Robots that can see, reason and act are no longer the hard part, Arm argues. The hard part is building autonomous machines that people will actually trust outside the lab, and that is the subject Arm is taking to RoboBusiness 2026, which runs October 20-21 in Santa Clara, Calif.
The talk, titled “Building Physical AI that Scales”, will be given by Dermot O’Driscoll, VP of go-to-market for physical AI at Arm. The session was announced by The Robot Report, which produces the event, on September 14, 2026. This edition also marks the 20th anniversary of RoboBusiness.
From capable to deployable
Arm’s starting point is that foundation models, vision-language-action systems and edge AI have widened what robots can sense, reason about and do. The company frames the next step as turning those abilities into systems that work reliably in the real world, citing examples such as intelligent prosthetics that restore mobility, robots that deliver critical medical supplies, and machines that help workers operate more safely and productively.
As robots grow more capable, Arm says engineering priorities shift toward system-level design. O’Driscoll’s session is set to cover:
- how intelligence is distributed across the robot, and across cloud and edge compute;
- how learned and deterministic behaviors interact;
- how compute platforms balance real-time responsiveness, efficiency and safety;
- how to keep systems adaptable, reliable and safe as they move into production.
O’Driscoll has spent more than 20 years at Arm in engineering and leadership roles spanning IT, EDA, CPU and SoC design. He holds a bachelor’s degree in Electronics Engineering and a master’s degree in Microelectronics from the University of Edinburgh.
The ecosystem behind the talk
The RoboBusiness slot follows a bigger move by Arm. On September 8, 2026, the company expanded Arm Total Design for Physical AI, a collaborative ecosystem of over 80 companies across the physical AI stack. Named participants include AWS, ECARX, Hugging Face, Liquid AI, NXP, PlusAI, PSYONIC, QNX, Qwen, Siemens and Unitree Robotics. The stated aim is to reduce integration complexity and shorten the path from design to deployment for autonomous systems.
Alongside it, Arm introduced a Robotics Capability Framework, meant as a common language for describing robots. It defines levels of increasing sophistication, from RL0 (reactive systems) to RL5 (self-improving systems), and ties real-world use cases to requirements such as latency, compute placement, memory, power constraints, determinism and safety. Arm puts the industries involved at trillions of dollars in global economic activity and a $200B annual compute opportunity in the 2030s.
O’Driscoll has already been the public voice of that effort. Welcoming location-intelligence firm ZaiNar as a founding partner, he said that “as physical AI moves toward deployment, autonomous systems need a shared understanding of the environments around them.” Expect the Santa Clara talk to put that system-level thinking in front of the commercial robotics developers who have to ship it.
Earlier on GismoLand: ARM Physical AI. Drew Henry maps out the chip giant’s robotics playbook.
