A stage-by-stage startup IP strategy for 2026: what to protect from pre-seed to Series B, what it costs, and why one well-chosen tactic beats an IP program.
Most startups do not need an IP program. They need one deliberate IP move per funding stage, run with the rigor you would give a product launch, on a budget that grows only when the round does. For most companies, that is the whole startup IP strategy through Series B.
Big portfolios need machinery (review committees, annual pruning, competitive monitoring). A twelve-person company with eleven months of runway does not, yet. Our approach for early companies: pick the single tactic that changes your odds at the next milestone, do it well, and defer everything else on purpose. Sometimes that is a patent filing. Sometimes it is a trade secret registry and a few access rules that cost almost nothing. One well-chosen move creates real advantage. A half-built program creates invoices.
The second idea we push on founders: treat the patent itself like a startup. It has a customer (an acquirer, an investor, a licensee, or a competitor you need to block), a budget, a timeline, and a kill decision. Founders who frame it that way run the filing with startup-level rigor instead of handing it off and forgetting it.
IP does two jobs for a startup: it helps you raise money, and it is downside insurance. On the first, the best evidence we know of is the patent “lottery” study in the Journal of Finance, which used the near-random assignment of examiners to isolate the effect of a first grant. Startups that drew a lenient examiner had, on average, 55 percent higher employment growth and 80 percent higher sales growth five years later, largely because the grant opened up funding from VCs, banks, and public investors. That is first grants at small companies, not a promise about your application, but the direction is clear. On the second, we wrote in July about a robotics company whose 220 patents were the only asset that outlived the company when it failed.
At pre-seed the IP work costs hours rather than dollars. Three things:
The sorting is where founders get stuck, so borrow a filter we like from Outlier Patent Attorneys’ 2026 founder questions: detectability. If a competitor could pull your product apart and see the invention, it is a patent candidate, because you could detect infringement. If it runs on your servers where nobody can see it, keep it a trade secret. We would add a second filter: would an investor or acquirer need to see it to believe you? If yes, a secret alone will not help you in a data room.
Trade secrets are not free, though. Under U.S. law a trade secret has to be subject to reasonable efforts to keep it secret, which in practice means a written list of the secrets, who has access, and an NDA before anything goes to a contractor or investor. J A Kemp’s February 2026 note on startup know-how covers the same basics, including exit confirmations from departing employees. In our experience that registry is the cheapest IP tactic a pre-seed company has. If it is all you do this year, it is a legitimate choice.
Whether your first provisional lands at pre-seed or seed follows your disclosure calendar. The round you just closed has little to do with it.
Timing first. File before any public disclosure you cannot control, and after the invention works well enough to describe in detail. The U.S. gives you a one-year grace period for your own disclosures; WIPO’s guidance is blunt that in most of the world anything made public before filing is prior art, and grace periods, where they exist, run 6 to 12 months. A demo day, a pitch competition, a conference talk, a public repo, a customer pilot without an NDA: each starts a clock or ends your foreign options. Work backward from the first one on your calendar.
The provisional is the right vehicle because it is cheap to file and buys twelve months. The USPTO fee is $325, or $130 for a small entity (fewer than 500 employees), or $65 for a micro entity, which in 2026 means, roughly, gross income under $251,190 for the company and each inventor and no inventor named on more than four earlier applications. A provisional is never examined, needs no formal claims, and expires at twelve months with no extension.
The cheapness is also the trap. A provisional only holds the early date for what it actually teaches, so a two-page sketch filed the night before demo day protects almost nothing; the founders we work with tend to discover that eleven months later, when the attorney sits down to write claims. Our rule is that the provisional should look like a patent: a complete technical description, real figures, the variations you can already see. Professionally drafted, that usually runs $2,000 to $6,000 on top of the government fee. Pick the one to three inventions the company cannot afford to lose and spend there.
Run your first patent like a startup: one clear customer, a budget you can defend, and a kill decision already on the calendar.
AI has changed the founder’s half of this job. A CTO with a drafting tool and an afternoon can now put together a description more complete than the invention disclosures we see most attorneys receive. The USPTO’s November 2025 inventorship guidance treats AI as a tool like any other, so the humans stay the inventors, and its April 2024 guidance on AI tools in practice created no new rules. Per Skadden’s summary, there is no general obligation to disclose that you used a tool, but whoever signs is responsible for every word, and pasting your invention into a consumer chatbot has confidentiality consequences. Our read: draft the disclosure yourself with AI help, then pay a patent attorney to turn it into the filing and tell you what is missing. A full piece on where AI drafting works and where it breaks is coming in a few weeks.
After you file, put two dates on the calendar. At month six, re-project the full cost of converting and make a preliminary go or no-go call, so conversion is never an emergency purchase against a fundraising deadline. At month eleven, be ruthless comparing what the provisional actually teaches against the claims you now want. We went through the post-filing checklist on Invent Anything episode 72: Your Provisional Patent Is Filed: Now What?
Before you convert anything, put real numbers on the next 30 months. The Patent Cost Calculator lets you pick the countries you are actually considering and see filing, prosecution, and maintenance costs side by side, a better basis for a board conversation than one attorney quote.
At Series A the provisionals come due and the spending gets real. Converting one to a U.S. utility application typically means $8,000 to $15,000 or more in attorney drafting, plus USPTO filing, search, and examination fees of $2,000 at the standard rate, $800 for a small entity, and $400 for a micro entity, plus office action responses at $1,500 to $3,000 or more each. Through grant, one U.S. patent commonly lands at $15,000 to $30,000 or more. Our budget guide puts the whole Series A stage, with second and third families, continuations, and first office actions, at roughly $50,000 to $150,000.
Not every provisional deserves that. This is the part of startup patent strategy where saying no matters most. Convert the ones an acquirer, investor, or competitor would care about in two years, not the ones the team was proudest of. When you drop one, do a short postmortem (thin draft, market moved, money ran out); the answer changes how you file next time.
Two things about timing surprise founders. First, the USPTO said in April 2026 that its unexamined inventory had fallen to 776,995 applications, from a peak of 837,928 in January 2025, and that it was close to eliminating cases that would otherwise wait more than 36 months. Real progress, and you will still probably raise your next round before a first office action arrives. If a granted claim would genuinely change a term sheet, Track One targets a final disposition within about twelve months, for $4,515, or $1,806 as a small entity. Second, a utility application usually publishes within about 18 months of the earliest priority date, and sometimes sooner, so conversion is when the trade secret decision becomes final; anything you meant to keep secret has to come out before filing.
Want options outside the U.S.? This is also when you file a PCT application, which keeps 159 countries open and defers the expensive part until national phase entry, usually at 30 months from your earliest filing date.
The 30-month PCT deadline is where startup patent budgets break, because it arrives all at once and per country. National phase entry commonly runs $3,000 to $8,000 or more per country, and our budget guide puts broad international coverage at 10 to 15 times a U.S.-only budget over the life of the patent, with companies at this stage commonly spending six figures a year. Pick countries by where you will sell, manufacture, or be copied, and let the rest lapse without guilt.
Series B is also when the one-tactic rule expires. With several families in flight you have a portfolio whether you wanted one or not, and it deserves an annual hour with the board: map what covers what, prune what no longer matches the product, watch competitors’ filings for where they are heading. Before the next raise or an exit, bundle the applications into one documented asset; a portfolio described with a method can be underwritten and a “patent pending” bullet on a slide cannot.
Michael Meyer’s March 2026 guide says most VCs expect at least one filed application before investing in a hardware company. We think that overstates it. In the diligence we support, investors reliably check ownership (do the assignments exist) and want a plain answer to “what stops a competitor from copying this?” A filing can be part of that answer. In the deals we see, biotech, hardware, and deep tech investors do expect filings; seed-stage software investors mostly expect you to have thought it through and to own what you built. Build first, file the week before you show it. Plenty of the investors we meet do not value patents at all, and it is more productive to find the ones who do than to convert the ones who do not.
Paul Graham’s 2006 essay argued that startups rarely get sued over patents and that patents matter later, in the “mating dance” with acquirers. Twenty years and a first-to-file system later, that still holds, and it is the point: the acquirer, the lender, and the Series C lead are your patent’s customers. File early enough that the asset exists when they show up.
The pushback we hear most, in founder forums like this recent thread, is that trade secrets are cheaper and easier than patents. For backend software, often true, provided you do the registry work above. A secret nobody wrote down, restricted, or covered with an NDA is unlikely to clear the reasonable-efforts test if you ever need to enforce it.
If you are pre-seed, get the assignments signed and the three-bucket list written this month. It costs a few hours and removes the diligence question you least want to answer live.
If you are three to six months from a raise or a launch, pick the one invention you cannot afford to lose and get a provisional drafted like a patent before the deck goes out. Provisionals already in flight? Run the month-six cost projection now. By month eleven there is nothing left to decide.
Talk with our team if you want a stage-appropriate IP plan with a realistic number next to it.
Work with ipCapital Group
From invention to monetization, our team has guided 2,000+ engagements across the full IP lifecycle. Start with a free 30-minute discovery call.
Written by
Seth Cronin