TL;DR
Section 174 of the US tax code governs how research and experimental costs, which explicitly include software development, are deducted.
From 2022 through 2024, every US business had to amortize those costs over five years (fifteen for foreign work) instead of deducting them in the year spent. Startups that had never paid tax suddenly owed it on money they had already spent on engineers.
In July 2025 that was reversed for domestic spend: US-based software development is deductible in the year incurred again. It was not reversed for foreign spend, which stays on the fifteen-year schedule.
For a founder deciding how and where to build, that is a real input. It is not a reason to change the decision on its own, and for a pre-revenue startup it may not matter this year at all. Ask your CPA with your numbers. This post is background, not advice.
What Section 174 actually is
Section 174 of the Internal Revenue Code covers “specified research or experimental expenditures.” Since 2022 the statute has said in plain terms that software development is included. Not just moonshot R&D: any development of software, including the ordinary product engineering a startup does every day.
That definition is the reason a tax provision aimed at research became a founder problem. If you paid people to build your product, you were doing Section 174 spend whether you knew it or not.
Before 2022
Research costs, including software development, were deductible in the year incurred. A startup that spent $500,000 on engineers deducted $500,000. Most startups had no profit to deduct it against, so the loss carried forward. Simple.
2022 to 2024: the amortization years
A provision of the 2017 Tax Cuts and Jobs Act, delayed for five years, took effect for tax years beginning after December 31, 2021. It required Section 174 costs to be capitalized and amortized: five years for domestic research, fifteen years for research performed outside the United States, with only a half-year of amortization allowed in year one.
A startup that spent $500,000 on domestic engineering in 2023 could deduct roughly $50,000 of it that year. The other $450,000 sat on the balance sheet. Companies that had been profitable on paper by a small margin, or that expected a loss, found themselves with taxable income they had not planned for.
The industry spent three years lobbying against it.
July 2025: the partial reversal
The One Big Beautiful Bill Act, signed in July 2025, created a new Section 174A. For tax years beginning after December 31, 2024, domestic research and experimental expenditures are once again deductible in the year incurred. Small businesses could elect to apply the change retroactively to 2022, and there were transition options for the amounts still sitting on balance sheets.
Foreign research and experimental expenditures were left where they were: capitalized and amortized over fifteen years.
That asymmetry is the whole reason this post exists.
Domestic versus foreign, side by side
| Domestic development | Foreign development | |
|---|---|---|
| Year-one deduction on $100k | $100,000 | ~$3,300 |
| Recovery period | Immediate | 15 years, half-year convention |
| Governing section | 174A (new, 2025) | 174 (unchanged) |
| Applies to | Work performed in the US | Work performed outside the US |
| What decides which | Where the work is done | Not where the vendor is incorporated |
The last row is worth a second look. What matters is where the research is performed. A US company paying a US vendor whose engineers sit abroad is, on the face of it, paying for foreign research. A foreign company with engineers physically in the US is closer to domestic. Your CPA will want to know where the people are, not where the invoice comes from.
What this means for a founder
If you are pre-revenue
Most early-stage startups are not paying income tax, because they have no taxable income. A deduction you cannot use this year is not worth anything this year, whether it is $100,000 or $3,300.
That does not make the difference disappear. Domestic losses carry forward and are available when the company becomes profitable. Foreign amortization also carries forward, just on its schedule. The gap shows up in the year you first owe tax, and its size depends on how much you spent abroad in the years before.
Ask your CPA: “We are pre-revenue. If we spend $X on offshore development this year, when does the Section 174 treatment actually cost us anything, and how much?”
If you are profitable
Then the difference is real and immediate. Deducting $100,000 this year against $3,300 this year is, at a 21 percent federal rate, a $20,000 swing in tax paid, before state tax. That has to go into the comparison of an offshore price against a domestic one.
Set against a build that is 40 to 60 percent cheaper offshore (the full breakdown is in what an MVP costs), the deduction timing rarely flips the decision. But it narrows the gap, and a founder should know by how much before choosing.
Ask your CPA: “Run the offshore quote and the domestic quote through our tax position for the next three years and tell me the real after-tax cost of each.”
If you raised on a SAFE and are burning
You are almost certainly pre-revenue and the analysis above applies. Whether you needed an MVP to raise at all is a separate question; having raised, the build spend is now a Section 174 question. But note one thing for the future: investors’ lawyers in a priced round or an acquisition will ask about capitalized Section 174 balances. A large foreign amortization balance is not a problem, but it is a line item, and a founder who can explain it in one sentence looks like a founder who knows their numbers.
What this does not mean
Because the rule is technical, it gets misrepresented in both directions.
It does not make offshore development a bad idea. It is a cash-timing difference, and for most startups a small one this year. Most US software companies use development outside the US in some form.
It does not apply only to “research.” Ordinary product engineering is Section 174 spend. Building your MVP counts.
It does not depend on the vendor’s contract type. Employee, contractor, agency, fixed-price, hourly: if it is software development, it is Section 174. The fixed-price versus hourly decision is unaffected.
It is not a reason to skip the CPA conversation. The specifics, including the retroactive election for small businesses and the transition rules for existing balances, are exactly the kind of thing that turns on your particular facts.
Section 174 is not the R&D tax credit
Founders mix these up constantly, and the confusion costs money in both directions.
Two different sections, two different things
Section 174 is about *deductions*: when you get to subtract development spend from income. That is everything above.
Section 41 is the *research credit*: a dollar-for-dollar reduction in tax, calculated as a percentage of qualifying research spend. For a startup with no income tax to reduce, the credit can instead be applied against payroll taxes, up to $500,000 a year, for up to five years. That is real cash for a pre-revenue company with US employees.
They interact, and a CPA handles the interaction. What a founder needs to know is that they are separate questions with separate answers.
Where offshore work sits for the credit
The credit has its own geography rule, and it is stricter. Qualified research for Section 41 must be conducted in the United States. Spend on a foreign development team does not generate the credit at all, whatever it does for the deduction.
So a US startup using an offshore team for the build and US engineers for anything else should track them separately: the domestic payroll may earn the credit, the offshore spend will not. A founder who lumps them together forfeits a credit they were entitled to on the domestic portion.
Ask your CPA: “Do we qualify for the payroll tax offset under the research credit, and which of our development spend counts?”
Why this matters more than it sounds
A pre-revenue startup with $400,000 of domestic engineering payroll might be looking at a payroll tax offset in the tens of thousands of dollars. That is more than the Section 174 timing difference on most offshore builds. If you only have one tax conversation this year, this is the one to have.
Does software development qualify for the R&D credit?
Often, but not automatically, and the test is specific. Work qualifies under Section 41 when it passes all four parts of the IRS test: it is intended to create a new or improved function, performance, reliability or quality (not just a new feature that uses known techniques); it relies on a hard science, and computer science counts; there is technical uncertainty at the outset about whether you can build it, how, or what the design should be; and you resolve that uncertainty through a process of experimentation, meaning you evaluated alternatives and tested them. Building a standard CRUD app with a known stack usually fails the third and fourth parts. Building a matching engine, a novel data pipeline, an ML feature or an integration nobody has done before usually passes them. Most MVPs contain some of each, and the credit applies to the portion that qualifies.
Software you build for your own internal use faces a higher bar (the IRS wants it to be genuinely innovative, involve significant economic risk, and not be commercially available). Software you sell, or that customers interact with, does not face that extra bar.
Contractor spend and the 65 percent rule
Payments to an outside development team can count, at 65 percent of the amount, if three things hold: the work is performed in the United States, you keep substantial rights to the results, and you bear the economic risk, meaning you pay whether or not the work succeeds. That last condition is a contract question. A fixed-price agreement where the developer only gets paid on delivery can shift the risk to them and disqualify the spend; a time-and-materials or milestone agreement where you pay for the work regardless usually keeps it. If you are choosing a contract shape with a domestic team and the credit matters, that is worth raising before signing, and the development contract should say who owns the output.
What to get from your development partner
The credit is claimed on documentation, and the documentation has to exist at the time the work is done, not be reconstructed at tax time. From a development team, in-house or contracted, that means: time recorded by person and by activity, so qualifying work can be separated from routine work; a short technical narrative for each qualifying piece describing the uncertainty at the start and what was tried; the alternatives considered and why they were rejected; and the test records that show the experimentation. A development partner who invoices “sprint 4, $18,000” has given you nothing to claim on. One who can hand over the sprint’s tickets, the design decisions and the test runs has given you the file. Ask for it before the build starts, because it costs nothing to keep and cannot be recreated afterwards.
Three founder situations
The pre-revenue SaaS with an offshore build
$60,000 fixed-scope build with a team abroad, no revenue, a SAFE in the bank. Section 174 puts the $60,000 on a fifteen-year schedule; this year’s deduction is about $2,000. This year’s taxable income is zero, so the deduction is worth nothing either way.
What the CPA does: notes the balance, sets up the amortization, and tells the founder it becomes relevant in the first profitable year. Twenty minutes. No change to the build decision.
The profitable company adding a product
A services business with $300,000 of taxable income decides to build a software product. Two quotes: $150,000 domestic, $65,000 offshore.
Domestic: the full $150,000 is deductible this year, reducing taxable income to $150,000. Offshore: about $2,200 is deductible this year, leaving taxable income at roughly $298,000. At a 21 percent federal rate the domestic deduction is worth about $31,000 in tax this year; the offshore one about $500.
The offshore build is still $85,000 cheaper on the invoice and roughly $54,000 cheaper after this year’s tax effect, with the remaining foreign deduction still to come over the following years. Cheaper, by less than the invoice suggests. The founder should know that number, and now does.
The hybrid team
Two US engineers on payroll, an offshore team doing the bulk of the build. Domestic payroll is expensed under 174A and may earn the Section 41 payroll credit. Offshore invoices are amortized and earn no credit.
What the CPA needs: separate invoices, separate cost centers, and invoices that say where the work was performed. What the founder gets: the credit on the domestic side, correctly classified spend on the foreign side, and no April surprises.
How to build with this in mind
Three practical habits, none of which require changing where you build.
Get invoices that say where the work was done
Your bookkeeper needs to classify the spend. An invoice that says “software development services” leaves them guessing. One that says “software development services performed in [country]” does not. Ask your vendor for the wording; a vendor that works with US companies will already do it.
Keep domestic and foreign spend separable
If you use both, a US contractor for some work and an offshore team for the rest, keep the invoices and the descriptions distinct. Mixed invoices create classification work for your CPA that you pay for by the hour.
Have the conversation before the contract, not in April
The whole point of knowing this is to price the decision correctly. A twenty-minute CPA call before you sign is cheaper than a surprise at year end, and it is one of the six checks in our guide to hiring an offshore team as a US company.
Where this sits in the build decision
The honest ranking of what decides an MVP build, for most founders: the price, whether the team can ship, the working rhythm, the contract, and then the tax timing. Section 174 is fifth. It is real, it should be priced, and it almost never changes the answer the first four produce.
Our comparison of offshore, nearshore and onshore covers the first three. The contract guide covers the fourth. This post covers the fifth so that it is known rather than discovered.
The short history, for the founder who wants to explain it in a meeting
- 2017: The Tax Cuts and Jobs Act schedules mandatory amortization of research costs, effective 2022, as a revenue offset. Nobody expects it to survive.
- 2022: It takes effect. Software development is explicitly included. Startups discover tax bills on money already spent on engineers.
- 2023 to 2024: Repeated bipartisan attempts to reverse it fail to pass. Companies amortize, or restructure, or both.
- July 2025: The One Big Beautiful Bill Act creates Section 174A, restoring immediate expensing for domestic research from tax years beginning in 2025, with retroactive elections for small businesses. Foreign research stays on fifteen-year amortization.
- Now: Domestic expensed, foreign amortized, and the distinction is where the work is performed.
That is enough to sound informed in a board meeting and not enough to do your own return.
Conclusion
Section 174 turned ordinary software development into a tax question in 2022, and the 2025 change fixed it for domestic work while leaving foreign work on a fifteen-year schedule. For a founder that means one thing: know the treatment of your build spend before you sign, price it, and let the bigger inputs decide.
For most pre-revenue startups it changes nothing this year. For profitable ones it narrows the offshore price gap without closing it. In neither case is it a reason to guess.
We work with US founders who have had this conversation with their CPA and US founders who have not, and we are happy to be the vendor that raises it first. Invoices describe where the work was done, the W-8BEN-E is ready, and the numbers are yours to run. The terms are on the offshore MVP development page. Start here.
*This post is general information for founders, not tax advice. The rules described changed in July 2025, include elections and transition provisions that depend on your facts, and may change again. Confirm your position with a CPA before relying on any of it.*
Frequently Asked Questions
What is Section 174?
Section 174 of the US Internal Revenue Code governs how research and experimental expenditures are deducted. Since 2022 the statute has explicitly included software development, so ordinary product engineering by a startup is Section 174 spend. Between 2022 and 2024 those costs had to be amortized over five years (fifteen for foreign work). From 2025, domestic costs are deductible in the year incurred again; foreign costs are still amortized.
Did the 2025 change fix Section 174 for startups?
For domestic software development, yes. The One Big Beautiful Bill Act created Section 174A, which restores immediate expensing for research performed inside the United States for tax years beginning after December 31, 2024, with retroactive options for small businesses. It did not change the treatment of research performed outside the United States, which remains on fifteen-year amortization.
Does Section 174 apply to offshore software development?
Yes. Software development performed outside the United States is a foreign research expenditure under Section 174 and is amortized over fifteen years rather than deducted in the year incurred. What determines the classification is where the work is performed, not where the vendor is incorporated or where the invoice is sent from.
How much of an offshore development invoice can I deduct this year?
Under fifteen-year amortization with a half-year convention, roughly 3.3 percent in year one. On $100,000 that is about $3,300, with the balance deducted over the following years. Domestic development of the same amount would be fully deductible this year. Confirm the exact figures with your CPA.
Does Section 174 matter if my startup is not profitable?
Not much this year. A deduction only reduces tax you would otherwise pay, and a pre-revenue startup with no taxable income pays none. Both domestic losses and foreign amortization carry forward, so the difference appears in the year the company first owes tax. Its size depends on how much was spent abroad in the years before.
Should I stop using offshore developers because of Section 174?
No, not on its own. It is a cash-timing difference that rarely offsets a 40 to 60 percent price gap, and for a pre-revenue company it is close to zero this year. It should be priced and known, and your CPA can do that in twenty minutes with your actual numbers. It should not decide where you build by itself.
Does the type of contract change the Section 174 treatment?
No. Employee, contractor, agency, fixed-price or hourly, if the spend is software development it is Section 174 spend. The domestic-versus-foreign distinction depends on where the work is performed, not on the contract structure.
What should my invoices say?
Where the work was performed. An invoice for “software development services performed in [country]” lets your bookkeeper classify the spend without guessing and supports both the Section 174 classification and, for foreign vendors, the no-withholding position. Vendors experienced with US clients will word invoices this way if asked.
Is software development really “research” under Section 174?
Under the statute as written since 2022, yes. Section 174 explicitly includes any amount paid or incurred in connection with the development of software. It does not require the work to be novel or experimental in the everyday sense. Building an MVP counts.
What should I ask my CPA about Section 174?
Three things. Whether your development spend this year is domestic or foreign, and what that means for this year’s return. Whether you should make any of the retroactive or transition elections available after the 2025 change. And, if you are choosing between an offshore and a domestic build, what the after-tax cost of each is over the next three years given your expected tax position.
Does software development qualify for the R&D tax credit?
When it passes the four-part test: a new or improved function or capability, reliance on computer science, technical uncertainty at the outset, and a process of experimentation to resolve it. Routine builds on a known stack usually do not qualify; novel engineering, data pipelines, ML features and first-of-their-kind integrations usually do, and most MVPs contain some of each. Contractor spend counts at 65 percent if the work is done in the US, you keep the rights and you bear the risk, and the claim depends on contemporaneous records of time, uncertainty, alternatives and tests from the development team.
