Arm has launched a new collaboration programme with more than 80 companies to speed up the development of physical AI-intelligent systems that perceive, reason, and act in the real world. The initiative, called Arm Total Design for Physical AI, targets the integration challenges holding back commercial deployment of robotics, autonomous vehicles, and industrial automation, sectors Arm believes represent trillions of dollars in economic opportunity.
Participants span the full physical AI technology stack, including AWS, ECARX, Hugging Face, Liquid AI, NXP, PlusAI, PSYONIC, QNX, Qwen, Siemens, and Unitree Robotics. Arm said the growing complexity of these systems means no single organization can build the required technologies alone. Deployable solutions demand tight integration of AI models, software, computing platforms, sensors, and actuators, while meeting strict safety, reliability, and scalability requirements.
Building on the cloud AI blueprint
The programme draws on Arm's experience with its Total Design programme for cloud AI infrastructure. Partners will collaborate across AI software, sensor technologies, computing hardware, virtual development platforms, and digital twins. The goal is to enable earlier testing and validation of physical AI applications, cutting the time from research and pilot projects to commercial rollout. For developers and IT teams working on embedded or edge systems, this kind of cross-ecosystem alignment directly reduces integration friction.
Arm Chief Architect Richard Grisenthwaite published a manifesto alongside the announcement. He argues that while robotics technology is advancing rapidly, the industry lacks a standardized way to describe and compare the capabilities of increasingly sophisticated machines. "This fragmentation makes robotic systems more difficult to design, integrate and scale," the manifesto states.
A common language for robotics capability
To address that fragmentation, Arm introduced a Robotics Capability Framework. The framework proposes progressive levels of robotic capability, linking real-world applications with specific system requirements such as latency, computing architecture, memory demands, power consumption, determinism, and safety. It is an initial step toward creating a common language for evaluating and communicating what a given robotic system can actually do.
The framework's initial design was shaped through consultation with a broad range of industry participants, including Anaxi Labs, ANYbotics, FMC³ Robotics, Fourier, GALBOT, Gravis Robotics, Lenovo, McKinsey, and Robotec.ai. Arm has invited wider industry participation as the framework evolves.
Why this matters for IT and development professionals
The shift toward physical AI and agentic systems-machines capable of performing tasks autonomously in the real world-will place new demands on software stacks, compute architectures, and integration pipelines. For developers and IT operations teams, the push for common standards and capability frameworks could reduce the bespoke engineering work currently required to connect AI models with physical hardware. Professionals building skills in areas like embedded AI, sensor fusion, and digital twins will be closest to the deployment opportunities as these technologies move from lab to production. The AI Learning Path for Software Developers offers a structured way to build those competencies, while the broader AI for IT & Development resources cover the operational and infrastructure skills that physical AI systems will demand at scale.
Your membership also unlocks:
