Patents, Moats, and What Actually Protects a Tech Startup

Why IP protection means something different in every corner of technology — and what investors should actually be evaluating

Prepared by Richstorm.co

Key Takeaways

• Roughly 40 to 55 percent of patented inventions never reach commercialization — a patent confirms novelty, not market viability.

• In hardware and materials science, patents function close to how they do in pharma — a concrete physical claim that is genuinely hard to design around.

• In software and AI, patents serve primarily strategic purposes — deterring litigation and enabling cross-licensing — not directly preventing a competitor from building something similar.

• Even in semiconductors, patents alone do not secure competitive leadership — manufacturing capability, design capability, lead time, and secrecy matter equally.

• The central investor question isn’t just whether a patent exists — it’s what combination of IP, capability, and structural advantage actually stops a well-funded competitor.

What a Patent Actually Confirms

A granted patent means a patent examiner has determined the invention is novel, non-obvious, and useful. That is the full scope of what the legal instrument verifies. It says nothing about whether anyone will buy the product, whether the market is large enough to justify investment, or whether the company can execute on turning the invention into a sustainable business.

Research on patent commercialization rates confirms this gap directly. Across multiple peer-reviewed studies, roughly 40 to 55 percent of patented inventions reach market launch. The remaining half never become commercial products — not because the underlying inventions were scientifically unsound, but because novelty and commercial viability are different questions answered by different processes. A patent office answers the first. The market answers the second, and the market’s answer is frequently no, regardless of how genuinely innovative the underlying invention is.

This matters for investors because a patent is often presented as evidence of a startup’s defensibility and market potential simultaneously. It is real evidence of neither. It is evidence of one narrower thing: that no prior public record of this specific invention existed at the time of filing. Everything beyond that requires a separate evaluation.

Why Patents Behave Differently Across Tech Categories

The most important thing to understand about patents in technology is that their strategic value is not uniform. It varies substantially depending on how physical and concrete the underlying invention is — and getting this wrong leads investors to either overweight a patent in categories where it provides little real protection, or underweight it in categories where it provides genuine competitive insulation.

In pharmaceutical investing, a patent on a drug molecule is close to the entire commercial moat. The molecule is the product, and a competitor cannot achieve the same therapeutic effect through a different molecular structure without developing a genuinely different drug — requiring its own full clinical trial program at massive cost and years of delay. In most of tech, the dynamics are different enough that directly applying this pharma-style weight to a software or AI patent is a category error.

Hardware and Materials Science: Closer to Pharma

Hardware, semiconductors, and materials-based technology behave meaningfully closer to pharma in patent terms. A specific chip architecture, sensor design, battery chemistry, or materials composition is a concrete physical claim that is genuinely difficult for a competitor to design around — they must find a different physical solution, which may require independent R&D and its own patent filing.

In semiconductors specifically, the importance of patents is real and documented — but with an important nuance that investors should hold alongside it. Research indicates that manufacturing capability, design capability, lead time, and secrecy each play roles in determining competitive position that are as significant as the patent itself. TSMC, Samsung, and Intel maintain massive patent arsenals, but their actual competitive leadership is inseparable from decades of accumulated manufacturing know-how that competitors cannot simply read from a patent filing and replicate. The patent protects the claim; the manufacturing capability is what makes the claim commercially unassailable.

For materials science specifically — a category where novel compositions, process innovations, and physical mechanisms are patentable and concrete — the absence of a patent in an early-stage startup warrants genuine scrutiny. This is the category where IP protection most closely resembles pharma’s model, and where the question of what specifically a competitor would have to do to build around the claim is a meaningful diligence question, not a formality.

Software: Patents Serve a Different Purpose

In software, the evidence on patent effectiveness as a competitive moat is genuinely mixed — and investors should hold that ambiguity rather than default to either extreme. Research from the NBER found that in software, patents are primarily used for strategic purposes: cross-licensing negotiations and deterring litigation, rather than directly preventing imitation. A separate large-scale study of U.S. IT firms found that software patents do have a positive effect on the patenting firm’s market value — but that rivals with sufficient R&D capability can substantially mitigate those effects.

Both findings are probably true simultaneously: patents provide some real legal leverage, but a determined, well-resourced competitor can usually find a path around a specific software claim.

This is structurally different from the pharma or hardware case, and the reason is implementation flexibility. Software solves problems through code, and the same functional outcome can almost always be achieved through more than one technical implementation. A patent on a specific method for doing something does not stop a competitor from achieving the same user-facing result through a different method.

This is why large tech companies treat patent portfolios primarily as defensive arsenals — deterrents against being sued by others, and tools for cross-licensing negotiations — rather than as the primary source of competitive protection. The real moat in software is usually elsewhere: proprietary data that took years to accumulate, network effects where the product becomes more valuable as more users join, or switching costs that make migration genuinely painful once a customer is embedded.

AI and Foundation Models: Trade Secrecy Over Patents

Frontier AI companies represent the sharpest departure from patent-as-moat thinking. The core training techniques behind leading AI systems are largely kept as trade secrets rather than patented — a deliberate strategic choice. A patent requires public disclosure of how the invention works in exchange for legal exclusivity. In AI, that disclosure would hand competitors a detailed roadmap to the approach faster than the legal protection would realistically defend against, given the speed at which the field moves.

This does not mean AI companies file no patents — AI-related patent filings have increased substantially in recent years, driven partly by large incumbents filing defensively and partly by a growing recognition that specific AI applications and implementations can be protectable. IBM has built substantial licensing revenue from its AI patent portfolio. But at the frontier model level specifically, the competitive advantage rests primarily on proprietary training data, compute scale, talent concentration, and the compounding advantage of operational experience — none of which a patent filing secures or reveals.

How Patent Weight Shifts by Tech Category

The Investor Question That Cuts Across All Categories

Strip away the patent question and one diligence question applies almost universally in tech investing: if a well-funded competitor saw this startup today, what specifically stops them from building something functionally equivalent within twelve months? The honest answer varies by category — in hardware and materials, a concrete patent claim and manufacturing complexity might be the real answer; in software, it is usually accumulated data, embedded customer workflows, or distribution relationships; in AI, it is usually proprietary training data and the compounding advantage of earlier operational experience.

Two other factors belong alongside the moat question regardless of category. Unit economics — whether the cost to acquire a customer is sustainably lower than that customer’s value over time — determines whether the business actually works at scale, independent of how defensible the technology is. And timing: a sound idea with a real patent pursued before the underlying market shift it depends on has actually materialized can fail for reasons that have nothing to do with the IP or the team.

What This Means for Evaluating a Tech Startup

A patent in a materials science or hardware startup is a meaningful data point that deserves the same diligence it would receive in a pharmaceutical deal: claim scope, how easily a competitor could design around the specific language, remaining patent life relative to the time needed to commercialize, and freedom-to-operate against existing third-party patents.

A patent in a software or AI startup is a different kind of signal. It is worth understanding — primarily because it tells you something about whether the company has thought about its IP strategy, and because it provides some defensive legal coverage — but it should not be treated as the primary evidence of commercial defensibility. The real defensibility question in these categories sits in the data, the network, the customer relationships, and the execution track record, none of which appears on a patent filing.

The consistent, honest conclusion across every category: a patent is real evidence of novelty at the time of filing. Whether that novelty translates into a commercial position that compounds over time is determined by a combination of IP, operational capability, market timing, and structural advantage that no patent document can confirm on its own.

RichStorm LLC is not a registered investment adviser, broker-dealer, financial institution, or licensed financial planning firm. It does not manage client assets, accept client funds, or provide personalized investment advisory services of any kind. This article is for informational and educational purposes only and does not constitute investment advice.