What is physical AI?
Physical AI is AI built into machines that do physical work, such as robots, drones, vehicles and industrial equipment, which must perceive, decide and act in real time.
Updated:
Definition
Physical AI describes artificial intelligence that acts in the physical world through a machine. It overlaps heavily with embodied AI.
The stack
- Compute on the device (sensing, inference and control).
- Models that turn perception into action.
- Hardware: actuators, hands, sensors.
- Safety and validation before deployment near people.
Why it matters
The term signals where AI investment is heading after chatbots: into factories, roads, construction and homes.
Frequently asked questions
What counts as physical AI?
Humanoids, robot arms, robotaxis, drones, warehouse robots and smart factory equipment: any AI whose output is physical action.
Why is physical AI harder than digital AI?
It needs real-time control, safety in human spaces, scarce training data, and hardware with limited power and weight.
Who is investing in physical AI?
Startups and large firms across chips, models, hardware and safety; for example Ubitium (processors), NEURA (models and robots) and SafeWorld (safety testing).
Coverage
The humanoid factory race: Korea, Shanghai and China's 10,000-unit plant
In one week Korea cleared 20 humanoids to work beside people, Shanghai set a goal of 8,000 factory humanoids by 2028, and UBTECH's 10,000-a-year plant came into focus. Here is what each move means.
NEURA's Neuron AI adds touch, sound and memory to the robot foundation model
NEURA Robotics launched Neuron AI, a 'perception-language-action' model that fuses 3D vision, touch, force, audio, robot state and language with memory, and will demo an 11-minute laundry task on its 4NE1 humanoid.
Promise Robotics signs D.R. Horton to robot-build parts for 1,250+ homes
Toronto's Promise Robotics will use its Factory-as-a-Service microfactories, running KUKA arms and domain-specific AI, to produce home components for America's largest homebuilder.
Ubitium raises $14M for one chip that senses, thinks and controls
The Düsseldorf startup's reconfigurable RISC-V processor aims to replace the DSP + NPU + CPU stack in robots and drones. Its first chip, UB108, works; the commercial UB410 is due in 2027.
AWS opens a Physical AI Toolchain with NVIDIA stack hooks
On Oct 8, 2026 AWS launched an open-source Physical AI Toolchain that stitches synthetic data, SageMaker training, simulation and edge deploy — with NVIDIA Isaac and Cosmos in the mix — after RoboMaker's 2025 shutdown.
Multiply Labs raises $75M Series B for robotic biomanufacturing
NantWorks led a $75M Series B for Multiply Labs; AstraZeneca and Teradyne joined as the San Francisco physical-AI company scales cell-and-gene and biologics manufacturing robots.
SafeWorld emerges with $12.2M to stress-test robots in simulation before they meet people
A Carnegie Mellon spin-out co-led by Shine Capital and a16z Speedrun runs thousands of simulated edge cases with reactive human models and issues third-party safety reports.
Teradyne invests in Bright Machines to robotize AI infra manufacturing
On Oct 5, 2026 Teradyne announced a strategic investment and collaboration with Bright Machines to combine UR cobots, board test and software-defined microfactories for AI data-center hardware.
Hitachi and FANUC team up on shop-floor Physical AI; customer rollout from FY2027
The Japanese giants will validate Physical AI inside Hitachi factories — picking, shape-varying parts, changeovers — then sell jointly from fiscal 2027, as industrial vendors race to productize the buzzword.
Newsletter
Robrief Weekly
One email a week with the robotics news that matters. Free, unsubscribe anytime.