Alphabet's CapitalG leads $600m round for US robotics startup

Physical Intelligence, a San Francisco-based robotics startup, has raised US$600 million in new funding, valuing the company at US$5.6 billion.

Alphabet’s independent growth fund CapitalG led the round, with participation from Index Ventures and T. Rowe Price.

Existing investors Lux Capital, Thrive Capital, and Amazon founder Jeff Bezos also took part, according to people familiar with the matter who asked not to be identified discussing private information.

Founded in 2024 by former Google DeepMind researchers and academics from Stanford and UC Berkeley, the company develops AI software for robots to perform various tasks.

It has tested its software on robotic arms for folding clothes, making coffee, and assembling boxes.

The company also recently released an AI vision model that uses reinforcement learning to improve performance as industry interest in robotics continues to rise.

🔗 Source: Bloomberg

🧠 Food for thought

Implications, context, and why it matters.

Few OEM deals and limited pilot info

Physical Intelligence ran tests where robotic arms assembled boxes in about 2.5 minutes. It has not announced partnerships with industrial robot OEMs (original equipment manufacturers) like ABB, FANUC or Universal Robots.
Telexistence, a retail robotics company in convenience stores, is working with Physical Intelligence to build error recovery for drink restocking (automatic fault handling during tasks) 1.
SAP, an enterprise software company, through its Embodied AI initiative (Project Embodied AI) cites up to 50% cuts in unplanned downtime plus up to 25% productivity gains 2. Pilots include NEURA Robotics, BITZER and Sartorius 2.
Focus on general-purpose foundation models like π0, a model trained on diverse robot types, could slow revenue versus rivals with faster narrow offerings 3.

Brownfield integrators stand to gain from AI manipulation demand

Warehouse robotics integrators will see more brownfield retrofits (upgrades to existing facilities) as operators want intelligent manipulation without replacing entire systems 4. Examples include Bastian Solutions, JR Automation and enVista.
Prime Robotics and other goods-to-person system providers (systems that bring items to human pickers) can add AI vision models to AMR fleets (Autonomous Mobile Robots) to take on picks that now need people 5.
Suppliers of fisheye cameras, pinhole cameras or varied end-effectors could get a lift as AgiBot (a robotics platform focused on multi-sensor manipulation) with Physical Intelligence built a model that integrates multiple sensor types and used it for complex tasks 6.