Definition
Physical AI (also called Embodied AI) refers to AI systems that interact with the physical world through sensors and actuators. It encompasses robotics, autonomous vehicles, and any AI system that perceives and acts in real-world environments.
Recent Breakthroughs
Atoms $1.7B (July 2026)
atoms (Travis Kalanick) raised $1.7B led by andreessen-horowitz; Food/Mining/Transport industrial AI holding (2026-07-23-atoms-1-7b-a16z-kalanick-robotics).
Japan FRONTia / Noetra (July 2026)
noetra (SoftBank/Sony/NEC/Honda consortium) launches full-scale R&D for Japan multimodal foundation models for robotics under frontia-project, planning ~27,500 nvidia-rubin GPUs (~140 MW AI factory; ops ~June 2028) (japan-noetra-frontia-rubin-ai-factory).
Physical AI Capital Week (June 2026)
Two landmark funding events signal capital migration from software AI to physical-world engineering:
- prometheus (June 11): 41B valuation — “artificial general engineer” for design/manufacturing (2026-06-11-prometheus-12b-funding)
- neura-robotics (June 10): Up to $1.4B Series C led by tether — cognitive robots + neuraverse platform (2026-06-10-neura-robotics-1-4b-series-c)
- Robotics sector raised $55.8B in 2026 YTD per Dealroom — nearly double prior year
NVIDIA Cosmos 3 (May 2026)
nvidia launched cosmos-3 at GTC Taipei — first open omnimodel unifying vision reasoning, world generation, and action prediction. Mixture-of-Transformers architecture with Nano (16B) and Super (64B) variants. Open weights, six SDG datasets, Cosmos Coalition ecosystem.
Isaac GR00T Reference Humanoid (June 2026)
isaac-groot platform bundles unitree H2 Plus, sharpa hands, Jetson Thor compute for standardized academic humanoid research.
Sony Ace Robot (April 2026)
First autonomous robot to achieve expert-level performance in competitive high-speed physical activity (table tennis):
- Model-free reinforcement-learning
- 8 degrees of freedom robotic arm
- 75% serve return rate against professionals
- Real-time ball tracking with 9 cameras
Industrial Applications
nvidia and partners advancing physical AI for:
- Factory automation
- Warehouse logistics
- Hazardous environment robots
- Autonomous vehicles
Key Technologies
| Technology | Application |
|---|---|
| reinforcement-learning | Learning motor control policies |
| Computer Vision | Real-time environment perception |
| Sim-to-Real | Bridging simulation and physical deployment |
| High-speed sensors | Tracking fast-moving objects |
Related
-
cuspai Concepts