Physical Intelligence raises $600M to advance robot foundation models

Physical Intelligence, a San Francisco-based robotics startup, has raised $600 million in Series B funding to advance its robot foundation models, bringing its total funding to $1.1 billion and a valuation of $5.6 billion. Led by CapitalG with participation from Lux Capital, Sequoia, Redpoint , Bond, and returning investors including Jeff Bezos and OpenAI , the company aims to create universally embodied AI—a single generalist intelligence capable of performing a wide range of real-world tasks across multiple robot platforms. Founded in 2024, Physical Intelligence develops vision-language-action (VLA) models and uses the RECAP (RL with Experience & Corrections via Advantage-conditioned Policies) approach to train robots. This allows machines to learn from demonstrations, corrective guidance, and autonomous experience, enabling them to efficiently handle tasks such as folding laundry, assembling boxes, operating machinery, or performing complex manufacturing operations. Recent tests show increased throughput and lower failure rates compared to imitation learning or traditional reinforcement learning alone. Impact on U.S. manufacturers : 1. Productivity – Robots can handle repetitive or complex tasks faster, increasing throughput. 2. Flexibility – Foundation models let robots adapt to new product lines without extensive reprogramming. 3. Labor efficiency – Reduces dependence on human labor for manual or hazardous tasks. 4. Quality & consistency – AI-driven robots decrease error rates and improve precision in manufacturing. 5. Scalability – Factories can scale operations quickly to meet fluctuating demand.