Physical Intelligence Funding History

Physical Intelligence is a US startup building foundation models for robotic manipulation.

Company Overview

  • Robotics

Founded 2024 · San Francisco, California, US

Official Website

Physical Intelligence Funding Timeline

  1. Announced Mar 12, 2024

    Seed

    US$70 million

    The seed investment backed research into general-purpose software that could control different types of robots.

    Led by Thrive Capital. Participation included Khosla Ventures, Lux Capital, OpenAI, Sequoia Capital.

    Source · maginative.com ↗
  2. Date under review

    Series A

    US$400 million

    The company raised the financing while developing software intended to control different kinds of robots.

    Led by Bezos Expeditions.

    Source · reuters.com ↗
  3. Date under review

    Series B

    $600M · $5.6B valuation

    The financing supported work on general-purpose AI software that helps robots learn and perform a wider range of tasks.

    Led by CapitalG.

    Source · bloomberg.com ↗

Physical Intelligence Founding Team

Physical Intelligence brought together an early team that included Karol Hausman, Sergey Levine, Chelsea Finn, Brian Ichter, Lachy Groom, and Quan Vuong. They worked on general-purpose robot control, testing whether one model could guide different robots through a range of physical tasks.

Physical Intelligence Company Development

A Model That Can Act Through a Robot

Physical Intelligence made its approach concrete with π0, a vision-language-action model for robot manipulation. The release showed how visual observations and task instructions could feed into actions, an additional step beyond a model that interprets what it sees.

Physical Intelligence Competitive Landscape

Adapting a Robot Policy to New Hardware

Physical Intelligence released π0 code and weights so labs can adapt a policy to their own robot and task data. Eka Robotics emphasizes force as an input and uses simulation to train for contact-heavy work such as inserting wires. For teams working on manipulation, the contrast is between adapting a released checkpoint and evaluating a system built around force feedback.