Funded Research: How Your Contracts Decide Who Gets the R&D Credit

Most R&D credit arguments are about engineering. This one is about paperwork nobody read for tax purposes.

If your company builds things for customers under contract - a development agency, a contract engineering firm, a defense supplier, a SaaS company running a paid pilot - the largest determinant of whether your engineering hours produce a credit is not how hard the problem was. It is what the signed agreement says about two things: who eats the cost if the work fails, and who gets to keep and reuse what you figured out.

The one-sentence exclusion

Work can clear all four parts of the four-part test and still produce nothing. IRC §41(d)(4)(H) removes from qualified research “any research to the extent funded by any grant, contract, or otherwise by another person (or governmental entity).”

Read plainly, that would wipe out nearly every services business in the country: every dollar an agency spends on engineering is, loosely, funded by the client paying the invoice.

But “funded” is a term of art, and it is much narrower than the plain reading. Treas. Reg. §1.41-4(c)(9) states the exclusion, then routes the analysis elsewhere: to determine the extent to which research is funded, §1.41-4A(d) applies. That regulation’s title says it governs years beginning before 1986, which throws people off. It is still the operative test, and where the modern cases go.

It asks two questions, and you have to answer both correctly:

1. Risk. Are the amounts payable to you contingent on the success of the research? Amounts contingent on success are treated as paid for the product or result rather than for the performance of research, and are not treated as funding.
2. Rights. Did you retain substantial rights in the research results? If the performer retains no substantial rights, the research is treated as fully funded, and none of the performer’s expenses are qualified research expenses.

Two mechanics matter. The regulation directs you to consider all agreements between the parties, so an MSA, a statement of work, and a separate IP side letter get read together. And the test runs contract by contract: a firm can have some engagements unfunded and others funded in the same year, for the same team. That is normal, not an anomaly.

There is also a middle case. Under §1.41-4A(d)(3), a performer who does retain substantial rights is still funded to the extent of the amounts it becomes entitled to by performing the research, with amounts contingent on success excepted. A single contract can be partially funded: a guaranteed base fee is funding, while success-contingent milestones layered on top are not. The statute’s “to the extent” means exactly that.

Prong one: who bears the risk

The regulation does not ask whether the project felt risky or whether you made money on it. Courts ask something narrower: if the work had failed to meet the contract’s requirements, who would have absorbed the cost?

The anchor case is Fairchild Industries, Inc. v. United States, 71 F.3d 868 (Fed. Cir. 1995, modified 1996). Fairchild held a fixed-price incentive contract with the Air Force to develop the T-46A trainer aircraft, and the credit at issue related to the full-scale development phase. The Federal Circuit worked through the payment machinery: the Air Force was obligated to pay only for work delivered and accepted; quality assurance provisions let it reject unacceptable work, require Fairchild to correct it at its own expense, or accept it at a reduced price; and unliquidated progress payments were repayable on a default termination, so they were not Fairchild’s to keep. The court’s reasoning tracked the regulation: where payment is contingent on the success of the research, the researcher bears the risk of failure. Held: not funded. (The case was decided under §44F, the statutory predecessor of §41, whose funded-research language is materially identical.)

Geosyntec Consultants, Inc. v. United States, 776 F.3d 1330 (11th Cir. 2015), is the other edge of the same blade. An engineering consultancy tested its position on six sample contracts; the fixed-price ones survived below. The two on appeal were capped cost-reimbursement agreements, under which Geosyntec billed labor and expenses up to an agreed maximum. The Eleventh Circuit affirmed those were funded. The reasoning is the useful part: they had no inspection or acceptance criteria and no rejection mechanism expressly conditioning payment on the success of the research, and monthly invoices were payable on submission unless a specific item was disputed. Clients owed money for hours worked, whatever the outcome. The court also made clear the inquiry is who bears the loss on failure, not whether the deal was profitable.

In Meyer, Borgman & Johnson, Inc. v. Commissioner, decided by the Eighth Circuit in 2024 affirming the Tax Court, a structural engineering firm lost on contracts that were fixed-price. The court distinguished proper performance from successful performance: obligations to comply with applicable codes or to work to a general standard of care do not mandate success. The agreements lacked the express terms courts have looked for, such as rejection language, payment limited to accepted work, or a refund if benchmarks were missed.

Put those together and the practical rule is clear: the label on the contract does not decide it. “Fixed price” is a strong starting position and cost-plus a weak one, but what carries the day is whether the document gives the customer a mechanism to refuse payment for work that does not meet a technical requirement.

Prong two: what counts as substantial rights

If you retain no substantial rights in the research, §1.41-4A(d)(2) treats it as fully funded no matter how much risk you took.

The binding authority is Lockheed Martin Corp. v. United States, 210 F.3d 1366 (Fed. Cir. 2000): a performer retains substantial rights where it may use the research results in its own business without paying for that use, even where its rights are not exclusive. Substantial rights do not mean sole rights. (§1.41-2(e)(3) makes the same point from the payer’s side: research can be performed on a payer’s behalf even though the payer lacks exclusive rights to the results.)

Populous Holdings, Inc. v. Commissioner (T.C. Docket No. 405-17) is the recent applied illustration. On cross-motions for summary judgment, the Tax Court issued an order in December 2019 for the taxpayer, an architectural design firm working under fixed-price contracts. On rights, the point was narrow and useful: the clients’ rights ran to ownership of the delivered documents, and nothing in the contracts prohibited the firm from using the research it had performed or required it to pay the client for that use. That was enough. Assigning the deliverable is not automatically fatal if you keep the right to use the underlying know-how. Note the posture, though: this is an unpublished order on summary judgment, not a published opinion. Tax Court orders are not treated as precedent, so it is persuasive reasoning only, and a position should rest on your contract terms rather than on the case.

The boundary on the other side is Dynetics, Inc. v. United States, 121 Fed. Cl. 492 (2015), where the Court of Federal Claims held the taxpayer had not carried its burden of showing substantial rights. Broad contract language vesting intellectual property in the other party did the damage, and the court treated the incidental benefits of doing the work, such as increased experience and skill, as short of substantial rights. That is the trap for services firms whose agreements contain a sweeping “all work product and all intellectual property therein shall be the sole property of Client” clause with nothing carved back.

What a reviewer actually looks for

Setting labels aside, here is the inventory a study works through per contract. General information, not legal advice, and no substitute for counsel drafting your agreements.

• Payment structure. Fixed-price and milestone deals are the friendliest starting point, because the price does not move when your costs do. Time-and-materials and cost-plus are hardest, because the customer is paying for effort. Capped cost arrangements read as cost-plus here, per Geosyntec.
• Acceptance, rejection, and rework. The highest-value paragraph in the agreement: objective acceptance criteria, a right to reject non-conforming work, and an obligation on you to correct it at your own cost. Fairchild turned on exactly this. A right to “inspect” with no rejection remedy does much less work, and invoices “payable on submission unless disputed” is the weak Geosyntec pattern.
• Warranty and remedy. A warranty that the deliverable will meet stated technical specifications, backed by repair, replacement, or refund, reinforces the risk position. One limited to professional standard of care generally does not.
• Termination and progress payments. Whether unliquidated progress payments must be repaid on a default termination, or whether the customer owes costs incurred to date regardless of outcome. The latter shifts risk to the customer.
• IP terms. Full assignment with no license back is the worst case. Assigning the deliverable while retaining the right to use the underlying methods, tools, and know-how is materially better. A perpetual license back is better still.

Government contracts add a data-rights layer on top of the payment analysis. On DoD work, DFARS 252.227-7013 and its companion clauses allocate technical-data rights by who paid for the development: unlimited rights for work developed exclusively with government funds, government purpose rights for mixed funding, and limited or restricted rights for data and software developed exclusively at private expense and properly marked. Civilian-agency work runs under FAR 52.227-14, a simpler scheme: the government generally takes unlimited rights in data first produced under the contract, with limited rights available for properly marked data developed at private expense. Under both, contractors commonly retain ownership while the government takes a license, and those markings interact with, rather than replace, the §41 rights analysis. §1.41-4A(d)(4) separately addresses independent research and development costs under the FAR.

The other side of the table

When a company pays someone else to do research for it, the same regulation decides whether the payer can claim. That route is contract research expenses. Under §41(b)(3) and Treas. Reg. §1.41-2(e), generally 65% of amounts paid to a non-employee for qualified research performed on the taxpayer’s behalf counts as a QRE, with higher percentages in narrow cases. But §1.41-2(e)(2) imposes conditions: the agreement must be entered into before the research is performed, must provide that the research is performed on the taxpayer’s behalf, and must require the taxpayer to bear the expense even if the research is not successful. If payment is contingent on success, the expense is treated as paid for the product or result rather than the performance of research. Our page on qualified research expenses covers how these amounts land in the calculation.

Notice the symmetry. The clause that makes a contract good for the performer, payment contingent on success, is the same clause that makes it bad for the payer. That is the design: the credit follows the party at economic risk.

The case where nobody gets it

§1.41-4A(d)(2) contemplates this directly. Where the performer retains no substantial rights and the payments to the performer are contingent on the success of the research, neither the performer nor the person paying for the research is entitled to treat any portion of the expenditures as qualified research expenditures.

Side by side, the mechanics are straightforward. The performer is out because the rights prong fails and the research is treated as fully funded. The payer is out because contingent payment is treated as buying a product or result rather than the performance of research, which is not what §1.41-2(e)(2) requires of a contract research expense.

Be careful how far you take this. It is a specific combination of facts, not a general rule, and it is why “make the contractor’s payment contingent and assign us all the IP” is a worse tax outcome than either party expects. Even where the payer clears the funding hurdle, the work still has to be qualified research in the payer’s hands, tested on the payer’s own business component and subject to the other §41(d)(4) exclusions. Our FAQ covers several adjacent questions.

A per-contract analysis, written down

Nothing above resolves at the company level. It resolves per agreement, on the words in that agreement, against case law that has moved in both directions since 1995. Two engagements with the same client, in the same year, on the same technology, can land differently because one had an acceptance-and-rejection clause and the other had “invoices payable on submission.”

So a defensible study for a contract-driven business runs the funded-research analysis per contract, records the conclusion with supporting clause references, then maps each contract to the business components and expenses it supports. Firms doing custom development in hardware and embedded systems tend to have the messiest version of this; our technology and hardware page covers what that looks like, and how it works lays out the sequence a study runs in.

Contracts are the one part of an R&D credit claim you can fix prospectively. The engineering already happened the way it happened. The next master services agreement has not been signed yet. One caution as you do: clauses only help where they describe the arrangement the parties actually perform. A recital of contingency that no one honors in practice is not a risk position.

Sources

• IRC §41 - Credit for increasing research activities (U.S. House, Office of the Law Revision Counsel), including §41(d)(4)(H) and §41(b)(3)
• Treas. Reg. §1.41-4A (Cornell LII), paragraph (d) on funded research
• Treas. Reg. §1.41-4 (Cornell LII), paragraph (c)(9) and its cross-reference to §1.41-4A(d)
• Treas. Reg. §1.41-2 (Cornell LII), paragraph (e) on contract research expenses
• Fairchild Industries, Inc. v. United States, 71 F.3d 868 (Fed. Cir. 1995); Geosyntec Consultants, Inc. v. United States, 776 F.3d 1330 (11th Cir. 2015); Lockheed Martin Corp. v. United States, 210 F.3d 1366 (Fed. Cir. 2000); Dynetics, Inc. v. United States, 121 Fed. Cl. 492 (2015); Populous Holdings, Inc. v. Commissioner, T.C. Docket No. 405-17 (order, Dec. 2019); Meyer, Borgman & Johnson, Inc. v. Commissioner, 100 F.4th 986 (8th Cir. 2024)
• DFARS 252.227-7013 - Rights in Technical Data (Acquisition.gov)
• FAR 52.227-14 - Rights in Data - General (Acquisition.gov)
• IRS - Research credit