What investors read in a quantum startup's patent portfolio
Investor counsel reads a quantum startup's patent families twice, in the data room and in the public record at the USPTO and EPO. How to spend a limited patent budget so both tell the same story.
A quantum computing startup at seed or Series A can realistically fund a handful of patent families. Counsel for investors can review those families in the diligence data room and, once they have published, in the public record at the United States Patent and Trademark Office (USPTO) and the European Patent Office (EPO). The portfolio has to protect what is proprietary, match the story in the pitch deck, and hold up at both offices.
What diligence counsel can already see
Patent applications publish 18 months after the earliest priority date. From then on the whole file is public, in Patent Center and the European Patent Register: the claims, every office action, every reply, every abandonment and every lapsed fee. Anything filed more recently can only be seen in the data room.
The data room carries the newest filings, which are usually the ones closest to the product. The public file carries the history. The two should tell the same story.
Publication also sets a deadline. An application abandoned before publication is generally never published, so a decision to drop a family should be made before the 18-month mark.
Put the money where the offices grant
Error-correction filings are granted at 94% of concluded applications at the USPTO and 88% at the EPO, and we found no EPO refusal in the class. Algorithm filings do worst at the EPO, at 62%, because the EPO asks whether the quantum hardware shaped the method.
For a hardware or control-stack company, the first families should cover the device and the parts of the error-correction cycle specific to its platform. For an algorithm company, file the method at the USPTO, file at the EPO only on claims built around the processor’s limits, and protect simulation outputs by contract and trade secret.
Buy fewer families and get them granted
83% of Google’s EPO grants came after one examination report or none. A long prosecution stays in the public file, along with every amendment that narrowed the claims.
Much of what can shorten examination is in the specification. Name the quantum computers that exist and say which claimed operations run on them. Define terms of art the way the field’s papers define them. State the quantum operating regime as a condition in the claim. Draft one specification for both offices from the first day.
Think before abandoning
Once an application has published, an abandonment is part of the record. At the USPTO, one examiner rejected 13 of 20 quantum software applications on the theory that the invention needs a fault-tolerant computer that does not yet exist. Nine applicants abandoned. The three who appealed won on appeal at the USPTO’s Patent Trial and Appeal Board and had patents within eight months, and every appellate reversal in the USPTO’s quantum record became a patent.
At the EPO, 22 of the withdrawals we reviewed lapsed on a missed fee, ten of them with an allowable claim already on the table. A docket and a written renewal policy prevent that.
Check the papers against the claims
Quantum startups publish to hire and to raise, and the EPO’s opposition division used Rigetti’s own 2012 paper to define a term in Rigetti’s patent against it. Before a preprint goes to arXiv, ask two things: does an application need to be filed first, and does the paper use any claimed term differently from the patent.
Keep a continuation pending
Keep one USPTO continuation open on the core family. It lets the company write new claims later, for the machine it actually builds, with the original filing date, as long as the original specification supports them. That is a reason to put more detail into the first filing, not less. Say so in the data room.
Clean up ownership
Many quantum startups begin with a university license, and some take government money with invention clauses attached. Quantinuum’s prospectus describes a policy of limiting reliance on university-owned IP, and a letter of intent with the US Department of Commerce, described in the same prospectus, would give the US government a license to inventions developed with the funding and keep that IP in US company ownership for ten years. Terms like those can limit who is able to buy the company.
Know who holds the claims nearest your design. At the error-correction layer, Google, IBM and Microsoft hold 293 of the 878 error-correction patents the USPTO has granted. A seed company may not need a clearance opinion yet, but it should know who holds those claims.
What the data room should say
A one-page summary does most of the work: each family, where it is filed, its status, what part of the product it covers, why anything pending is still pending, and what was deliberately kept as a trade secret instead of filed. The aim is a patent file that raises no question the company cannot answer.
Source: CipherLaw; based on the USPTO’s records for 11,306 quantum applications, the EPO’s for 5,590, the 34 quantum appellate decisions of the USPTO’s Patent Trial and Appeal Board, and the 31 EPO refusals and 39 EPO withdrawals.