Article

The economics of no

Most university inventions never earn back what it costs to patent them, and nobody can tell at the start which few will. The work of a technology transfer office is deciding, gate by gate, when to stop paying -- and in 2025-26 that decision became more expensive to get wrong.

university technology transfer -- what the office does, why most of its patents lose money, and where in the timeline an assessment pays · 2026-10-16 · 37 sources

A technology transfer office is the university department that moves what its researchers invent into commercial hands. An inventor files a disclosure. The office decides whether to patent it. It markets the patent to companies, or helps the inventor found a startup. It negotiates a licence and splits whatever royalties arrive among the inventor, the department and the university. That is the brochure. Read from the numbers, the job is something else: the craft of saying no at the right price. Most disclosures will never be licensed, most patents will never earn back their fees, and no one can pick out in advance the handful that pay for everything. So the decision that matters is seldom "is this invention a winner?" It is "is this worth paying for the next gate?" -- a question that has just become costlier to answer badly.

The funnel

American universities spent $112 billion on research in fiscal 2025 and reported 28,278 invention disclosures. Those disclosures produced 13,729 new US patent applications, 7,827 issued patents, 10,277 licences and options, 1,095 startups and 761 new products.1AUTM, 2025 US Licensing Activity Survey infographic (FY2025 data): $112B research expenditure; 28,278 invention disclosures; 13,729 new US patent applications; 7,827 US patents issued; 10,277 licences and options; 1,095 startups formed; 761 new products. https://autm.net/AUTM/media/SurveyReportsPDF/AUTM-2025-US-Infographic.pdf (T1 -- the survey owner's own data release) Applications per disclosure ran at 0.49, down from 0.55 the year before, when, on relayed counts, 26,196 disclosures produced 14,432 applications.2FY2024 totals -- 26,196 disclosures, 14,432 new US patent applications, 7,968 patents issued, 9,507 licences and options, 951 startups -- as relayed by two secondary summaries of the AUTM FY2024 infographic, whose original URL now returns 404; cross-checked against AUTM's stated FY2025 year-on-year changes (+7.9% disclosures). Ratios 0.49 and 0.55 are our derivation. (T3 -- relayed figures, original not retrieved) That ratio is not the share of disclosures patented, since one disclosure can spawn several filings and filings lag. The direction, though, is a choice.

The cohort evidence agrees. Stanford's office tells its inventors that about half of disclosures get an application filed and 20-25% are eventually licensed.3Stanford Office of Technology Licensing, FY2015 overview: about half of disclosures lead to a patent application; 20-25% of disclosures are licensed; Google licence, $341 million; triage question "evolutionary or revolutionary?" https://web.stanford.edu/group/OTL/documents/fy15OTL_overview.pdf (T1 -- the office's own record) Of the patents granted to the 58 most productive American universities between 2002 and 2010, 29.7% were licensed out; the rest found no licensee.4Caviggioli F, et al. Technological Forecasting and Social Change, 2020: of patents granted 2002-2010 to the 58 US universities in the sample, 29.7% were licensed. https://www.sciencedirect.com/science/article/pii/S0040162520310155 (T2 -- peer-reviewed analysis) A single year's ratios are not a cohort -- a disclosure from 2025 may find its licensee in 2031 -- which is why the cohort shares are the ones to remember. Roughly half of what comes in the door is never filed, and of what is filed, well under half is ever licensed.

Who pays for the office

Licensing income obeys a power law. Stanford's office, by its own account in 2011, had handled some 8,300 disclosures since its founding. Three of them produced 67% of its cumulative income, and 68 had earned $1 million or more. "It took us 15 years to break even," its director told a National Academies audience.5Ku K, Stanford OTL, remarks to a National Academies symposium, January 13, 2011: ~8,300 cumulative disclosures; 3 inventions produced 67% of cumulative income; 68 cases earned $1 million or more; "It took us 15 years to break even." https://sites.nationalacademies.org/cs/groups/pgasite/documents/webpage/pga_060747.pdf (T1 -- the office's own account, in the symposium record) The University of California told its Regents in January 2012 that fewer than 0.2% of disclosures accounted for 80% of revenue, that it is "almost impossible to predict which inventions have blockbuster potential at the time they are disclosed", and that an invention that does pay takes ten years or more to do so.6University of California Office of the President, item F9 to the Regents, January 19, 2012: fewer than 0.2% of disclosures make up 80% of revenue; "almost impossible to predict which inventions have blockbuster potential at the time they are disclosed"; ten or more years to revenue. https://regents.universityofcalifornia.edu/regmeet/jan12/f9.pdf (T1 -- institutional record)

Those are the winners. Across the system, a Brookings study of 155 AUTM-reporting offices found that fewer than 13% earned enough licensing income to cover their operating costs, and that about 75% of gross licensing income between 2003 and 2012 went to 16 universities, a tenth of the sample.7Valdivia WD, "University Start-Ups: Critical for Improving Technology Transfer", Brookings, November 2013: of 155 AUTM-reporting offices, fewer than 13% generated enough licensing income to cover operating costs; about 75% of gross licensing income 2003-2012 went to 16 universities (10% of the sample). https://www.brookings.edu/wp-content/uploads/2016/06/Valdivia_Tech-Transfer_v29_No-Embargo.pdf (T2 -- think-tank analysis of AUTM data) We found no newer national study. An earlier survey, from about 2009, put it more bluntly: more than half of programmes earned less than they cost, 16% were self-sustaining, and the typical office spent about 0.6% of its institution's research budget on technology transfer, 45% of that on patents.8Abrams I, Leung G, Stevens AJ, "How are U.S. Technology Transfer Offices Tasked and Motivated -- Is It All About the Money?", undated (c. 2009): more than half of programmes earn less than they cost; 16% self-sustaining; offices spend about 0.6% of research budgets on technology transfer, 45% of it on patent costs. The copy read carries no date. https://www.wpi.edu/sites/default/files/docs/Offices/Intellectual-Property/How_are_US_Academic_Licensing_Offices_Organized_Tasked_Financed_and_Motivated_--_Final.pdf (T2 -- practitioner survey) A 2024 case study of one mid-ranked R1 university, counting faculty time, found every category of cost exceeded intellectual-property income; the faculty's opportunity cost alone was more than 33 times what the patents earned. One institution, and an open-source advocate's pen: a provocation rather than a benchmark.9Pearce JM, Transfer technologii bulletin, October 2024, doi:10.52036/ttb2024211: full-cost case study of a mid-ranked R1 university, including faculty time; every cost category exceeded IP income; faculty opportunity cost more than 33 times income. Author's summary in The Conversation, April 23, 2025. https://theconversation.com/us-universities-lose-millions-of-dollars-chasing-patents-research-shows-244270 (T2 -- peer-reviewed single-institution case study; T4 for the summary)

The usual reply is that offices are judged on impact, not profit, and the reply is sound. The Bayh-Dole Coalition, an advocacy group, puts average gross licensing income at $3.2 billion a year for 2021-2025 and the contribution of university licensing to GDP since 1996 at up to $1.7 trillion in 2025 dollars.10Bayh-Dole Coalition, "The Economic Impact of the Bayh-Dole Act", September 2026: average gross licensing income $3.2 billion a year, 2021-2025; contribution to GDP 1996-2025 up to $1.7 trillion (2025 dollars). https://bayhdolecoalition.org/wp-content/uploads/2026/09/The-Economic-Impact-of-the-Bayh-Dole-Act-Report.pdf (T2 -- advocacy-commissioned economic report) Both things are true at once: the system pays handsomely, and most of its offices are a mission expense. Which leaves the live question where it was: cost control at the margin -- which filings to make, and which to let go.

The tail

The famous deals are what the averages are made of, and each was unguessable at disclosure. Emory sold its royalty stream on emtricitabine in 2005 for $540 million, then the largest such sale by an American university; its office's cumulative revenue now exceeds $1.4 billion.11Emory Office of Technology Transfer, timeline: 2005 sale of emtricitabine (Emtriva) royalty interest for $540 million, "the largest" such transaction in US university history at the time; cumulative OTT revenue over $1.4 billion. Emory's own figure is used in preference to the widely repeated $525 million. https://ott.emory.edu/about/timeline/index.html (T1 -- the office's own record) Stanford's licence to Google returned $341 million.3Stanford Office of Technology Licensing, FY2015 overview: about half of disclosures lead to a patent application; 20-25% of disclosures are licensed; Google licence, $341 million; triage question "evolutionary or revolutionary?" https://web.stanford.edu/group/OTL/documents/fy15OTL_overview.pdf (T1 -- the office's own record) Northwestern sold a portion of its royalty interest in Lyrica to Royalty Pharma in December 2007 for $700 million in cash; Richard Silverman's laboratory had synthesised the molecule, pregabalin, in 1989.12Royalty Pharma and Northwestern University joint release, December 18, 2007: Royalty Pharma acquired "a portion" of Northwestern's royalty interest in Lyrica (pregabalin) for $700 million in cash; pregabalin synthesised by Richard Silverman in 1989. https://www.fiercebiotech.com/biotech/royalty-pharma-acquires-a-portion-of-northwestern-university-s-royalty-interest-lyrica-r (T1 -- the parties' own release, as carried by the trade press) Columbia's Axel co-transformation patents earned $790 million in licensing revenue, followed by costly litigation over an attempt to extend their life.13Colaianni A, Cook-Deegan R. "Columbia University's Axel Patents: Technology Transfer and Implications for the Bayh-Dole Act." Milbank Quarterly, 2009: $790 million in licensing revenue; litigation over the attempt to extend patent life. https://pmc.ncbi.nlm.nih.gov/articles/PMC2750841/ (T2 -- peer-reviewed case history)

The most recent entry is the most instructive. In April 2025 BioNTech agreed to pay the University of Pennsylvania up to $467.0 million, including $400.0 million as royalties for 2020 to 2023, with a low single-digit royalty continuing from 2024.14BioNTech SE, Form 6-K, April 2, 2025: agreements with the University of Pennsylvania under which BioNTech pays up to $467.0 million, including $400.0 million as royalties for 2020-2023, plus a low single-digit royalty from 2024. https://www.sec.gov/Archives/edgar/data/1776985/000177698525000017/form6-kpennagreements2apr2.htm (T1 -- SEC filing) The patent behind that cheque was granted on October 2, 2012.15USPTO Open Data, application 11/990,646, US patent 8,278,036: granted October 2, 2012. The USPTO record gives an effective filing date of February 19, 2008; Google Patents shows a 2005 priority claim; only the grant date is used here. (T1 -- patent office record, read October 16, 2026) It reached Moderna by a route no office could have drawn in advance: an exclusive Penn licence to a small company called mRNA RiboTherapeutics, and a sublicence through Cellscript.16Moderna, Inc., Form S-1, Exhibit 10.8 (sublicence agreement effective June 26, 2017): the Penn patents were exclusively licensed to mRNA RiboTherapeutics and sublicensed through Cellscript. https://www.sec.gov/Archives/edgar/data/1682852/000119312518323562/d577473dex108.htm (T1 -- SEC filing)

Survivorship is the obvious objection: citing the winners teaches nothing about the 99% that lost. Quite so. The tail is the reason an office cannot simply stop filing. The 99% are the reason it cannot afford to file everything.

Where the money is spent

An office does not make one patent decision. It buys a cheap option and then decides, at each of a series of rising gates, whether to keep paying. A provisional application at disclosure starts a 12-month clock. At 12 months the office must convert to a full US application or an international (PCT) filing, or let the provisional lapse. At 30 months it must choose the countries in which to pursue national patents, each with its own fees.17University of North Carolina Office of Technology Commercialization, "Patent Process Outline", August 22, 2024: provisional, 12-month conversion, 30-month national phase; triage questions -- competitive superiority, defined market, interested company, prior art, sponsor obligations. https://innovate.unc.edu/office-of-technology-commercialization/inventors/patent-process-outline/ (T2 -- institutional guidance)

Drexel told its inventors in February 2026 that a US patent costs $10,000-40,000 to obtain, that "most costs occur after initial filing", that issue takes five to six years, and that it will file abroad only when a licensee is already in hand. It also said it was cutting its portfolio back, mostly of older, unlicensed cases.18Drexel University Applied Innovation, "Processes and Guidelines", February 3, 2026: $10,000-40,000 per US patent; "most costs occur after initial filing"; 5-6 years to issue; foreign filing only with a licensee in place; portfolio being reduced, "most older and unlicensed". https://drexel.edu/applied-innovation/announcements/2026/February/processes-guidelines (T1 -- the office's own policy statement) Since most of the money lands after the first filing, the provisional is the cheap decision and the valuable ones are the second and third. An office that is careful at disclosure and casual at 30 months has its diligence pointed the wrong way.

Offices know this and triage accordingly. Emory's evaluation report assembles a technology summary, a market summary and a full prior-art search before a filing decision;19Emory Office of Technology Transfer, "I've submitted my invention -- now what?", November 4, 2024: the Commercialization Evaluation Report comprises a technology summary, a market summary and a full IP search. https://scholarblogs.emory.edu/techtransfer/2024/11/ive-submitted-my-invention-now-what/ (T2 -- institutional guidance) UNC asks about competitive superiority, a defined market, an interested company, prior art and what the sponsor is owed.17University of North Carolina Office of Technology Commercialization, "Patent Process Outline", August 22, 2024: provisional, 12-month conversion, 30-month national phase; triage questions -- competitive superiority, defined market, interested company, prior art, sponsor obligations. https://innovate.unc.edu/office-of-technology-commercialization/inventors/patent-process-outline/ (T2 -- institutional guidance)

The licensee's side of the gate is slower still. One patent that one of us later licensed as a founder was filed on October 25, 2011 and granted on August 23, 2016.20USPTO Open Data, application 13/281,287, US patent 9,426,429: filed October 25, 2011; granted August 23, 2016. (T1 -- patent office record, read October 16, 2026) The exclusive licence, negotiated by that founder, came after the grant -- more than five years after the first filing -- with the freedom-to-operate work after that.21BigBio house record, data/career_history.md: exclusive licence of three university patents, including US 9,426,429, taken by a founder-led startup; the founder negotiated the terms and drove the freedom-to-operate work. The licence year is recorded inconsistently in the house record and is omitted. (T1 -- first-party record) The office holding that docket paid every gate on a technology whose buyer did not yet exist.

The counter-argument writes itself from the UC memo: if blockbusters are almost impossible to predict, no assessment can beat the odds. True, and beside the point. An assessment at the 12- or 30-month gate does not predict winners. It raises the price of a lazy yes, and it makes an early no defensible to an inventor who wants to know why not mine. Across a portfolio where most cases lose, those savings compound.

What changed

The fiscal 2025 numbers show offices already doing this harder. Disclosures rose nearly 8%, licences more than 7%, options 10% and licences to startups 15% -- while new US patent applications fell 4% and foreign filings fell 40%. AUTM's own gloss was "a more disciplined approach to intellectual property investments".22AUTM, 2025 Licensing Activity Survey page: invention disclosures up nearly 8%; licences up more than 7%; options up 10%; licences with startups up 15%; new US patent applications down 4%; foreign filings down 40%; "a more disciplined approach to intellectual property investments". Read October 16, 2026. https://autm.net/surveys-and-tools/surveys/licensing-survey/2025-licensing-survey (T1 -- the survey owner's own summary)

The money that funds the office is moving the other way. In fiscal 2024 gross licensing income fell 24% and running royalties 41%; licensing staff numbers fell by more than 4%, and equity cash-outs rose 25%.23AUTM, 2024 Licensing Activity Survey page: gross licensing income down 24%; running royalties down 41%; licensing staff down more than 4%; equity cash-outs up 25%. https://autm.net/surveys-and-tools/surveys/licensing-survey/2024licensingsurvey (T1 -- the survey owner's own summary) The fall came off a high base: licence income was $3.6 billion in 2023 after a $3.8 billion peak in 2022,24AUTM, 2023 Licensing Activity Survey page: licence income $3.6 billion in FY2023, after a $3.8 billion peak in FY2022. https://autm.net/surveys-and-tools/surveys/licensing-survey/2023-licensing-survey (T1 -- the survey owner's own summary) a year in which Penn alone reported more than $1 billion25Philadelphia Inquirer, March 23, 2023: Penn reported more than $1 billion in licensing revenue for FY2022, mostly from mRNA vaccine royalties. "About a quarter" of the national total is our derivation against [^24]. https://www.inquirer.com/health/penn-covid-vaccine-licensing-revenue-rna-pfizer-moderna-20230323.html (T3 -- news report) -- about a quarter of the national total.

Then Washington leaned on three sides of the business at once. On overhead: the NIH capped indirect-cost recovery at 15% on February 7, 2025, a cut that was enjoined in April 2025,26Congressional Research Service, Insight IN12516, December 9, 2025: NIH notice capping indirect costs at 15% issued February 7, 2025; permanently enjoined April 4, 2025. https://www.congress.gov/crs-product/IN12516 (T1 -- CRS summary of the agency action and court order) the injunction upheld by the First Circuit in January 2026,27STAT, April 8, 2026, and FABBS, April 2026: First Circuit affirmed the injunction in January 2026; the administration did not seek further review. https://www.statnews.com/2026/04/08/trump-administration-drops-nih-indirect-costs-court-challenge/ (T3 -- news report) and abandoned when no Supreme Court petition arrived by the April 6, 2026 deadline -- after which the fiscal 2027 budget request proposed the same cap again.28NACUBO, April 13, 2026, and APLU, April 14, 2026: no Supreme Court petition filed by the April 6, 2026 deadline; the FY2027 budget request again proposes a 15% cap. https://www.nacubo.org/News/2026/4/Court-Fight-Over-NIH-Indirect-Costs-Cap-Ends (T2 -- association reporting) Technology transfer offices are largely paid from institutional overhead, which is why AUTM read the research cuts as cuts to itself,29AUTM advocacy update, "Now is the time for tech transfer to shine", April 9, 2025: cuts to research funding and indirect-cost recovery read as cuts to technology transfer. https://autm.net/about-tech-transfer/advocacy/autm-advocacy-update/now-is-the-time-for-tech-transfer-to-shine (T2 -- trade association statement) and why the University of Washington's CoMotion let three people go in September 2025, leaving 52.30GeekWire, September 11, 2025: University of Washington's CoMotion laid off three employees; 52 staff remain. https://www.geekwire.com/2025/university-of-washingtons-comotion-tech-transfer-office-lays-off-3-employees/ (T3 -- news report) On royalties: the Commerce Secretary said in September 2025 that the taxpayer "should get half the benefit" of federally funded university patents,31Axios, September 10, 2025: Commerce Secretary Howard Lutnick said the US "should get half the benefit" of federally funded university patents. https://www.axios.com/2025/09/10/howard-lutnick-university-patents-research (T3 -- news report) and NIST followed with a funding solicitation under which, by one account, awardees may be required to share revenue with Commerce; we have read the account, not the solicitation.32AIP FYI, October 21, 2025: NIST funding solicitation under which awardees may be required to share patent revenue with the Commerce Department. The solicitation itself was not read for this article. https://www.aip.org/fyi/commerce-moves-to-take-cut-of-research-patents (T3 -- news report) On title: Commerce opened a Bayh-Dole compliance review of Harvard's patents in August 2025 and raised the prospect of march-in rights; the outcome is not public.33Reuters, August 8, 2025: Commerce Department letter opening a Bayh-Dole compliance review of Harvard's federally funded patents and raising march-in rights. Outcome not found in the public record as of October 16, 2026. https://www.reuters.com/legal/government/harvard-patents-targeted-by-trump-administration-2025-08-08/ (T3 -- news report) The draft framework that would make price a factor in march-in decisions has meanwhile sat with NIST since December 2023 with no timeline for finalisation.34Government Accountability Office, GAO-26-107885, February 18, 2026: NIST's December 2023 draft march-in framework, which would allow price to be considered, has no timeline for finalisation. https://files.gao.gov/reports/GAO-26-107885/index.html (T1 -- federal audit report)

Fairness requires the other column. Nothing on royalties or march-in has been enacted; the indirect-cost cap lost in court; and Congress, the Bayh-Dole Coalition notes, considered and rejected a federal payback clause in 1980.35Bayh-Dole Coalition statement on the Commerce Secretary's profit-sharing proposal, September 10, 2025: Congress rejected a federal payback provision when enacting Bayh-Dole in 1980. https://bayhdolecoalition.org/bayh-dole-coalition-releases-statement-on-secretary-lutnicks-patent-profit-sharing-proposal/ (T2 -- advocacy statement) All true. But an office setting a 30-month budget today has to price the chance that the rules change before the patent issues, and that chance is no longer zero. This is what a more expensive decision looks like from the inside: a wider error bar on every line of the budget.

What an outside read adds

The offices already assess; what an outside assessment adds is independence and depth at the gates where the money is. A licensing associate carrying something like 40 dockets -- one practitioner's figure, not a survey -- cannot give each a market study.36Precision Federal, "Working with a university tech transfer office", April 28, 2026: a licensing associate may carry about 40 dockets. One practitioner's figure; no survey basis. https://precisionfederal.com/insights/working-with-a-university-tech-transfer-office (T4 -- commentary) An inventor cannot be asked to grade his own enthusiasm. An independent, sourced read on market, buyer and freedom to operate, written as keep-or-kill and priced against the next gate, is a different instrument from an in-house form.

We have built one. A university technology transfer office engaged us under a confidentiality agreement to assess a single invention disclosure; we delivered the result as a decision with an editable market model behind it, and benchmarked it head to head against an incumbent consultant's market assessment and a commercial report vendor's product on the same input. The work was unpaid and the office has reported no result, so we claim nothing here about who came out ahead.37BigBio house record, data/career_history.md: a university technology transfer office engaged BigBio under a CDA, July 2026, to assess one invention disclosure; deliverable a decision with an editable market model; benchmarked side by side against an incumbent consultant's market assessment and a commercial report vendor's report on the same input; unpaid; no outcome reported. (T1 -- first-party record) The comparison is the point: there are now three ways to buy a no, at three prices, and a director can put them side by side.

The objection is the budget: consultants and report vendors already sell this, and the office's budget is the thing being cut. Price it against the gate. A sourced no at 12 months should cost less than a national-phase filing that never licenses, and Drexel's portfolio cull is what the alternative looks like a decade on.18Drexel University Applied Innovation, "Processes and Guidelines", February 3, 2026: $10,000-40,000 per US patent; "most costs occur after initial filing"; 5-6 years to issue; foreign filing only with a licensee in place; portfolio being reduced, "most older and unlicensed". https://drexel.edu/applied-innovation/announcements/2026/February/processes-guidelines (T1 -- the office's own policy statement) The craft was always saying no cheaply; what has changed is the price of a careless yes. The question an office brings to any of these instruments is the one we try to answer: is this invention worth licensing, and to whom?

Footnotes

  1. AUTM, 2025 US Licensing Activity Survey infographic (FY2025 data): $112B research expenditure; 28,278 invention disclosures; 13,729 new US patent applications; 7,827 US patents issued; 10,277 licences and options; 1,095 startups formed; 761 new products. https://autm.net/AUTM/media/SurveyReportsPDF/AUTM-2025-US-Infographic.pdf (T1 -- the survey owner's own data release) ↩

  2. FY2024 totals -- 26,196 disclosures, 14,432 new US patent applications, 7,968 patents issued, 9,507 licences and options, 951 startups -- as relayed by two secondary summaries of the AUTM FY2024 infographic, whose original URL now returns 404; cross-checked against AUTM's stated FY2025 year-on-year changes (+7.9% disclosures). Ratios 0.49 and 0.55 are our derivation. (T3 -- relayed figures, original not retrieved) ↩

  3. Stanford Office of Technology Licensing, FY2015 overview: about half of disclosures lead to a patent application; 20-25% of disclosures are licensed; Google licence, $341 million; triage question "evolutionary or revolutionary?" https://web.stanford.edu/group/OTL/documents/fy15OTL_overview.pdf (T1 -- the office's own record) ↩ ↩2

  4. Caviggioli F, et al. Technological Forecasting and Social Change, 2020: of patents granted 2002-2010 to the 58 US universities in the sample, 29.7% were licensed. https://www.sciencedirect.com/science/article/pii/S0040162520310155 (T2 -- peer-reviewed analysis) ↩

  5. Ku K, Stanford OTL, remarks to a National Academies symposium, January 13, 2011: ~8,300 cumulative disclosures; 3 inventions produced 67% of cumulative income; 68 cases earned $1 million or more; "It took us 15 years to break even." https://sites.nationalacademies.org/cs/groups/pgasite/documents/webpage/pga_060747.pdf (T1 -- the office's own account, in the symposium record) ↩

  6. University of California Office of the President, item F9 to the Regents, January 19, 2012: fewer than 0.2% of disclosures make up 80% of revenue; "almost impossible to predict which inventions have blockbuster potential at the time they are disclosed"; ten or more years to revenue. https://regents.universityofcalifornia.edu/regmeet/jan12/f9.pdf (T1 -- institutional record) ↩

  7. Valdivia WD, "University Start-Ups: Critical for Improving Technology Transfer", Brookings, November 2013: of 155 AUTM-reporting offices, fewer than 13% generated enough licensing income to cover operating costs; about 75% of gross licensing income 2003-2012 went to 16 universities (10% of the sample). https://www.brookings.edu/wp-content/uploads/2016/06/Valdivia_Tech-Transfer_v29_No-Embargo.pdf (T2 -- think-tank analysis of AUTM data) ↩

  8. Abrams I, Leung G, Stevens AJ, "How are U.S. Technology Transfer Offices Tasked and Motivated -- Is It All About the Money?", undated (c. 2009): more than half of programmes earn less than they cost; 16% self-sustaining; offices spend about 0.6% of research budgets on technology transfer, 45% of it on patent costs. The copy read carries no date. https://www.wpi.edu/sites/default/files/docs/Offices/Intellectual-Property/How_are_US_Academic_Licensing_Offices_Organized_Tasked_Financed_and_Motivated_--_Final.pdf (T2 -- practitioner survey) ↩

  9. Pearce JM, Transfer technologii bulletin, October 2024, doi:10.52036/ttb2024211: full-cost case study of a mid-ranked R1 university, including faculty time; every cost category exceeded IP income; faculty opportunity cost more than 33 times income. Author's summary in The Conversation, April 23, 2025. https://theconversation.com/us-universities-lose-millions-of-dollars-chasing-patents-research-shows-244270 (T2 -- peer-reviewed single-institution case study; T4 for the summary) ↩

  10. Bayh-Dole Coalition, "The Economic Impact of the Bayh-Dole Act", September 2026: average gross licensing income $3.2 billion a year, 2021-2025; contribution to GDP 1996-2025 up to $1.7 trillion (2025 dollars). https://bayhdolecoalition.org/wp-content/uploads/2026/09/The-Economic-Impact-of-the-Bayh-Dole-Act-Report.pdf (T2 -- advocacy-commissioned economic report) ↩

  11. Emory Office of Technology Transfer, timeline: 2005 sale of emtricitabine (Emtriva) royalty interest for $540 million, "the largest" such transaction in US university history at the time; cumulative OTT revenue over $1.4 billion. Emory's own figure is used in preference to the widely repeated $525 million. https://ott.emory.edu/about/timeline/index.html (T1 -- the office's own record) ↩

  12. Royalty Pharma and Northwestern University joint release, December 18, 2007: Royalty Pharma acquired "a portion" of Northwestern's royalty interest in Lyrica (pregabalin) for $700 million in cash; pregabalin synthesised by Richard Silverman in 1989. https://www.fiercebiotech.com/biotech/royalty-pharma-acquires-a-portion-of-northwestern-university-s-royalty-interest-lyrica-r (T1 -- the parties' own release, as carried by the trade press) ↩

  13. Colaianni A, Cook-Deegan R. "Columbia University's Axel Patents: Technology Transfer and Implications for the Bayh-Dole Act." Milbank Quarterly, 2009: $790 million in licensing revenue; litigation over the attempt to extend patent life. https://pmc.ncbi.nlm.nih.gov/articles/PMC2750841/ (T2 -- peer-reviewed case history) ↩

  14. BioNTech SE, Form 6-K, April 2, 2025: agreements with the University of Pennsylvania under which BioNTech pays up to $467.0 million, including $400.0 million as royalties for 2020-2023, plus a low single-digit royalty from 2024. https://www.sec.gov/Archives/edgar/data/1776985/000177698525000017/form6-kpennagreements2apr2.htm (T1 -- SEC filing) ↩

  15. USPTO Open Data, application 11/990,646, US patent 8,278,036: granted October 2, 2012. The USPTO record gives an effective filing date of February 19, 2008; Google Patents shows a 2005 priority claim; only the grant date is used here. (T1 -- patent office record, read October 16, 2026) ↩

  16. Moderna, Inc., Form S-1, Exhibit 10.8 (sublicence agreement effective June 26, 2017): the Penn patents were exclusively licensed to mRNA RiboTherapeutics and sublicensed through Cellscript. https://www.sec.gov/Archives/edgar/data/1682852/000119312518323562/d577473dex108.htm (T1 -- SEC filing) ↩

  17. University of North Carolina Office of Technology Commercialization, "Patent Process Outline", August 22, 2024: provisional, 12-month conversion, 30-month national phase; triage questions -- competitive superiority, defined market, interested company, prior art, sponsor obligations. https://innovate.unc.edu/office-of-technology-commercialization/inventors/patent-process-outline/ (T2 -- institutional guidance) ↩ ↩2

  18. Drexel University Applied Innovation, "Processes and Guidelines", February 3, 2026: $10,000-40,000 per US patent; "most costs occur after initial filing"; 5-6 years to issue; foreign filing only with a licensee in place; portfolio being reduced, "most older and unlicensed". https://drexel.edu/applied-innovation/announcements/2026/February/processes-guidelines (T1 -- the office's own policy statement) ↩ ↩2

  19. Emory Office of Technology Transfer, "I've submitted my invention -- now what?", November 4, 2024: the Commercialization Evaluation Report comprises a technology summary, a market summary and a full IP search. https://scholarblogs.emory.edu/techtransfer/2024/11/ive-submitted-my-invention-now-what/ (T2 -- institutional guidance) ↩

  20. USPTO Open Data, application 13/281,287, US patent 9,426,429: filed October 25, 2011; granted August 23, 2016. (T1 -- patent office record, read October 16, 2026) ↩

  21. BigBio house record, data/career_history.md: exclusive licence of three university patents, including US 9,426,429, taken by a founder-led startup; the founder negotiated the terms and drove the freedom-to-operate work. The licence year is recorded inconsistently in the house record and is omitted. (T1 -- first-party record) ↩

  22. AUTM, 2025 Licensing Activity Survey page: invention disclosures up nearly 8%; licences up more than 7%; options up 10%; licences with startups up 15%; new US patent applications down 4%; foreign filings down 40%; "a more disciplined approach to intellectual property investments". Read October 16, 2026. https://autm.net/surveys-and-tools/surveys/licensing-survey/2025-licensing-survey (T1 -- the survey owner's own summary) ↩

  23. AUTM, 2024 Licensing Activity Survey page: gross licensing income down 24%; running royalties down 41%; licensing staff down more than 4%; equity cash-outs up 25%. https://autm.net/surveys-and-tools/surveys/licensing-survey/2024licensingsurvey (T1 -- the survey owner's own summary) ↩

  24. AUTM, 2023 Licensing Activity Survey page: licence income $3.6 billion in FY2023, after a $3.8 billion peak in FY2022. https://autm.net/surveys-and-tools/surveys/licensing-survey/2023-licensing-survey (T1 -- the survey owner's own summary) ↩

  25. Philadelphia Inquirer, March 23, 2023: Penn reported more than $1 billion in licensing revenue for FY2022, mostly from mRNA vaccine royalties. "About a quarter" of the national total is our derivation against 24AUTM, 2023 Licensing Activity Survey page: licence income $3.6 billion in FY2023, after a $3.8 billion peak in FY2022. https://autm.net/surveys-and-tools/surveys/licensing-survey/2023-licensing-survey (T1 -- the survey owner's own summary). https://www.inquirer.com/health/penn-covid-vaccine-licensing-revenue-rna-pfizer-moderna-20230323.html (T3 -- news report) ↩

  26. Congressional Research Service, Insight IN12516, December 9, 2025: NIH notice capping indirect costs at 15% issued February 7, 2025; permanently enjoined April 4, 2025. https://www.congress.gov/crs-product/IN12516 (T1 -- CRS summary of the agency action and court order) ↩

  27. STAT, April 8, 2026, and FABBS, April 2026: First Circuit affirmed the injunction in January 2026; the administration did not seek further review. https://www.statnews.com/2026/04/08/trump-administration-drops-nih-indirect-costs-court-challenge/ (T3 -- news report) ↩

  28. NACUBO, April 13, 2026, and APLU, April 14, 2026: no Supreme Court petition filed by the April 6, 2026 deadline; the FY2027 budget request again proposes a 15% cap. https://www.nacubo.org/News/2026/4/Court-Fight-Over-NIH-Indirect-Costs-Cap-Ends (T2 -- association reporting) ↩

  29. AUTM advocacy update, "Now is the time for tech transfer to shine", April 9, 2025: cuts to research funding and indirect-cost recovery read as cuts to technology transfer. https://autm.net/about-tech-transfer/advocacy/autm-advocacy-update/now-is-the-time-for-tech-transfer-to-shine (T2 -- trade association statement) ↩

  30. GeekWire, September 11, 2025: University of Washington's CoMotion laid off three employees; 52 staff remain. https://www.geekwire.com/2025/university-of-washingtons-comotion-tech-transfer-office-lays-off-3-employees/ (T3 -- news report) ↩

  31. Axios, September 10, 2025: Commerce Secretary Howard Lutnick said the US "should get half the benefit" of federally funded university patents. https://www.axios.com/2025/09/10/howard-lutnick-university-patents-research (T3 -- news report) ↩

  32. AIP FYI, October 21, 2025: NIST funding solicitation under which awardees may be required to share patent revenue with the Commerce Department. The solicitation itself was not read for this article. https://www.aip.org/fyi/commerce-moves-to-take-cut-of-research-patents (T3 -- news report) ↩

  33. Reuters, August 8, 2025: Commerce Department letter opening a Bayh-Dole compliance review of Harvard's federally funded patents and raising march-in rights. Outcome not found in the public record as of October 16, 2026. https://www.reuters.com/legal/government/harvard-patents-targeted-by-trump-administration-2025-08-08/ (T3 -- news report) ↩

  34. Government Accountability Office, GAO-26-107885, February 18, 2026: NIST's December 2023 draft march-in framework, which would allow price to be considered, has no timeline for finalisation. https://files.gao.gov/reports/GAO-26-107885/index.html (T1 -- federal audit report) ↩

  35. Bayh-Dole Coalition statement on the Commerce Secretary's profit-sharing proposal, September 10, 2025: Congress rejected a federal payback provision when enacting Bayh-Dole in 1980. https://bayhdolecoalition.org/bayh-dole-coalition-releases-statement-on-secretary-lutnicks-patent-profit-sharing-proposal/ (T2 -- advocacy statement) ↩

  36. Precision Federal, "Working with a university tech transfer office", April 28, 2026: a licensing associate may carry about 40 dockets. One practitioner's figure; no survey basis. https://precisionfederal.com/insights/working-with-a-university-tech-transfer-office (T4 -- commentary) ↩

  37. BigBio house record, data/career_history.md: a university technology transfer office engaged BigBio under a CDA, July 2026, to assess one invention disclosure; deliverable a decision with an editable market model; benchmarked side by side against an incumbent consultant's market assessment and a commercial report vendor's report on the same input; unpaid; no outcome reported. (T1 -- first-party record) ↩