"Do you have a patent?" is the question every founder expects. It is also the least useful one. Patent due diligence for investors starts where that question ends: who owns the patent, whether it will survive a challenge, whether it covers the product, and how much protection is left by the time revenue arrives. The guides that rank for this topic are written by law firms for generic tech deals and acquisitions. They are thorough and they are right. They also miss what makes deep tech different. The patent is often licensed in from a university, the inventors are academics who publish for a living, the research was paid for with public money, and most of the portfolio is still pending. We see this from the builder side. EX EPIC finances, patents and deploys deep tech on a capital track record, with 24 patent families filed through EX IX and around 100 more validated in pipeline. Every one of those families has been through the check below, run by us, before anyone else's money touched it. ## Why patent diligence is different in deep tech Fish & Richardson's Matthew Colvin opens a paper on patent diligence for VCs and PEs with Shark Tank: the sharks always ask about the patent, the discussion is brief and superficial, and viewers learn that owning one is a box to tick. Real diligence, the paper argues, means probing the claims, the portfolio, the company's patent practices and its infringement risk. In software, a weak patent is an annoyance. The product moat is distribution and speed. In deep tech the patent is frequently the main asset for years, because there is no revenue yet to diligence instead. A battery chemistry or a gasifier design with a broken chain of title is not a company with a legal problem. It is a legal problem with a lab attached. ## Size the review to the round Not every cheque deserves a full patent opinion. Orrick lays out three levels of IP diligence tied to the deal: - Sanity check for seed and Series A: assignments, employment agreements, in-licences, the references examiners cited, and estimated expiry dates.
- Standard check for later rounds, often with a board seat: validity, enforceability and whether the claims cover the lead product, plus a look at known competitors' patents.
- Clearance check for mid-stage companies with commercial products: prosecution histories, deeper freedom-to-operate work and any pending litigation. PatentAuditor puts a quick-look review of 5 to 15 patents at 1 to 2 weeks, which fits inside most Series A timelines. The ten points below are ordered so that a seed investor can stop after the first five and still have caught the deal killers. ## The 10-point check, ordered by what kills deals ### 1. Does the company own or exclusively license the patents? Start with the chain of title: an unbroken line of signed and recorded assignments from every named inventor to the company. Cooley's investor request list asks for every assignment, every inventor's employment and consulting agreement, any co-ownership, and any liens recorded against the patents. If the documents look thin, our guide to the assignment of patent rights agreement shows what a clean one contains. Co-ownership deserves special attention. Under US law each co-owner can license the patent without accounting to the others, so a single unassigned co-inventor at a partner lab can hand your competitor a licence. Inventors who left also matter: Fish & Richardson's investor questions include who the inventors are and whether they are still at the company, because departures tend to leave gaps in title. ### 2. Is the university licence fit to build on? Most deep-tech spinouts do not own their founding patent. They license it, which is why a startup built around a patent is usually a startup built around a licence. Read that licence as closely as the patent. University policies are not written with your exit in mind. The University of Edinburgh's policy states that IP will normally be licensed, not assigned, and that licences will generally be non-exclusive, limited by field of use or territory. It also rules out "pipeline" rights to IP developed later, offering options on future improvements only for additional payments. So check four things: exclusivity, field of use against the actual business plan, rights to improvements from the lab, and the termination triggers. If the deal is still at option stage, the terms of a university license option agreement set most of these. ### 3. Did anyone publish before filing? Academics publish, present and post. In Europe and China, an inventor's own publication before filing destroys novelty, apart from narrow exceptions such as international exhibitions. The US and Japan give a one-year grace period. A paper presented in March and filed in November can therefore leave a valid US patent and no European rights at all. Ask for the inventors' publication list, conference talks, posters and theses, then compare dates against priority dates. Cooley's list asks the same question, down to slide decks and customer conversations not under NDA. For a company selling into Germany or Singapore, a missing EU patent is not a footnote. ### 4. Is there government funding in the chain? US federally funded inventions fall under Bayh-Dole, and the obligations are strict. The University of Utah's technology licensing office lists them: notify the funding agency within 60 days, elect title within 2 years, file within a year of election. It warns that missing the first two deadlines can result in the government claiming ownership permanently. It also applies when a grant only paid a graduate student who worked on the technology. Ask which grants funded the work at conception and at reduction to practice, and ask for proof the reporting was done. European public funding carries its own conditions, so the question applies there too. ### 5. Are the patents alive? Basic, and missed more often than it should be. Distinguish granted patents from pending applications, which give no enforceable rights until they issue. Check that US maintenance fees were paid at 3.5, 7.5 and 11.5 years after grant, and that no application was abandoned on a missed deadline. Patentext lists abandoned applications and thin provisionals among the issues that surface most often. University portfolios add a twist. Radboud University's inventor manual says it enters the expensive national phase, 2.5 years after first filing, only when a transfer is agreed or being negotiated. A family that looked global at PCT stage may have quietly narrowed to two countries. ### 6. Do the claims cover the product, and are they hard to design around? Map the independent claims to the product the company will actually sell, and then to what a competitor could sell instead. Fish & Richardson's paper flags the signs of weak patents: narrow claims that are easy to design around, long backgrounds that read as admitted prior art, and prosecution closed without a continuation to draft new claims later. Good portfolios keep a pending application alive so future claims can be written with real competitor products in view. ### 7. How would the claims survive a validity challenge? Grant is not the end of the argument. At the USPTO's Patent Trial and Appeal Board, 47 percent of instituted claims were found unpatentable in fiscal 2025. Read the prosecution history for how hard the examiner pushed and what was given up, and run a prior art search that includes academic literature, which examiners often miss and which deep-tech fields generate in volume. ### 8. Is there freedom to operate? A patent lets you stop others. It does not let you practise your own invention if someone else's patent covers part of it. Ask whether a freedom-to-operate analysis exists, who wrote it and when it was last updated. Expect resistance to sharing it: Fish & Richardson notes that disclosure can waive attorney-client privilege, and that pushback does not necessarily mean bad results. A summary letter from counsel is a reasonable compromise. ### 9. Does the family match the markets? Patent rights are territorial. Line up the countries where the family is filed or still pending against where the company will manufacture, sell and face competitors. A climate-hardware company that files only in the US and then builds its first plant in Southeast Asia has protection where it has no operations. ### 10. How much term is left when revenue starts? Patent term counts from filing, not from first sale, and deep tech eats into it early. Prosecution alone routinely takes 18 to 36 months, and pilots take longer. Lay the realistic lab to market timeline against the expiry dates. If the protection ends three years after first revenue, the patent is a head start, not a moat, and should be priced that way. ## Red flags that should change the price, not just the paperwork Most findings fall into three bins: 1. Fix before close. Unrecorded assignments, missing contractor IP clauses, overdue maintenance fees. Make them conditions precedent.
- Reprice. A non-exclusive or field-limited licence, lost European rights from early publication, narrow claims, short remaining term. These change what the asset is worth, and the right patent valuation work quantifies how much.
- Walk. A co-owner who will not assign, unresolved government title, or a blocking third-party patent with no design-around and no licence on offer. The BVCA's guide makes a broader point: traditional diligence is confirmatory and legal, and it rarely asks whether the patents match the main revenue lines. A qualitative IP business review of the landscape and competitors catches what the legal review leaves out. In deep tech that mismatch is a leading reason good science fails to become a good company, a pattern covered in why deep tech startups fail. ## How EX EPIC runs this check on its own patents We run these ten points on ourselves, before we invest, file, or bring a technology into the portfolio. Three habits come from doing it on the operating side. We check the deployment plan before the claims. Zero-X has installed waste-to-energy systems in 11 countries with 200+ units deployed. Point 9 is not abstract for us: we file where we build and where competitors manufacture, and we drop countries we will never operate in. We treat publication as a filing deadline. Researchers we work with keep publishing. The rule is simple: file first, then publish, so point 3 never comes up. We decide per field of use. The same family can carry an operating company in one market and be licensed or sold in others. That choice, covered in our license or build test, is why point 2 matters so much. A licence too narrow for the plan is a valuation problem, not a legal one. An investor who runs these ten points will not need to be a patent attorney. They will know which questions to escalate, what to pay for, and when a patent is a moat rather than a press release. ## FAQ ### How long does patent due diligence take for a venture investment? For an early-stage deal, a quick-look review of 5 to 15 patents takes roughly 1 to 2 weeks and covers ownership, status and a high-level claim read. A full clearance review with freedom-to-operate and validity work takes considerably longer and is usually reserved for later rounds. ### Is a pending patent application worth anything to an investor? It gives no enforceable rights until it is granted, so it cannot stop a competitor today. It still has value: a pending application keeps continuations open, which lets the company draft future claims aimed at real competitor products, and its prosecution history tells you how examiners view the invention. ### Should an investor pay for their own freedom-to-operate search? At seed, usually not. Review the company's own analysis, or a counsel summary of it. Where a single product carries most of the value and a known competitor holds patents nearby, commissioning an independent view before a large round is cheap insurance.