Updated
Updated · Ars Technica · Oct 8
Nvidia Expands Halos to Robotics as Physical AI Nears $10 Billion in Annual Revenue
Updated
Updated · Ars Technica · Oct 8

Nvidia Expands Halos to Robotics as Physical AI Nears $10 Billion in Annual Revenue

3 articles · Updated · Ars Technica · Oct 8

Summary

  • Nvidia has widened its Halos safety system from autonomous vehicles to robots, aiming to make warehouse, factory and surgical machines safer to deploy around people.
  • Safety is emerging as the next bottleneck for physical AI, Nvidia said, even as AI models and robot hardware become capable enough for broader real-world use.
  • Halos for Robotics, announced in June 2026, builds on the 2025 Halos launch for self-driving and other autonomous vehicles with hardware and software guardrails.
  • IGX Thor sits at the hardware core of the system, pairing robotics computing with an independent processor dedicated to safety-related workloads on the same silicon.
  • The push supports a business Jensen Huang said was already generating nearly $10 billion in annual revenue, underscoring Nvidia’s broader bet beyond AI data centers.

Insights

With a massive physical AI empire, is Nvidia building a vital safety standard or creating an inescapable monopoly for future robotics?
When a learning robot alters its behavior post-deployment, can a static hardware chip really guarantee human safety in a dynamic workspace?