Patent Strategy for Deep Tech Startups

Almost every published patent strategy for deep tech startups is a software patent strategy wearing a hardware jacket. File a provisional, convert at month twelve, show the filing at Series A, ship the product. That sequence assumes the product reaches the market inside the patent timeline. In deep tech it does not. Long development cycles and high upfront investment delay commercial returns, which is why an IP strategy has to reflect how value is actually created and where the technical advantage sits, not where the funding round sits.

The consequence is economic, not legal. The term runs from the filing date. If a reactor, a sensor array or a bioprocess needs seven years of engineering before the first paying installation, a filing made at concept stage spends a third of its protected life before it earns anything. Choosing a filing date is a capital allocation decision.

We write this from the builder side. EX EPIC finances, patents and deploys breakthrough science, and the portfolio runs 24 patent families filed with roughly 100 more validated in the pipeline, alongside 200 plus energy units deployed across 11 countries. The playbook below is the one that survives contact with hardware that has to be manufactured, shipped, installed and serviced somewhere real.

Why the standard startup patent playbook breaks in deep tech

Three assumptions carry the generic advice, and deep tech breaks all three.

The first is that disclosure risk arrives at a demo day. In deep tech it arrives much earlier and from directions a software founder never faces: conference abstracts, journal submissions, consortium deliverables, and above all the supply chain. Detailed technical specifications go to contract manufacturers, ODMs and component suppliers long before launch, and NDA coverage in those relationships is frequently incomplete. One unprotected disclosure in a jurisdiction with no grace period ends international rights quietly, months before anyone notices.

The second is that twelve months is enough time to know what you have. Prototyping cycles alone commonly run eighteen to thirty six months, so the moment your provisional expires you are usually still learning what the invention really is.

The third is that a patent is a fundraising artefact. It is, but that is the smaller half. For a company whose revenue depends on installed hardware in specific countries, the patent is the instrument that decides whether a competitor can build the same machine in the same factory town.

Step 1: File on the mechanism that is your moat

Before spending anything, answer one question precisely: what single technical mechanism creates the advantage a well funded competitor would find genuinely hard to replicate? The test is clean. If a competitor copied only that mechanism and built everything else independently, could they replicate your core value? If they would have to copy twenty other things too, you are looking at peripheral features rather than the moat.

This matters on a startup budget because patenting a peripheral feature costs exactly what patenting the moat costs. A granted patent on your moat creates a barrier. A granted patent on a secondary feature creates paperwork with renewal fees attached.

Claim the method, not the output

Founders describe what the system produces: it removes nitrogen, it stabilises the reaction, it predicts failures. That is sales language. The patentable value is in how you do it, step by step, and why that route differs from the standard one. Patents are strongest when they claim a method that holds across many settings rather than the specific embodiment you happen to have built this quarter. That is also what makes a filing survive a pivot, which deep tech does more often than it admits, usually into an adjacent feedstock, substrate or application.

A targeted prior art search in the moat area, before drafting, is the cheapest insurance in the sequence. It tells you what you must design your claims around and where the white space is, and it costs a fraction of what an office action response costs later.

Step 2: Sequence filings against the development clock

The best European practice here is staged, not single shot. File an initial application when the technical concept is stable, then use the twelve month priority window to add the variants, failure modes and narrower fallback positions that implementation actually taught you, keeping that later material clearly separated in the document so it cannot create added matter and priority loss problems. Waiting for technical maturity is what turns a concept patent into one that withstands a prior art attack. Filing early is what stops someone else owning the date. Staging is how you get both.

The discipline inside the first filing is to write past your current prototype. Draft it to cover what you have built, what you might build, and the underlying principle the design rests on, including alternative embodiments and materials substitutions you have not yet validated. Subject matter that is not in the first filing gets no priority benefit from it, so a narrow first document quietly loses the date on every feature your product ships with.

Then set the calendar. The twelve month Paris Convention deadline to decide where else to file is non extendible, and a Patent Cooperation Treaty filing pushes national phase entry out to thirty months from the original priority date. For a technology that will still be pre revenue at month thirty, that extension is not a delay tactic. It is the only mechanism that lets market evidence catch up with filing decisions.

In practice the trigger events for a deep tech filing programme are technical, not financial: a stable concept, a validated subassembly, a process that hits target yield, a control routine that holds on real equipment. Each one is a filing decision. The pitch deck is not on that list.

Step 3: Choose jurisdictions by where you build and deploy

Five offices, the US, the EPO, Japan, South Korea and China, account for 89% of global patent filings, and most US companies follow that order by default. Default is the wrong instinct for deep tech hardware. The questions that should decide the map are where your competitors manufacture, where enforcement would actually function, and where the equipment will physically sit.

Deployment geography is a filing input because hardware is copied where it is visible. A system installed in a country with no local protection is a working reference model for anyone who can visit it. Our own portfolio runs across eleven countries, and that footprint changes which national phase entries are worth their translation costs and which are vanity. So does manufacturing: if your tooling and your suppliers sit in one jurisdiction, that jurisdiction matters whether or not you will ever sell a unit there.

The scale of the route is worth knowing. WIPO recorded 273,900 PCT filings in 2024, which is why the provisional to PCT bridge has become the standard structure for hardware companies with international ambitions rather than an exotic option.

Step 4: Split patents and trade secrets by detectability

The sharpest test in deep tech IP is not novelty, it is detectability. If a competitor cannot determine how the technology works, you will struggle to detect or prove infringement, which makes the patent expensive theatre. Patents suit what is discoverable by product inspection, disclosed to third parties, or likely to be independently developed. Drafted with detectability in mind, claims can be aligned to externally observable behaviour or measurable system outputs, which is what makes them enforceable in practice.

Trade secrets suit the other half of the deep tech advantage, and in hardware that half is large: manufacturing methods, calibration routines, materials processing steps, tuning parameters, test methodology. These often deliver the real gains in yield, reliability and cost while remaining invisible in the finished unit. Publishing them in a patent hands a competitor the recipe in exchange for a right you cannot police.

Two caveats keep this from becoming a reason to file nothing. Physical hardware can be reverse engineered, expensively, through techniques like X ray imaging and electron microscopy, so secrecy is not permanent where the upside is large enough. And process know how walks out when people do. Patents covering the core technical concept, even while implementation details stay confidential, are the backstop against employee mobility. Trade secret protection in the EU also depends on documented access controls and confidentiality obligations under the Trade Secrets Directive, which means it is an operating discipline rather than a decision.

Step 5: Fix ownership before it costs you a round

The failure mode that surfaces in diligence and has no quick remedy is unassigned invention. Startups regularly discover late that key inventors, early freelancers or university affiliated contributors, never transferred their rights, which creates enforceability uncertainty and delays financing at exactly the wrong moment. Employed inventors are generally obliged to report service inventions to their employer, which is precisely why a professor's contribution to your filing can belong to their institution rather than to you.

This is structural in deep tech because so much of it starts in a lab. If your technology came out of a research group, the assignment chain is part of the asset, and it belongs on the same page as the filing plan. It is the least glamorous line item in commercialising university research and the one that most often stalls a round.

What a deep tech patent programme costs

These are published third party ranges, not quotes, and drafting rates vary widely by jurisdiction and technology.

The five jurisdiction figure is maximum coverage, not the minimum that works. A budget of 15,000 to 30,000 USD, spent in the right order, buys the priority dates and prosecution quality that investors and acquirers actually check, while keeping the expensive filings optional until capital exists. In the UK, high quality drafting by an experienced attorney can exceed GBP 10,000 per application, and Innovate UK grants commonly carry around GBP 7,500 for IP work, which is often enough to fund an initial filing. Many companies run a tiered approach: invest heavily in genuinely foundational technology, spend lightly on implementation level work that will be superseded anyway. That tiering is the heart of any workable patent portfolio strategy at pre revenue scale.

How investors and grant bodies read your IP position

The signal is measurable. A joint EPO and EUIPO study found that European startups filing for patents and trade marks in the seed or early growth stage were up to 10.2 times more likely to secure funding, that European patents were associated with a 5.3 times higher likelihood of early stage funding against 3.8 times for national rights only, and that filing was linked to more than twice the likelihood of a successful investor exit. The study singles out deep tech, whose long lead times and heavy investment make patents the instrument that attracts patient capital.

The stage curve tells you what normal looks like. Patent filings sit at 10% of startups at seed, 28% at early growth and 44% at late stage rounds. The jump between seed and Series A is where expectations harden from "we intend to file" into a portfolio with prosecution history behind it.

A useful way to read your own position: a filing programme is evidence of technical capability that a deck cannot fake, which is a large part of why patents function as IP as an asset class rather than as legal overhead. It is also the difference between describing your technology and owning it, which is the line that separates real deep tech from what separates deep tech from high tech in an investor's notes.

The filing sequence, condensed

1. Name the single mechanism that is the moat. Everything else waits.
2. Run a targeted prior art search in that area before any claim is drafted.
3. File the first application when the concept is technically stable, and always before any non confidential disclosure, including supplier and consortium disclosure.
4. Draft it to cover what you built, what you might build and the underlying principle, with at least one detailed implementation.
5. Use the twelve month window to file a follow up carrying the variants and fallbacks implementation taught you, kept separate from the priority disclosure.
6. Decide the PCT question before the twelve month deadline, which does not move.
7. Map jurisdictions to manufacturing sites, deployment sites and enforceability, not to a default top five.
8. Route non detectable process know how into a documented trade secret programme instead of a publication.
9. Clear the assignment chain for every contributor, especially academic ones.
10. Re-open the portfolio at every TRL step, abandoning what the roadmap has left behind.

FAQ

When should a deep tech startup file its first patent? When the technical concept is stable enough to describe with at least one detailed implementation, and always before any non confidential disclosure. Most jurisdictions outside the US offer no grace period at all, so a conference paper or an unprotected supplier conversation can end patentability abroad before the company knows it happened.

Does the patent expire before a deep tech product reaches market? Often a large part of it does, because the term runs from the filing date while commercial returns are delayed by the development cycle. That is the argument for staged filings and a live family with applications still pending, rather than one early application treated as finished business.

How many patents does a deep tech startup need before Series A? Published guidance points at a handful of applications spanning the system and its most novel components rather than a single filing, with expectations rising sharply between seed and Series A. Breadth across the technology stack reads as strategy; one filing reads as a checkbox.

Do EU grants and public funding change patent timing? Yes, in both directions. Grant milestones, deliverables and consortium reporting are disclosure events that set hard filing deadlines, and some programmes carry a small dedicated IP budget that can cover an initial filing.

Should a deep tech startup file a PCT application or file nationally? The PCT defers national phase cost and buys time to see which markets and manufacturing sites actually matter, at the price of a later grant. Direct national filing buys speed and earlier enforceable rights for a higher upfront spend. Cash constrained companies with unsettled geography usually take the PCT.