## The Android of Robotics Physical Intelligence isn't building robots. They're building the brain that goes inside every robot. Their core product line — pi-0, pi-0.5, and the latest pi-star-0.6 — are Vision-Language-Action (VLA) foundation models trained across 7 robotic platforms on 68 different tasks. Accept a natural language command, translate it into physical manipulation. The key: one model works on any hardware. The "Android of robotics" is not a metaphor — it's the business model. ## What the Robot Brain Can Do The demos that turned heads: folding laundry from a hamper (unscripted, not teleoperated), assembling cardboard boxes, making espresso with a professional machine, grocery bagging, table bussing. These are dexterous, multi-step tasks that no prior robot learning system had achieved. The latest model, pi-star-0.6, uses a novel reinforcement learning method called RECAP. The robot watches human demos, gets corrected by a teleoperator, then practices autonomously thousands of times. On hard tasks, RECAP more than doubles task throughput and halves failure rates. ## The Team Six co-founders, nearly all from the academic frontier of robot learning: - **Karol Hausman** (CEO) — Former Staff Research Scientist at Google DeepMind, adjunct professor at Stanford - **Sergey Levine** (Chief Scientist) — UC Berkeley, one of the most-cited researchers in deep RL - **Chelsea Finn** — Stanford associate professor, pioneer in meta-learning and sim-to-real transfer - **Brian Ichter** — Ex-Google Research robotics - **Lachy Groom** (COO) — Former Stripe product leader turned prolific angel investor - **Quan Vuong** — Former Google DeepMind researcher This is perhaps the most credentialed robotics AI team ever assembled in a single company. ## The Numbers $1.07B raised across two rounds: $400M Series A and $600M Series B (Nov 2025) led by CapitalG (Google's growth fund). $5.6B valuation. Jeff Bezos, Thrive Capital, Lux Capital, Sequoia, NVIDIA all on the cap table. Founded in 2024. Headquartered in San Francisco. ## The Bet Against Vertical Integration Figure AI builds the complete robot. Tesla builds the complete robot. PI builds the intelligence layer and licenses it to everyone else. Their pi-0 base model is open-sourced — the same playbook that made Android the dominant mobile OS. The risk: no hardware means no control over deployment or data collection. The advantage: if robots become as diverse as smartphones, the universal brain wins. Every robot maker becomes a potential customer instead of a competitor. The physical AI market hit $3.1B in 2025 and is projected to reach $83.6B by 2035. PI is betting they'll be the intelligence inside most of it.