Startups with more than 30 patents exited more than 80% of the time in one angel group's portfolio.

Chris McKenna and Chethan Srinivasa of Foley & Lardner LLP recently tied patent data to funding outcomes in physical AI in The National Law Review. The numbers, and the limits of the underlying studies, merit consideration by founders and investors in autonomous systems.

Robotics and physical AI startups raised $27.6 billion in 2025, more than double 2024, per PitchBook, and $16.3 billion in Q1 2026 alone.

What the Correlation Data Shows

  • PitchBook (2011–2020): median valuations for patent startups averaged 17% higher at seed, 30% at early stage, 51% at late stage, and 93% at the angel stage. Counts include — and do not distinguish between — companies with granted patents or pending applications only.
  • TCA Venture Group, formerly Tech Coast Angels (281 outcomes, 1997–2023): exit probability exceeded 80% above 30 patents. Startups with over 51 patents returned an average of 38.7 times TCA's investment across exits and shutdowns, versus 1.4 times with no patents.
  • European Patent Office and EUIPO - European Union Intellectual Property Office (2023, Europe): startups with prior patent applications but no trade mark filings had 2.9 times higher odds of seed funding, and 6.4 times at early stage, than startups with no IP filings.

A 2021 Srinivasa and Pakin Pongcheewin study found about 20 U.S. patent assets at each of three AI startups Apple acquired at values of $100 million or more (Drive.ai, Xnor.ai, and Emotient). Apple's price was publicly reported only for Xnor.ai, about $200 million.

Correlation vs. Causation in Patent Value

The authors treat the studies as correlation, not causation. Filers may also have stronger teams, more capital, or more mature products. TCA also counts a failing company's sale of its patents as an exit, so part of the higher exit rate can reflect residual patent value.

Detectability and § 101 Eligibility in Physical AI

The authors' framework maps filings to milestones (provisionals at pre-seed, utility filings at seed and Series A, international filings as markets expand) and flags detectability in diligence: patents on internal algorithms or embedded control systems can be hard to enforce if infringement cannot be observed or reverse-engineered.

The article also does not directly discuss patent eligibility. Recentive Analytics v. Fox Corp. (Fed. Cir. 2025), cert. denied, held that patents claiming no more than generic machine learning applied to a new data environment, without disclosing technical improvements to the models, are ineligible under § 101. Claims directed to sensing, control, and actuation can still recite a technological improvement in the machine's operation, apart from the model.

Read together, detectability and eligibility favor similar physical AI drafting choices, including claiming features of the machine that can be observed.

Sources

  1. McKenna & Srinivasa, The National Law Review (Sept. 21, 2026): https://natlawreview.com/article/can-patent-filings-help-physical-ai-companies-raise-capital
  2. PitchBook, Q4 2025 Robotics & Physical AI VC Trends: https://pitchbook.com/news/reports/q4-2025-robotics-physical-ai-vc-trends
  3. PitchBook, Q1 2026 Robotics & Physical AI VC Trends: https://pitchbook.com/news/reports/q1-2026-robotics-physical-ai-vc-trends
  4. PitchBook Analyst Note, Introducing PitchBook Patent Research (Feb. 2023): https://files.pitchbook.com/website/files/pdf/Q1_2023_PitchBook_Analyst_Note_Introducing_PitchBook_Patent_Research.pdf
  5. John Harbison, TCA Venture Group, Do Patents Affect Outcomes in Early Stage Investing? (Apr. 22, 2024): https://tcaventuregroup.com/15119-2/