Skip to main content

Physical Intelligence

Category: Robotics / Embodied AI

Building general-purpose AI foundation models for robotics that control any robot to perform any task in the physical world. Physical Intelligence was founded in 2024. The company is led by Karol Hausman. Based in San Francisco, United States. Team size: 80. Total funding raised: $1,070,000,000. Latest round: Series B (Nov 2025). Key investors include ["Jeff Bezos","OpenAI","Thrive Capital","Lux Capital","Sequoia Capital","Khosla Ventures","Bond Capital","CapitalG","Redpoint Ventures","Index Ventures","T. Rowe Price"].

Founded
2024
Headquarters
San Francisco, United States
Team size
80
Total funding
$1,070,000,000

Value proposition

A single generalist robot foundation model that controls diverse robot types (7+ platforms) across 50+ tasks without task-specific programming, enabling rapid deployment, adaptation, and hardware-agnostic intelligence.

Products and solutions

["π0 (open-sourced generalist robot policy)","π0-FAST (5x faster training tokenizer)","π0.5 (mobile manipulator with open-world generalization)","π*0.6 (RL-enhanced model learning from experience)","Multi-Scale Embodied Memory (MEM)","Real-time action chunking systems"]

Unique value

Building a single generalist robot foundation model—'ChatGPT for robots'—that is hardware-agnostic, controlling any robot across 50+ tasks and 7+ robot platforms without task-specific engineering.

Target customer

Robot manufacturers, logistics and warehousing companies, manufacturing facilities, enterprise automation sectors, and long-term service robotics applications in homes.

Industries served

["Robotics & Embodied AI","Logistics & Warehousing","Manufacturing","Defense & Aerospace","Healthcare Robotics","Service Robotics"]

Technology advantage

Novel flow-matching vision-language-action (VLA) architecture that combines internet-scale semantic knowledge with continuous action outputs; trains on large-scale multi-task, multi-robot data; open-source releases accelerate ecosystem adoption.

How they differentiate

Pure software-first approach building hardware-agnostic foundation models for robots ('ChatGPT for robots'); focuses on general-purpose robot intelligence across 7+ platforms and 50+ tasks with novel flow-matching VLA architecture; open-sources models to accelerate ecosystem adoption; ended OpenAI partnership to develop proprietary models

Main competitors

["Skild AI","Figure AI","NVIDIA"]

Key partnerships

["WEKA (high-performance data infrastructure)","Oracle Cloud Infrastructure","Academic affiliations: Stanford University, UC Berkeley","Strategic investors: Jeff Bezos, OpenAI, Thrive Capital, Sequoia Capital, Khosla Ventures"]

Notable customers

["AgiBot (manufacturing pilot)","Longcheer Technology (manufacturing deployment)","WEKA (infrastructure partner)","Oracle Cloud Infrastructure"]

Major milestones

["Founded January 2024","π0 generalist robot policy release (Oct 2024)","$400M Series A at $2.4B valuation (Nov 2024)","Open-sourced π0 model with weights (Feb 2025)","$600M Series B at $5.6B valuation led by CapitalG (Nov 2025)","Manufacturing pilot with AgiBot and Longcheer Technology (2025)","Discussions for $1B additional funding at $11B+ valuation (Mar 2026)"]

Growth metrics

Pre-revenue R&D stage; 80+ employees; 7+ robot platforms supported; 50+ tasks demonstrated; π0, π0.5, π*0.6 model releases; open-sourced π0 (Feb 2025)

Market positioning

Pre-revenue R&D-stage foundation model provider competing to become the dominant 'robot brain' platform layer for robotics; highest-funded pure-play robotics AI startup

Geographic focus

San Francisco Bay Area (primary), North America, global research partnerships and manufacturing pilots

Patents and IP

No registered patents disclosed as of latest update; primary IP resides in proprietary model architectures, training methodologies, and data pipelines.

About Karol Hausman

Co-founder and CEO of Physical Intelligence. Previously served as Staff Research Scientist at Google DeepMind (2017-2024) and Adjunct Professor at Stanford University (2021-present). Expertise in machine learning, robotics, and reinforcement learning with a PhD in Computer Science from University of Southern California (2012-2017). Pioneer in robot learning research with over 100 publications and significant contributions to RT-X, a large-scale robotics dataset collaboration.

Official website: