The sponsor's guide to university research agreements: IP, publication, overhead, and time to signature

If your company is about to fund research at a university, the agreement will assert four things you may not expect. The university will usually own what its people invent, and offer you a license or an option. Its researchers will publish, after you've had a limited time to review. Your budget will carry the university's indirect-cost rate, often more than half of the direct costs. And it will take months to sign, unless a previous agreement is already in place that can shorten that window. This guide explains each of those, why the university holds the position it does, what's open to negotiation, and what the evidence says shortens the clock. It's written for the sponsor, with asides for the faculty member on the other side of the table.

We're not a law firm, and this isn't legal advice. It's the practitioner's map of the territory, so that the first meeting with the university's contracts office is about terms rather than positions.

Which agreement are we talking about?

A sponsored research agreement (SRA) is the contract under which a company pays a university to carry out a defined piece of research, with terms on scope, budget, intellectual property, publication, confidentiality, etc. It's one of a family, and the first question a sponsor should settle is which member of the family fits the situation.

Agreement	What it's for	How it differs from an SRA
Gift	Money given to support a lab or a field, with no deliverables and no rights	No scope, no IP terms, no report beyond stewardship. If you want results or rights, it isn't a gift
Grant	Funding awarded under a program's rules, usually by a government agency or foundation	The funder's terms govern, not a negotiated contract. Federal grants carry Bayh-Dole rights to inventions
Contract research (fee for service)	The university performs a defined test or service to the sponsor's specification	Little or no new invention is expected, and the sponsor often gets the results outright. Most universities limit this to work that isn't research in the academic sense
Collaborative research agreement	Both sides contribute effort and resources to a shared project	Each side brings background and may own some of the results. IP terms are reciprocal rather than one-way
Master research agreement	A framework signed once, with individual projects added as statements of work	The IP, publication, and liability terms are settled in advance. Each project then needs only a scope and a budget
Nondisclosure agreement (NDA)	Lets the two sides talk about a possible project	No research, no money. Often the first agreement, and a diagnostic of how the rest will go
Material transfer agreement (MTA)	Moves materials (cells, compounds, samples, prototypes) between the parties	Governs use of the material, not the research around it. Biological materials usually move under the Uniform Biological MTA
Data use agreement (DUA)	Moves data, often with privacy or confidentiality conditions	Governs what the recipient may do with the data and for how long
Cooperative research & development agreement (CRADA)	Cooperative research with a federal laboratory	Governed by its own statute (15 U.S.C. 3710a), with the lab's standard terms and a different IP default
License	Grants rights to use existing university IP	Comes after the invention exists. An SRA may contain an option to negotiate one

Most of what follows is about the SRA, with the master agreement as the shortcut and the ancillary agreements (NDA, MTA, DUA) as the pieces that travel with it. The glossary has an entry for each, written for both sides.

Who owns the IP, and why isn't it the sponsor?

This is the question university contracts offices get asked most, usually in the form "we're paying for it, so why don't we own it?" The university's answer has three layers.

The law, where federal money is involved. Under the Bayh-Dole Act, a university that makes an invention with federal funding may elect to keep title, subject to a license to the government and other conditions (35 U.S.C. 202 and the standard clause at 37 CFR 401.14). Most research labs run on a mix of funding, and once a federal dollar touches a project, the university has obligations it can't contract away to a sponsor. The Supreme Court added a wrinkle in 2011: Bayh-Dole doesn't automatically vest title in the university, because rights start with the inventor, so universities rely on assignment agreements with their researchers (Stanford v. Roche, 563 U.S. 776). The practical consequence for a sponsor is that a university can only give you what its inventors have assigned to it, and a well-run office will have those assignments in place.

The policy, everywhere else. Even without federal money, nearly every US research university's IP policy says the institution owns inventions made by its employees in the course of their work, and licenses them. That policy exists partly because of Bayh-Dole and partly because of what predates it. Mowery, Nelson, Sampat, and Ziedonis found that university patenting and licensing were already growing before 1980, and that the Act was "an important, but not a determinative, factor" in what followed (Research Policy, 2001, with the quoted phrase from the authors' companion chapter). Thursby and Thursby, in Management Science, found that the growth in licensing came mainly from universities and faculty becoming more willing to license, and from business relying more on external research, rather than from any change in what faculty studied. The point for a sponsor is that institutional ownership isn't a bargaining position invented for your deal. It's the operating model.

The practical reason. A university can't price an invention that doesn't exist yet, and it can't assign away something a graduate student might make next year under a different grant. What it can do is commit, in advance, to the terms on which you'll get rights when something is invented.

That's what the four common pathways do:

1. University ownership with an option. The university owns what its people invent. The sponsor gets a time-limited option to negotiate a license, often exclusive, at terms to be agreed. This is the default at most institutions.
2. Field-of-use license. The sponsor gets rights in the field it cares about, and the university keeps the rest. Narrower, and usually cheaper and faster, because it leaves the university free to license elsewhere.
3. Pre-priced menus. A growing number of universities publish fixed terms so the IP negotiation is a selection rather than a negotiation. The University of Wisconsin-Madison's Badger IP Industry Advantage offers three: a no-upfront-fee route where the license is negotiated when an invention appears, a $10,000 option to a non-exclusive license in a field of use, and a pre-paid exclusive license at 10% of the project budget (or $15,000) for one field of use or 15% (or $22,500) for all fields, royalty-free unless annual sales pass $20 million. Stony Brook's CLEAR-IP offers a pre-paid non-exclusive royalty-free license for the greater of $10,000 or 15% of the project budget, an exclusive license for the greater of $15,000 or 20%, and a traditional route. The University of Minnesota's MN-IP sets a one-time upfront fee, royalty-free non-exclusive licenses, and a 1% royalty on exclusive licenses only once product sales pass $20 million a year.
4. Specific language for the new cases. Software, data, and AI-assisted work don't fit the patent-shaped defaults. The USPTO's revised inventorship guidance of November 2025 restates that only natural persons can be inventors, so an agreement for AI-heavy work should say who the human inventors are expected to be and how AI tools will be documented. Data ownership and use need their own clause (the UIDP Contract Accords have one), separate from inventions.

Two things to put in front of your counsel before the first draft. First, what you actually need: freedom to operate in your field, an exclusive position, or just the results. Sponsors who ask for everything get the slowest negotiation and often the least. Second, which pathway the university already publishes. If it has a pre-priced menu, the fastest deal is to pick from it.

In a study of 38 Advanced Technology Program projects, Hall, Link, and Scott found that of the 25 projects with no university partner, 12 said intellectual property issues had been an insurmountable barrier to working with one. None of the 13 projects that did have a university partner said so. The IP conversation is where sponsors most often walk away, and it's what a published pathway is designed to shorten.

Translation Box: "deliverable"

On the academic side: the research plan, the reports, the publications, the trained students. A deliverable is what the process of science produces, and a result that disproves the hypothesis is still a deliverable.

On the industry side: a specified output on a date, often a prototype, a dataset, or a recommendation, with acceptance criteria. "Deliverable" implies it works.

Write both into the statement of work. Name the reports, the review meetings, and any prototypes separately, and say what happens if the experiments lead somewhere else.

Background IP, foreground IP, and the inventory to keep

Agreements divide IP into what each side brings (background) and what the project creates (foreground, or "project IP"). The model agreements all do this, and they differ on improvements to background and on who may use what after the project ends. The common failure is that neither side lists its background before signing, so the first dispute is about whether a result was invented or merely applied.

The fix is a one-page background-IP inventory, filled in by each side before the term sheet: the patents, software, data, materials, and know-how each party will bring, what the other side may do with each during the project, and what happens after. Our article on background IP versus foreground IP walks through the clause language in the Lambert, DESCA, and US federal frameworks, and comes with the inventory as a free download.

Can the sponsor control publication?

No. Not at a US research university, and not in most of the world. The university's researchers will publish, and their graduate students will defend theses, because that's what the institution exists to do and because its tax status and federal funding depend on it. What a sponsor can get is a limited review period before submission, to check for its confidential information and to file patents, and sometimes a short further delay for that filing.

The standard is well documented. In a survey of 107 academic medical centers, Mello, Clarridge, and Studdert found that 96% would accept a provision letting the sponsor review manuscripts for a limited time before publication, while 89% considered sponsor authority to revise a manuscript impermissible and 93% said the same of sponsor authority to decide whether results are published (NEJM, 2005). Asked the longest review period they'd allow, 44% said 60 days and 31% said 90, with 14% at 30 days and only 7% beyond 90. In the same survey, the terms most often rated "very difficult" to negotiate for multicenter trials were ownership of inventions and IP (31%) and ownership of the data (25%), and the disputes institutions reported with sponsors in the previous year were about payment (75%), intellectual property (30%), and control of or access to data (17%). An earlier survey of 108 medical schools found a median of only 1% of site agreements complied with the international editors' standard that authors have access to all the trial data (Schulman et al., NEJM, 2002), which is why data-access clauses have tightened since.

What to ask for: a review window (30 days is common, 60 is where most institutions draw the line, 90 is the longest), a further delay only if a patent filing is needed, the right to have your confidential information removed, and nothing perceived as a veto. A dissertation carve-out, so a student's defense isn't held hostage to a filing decision, is a sign the agreement was drafted by people who've done this before. Our article on whether a sponsor can block publication sets the review and delay terms of the UIDP Accords, the Lambert agreements, DESCA, and the NIH policies side by side, with the faculty view of the same clause.

Why is there an indirect-cost rate, and is it negotiable?

Every university budget carries facilities and administrative costs, usually called F&A or indirect costs: the buildings, utilities, compliance, and administration that a project uses but doesn't buy directly. Federal agencies pay them at a rate negotiated with each university under the Uniform Guidance (2 CFR 200.414). The Congressional Research Service reports the median negotiated rate rising from 43% in the 1980s to about 56% in 2024 (R48540). The National Institutes of Health tried to cap the rate at 15% in February 2025. A federal court vacated the policy that April, the First Circuit affirmed in January 2026, and the government let the deadline to ask the Supreme Court pass in April 2026, so as of this writing the negotiated rates stand.

For a sponsor the rate is applied to most direct costs, so a $300,000 project at a 56% rate costs about $468,000 depending on what the base excludes (equipment and large subcontracts usually sit outside it). Universities charge industry sponsors their negotiated rate, or a published industry rate, for a reason they state plainly: if industry work paid less than cost, the university would be subsidizing a commercial sponsor with money meant for its own mission. Some institutions do publish reduced rates for specific programs (several waive or cut the rate on Phase I SBIR and STTR subawards, for example), and some accept a lower rate for a gift-like project with no deliverables and no rights. The rule is that the sponsor can't negotiate the rate down, but the university can choose to charge less, so ask what's published before you ask for an exception. The article on indirect costs on industry-sponsored research has a worked example and the list of what's negotiable.

Translation Box: "overhead"

On the academic side: facilities and administrative costs recovered at a rate audited and negotiated with the federal government, and the subject of a running fight about whether even that rate covers the real cost.

On the industry side: overhead is a line the finance team tries to cut, and a 56% addition looks like a markup.

It isn't a margin. It's the cost of running the facilities where the project happens, priced the way the university's largest funder requires. Budget for it from the first conversation, and negotiate the scope rather than the rate.

How long will this take, and what makes it faster?

Months, usually. UIDP asked its members in 2022 how long sponsored research agreements take: 44% said three to six months, 22% said under three, and about one in five said ten months or more. The two sides didn't agree with each other. Of the university respondents, 71% said agreements close within six months, while 54% of the industry respondents believed seven months or more. The survey was small (64 respondents, three quarters of them from universities), but the gap is the finding: each side thinks the other is slow, and each is measuring from a different starting line.

The evidence on what shortens the clock is more positive than the evidence on how long it usually takes. Across five University of California medical campuses, contracts finalized under a pre-negotiated master agreement took 39 days on average against 73 without one, a 47% reduction through the same legal offices on the same kind of agreement (Tran and colleagues, Research Management Review, 2017). A separate multisite study found that a standard clinical-trial agreement template saved an average of 48 days against custom negotiation, and the Federal Demonstration Partnership's template saved 57 days against a baseline of more than 100 days for a negotiated contract (Lawrence and colleagues, Journal of Clinical and Translational Science, 2023). An earlier observational study of 568 contracts at 29 academic medical centers, almost all of them industry-sponsored trials, found a median of 39 days to final negotiation and 91 days to full execution (Kiriakis and colleagues, Clinical and Translational Science, 2013). Those are clinical-trial numbers, and an SRA for engineering research isn't the same instrument, but the mechanism carries over: time is spent on the terms that haven't been settled before, not on the ones that have.

So the sequence that works:

1. Sign an NDA first, and watch how it goes. It's the smallest agreement in the family and a reliable diagnostic. If the NDA takes six weeks, plan for the SRA accordingly.
2. Agree on a term sheet before anyone drafts. One document: scope, money, the IP pathway (by name, if the university publishes one), the publication window, confidentiality, and who signs. Hand that to counsel, not a blank page.
3. Ask who has signature authority. At most universities it isn't the faculty member and it isn't the department chair. It's the sponsored programs office or a named institutional official, and the agreement waits in that queue. Find out where the queue is and how long it is.
4. If you'll fund more than one project, negotiate a master agreement once. The University of Minnesota signed a master research agreement with Cisco in about three months, with funding to deploy on a deadline, because its IP options were already published and the framework was waiting (UMN, 2021). Projects then go under it as statements of work.
5. Use the published standards. The UIDP Contract Accords are a set of agreed positions on the clauses that take longest (publication, IP, data, liability, budgeting). Many US contracts offices treat them as the reasonable middle. The UK's Lambert Toolkit and the EU's DESCA model consortium agreement do the same job in their jurisdictions, and both are free.

For the research office reading this, the same evidence is the case for measuring your own cycle times by agreement type and tracking every negotiation against a plan.

Typical ranges, in our experience, once the terms above are settled: an NDA in one to four weeks with templates, an MTA in one to four weeks, an SRA in two to six months, and a license in three to twelve months (two to four weeks where the university runs an expedited process). Treat these as practitioner estimates, not benchmarks.

A CRADA with a federal laboratory is the outlier, and the published figures run from weeks to more than a year depending on the agency and the terms. NOAA says its CRADAs take one to two months on average. The EPA quotes about three months for a basic agreement, but its own reports put the average elapsed time at four to 13 months from FY2019 to FY2023. The National Cancer Institute expects negotiations to take no longer than six months once the parties agree on the goals. At the Department of Energy's national laboratories, the GAO measured an average of about 7.5 months in the early 1990s, the department set a 60-day target in 2012, and Sandia reported in 2024 that 78% of agreements under its new master scope of work closed in 30 days or less. What stretches a DOE agreement past a year is the terms a sponsor asks to change. Sandia says a CRADA with a company that isn't US-owned adds two to three months of review, the GAO found delays of up to six months where the US Trade Representative had to be consulted, and a request to waive the US manufacturing requirement goes to the department's general counsel, is rarely granted in full, and has no published clock beyond a 30-day window for the contracting officer to act once a decision is made. In our own experience, CRADAs that needed a manufacturing waiver ran 12 months or more. Plan on the agency's published range if you can live with its template, and on a year or more if you can't.

What are the red flags?

Some clauses are worth a second look from either side: a sponsor veto over publication, an open-ended confidentiality term that never expires, assignment of all results without reference to federal funding, a liability or indemnity clause that one side can't lawfully give (public universities usually can't indemnify), a non-compete on the researcher, and silence on what happens to funded students if the project ends early. Our free Contract Red Flags Kit lists 54 clauses across the checklist, consortium, and materials-transfer cases, with the balanced position and the sponsor's legitimate concern for each.

Three things that surprise first-time sponsors

Export control. Research done in the open, with no restriction on publication, usually qualifies as fundamental research and sits outside export-control licensing (15 CFR 734.8). The moment a sponsor asks for a publication veto or restricts who may work on the project, that exclusion can be lost, and the university may have to treat the project as controlled. This is one of the strongest reasons universities refuse publication control, and a sponsor with export-sensitive technology should raise it on day one.

Antitrust, for consortia. If you're joining a research consortium with competitors, the National Cooperative Research and Production Act judges a notified research joint venture under the rule of reason rather than per se, and limits damages to actual rather than treble damages for conduct within a filed notification (15 U.S.C. 4301-4306). Gundlach and Mohr's analysis in the Journal of Public Policy and Marketing is a readable account of why intermediate forms of collaboration confused antitrust analysis in the first place. Ask whether the consortium has filed.

March-in rights. Sponsors sometimes worry that the government can take back federally funded inventions. It can, in principle, under Bayh-Dole's march-in provision. In practice the Government Accountability Office reported in February 2026 that agencies have never exercised them, and cited research finding that of 883 new drugs approved from 1985 to 2022, only 18 were covered entirely by government-interest patents (GAO-26-107885, citing Ouellette and Sampat, JAMA Health Forum, 2024). It belongs in your risk register as a line, not a page.

After signing

The people who negotiate the agreement are rarely the people who run the project, and the hand-off is where good agreements go quiet. Put the agreement's dates (the publication-review window, the confidentiality term, any IRB approval, the NDA and contract expiry) somewhere the project team will see them, name who owns each one, and hold the first kickoff meeting within a month of signature. Our guide to managing a university-industry research collaboration covers the governance, and the tutorial on tracking research contracts and proposals shows how a project on Helikon records the agreement, its signed copy, its expiry, and the reminders, with the audit trail kept.

In the Helikon Method, this is the Unite phase: the relationship is designed, written down, and signed in a way the Steward phase can actually run. The sponsors who do it well arrive at the contracts office with a term sheet, a named IP pathway, a publication window they can live with, and a budget that already includes the overhead. The rest is paperwork that can move quickly.