Valuing AI Intellectual Property: A Practitioner's Framework

Valuing intellectual property at the seed stage is one of the most consequential and least systematized activities in venture capital. Most seed investors lack the technical background to meaningfully evaluate patent applications; most IP attorneys lack the commercial judgment to translate legal assessments into investment theses. At NL Patent AI Capital, we have spent years developing a practitioner's framework for AI IP valuation that bridges this gap — combining rigorous technical assessment with commercial due diligence in a structured process that we apply consistently across every prospective investment.

This piece shares the core elements of that framework. It is not a comprehensive academic treatment of IP valuation methodology — there is extensive literature on that. It is a practical guide to the questions we ask, the evidence we seek, and the judgments we make when we are sitting across the table from a deep tech AI founder who has filed their first patent applications and is trying to explain why their intellectual property is worth the valuation they are seeking.

The Four Dimensions of AI IP Value

We assess AI IP across four dimensions, each of which contributes independently to our overall valuation judgment. No single dimension is sufficient on its own; all four must be evaluated and weighted together.

Dimension 1: Technical Novelty and Non-Obviousness

The foundational legal question for any patent is whether the claimed invention is both novel (not disclosed in the prior art) and non-obvious (not a trivial combination or modification of existing ideas). For AI patents, this is simultaneously the most important and the most difficult dimension to evaluate, because the pace of AI research publication means that the relevant prior art is voluminous, recent, and sometimes obscure.

Our standard practice is to engage an independent technical expert — typically a senior researcher with domain expertise in the relevant AI subfield — to conduct a prior art search and novelty assessment before we make any investment decision. This is not the same as what a patent attorney does during prosecution: it is a substantive technical judgment about whether the claimed invention represents a genuine advance beyond what the field already knew at the time of filing. We have found that roughly one-third of AI patent applications that look impressive on a superficial review contain claims that are clearly anticipated by prior art that the applicant's counsel apparently did not find — a problem that is particularly common in academic spinouts whose founders are aware of their own research but less systematically aware of work being done at other institutions simultaneously.

The novelty assessment produces a confidence score that we apply to every element of our commercial valuation. A patent application with compelling claims but low confidence of surviving examination is worth a fraction of what the same claims would be worth with high examination confidence. This adjustment is rarely done explicitly by seed investors, but it is essential for accurate IP valuation.

Dimension 2: Claim Breadth and Scope

A technically novel invention that is claimed too narrowly provides limited commercial protection. This is unfortunately common in AI patents, where the combination of technically specialized patent attorneys and research-focused founders often produces applications that claim the specific embodiment developed in the lab rather than the broadest version of the inventive concept that the prior art will allow.

We evaluate claim breadth by conducting a "design-around" analysis: could a technically sophisticated competitor achieve substantially the same commercial result by implementing a variation of the invention that the current claims would not cover? If the answer is yes — and if the variation is not itself independently unobvious — we assess the breadth gap and estimate the cost and feasibility of broadening the claims through continuation practice or new filings to close that gap.

Claim breadth in AI is a genuinely difficult challenge because many of the most valuable AI innovations are methods — sequences of computational steps — rather than physical devices or chemical compounds. Method claims must be specific enough to survive examination but broad enough to cover the commercially important variations. Striking this balance requires both technical depth and patent prosecution experience, and the two are rarely found in the same person. Our investment process includes a claim improvement session with specialized AI patent counsel as a standard post-investment deliverable for every portfolio company.

Dimension 3: Commercial Relevance and Market Sizing

A technically novel, broadly claimed patent for a solution to a problem no one wants solved is worth nothing commercially. This may seem obvious, but the AI research world produces an extraordinary volume of technically impressive work that solves well-defined problems with little or no commercial demand. The seed-stage AI companies that build IP around commercially irrelevant problems are a consistent source of heartbreak for investors who were dazzled by the technical elegance and missed the market size question.

Our commercial relevance assessment proceeds through three sub-questions. First, what is the size and growth rate of the market that would benefit from the patented technology? We require credible bottom-up market analysis, not the recycled top-down TAM figures from industry analyst reports that populate most pitch decks. Second, what is the pathway from the patented technology to commercial revenue — through direct product sales, licensing to established players, or acquisition by a strategic buyer? We evaluate the plausibility and time horizon of each pathway separately. Third, what fraction of the commercial value in this market is attributable to the specific innovation covered by the patent, versus other factors (brand, distribution, data, customer relationships)?

This third sub-question is where most IP valuations fail at the seed stage. Investors often implicitly attribute all of a company's value to its patent portfolio, when in reality the patents may represent only a modest fraction of the value — with the rest coming from non-IP-protected factors that are equally or more important. Conversely, in categories where patents are genuinely blocking — where a competitor cannot enter the market at all without licensing the core patents — the IP may represent the majority of the company's enterprise value.

Dimension 4: Defensive Value and Blocking Power

Even patents that are difficult to monetize through direct licensing or product sales can have significant value as defensive assets. A well-constructed AI patent portfolio can deter competitor patent attacks (cross-license situations), block imitators from entering a market, and provide leverage in M&A negotiations with strategic acquirers who covet the IP coverage.

We assess defensive value by mapping the patent portfolio against the competitive landscape: how many credible competitors exist who could build competing products, and how many of them would be required to either license or design around the portfolio company's patents to do so? The answer to this question depends heavily on claim breadth (Dimension 2) and on the technical sophistication of potential competitors' development teams.

Blocking power — the ability to prevent competitors from entering a market entirely — is rare but enormously valuable when it exists. It is most commonly found in patents that cover foundational methods required for any implementation of a particular technology, rather than specific optimizations or applications. Companies that have filed and obtained foundational method patents in emerging AI categories before those categories became commercially significant can hold extraordinary leverage over all subsequent entrants. This is the scenario we are most actively seeking to identify and back at the seed stage.

The IP Maturity Curve and Investment Timing

AI patents follow a predictable maturity curve, and the appropriate investment strategy varies significantly depending on where in that curve a company's IP portfolio sits. At the earliest stage — provisional applications and PCT filings before national phase entry — the IP has maximal flexibility but minimal legal certainty. The claims can still be significantly reshaped, but nothing has been granted and the prior art landscape has not been formally tested.

Between PCT filing and national phase entry, the company has a defined window (typically 30 months from the earliest priority date) to decide which national jurisdictions to pursue. This is often where IP strategy mistakes are made — companies choose jurisdictions based on where their customers are today rather than where they expect IP enforcement to matter most over the twenty-year patent term. Our investment process includes a jurisdiction strategy review for all portfolio companies at this stage.

After national phase entry but before examination, the company faces the examination process in each jurisdiction. US examination for AI patents typically takes two to four years; European and Chinese examination timelines vary. During this period, the claims are being shaped by examiner objections and applicant responses. Companies that are not actively managing their prosecution strategy during examination — responding to office actions promptly, arguing claim scope aggressively, and using continuations strategically to preserve options — are leaving significant value on the table.

Red Flags: When AI IP Valuation Should Give You Pause

As important as knowing what good AI IP looks like is recognizing the warning signs that indicate IP that is likely to be less valuable than it appears. We have identified several patterns that should prompt additional scrutiny.

First, patents filed only in the United States with no international strategy. This may indicate either budget constraints (understandable at the earliest stage) or a failure to think globally about the commercial significance of the invention (more concerning). A company that has genuinely foundational AI IP should be filing PCT applications to preserve its international options.

Second, claims that are primarily directed to software-implemented methods without hardware tie-in, filed only in the US context. The Alice/Mayo eligibility framework continues to create significant uncertainty for pure software and mathematical method claims. AI patents that rely on specific hardware implementations or that can be reframed in terms of specific technical effects on computer systems are much more likely to survive examination and subsequent validity challenges.

Third, a large number of filed applications with no granted patents and no office actions after more than two years. This often indicates a prosecution strategy problem — either the applications are so narrowly drafted that they are sailing through examination without challenge (suggesting they may be too narrow to have commercial value), or they are stuck in examination limbo because the claims are overly broad and the prosecution team has not found a viable path to allowance.

To understand how we apply this framework in practice, visit our Portfolio page or reach out through our Contact page to discuss a specific company or technology.