As of Sep 22, 2026
All companies

Robotics foundation modelsFounded

Physical Intelligence

Builds AI models that teach robots a range of tasks.

Karol Hausman
Karol Hausman
Brian Ichter
Brian Ichter
Chelsea Finn
Chelsea Finn
Lachy Groom
Lachy Groom
Sergey Levine
Sergey Levine

Business performance

Read the revenue model ↗
Valuation
$5.60B
ARR
Not disclosed
Annual revenue
Not disclosed
Users / customers
Not disclosed
Net income
Not disclosed
Revenue multiple
Unavailable

Reading the numbers

Demo success and field throughput are research and operating measures. Without disclosed fees and support costs, they are not converted into revenue or profit.

01What it sells and who buys it

Physical Intelligence develops general-purpose models for robot manipulation. It works toward providing an intelligence layer for companies deploying robots rather than relying solely on selling its own finished robot.

02How the business makes money

Model collaborations with robot-deployment companies are the documented commercialization route. Partners operate revenue-generating robots, but the cited material does not disclose payments to Physical Intelligence or its revenue.

03Marketing and customer acquisition

Physical Intelligence connected research to partners’ laundry and logistics work.

  1. 01

    Apply a general manipulation model to robots

    The models support manipulation across robots, aiming to supply intelligence to deployment companies rather than only sell a single finished machine.

    Source
  2. 02

    Financing the next stage

    Reporting in November 2025 described a $600 million round at a $5.6 billion valuation. Financing supports model research, robot data and training infrastructure.

    Source
  3. 03

    How distribution and adoption expanded

    Research and demonstrations introduce the models, followed by validation at customer sites with Weave and Ultra. Laundry folding and order-packing results show autonomy and intervention improvements. The published outcomes invite other robot companies to evaluate partnerships.

    Source

How the growth mechanism works

  1. Publish research and demos
  2. Integrate a partner robot
  3. Measure field outcomes
  4. Expand to adjacent tasks

Weave and Ultra connect the model to real operators. Reduced human intervention and throughput support evaluation, though a partnership case study is not a revenue disclosure.

04Funding and expansion

Reporting in November 2025 described a $600 million round at a $5.6 billion valuation. Financing supports model research, robot data and training infrastructure.

↗Further reading

  1. 01pi.websiteRead source
  2. 02bloomberg.comRead source
  3. 03physicalintelligence.companyRead source
  4. 04indexventures.comRead source
  5. 05news.bloomberglaw.comRead source
  6. 06forbes.coRead source
  7. 07sequoiacap.comRead source
  8. 08brianichter.comRead source
  9. 09ai.stanford.eduRead source
  10. 10x.comRead source
  11. 11people.eecs.berkeley.eduRead source