# Provisional vs. Non-Provisional Patents: When to File Each for Deep-Tech Inventions
For solo inventors and small technical teams working on hardware, software, or hybrid systems, the choice between filing a provisional patent application (PPA) and jumping straight to a non-provisional (utility) patent is one of the highest-leverage decisions you'll make in your IP strategy. Get it wrong, and you'll either waste money filing too early on half-formed ideas, or miss critical deadlines and lose protection windows. Get it right, and you've bought yourself breathing room, captured a priority date, and preserved your options without burning cash.
The short version: provisional patents exist to buy time and lock in a priority date cheaply. Non-provisional patents exist to get actual legal protection. Most deep-tech inventors benefit from filing provisional first—if they file at all.
What a Provisional Patent Actually Does
A provisional patent application is a simplified, inexpensive filing with the US Patent Office that establishes a priority date for your invention. It costs between $110 and $275 (depending on entity size), requires no claims section, no oath or declaration, and no formal examination. You just describe your invention thoroughly in a specification—drawings, pseudocode, CAD models, schematics, test results, everything that supports the technical novelty of what you've built.
The critical advantage: that priority date. If you file a provisional on January 15, 2026, and someone else independently invents the same thing and files a non-provisional on February 1, 2026, your invention has the earlier priority date. In examination, in litigation, in patent offices worldwide—you're first.
A provisional patent lasts exactly 12 months. At month 11, you must either file a non-provisional claiming priority to that provisional, or your provisional expires and the priority date is gone forever.
What it does NOT do: A provisional patent does not give you legal protection. You cannot sue anyone for infringement based on a provisional alone. The claims that define what you've actually patented don't exist yet. You're not patented; you're "patent pending," which is a signal to the market and a placeholder in the system, nothing more.
What a Non-Provisional Patent Does
A non-provisional (utility) patent application is the real thing. It includes a detailed specification, drawings, claims (the legal definition of what you've patented), an abstract, and background. It triggers examination by a patent examiner at the USPTO. The examiner will search prior art, ask questions, reject claims, and force you to narrow, amend, or defend your invention's novelty and non-obviousness.
If your non-provisional is approved, you get a patent: a government-issued property right that lasts 20 years from the filing date (or from the priority date if you claimed priority from a provisional). During that 20 years, the patent holder can exclude others from making, using, selling, offering to sell, or importing the patented invention in the US.
Non-provisionals are expensive and slow. Filing costs $900–$4,500 depending on entity size. Prosecution (examination, rejections, amendments, appeals) typically takes 2–5 years and costs $5,000–$20,000 or more if claims are heavily contested or if you hire an agent to respond to office actions.
The Solo Inventor's Path: Provisional → Non-Provisional
For most solo deep-tech inventors, the optimal sequence is:
- File a provisional when you have a clear, reproducible description of your invention. You don't need working hardware. You don't need a prototype. You need a specification detailed enough that someone skilled in your field could understand what you've built and why it's novel.
For software: pseudocode, architecture diagrams, and examples of novel algorithms or data structures.
For hardware: CAD models, schematics, simulation results, and a description of the design's unique aspects.
For hybrid systems: all of the above, plus how the components interact.
- Use that 12-month window to validate, build, test, and learn. If your invention doesn't work, file nothing else and let the provisional expire. The $275 is a cheap lesson. If it works better than you expected and looks genuinely defensible, move to step 3.
- Before month 11, file a non-provisional claiming priority to the provisional. If you've learned new things about the invention (edge cases, parameter ranges, design variations that improve performance), include them in the non-provisional. You can claim priority to the provisional even if the non-provisional discloses more.
The advantage here is optionality with a safety net. You've locked in a priority date for less than $300. You've given yourself a full year to validate before spending $5,000–$20,000 on prosecution. And the non-provisional benefits from your year of validation—your specification is stronger, your understanding is deeper, and your claims are more likely to survive examination.
When NOT to File Provisional
There are cases where filing a provisional is a waste:
- Trade secrets with indefinite value. If your invention is a manufacturing process, an algorithm, or a system architecture that will remain secret and competitively valuable for more than 20 years, patenting it is counterproductive. A patent expires; a trade secret doesn't. Keep it confidential instead.
- Inventions with very short market windows. If you're inventing something in a market that moves at six-month cycles (certain mobile software, graphics libraries, fast-fashion manufacturing processes), by the time your patent issues, the market may have moved on. Evaluate: are you protecting market position, or locking up the idea for its own sake?
- Foundational research that's not yet inventive. If you're describing pure theory or mathematical exploration without a clear, practical application, examiners will reject your claims as abstract. Wait until you've tied the theory to a concrete, reproducible implementation.
- Obvious variations on existing work. If your invention is a minor tweak to published prior art (a 5% parameter adjustment, a standard integration of existing components), file provisionals only to learn from office action rejections—don't plan to defend them.
When to Skip Provisional and Go Straight to Non-Provisional
In rare cases, skipping the provisional and filing non-provisional directly is the right call:
- You're well-capitalized and know exactly what you're patenting. If you've had the invention fully developed, tested, and validated before filing, and you have the budget to prosecute confidently, the provisional adds process overhead without value. Go straight to non-provisional.
- You need protection NOW and can't wait 12 months. If you're about to license, raise institutional capital, or enter a competitive market where early filing dates matter, file non-provisional immediately. The provisional only buys time; it doesn't buy rights.
- Your invention is narrow, well-defined, and unlikely to change. If you've invented a very specific chemical compound, a precise mechanical assembly, or a tightly-scoped algorithm, and you're confident it won't be improved or modified, the extra validation year the provisional offers is less valuable.
Even in these cases, many sophisticated inventors still file provisional first—the cost is trivial compared to the optionality it preserves.
International Considerations
If you're thinking beyond the US, understand the cost/benefit of provisional patents in your jurisdictions of interest:
- PCT (Patent Cooperation Treaty): You can file a PCT application claiming priority to your US provisional. This buys you an additional 18 months to decide which countries to enter (Japan, EU, China, South Korea, etc.). The PCT doesn't grant patents; it's a single filing that buys time in multiple jurisdictions.
- Provisional has no international equivalent: Only the US offers cheap, simplified provisional filings. If you file a provisional in the US, you MUST file a corresponding application in other countries within 12 months to preserve priority there.
For deep-tech inventors with global aspirations, the sequence is often: US provisional → validate → US non-provisional + PCT application (same filing) → enter specific countries at month 18.
Building a Defensible Specification
Regardless of whether you file provisional or non-provisional, the strength of your specification is everything.
Include:
- A detailed description of the problem your invention solves and why existing solutions are inadequate.
- The exact mechanism or approach your invention uses, with sufficient detail that an engineer could implement it from your description.
- Drawings, diagrams, or pseudocode showing the structure, flow, or architecture.
- Examples and test results demonstrating that the invention works and produces claimed benefits.
- Design variations and alternate embodiments that still achieve the same core innovation.
- Parameter ranges and tolerance levels for any numerical or physical values critical to the invention.
A weak specification—vague descriptions, hand-wavy benefits, no evidence that the invention actually works—will result in rejections during examination that are hard to overcome. You can't add significant new technical content after filing. What you file in your specification is what you've got to work with.
Timing Your Non-Provisional Filing
If you're using the provisional → non-provisional path, file your non-provisional between month 10 and month 11 of your provisional's life. Don't file at month 12—you risk missing the deadline and losing priority.
Your non-provisional should:
- Explicitly claim priority to the provisional (this is a single line: "This application claims priority to US Provisional Patent Application No. [number], filed [date]").
- Include everything in the provisional, plus new material learned over the past 12 months.
- Use the same inventive concept so the priority claim is valid. You can't add a wholly new invention and claim priority to the old provisional.
The Reality Check
Filing patents is not the same as having defensible IP. A granted patent that covers only obvious variations of existing work is expensive wallpaper. A patent that genuinely protects a novel, non-obvious, and commercially important invention is a moat.
Before you file—provisional or otherwise—be honest: Is this invention genuinely novel? Would a skilled engineer in your field have naturally arrived at this solution? Does it solve a real problem, or is it an intellectual exercise?
Use the provisional filing as a decision checkpoint, not as the start of a multi-year commitment you're forced to honor. Many deep-tech inventors file provisional applications on three ideas, file non-provisional on one of them, and ultimately patent zero. That's fine. The provisional gave you cheap optionality.
The goal isn't to maximize patent count. It's to protect genuinely defensible innovations while spending money strategically.
File when you have something real. Validate during the provisional year. Commit to non-provisional only on inventions worth defending.