timeline. The credit’s value depends on tax capacity, because it’s nonrefundable. If you don’t have a bill yet, the payroll offset is the mode that exists for you. Loss positions bring the 20-year NOL carryforward into play, which is the same logic that once made the 174(b) deferral election worth considering.
• Recalibrate estimated tax payments to the law as it exists now, not to schedules prepared under 2022 through 2024 rules. Your quarterly numbers may have quietly changed.
• Treat the 280C election as the hinge. It’s where the dollar-by-dollar overlap question gets decided. Run both scenarios against your own numbers and see which pays more.

There’s a fuller walkthrough of the expensing-versus-amortizing side of this in our breakdown of whether software development is tax deductible, if you want the Section 174 mechanics with plain-English examples.

Closing

The deduction and the credit were never rivals. They’re two rules over the same dollars, and the real work is fourfold: timing (Section 174A), qualification (Section 41), coordination (Section 280C), and documentation (one evidence set, built at the source).

Where things stand: domestic expensing is back, permanently. Foreign R&E still amortizes over 15 years. The four-part test hasn’t budged. The rules are legible now. The trick is making sure the advice you’re following is legible too.

Frequently Asked Questions