Arm's Agentic AI: Architecting Autonomous Future

Arm is expanding its focus into artificial intelligence, specifically agentic and physical AI. These technologies allow machines to perceive, decide, and act autonomously. What's interesting is this shift requires new ways of architecting computing, especially at the edge. Systems need to operate with low latency, limited power, and high reliability. Arm's work helps AI run efficiently on devices for applications like robotics and autonomous systems. Suraj Gajendra, from Arm’s Physical AI Business Unit, explains the company is addressing architectural, safety, and ecosystem challenges for agentic AI. Here's the thing: as AI moves from generating on-screen outputs to controlling machines, the architecture must be designed around the full sensing-to-decision-to-actuation loop. This means CPUs need to support deterministic, real-time control with strong security, and NPUs must provide efficient inference for complex AI models. The bottom line is that physical AI will be defined by how efficiently and safely the entire system closes this loop under real-world constraints. This is important because it will shape how intelligent machines interact with our physical world.