Skip to main content
Software R&D Credit & Section 174

Software R&D Tax Credit (Section 41) & 174 Guide

By Zach CardozaPublished August 24, 2025Updated June 9, 2026
How qualifying custom development can earn the Section 41 R&D credit, how Section 174 changes when you get to deduct the cost, and why the SaaS subscription you just bought does not count.

Overview

When your team builds custom software and solves real technical problems doing it, those costs can earn a dollar-for-dollar federal credit under Section 41. The catch since 2022 is Section 174, which makes you spread the deduction over five years instead of taking it all at once. The license fee for an off-the-shelf tool earns you neither, because you did no experimentation. You just bought a product.

What Counts as Software R&D

The work has to involve real technical uncertainty that you resolve by trying things. If you knew exactly how to build it before you started, it probably does not qualify. Software you sell or license only has to pass the core four-part test. Software your team uses internally has a higher bar to clear.
Four-Part Test
It relies on hard science or engineering, sets out to remove a technical unknown, improves a product or process, and gets there through iteration, prototypes, and testing.
Internal-Use High Threshold
Software your team uses internally also has to be genuinely innovative, carry real economic risk, and not be something you could have bought and lightly configured.
External-Use Software
Software you sell, license, or market to customers usually only has to clear the four-part test, not the internal-use bar.

Common Exclusions

Plenty of real work does not qualify, and claiming it anyway is how you fail an audit. Routine adaptation with no unknown to solve, cosmetic changes, support after launch, market research, plain data collection, anything built outside the United States, and work a customer or grant paid you to do all fall outside the credit.

Why Off-the-Shelf Software Does Not Qualify

A license or SaaS fee is the cost of buying something, not the cost of inventing it. Your team never ran the experiment, so the Section 41 tests are never met. Those costs follow ordinary software recovery rules instead, like Rev. Proc. 2000-50 or Section 167 depreciation. This is a real edge for building over buying that most people leave out of the math.

Section 174 vs Section 41

These two run at the same time and people mix them up constantly. Section 174 is the rule that makes you capitalize and amortize the cost, five years for domestic work, fifteen for foreign, with a half-year in year one. Section 41 is the credit that still comes off your tax bill on those same qualified costs. One controls timing, the other hands you money. You can get both.

Qualified Research Expenses (QREs)

The credit is built from a few cost buckets. In-house technical wages, supplies, the cloud and compute you burned through testing, and a slice of what you paid contractors. The whole thing stands or falls on records. If you cannot show who spent which hours on what, the claim gets thin fast in an audit.
Technical Wages
Engineers, designers doing technical work, and the product and QA people directly supporting the experimentation.
Supplies and Cloud Usage
Physical materials and the compute you actually consumed running development and test cycles.
Contract Research Portion
A share of qualifying contractor invoices counts, capped at 65 percent of what you paid them.
Allocation Discipline
Keep the qualifying build work separate from maintenance, configuration, data migration, and support, because only the first one counts.
Lookback Opportunity
You can amend prior open tax years and claim credits you missed, as long as you can back up the expenses now.

How to Claim the Credit

A small, repeatable habit through the year beats reconstructing everything at tax time. Decide what qualifies, keep the evidence as you go, track the hours, and bring in someone who does this for a living before you file. The contemporaneous record is the whole game. Notes written after the fact carry far less weight.
Evaluate Eligibility
Run each project against the four-part test, and the internal-use bar where it applies, before you assume it counts.
Capture Experiment Artifacts
Hold on to the prototypes, the designs you rejected, the test results, and the notes on why you chose what you chose.
Track Time and Costs
Attribute hours and resource use to the qualifying work as it happens, not from memory in March.
Separate Non-Qualifying Spend
Pull out the licenses, the routine support, and the data cleanup, so they never end up in the claim by accident.
Prepare Form 6765
Total the qualified expenses, make the elections, and write down the method you used to get there.
Apply Section 174 Rules
Capitalize and amortize the required research costs in parallel, because the credit does not exempt you from the timing rule.
Involve Specialized Advisors
A specialist review makes the method stronger and the claim much easier to defend if anyone asks.

Why This Favors Building

The credit comes off your tax bill dollar for dollar, the amortization softens the hit over time, and at the end you own a tool built around how you actually work. A subscription gives you none of that. It is rent, and it qualifies for nothing.

References & Source Materials

The primary IRS and regulatory sources, if you want to read the rules yourself rather than take our word for it.

Evaluate Your Opportunity

We help Central Valley businesses scope which work qualifies, keep documentation that holds up without drowning your team in it, and coordinate with your tax advisor to claim what you have earned.

Frequently Asked Questions

Common questions about claiming the R&D tax credit for software development and what expenses qualify.
Work where you had to solve a technical unknown by trying things. Custom algorithms, performance work, a genuinely hard integration, a new interface pattern. The test is whether you knew how to build it before you started. Routine configuration and standard features do not qualify, but a real solution to a real technical problem usually does.
The federal credit is roughly 6 to 14 percent of qualified expenses for most businesses, and many states add another 5 to 25 percent on top. On a $100,000 qualifying project that is often $15,000 to $30,000 back. It offsets your tax bill, and qualified small businesses can take it against payroll tax even with no income tax owed.
Keep it as you go. Project plans, technical specs, test records, time tracking, the code history, and evidence that you actually experimented. Write down the problem, the unknowns, the approaches you tried, and how you resolved them. Records made during the work hold up. Records reconstructed the week before filing do not carry the same weight.
Both, with limits. Internal costs cover wages, benefits, and supplies for the qualifying work. Contractor costs are capped at 65 percent of what you paid, unless the contractor is effectively your employee or a related party. Cloud and tooling consumed directly in the experimentation can count too.
Since 2022 you have to capitalize software development costs and deduct them over five years, fifteen for foreign work, instead of all at once. That does not touch your eligibility for the credit. You still claim Section 41 on the qualifying work. It just means the credit often shows up before the full deduction does, which is a cash-flow timing thing to plan around.
Qualified research removes a technical uncertainty through experimentation. Standard features, routine fixes, and adapting something that already exists generally do not, because there was no real unknown. The line is whether you created new capability by systematically figuring something out. Custom work against a novel technical problem usually lands on the qualifying side.
Yes, and the payroll election is why. A qualified small business can apply up to $250,000 a year of R&D credit against payroll taxes even with no income tax liability, so you get cash back while you are still pre-profit. For a company investing in real custom development, that is money worth claiming.

Ready to move forward?

Start with structured discovery and a clear path to execution.