Only one or two of every hundred issued patents qualifies as lending collateral. The lender sees that within an hour. The founder is usually surprised.
A founder brings a portfolio to a lender for an IP-backed loan. The patents have been written, prosecuted, and granted. The plaques are on the wall. The portfolio looks like an asset.
The lender opens the file and looks at five things. None of them are the inventions.
By the end of the first hour, the lender has already decided whether the loan is possible. The founder is usually still describing what the patents are for. The conversation that follows is the gap.
These are the six things the lender checks, in the order they check them. Each one is the dispositive failure mode for ninety-eight percent of portfolios.
1. Who actually owns the patent
The first thing the underwriter pulls is the chain of title. Not the abstract, not the claims — the assignment record.
What they want to see: every named inventor signed an assignment to the company. Every assignment was recorded at the USPTO. Any prior security interest from an earlier lender has been released. If the company acquired any patent families through M&A, the assignments transferred correctly and were recorded by the surviving entity.
The patent the company thinks it owns may not be the patent the company actually owns. A missing inventor signature, an unrecorded assignment, a prior lien from a bridge round — any one of these means the company cannot pledge what it is trying to pledge.
This is dispositive. Title defects cannot be papered over in a closing memo.
The mechanical fix is fast. When the signed paperwork exists, an assignment is filed at the USPTO one day and recorded the next. The slow part is not the patent office. The slow part is human.
The signature may be sitting with an inventor who left the company two years ago and stopped returning calls. It may be sitting with a co-founder who knows the company needs the patent for the loan and is leveraging that knowledge into a settlement number. It may be sitting with an inventor who was never properly identified as a co-inventor in the original filing — and who now claims the entire patent.2
These are not rare cases. Inventors who withhold assignments as leverage have a name for it: patent blackmail. The pattern is common enough that the absence of clean, recorded assignments is the single most common reason a portfolio fails diligence — not just for lending, but for every M&A transaction, every financing round, and every license negotiation where the counterparty bothers to check.14
The lender looks at title first because if title is broken, nothing else matters. And the fix is not measured in days at the USPTO. It is measured in months of leverage negotiation with people who do not want to cooperate.3
2. Whether the patent reads on revenue — anyone’s revenue
The second thing the lender works through is whether the patent reads on a product that earns money. Crucially, the revenue does not have to be the company’s own.
This follows from how patent damages work under 35 U.S.C. § 284. The damages base in an infringement action is the infringer’s sales of the accused product, multiplied by a reasonable royalty. A patent that the patent owner has never practiced — or stopped practicing years ago — but that a competitor is unknowingly infringing has the competitor’s revenue as its damages base. Sometimes a much larger number than the patent owner ever generated themselves.5
That changes how the lender values the portfolio. The collateral question is not “does this patent protect the company’s current revenue line?” It is “does this patent read on a product anyone is currently selling at meaningful scale?” The same principle that governs every asset-backed loan applies: the lender sizes against what the asset can produce in licensing, enforcement, or sale, regardless of who is doing the producing.
Old patents are often the hidden gems. The company filed at the front of a product cycle. The company then pivoted, killed the product line, stopped paying attention. The patent went onto the maintenance-fee schedule as a line item nobody re-evaluated. Meanwhile the market kept building. Competitors entered the space. Some of them ended up shipping products that read on the abandoned patent without knowing it — because nobody at the patent owner has done the analysis to check.
When the lender opens those patents, reads the claims, and looks at what competitors are actually shipping today, the most valuable patent in the portfolio is often the one the founder has not thought about in five years. The CFO is paying maintenance fees on it out of habit. The lender finds two or three identifiable infringers. The collateral value of that one patent can exceed the collateral value of the entire portfolio of patents that read on the company’s own current products.6
The lender asks which patents read on which products — the company’s own and competitors’. The answer is sometimes uncomfortable. It is sometimes a windfall. It is always relevant to the size of the loan.
3. What the prosecution record actually shows
Before the lender looks at the claims, they read the prosecution record. The front page of every patent lists the references cited — but the count alone is not the story. The composition is, and so is the file history that produced it.
References on a patent come in two categories. Examiner-cited references are the prior art the USPTO examiner found through their own search, marked on the patent face with an asterisk. Applicant-provided references are the prior art the applicant or their attorney disclosed to the examiner through an Information Disclosure Statement (37 CFR 1.97-1.98), with no asterisk. The two tell different stories.
Examiner-cited count is a tell about examination quality. A typical examination produces three to five examiner-cited references. One reference is a red flag — it suggests the examiner ran a short search, found nothing close, and allowed the application without serious engagement. From the lender’s perspective, the patent has not been pressure-tested by the agency that issued it, and the validity exposure is high. Ten or more examiner-cited references is also worth a second look. It usually means the application was difficult — the examiner cited art repeatedly across multiple office actions, the claims went through several rounds of amendments, or both. The patent may have eventually issued, but the path it took to get there is recorded in the file wrapper, and that history is fair game in any later validity challenge.
Applicant-provided references work the other direction. They show that someone — the applicant’s attorney, an outside searcher, in-house counsel — actually did a prior-art search before filing. Each disclosed reference strengthens the patent against later invalidity arguments: a defendant in litigation cannot credibly argue that the applicant withheld prior art the examiner should have seen. The art was on the table. The examiner addressed it (or addressed it implicitly by allowing the claims over it). The defendant has to find new art that materially changes the picture, which is much harder. The statutory presumption of validity under 35 U.S.C. § 282 is much harder to overcome when the applicant disclosed broadly and the examiner engaged.
When the file history is thick, read the file wrapper. If a patent went through many office actions and many rounds of back-and-forth, the lender pulls the prosecution history and reads every amendment, every argument, every characterization the applicant made on the record. Statements made during prosecution bind the patent owner in litigation under the doctrine of prosecution history estoppel. The applicant who narrowed a claim to overcome a § 103 obviousness rejection has surrendered the territory between the original and amended claim — under Festo, that surrender is nearly impossible to reclaim later. The applicant who characterized their invention narrowly to distinguish prior art has handed every defendant a roadmap for design-around. Every statement on the record can come back to haunt the patent owner in court.
A patent the lender can read in fifteen minutes — short examiner record, reasonable references, clean prosecution — is a patent the lender can lend against. A patent that requires three hours of file-wrapper analysis to understand what the applicant actually surrendered is a patent the lender may walk away from regardless of what the claims appear to say on their face.7
4. What the claims actually block
The third check is whether the claims block a competitor’s product — or only the company’s own.
A patent claim is a fence. The question is which products the fence keeps out. If the fence is built around the company’s own implementation — its specific architecture, its specific database schema, its specific manufacturing sequence — then a competitor whose product achieves the same commercial result through a different implementation walks straight past it.8
The reason claims default to the company’s own implementation is structural. The inventor describes what they built. The attorney writes claims that read on what was described. Neither one is in a position to write claims that read on the competitor’s product. The inventor is too close to their own constraints to separate them from the market’s. The attorney does not know who the competitors are or what technology stacks they run.
The deeper problem is that the inventor’s constraints are not universal. Every inventor works inside specific obstacles — legacy systems, existing hardware, manufacturing processes, software-stack limitations. The solution is clever precisely because it overcomes a specific obstacle. But the competitor does not share those obstacles. A patent built around a workaround to a legacy database architecture is commercially useless. Nobody else has that company’s legacy database. The patent constrains a problem only the patent owner has.9
The doctrine of equivalents does not save the claim. It is supposed to cover implementations that perform substantially the same function in substantially the same way to achieve substantially the same result. But the Supreme Court in Festo Corp. v. Shoketsu Kinzoku Kogyo Kabushiki Co., 535 U.S. 722 (2002), held that any narrowing amendment during prosecution creates a presumption of surrender for the territory between the original and amended claims — a presumption that is nearly impossible to overcome at trial. Almost every patent application gets amended during prosecution. Which means almost every issued patent has surrendered most of the equivalents ground its drafter might have hoped to retain. A claim that has been amended is a claim that protects only what it literally describes.
The test the lender runs is operational. Stand the claim next to the competitor’s product as it actually ships. Walk through each element of the independent claim in order. Does the product perform each step, contain each component, satisfy each limitation — using publicly available evidence? If a single element is missing, the competitor does not infringe. This is the “all elements rule,” and it is why every word in a long, over-specified claim is one more thing a competitor can be missing.10
Amazon’s One-Click patent (US 5,960,411) is the canonical positive example. Online checkout was a contradiction: merchants wanted orders but did not want returns. The conventional answer was friction — confirmation buttons, shopping carts, “are you sure?” prompts. Amazon went the other direction. One button. One click. Purchase complete; solve returns downstream instead of preventing the order. The patent did not describe Amazon’s specific server architecture or session-tracking schema. It described the inventive shift — purchase-with-a-single-action — at the level any competitor running an e-commerce checkout would have to either copy or work around. Barnes & Noble could not work around it. They spent millions in legal fees trying.11
If the claim reads on the patent owner’s product but on no competitor’s product, the patent is a defensive instrument at best and an expensive disclosure document at worst. It does not generate licensing revenue while the company is operating, and it does not produce recovery if the company defaults. It is not collateral.
5. Whether infringement can be detected from outside
The fifth check is detectability. Can someone prove infringement using publicly available information — product documentation, marketing materials, API specifications, physical inspection, network behavior — without filing a lawsuit first?
This is the most-failed criterion for software patents. The instinct that produces allowed claims — describing the internal process steps because they are technically novel — produces unenforceable claims. The patent office allows the claim because the internal steps are clearly patentable. The lender rejects it because proving infringement requires source code, server logs, or internal documentation — none of which are available without the discovery that only begins after a lawsuit is filed.12
Patent litigation costs run into the millions before discovery even concludes. The AIPLA Report of the Economic Survey, the field’s standard cost reference, has reported median patent-litigation costs through the end of discovery between roughly $1.5M and $5M depending on the amount at stake. Lenders do not want collateral that requires that level of investment, and that much elapsed time, before anyone knows whether the patent is worth anything.
The test is concrete. Stand in front of the accused product. Can you build a claim chart from what you can see and touch? If yes, the patent is detectable. If no, it is not collateral, regardless of how novel the underlying invention.13
6. Whether the patent will survive an IPR
The fifth check is validity exposure. Inter Partes Review is the procedure by which a defendant challenges a patent’s validity before the Patent Trial and Appeal Board. The challenge is cheaper than litigation, faster, and substantially more successful for the defendant than for the patent owner.14
A patent that issued with a weak prior art search — many do — is a patent that may not survive the first IPR petition filed against it. From the lender’s perspective, that means the collateral can be destroyed by a well-funded defendant after default. The patent that produced no enforcement leverage during the borrower’s operating life is the same patent that produces no recovery in workout.
The underwriter wants to see the prosecution history, the prior art the examiner considered, the office actions, the amendments. They want to know what the examiner missed — and whether those gaps would surface in front of the PTAB.
This is also why portfolio depth matters. A single patent with IPR exposure is fragile collateral. Several patents covering the same technology from different angles is a portfolio that survives the first invalidity attack.
What the lender does not look at first
Notice what is not on this list: the cleverness of the invention. The credentials of the inventor. The novelty of the underlying technology. The number of patents in the portfolio. The number of citations the patent has received.
Those are things founders care about. They are not what the lender checks first.15
The lender checks the five things above because those are the things that determine whether the patent can be turned into cash in a workout, and whether the patent can produce licensing revenue while the company is operating.
The mismatch is not arbitrary. It comes from the lender’s job: they have to know, before the loan closes, that the collateral has independent value if everything else goes wrong. The patent that is brilliant but undetectable, brilliant but invalid, brilliant but uncovered by any revenue — those patents do not produce recovery. They are not collateral.
The six checks are the same six checks for enforcement
Here is the unsubtle pattern. The six things the lender checks are the same six things enforcement counsel checks at the start of an infringement case. They are the same six things an acquirer’s IP counsel checks during diligence. They are the same six things a Chief IP Officer checks every quarter.
The portfolio that survives lending underwriting is the portfolio that produces enforcement leverage. The portfolio that produces enforcement leverage is the portfolio that protects revenue from competitors. The portfolio that protects revenue from competitors is the one that holds up in acquirer diligence and produces capital using the IP.
The six checks define investment-grade patents. Everything else is filing activity.
The fix is upstream
If a portfolio fails lending underwriting in the first hour, the right reaction is not to find a different lender. It is to fix the portfolio. The six things the lender checked are the six things that determine whether the patents are doing business work or sitting on a balance sheet as expensive paper.
The work to fix the portfolio is the work a Chief IP Officer does upstream of every filing: chain of title at signing, revenue mapping at filing, prior-art depth at examination, claim drafting against competitor products, detectability tested before allowance, validity searched before issue. When that work is done, the portfolio survives lending underwriting because it was built to survive enforcement.16
If the portfolio survives enforcement, the loan is the easy part.
Further reading
- 1. Using Patents as Blackmail — What patent blackmail actually looks like — the departed inventor or disgruntled co-founder who holds the unsigned assignment as leverage when the company needs it most. ↩
- 2. Filing Patents After an Employee Leaves — The cooperation problem the company creates for itself when filings happen after the inventor is already out the door. ↩
- 3. Lots of Inventors Means Lots of Problems — Why multi-inventor patents multiply the cooperation risk — every additional named inventor is another signature you may someday need to chase. ↩
- 4. Assignment Due Diligence — Unassigned Assets — Why unrecorded assignments are the most common — and most expensive — defect found in IP diligence, and what a clean chain of title actually looks like on paper. ↩
- 5. How Patent Licensing Works — The mechanics that make the infringer’s revenue — not the patent owner’s — the actual damages base. Royalty calculation, established licensing programs, and why a patent owner who never practiced the invention can still collect from those who did. ↩
- 6. How to Find a Realistic Patent Value — The valuation approach that surfaces the hidden-gem patents — by reading the claims against current competitor products rather than the company’s own historical roadmap. ↩
- 7. How to Spot Bad Patents — The front-page indicators of patent quality — reference count, prosecution depth, claim length — and what each one tells you in the first 30 seconds. ↩
- 8. Your Attorney Drafted Claims on Your Product — Not Your Competitor’s — The drafting problem behind §4 here — why claims default to describing the company’s own implementation, and the test for whether yours have the same defect. ↩
- 9. Stop Patenting Your Invention. Start Patenting Your Competitor’s Product. — The reframe in full — and worked examples of what ‘patent the competitor’s product’ looks like in actual claim drafting. ↩
- 10. Every Word Hurts You: Patent Claims — Why long, narrow independent claims are commercially worthless — and the test for whether each word in your claim is doing real work. ↩
- 11. Valuable Patents Solve Contradictions — The structural reason the Amazon One-Click patent had real teeth — patents that resolve a market contradiction are the ones that compound into business value. ↩
- 12. Detectability Is a Key Factor for Patent Value — Why patents on internal process steps fail this test, and how to write claims that prove infringement from outside the accused product. ↩
- 13. Method Claims and Undetectability — The specific failure mode for method claims — undetectable manufacturing or software processes that ‘taught competitors how to do it’ instead of protecting the company. ↩
- 14. Inter Partes Review Was Supposed to Help Independent Inventors — And It Did — How IPR actually works in practice, who uses it against whom, and why validity exposure is the failure mode lenders weight most heavily. ↩
- 15. They say “Get More Patents” but they really mean something else — The gap between what founders measure in a patent program (count, novelty, citations) and what the program is supposed to produce (capital, leverage, options). ↩
- 16. Fractional Chief IP Officer — The role that runs the upstream work this post describes — chain of title at signing, revenue mapping at filing, claim drafting against competitor products — without the cost of a full-time hire. ↩
