R&D Tax Credit and Section 174 Guide 2026: What Founders and Small Businesses Miss

The R&D Credit and Section 174 Are Not the Same Thing

If you build products or write software in the United States, you’ll hear two phrases from your accountant almost every year: “the R&D credit” and “Section 174.” The trouble is how many founders treat them as one idea. My read is that untangling them is the first job, because getting this wrong quietly leaves money on the table.

Here’s the plain version. The Section 41 research credit is a dollar-for-dollar reduction of your tax bill. Spend $100 on qualifying work at a 10% effective rate and your tax drops by roughly $10. Section 174 is not a giveaway at all — it’s a timing rule that decides when you get to deduct those research costs. It doesn’t cut your tax; it decides whether the write-off happens all at once or gets stretched across years.

They are distinct, but they ride on the same spending. The same engineer’s salary and the same lab supplies feed the credit calculation under Section 41 and the deduction-timing rules under Section 174. Handle one and ignore the other, and you’ve used half the toolkit.

This piece is an informational guide for people running or planning a U.S. business. It is not tax advice for your specific situation. The goal is to walk into a conversation with a CPA knowing which questions to ask.

👉 If you’re also thinking about how you pay yourself, the S-corp reasonable salary and payroll tax guide rounds out the picture.

---

What Exactly Does the Section 41 Credit Reduce?

Section 41 goes by the formal name “Credit for Increasing Research Activities.” The idea is that the government hands back part of what a company spends solving technical problems inside the United States.

The word to sit with is credit, as opposed to deduction. A deduction lowers taxable income, so you save only your marginal rate on each dollar. A credit comes off the calculated tax itself, so the benefit is far more direct. That distinction is why the R&D credit is worth chasing even when a deduction already exists.

A common misconception is that this belongs to white-coat labs. It doesn’t. Software development, manufacturing process improvements, new formulations, hardware design, and building automation all show up in real claims. The work doesn’t have to be groundbreaking for the industry. It has to be new or uncertain for your company. That threshold is much lower than most founders assume.

---

Why Every Project Has to Pass the Four-Part Test

Eligibility comes down to the IRS four-part test. An activity has to clear all four gates. Fail one and that project is out.

Test	Requirement	The real question
Permitted purpose	Improve a business component’s function, performance, reliability, or quality	What were you trying to make better?
Technological in nature	Rely on hard science — engineering, physics, biology, computer science	Did it lean on a hard science?
Elimination of uncertainty	Real doubt about design, method, or capability at the start	Did you already know the answer?
Process of experimentation	Trial and error, modeling, simulation to evaluate alternatives	Did you test more than one approach?

The gates that trip people up are the third and fourth. If you simply implemented a known, settled method, there was no uncertainty, and the work fails. But if you didn’t know which architecture would perform and you compared and tested several, you satisfy the experimentation requirement. Whether the project succeeded is irrelevant. Failed projects can qualify too — that surprises people, so it’s worth repeating.

The exclusions are just as clear. Routine quality control after commercial release, market research, management studies, cosmetic-only changes, research performed outside the U.S., and funded research paid for by someone else are, as a rule, off the table.

---

Which Dollars Count as Qualified Research Expenses?

The size of the credit ultimately rides on your qualified research expenses, or QREs. They fall into three buckets.

First, wages. This covers not only the people doing the research but also those who directly supervise or directly support it. An engineer writing code, a lead steering the technical approach, a tester running experiments — all are candidates. Wages are usually the largest slice of QREs by far.

Second, supplies. Materials consumed in research or prototyping qualify. Land and depreciable property do not. The material you burned through building a prototype counts; the 3D printer you built it on does not.

Third, contract research. If you paid an outside party to do the work, only 65% of the amount counts (75% for a qualified research consortium). The key condition: your company has to bear the economic risk and keep rights to the results. If the contractor gets paid regardless of outcome, it’s funded research on their side and doesn’t qualify for them.

Cloud computing rental used directly for development and testing can also count as a QRE under certain conditions. For a modern startup running on rented infrastructure instead of owned servers, that’s an easy item to overlook.

---

Regular Method or the ASC — Which Wins?

Once you’ve pinned down QREs, you pick a calculation method. There are two roads.

Item	Regular method	ASC
Rate	20% of QREs above a base amount	14% of the amount above 50% of the prior 3-year average
No prior QREs	Complex, can be unfavorable	6% of current-year QREs
Data needed	Base-period ratios reaching back to the 1980s	Just the last three years of QREs
Practical difficulty	High	Low

The regular method demands old base-year data and is fiddly to compute, but for a mature company with a low base amount it can deliver a larger credit. The Alternative Simplified Credit needs only three years of data and is what most filers reach for. For a brand-new company with no research history, the ASC’s 6%-of-current-year figure is effectively the default.

My advice is blunt: run both and take the bigger one. Decent tax software and specialty firms compute the two in parallel and hand you the winner.

---

The Startup Payroll-Tax Offset: Saving Cash Without Any Profit

This is the part that matters most to early-stage startups. The R&D credit normally reduces income tax, but a company burning cash has no income tax to reduce, so the credit looks useless. That’s the gap the payroll-tax offset election fills.

A qualified small business can apply part of its R&D credit against the employer’s share of payroll taxes instead of income tax. Even with no profit, the cash you send to the IRS every quarter shrinks, which is a genuine boost to early runway.

The core rules and limits:

• Who qualifies: a company with gross receipts under $5 million for the year that has had receipts for five years or less — in other words, an early-stage business, not a mature one.
• The cap: since the Inflation Reduction Act, up to $500,000 per year. The first $250,000 hits the employer’s 6.2% Social Security tax, and the additional $250,000 hits the 1.45% Medicare tax.
• The mechanics: you elect the offset on Form 6765 with your income tax return, then claim it on the next quarterly payroll return (Form 941) by attaching Form 8974.

One thing people miss: the offset applies to payroll taxes after the election, on the following returns. It doesn’t refund payroll taxes you’ve already paid. So the discipline for a startup is to make the election, on time, every single year.

---

Why Section 174 Caused So Much Noise

Now the other side of the coin: Section 174. For decades, U.S. businesses could deduct research costs in full in the year they were spent. The 2017 tax law changed that.

Starting with tax years beginning in 2022, specified research and experimental (R&E) expenditures could no longer be expensed immediately. They had to be capitalized and then amortized — five years for domestic research, fifteen for foreign — and even then, a half-year convention meant only half the amortization landed in year one.

Why the uproar? Picture a startup that spent $1 million developing software with no revenue yet. Under the old rules, that $1 million was expensed and there was no taxable income. Under the new rule, only about $100,000 of domestic cost got deducted in year one, and the remaining $900,000 was treated as an asset not yet spent. The result was a company that had already burned the cash but showed a paper profit and owed tax on it. That is a genuinely strange place to be.

Software and biotech startups, where payroll is the dominant cost, took the hardest hit. Plenty of them watched their tax bills spike relative to revenue.

👉 If you want to see how U.S. tax structure works on the investment side too, the U.S. stock capital gains tax guide is a useful companion.

---

How OBBBA Softened Section 174

Relief arrived in 2025. The law commonly called OBBBA, enacted in July 2025, took most of the sting out of Section 174.

Three points do the work.

First, domestic expensing came back. Under the new Section 174A, businesses can again deduct domestic R&E in full for tax years beginning in 2025 — effectively a return to the pre-2017 treatment.

Second, foreign R&E stays put. Research performed abroad still amortizes over fifteen years. Splitting domestic from foreign spending in your records now matters more than ever.

Third, a small-business retroactive fix. Companies under the gross-receipts threshold (roughly $31 million) got a path to apply immediate expensing retroactively to 2022. Other companies received an option to recover the unamortized domestic amounts stacked up in 2022 through 2024 over one or two years.

So the timeline looks like this: 2022 through 2024 was the forced-capitalization window, 2025 restored domestic immediate expensing, and small businesses got room to reach back. If you filed returns in the 2022–2024 stretch, it is genuinely worth checking whether an amended return puts cash back in your pocket.

---

How the Section 41 Credit and Section 174 Move Together

Now that each provision is clear, look at where they interlock. A common adjustment shows up in practice: taking the Section 41 credit on the same spending you also deduct in full would be a double benefit. So the code, through Section 280C, either reduces your deduction (or the capitalized amount) by the credit you claimed, or lets you elect a reduced credit instead.

The wording is clunky, but the point is simple. You can’t take both the full credit and the full deduction on the very same dollar. A good preparer decides whether to elect the reduced credit or trim the deduction and take the full credit, based on your company’s rate situation.

Here’s a final side-by-side of what each provision actually is.

Item	Section 41 R&D credit	Section 174 R&E capitalization
Nature	A credit that cuts tax directly	A timing rule for the deduction
Effect	Reduces tax, dollar for dollar	Defers or controls when you deduct
2025 status	Permanent, always available	Domestic immediate expensing restored
Startup wrinkle	Payroll offset ($500K/year)	Small-business retroactive relief
Key forms	Form 6765, 8974, 3800	Accounting method change (Form 3115)

---

How Much Documentation Do You Really Need?

The most common reason an R&D claim collapses isn’t failing the test — it’s weak support. The IRS wants contemporaneous documentation that backs the number. Estimates reconstructed after the fact carry little weight.

Practically, assemble:

• Project-level time allocation for wages (timesheets or a defensible basis for estimates)
• Payroll registers and W-2 data
• Supply purchase and consumption records
• Contract research agreements, with the clauses that show risk-bearing and rights retention
• Experimentation evidence: lab notebooks, design docs, code commit history, meeting notes, test results
• The filings themselves: Form 6765 for the calculation, Form 8974 for the payroll offset, and Form 3800 to gather general business credits

Time allocation is the piece examiners fight over most. If you claim 80% of an engineer’s salary is qualified, you need to explain where that 80% came from.

---

The Mistakes That Keep Recurring — and How to Dodge Them

Let me close with the errors I see over and over. I’d run this list before filing.

Under-claiming. The most frequent miss is excluding software or process improvement because it doesn’t feel like “research.” If it clears the four-part test, the label on the door doesn’t matter.

Over-claiming. The opposite trap: sweeping in routine post-release maintenance, simple bug fixes, and cosmetic tweaks. Those crumble under audit. Treat borderline activities conservatively.

Treating 174 and 41 as unrelated. Handling one and forgetting the other. If you took the credit, apply the Section 280C adjustment; if you carry an unamortized balance, check whether OBBBA relief applies.

Skipping the payroll election. A loss-year startup that computes the credit but never elects the offset just watches the cash benefit evaporate. Make it a checklist item every filing.

Documenting after the fact. Scrambling to build support at year-end produces thin records. Building time logs and experimentation evidence while the work happens is what determines how you fare in an audit.

Whatever tax you save eventually cycles back into the business or into investments. If you’re weighing where to put freed-up cash, the AI stocks investment guide 2026 and the SCHD dividend ETF guide 2026 are worth a read.

---

Keep Reading

• 👉 S-corp reasonable salary and payroll tax guide 2026
• 👉 U.S. stock capital gains tax guide 2026
• 👉 AI stocks investment guide 2026
• 👉 SCHD dividend ETF guide 2026

---

This article is general information about U.S. federal tax rules and is not tax or legal advice. Tax law and the rules governing credits and amortization change often, and how they apply depends on your specific business. Consult a licensed CPA or tax professional before making any filing or tax-planning decision.