Physical Intelligence (Pi), the San Francisco-based robotics AI company founded by former Google DeepMind and Stanford researchers, is in advanced discussions to raise its next funding round at an $11 billion valuation — a figure that would make it one of the most valuable private robotics companies in the world.
What Physical Intelligence Does
Pi's core product is a general-purpose robot control foundation model called π0 (pi-zero) that can be fine-tuned to control different robot hardware for different tasks. Unlike traditional robot programming, which requires task-specific code for every application, π0 learns from demonstration data and can generalize across tasks it has never explicitly been trained on.
The model has demonstrated capabilities including:
- Folding laundry (multiple garment types, without task-specific training)
- Loading dishwashers (handling varied dish shapes and placement patterns)
- Assembly tasks (complex multi-step manipulation)
- Kitchen tasks (food preparation using standard tools)
What makes π0 commercially significant is its claimed transferability: the same model weights can be adapted to different robot hardware platforms by retraining on a much smaller task-specific dataset. This potentially commoditizes the most expensive part of robot deployment — programming.
Funding History and Investor Profile
Pi raised $400 million in late 2024 at a $2.4 billion valuation, with investors including Jeff Bezos, Khosla Ventures, and OpenAI. The reported $11 billion valuation at this new round would represent a 4.6x jump in 18 months — rapid even by AI startup standards.
The valuation is driven by a combination of commercial traction (multiple robot OEM partnerships disclosed) and the broader investor conviction that general-purpose robot AI will be a platform business similar to operating systems or cloud computing.
Competition and Market Dynamics
Pi is competing with several well-funded players in the robot foundation model space:
| Company | Model | Valuation / Status |
|---|---|---|
| Physical Intelligence | π0 | ~$11B (talks) |
| 1X Technologies | EVE/NEO models | ~$500M raised |
| Covariant | RFM-1 | Acquired by Amazon 2024 |
| Google DeepMind | RT-2, GROOT | Part of Alphabet |
| Figure AI | Helix model | $2.6B raised |
Notably, Amazon's acquisition of Covariant's team suggests that large tech companies view robot foundation models as strategically important enough to acquire rather than build independently.
Implications for Robot Buyers
For manufacturers evaluating robot investments in 2026, the rise of foundation models has a near-term practical implication: reprogramming costs are likely to fall significantly over the next 3-5 years as these models mature. Robots that today require expensive custom programming for each product change may, by 2028-2030, be reprogrammable with minimal engineering time.
This should be factored into 10-year TCO models — see robot total cost of ownership guide for how to model programming cost over a robot's lifetime.