From photon to actuation: How physical AI is redefining embedded compute
Arm outlines how physical AI systems integrate perception, reasoning, and actuation under strict power and safety constraints, highlighting a $200 billion compute opportunity by the 2030s.
Embedding AI into industrial machines, robots, and vehicles requires systems that perceive surroundings, reason, coordinate movements, and respond quickly within tight power, weight, heat, and cost limits. These physical AI systems must translate intelligence into reliable physical action, such as a robot grasping an object without breaking it, while managing real-time adjustments via sensors and actuators. The challenge extends beyond AI models to building complete systems that integrate perception, reasoning, and actuation under strict operational constraints. Safety and security are foundational to ensuring predictable, reliable operation essential for trust in autonomous machines.
Arm identifies three interconnected computing planes—perception-driven, interaction-driven, and actuation execution—supplemented by cloud compute for fleet management and training. At the perception and interaction planes, Arm Zena Compute Subsystems integrate processors, AI accelerators, memory, and interconnects, while collaborations with Cadence simplify integration and verification. Components in the actuation execution plane require fast, predictable responses to translate AI decisions into physical action, where embedded processors play a critical role.
Arm and STMicroelectronics are shaping the next generation of the Cortex-M roadmap to meet growing sensing, processing, and control demands in intelligent machines. At the STM32 Summit, STMicroelectronics introduced a converged STM32 platform for physical AI and a new series of Cortex-M-based STM32A microcontrollers designed for autonomous systems requiring high predictability and redundancy. The partnership combines Arm’s processor architecture with STMicroelectronics’ embedded expertise to bring intelligence closer to where machines interact with the physical world.
The Cortex-M portfolio, including processors like Cortex-M55 and Cortex-M85, powers billions of devices and delivers advanced signal processing and machine learning capabilities to microcontroller-class systems. As intelligence moves closer to sensors and actuators, Cortex-M provides precise, predictable processing to translate AI decisions into physical action. Arm emphasizes ecosystem collaboration as critical to accelerating innovation in physical AI, with continued investment in technology roadmaps and partnerships to unlock economic and productivity opportunities.