NVIDIA has introduced Halos OS, a new operating system designed for robotics that extends the safety standards of autonomous vehicles to industrial robots and physical AI systems. The platform, built on NVIDIA's IGX Thor computing module and the Halos Core software stack, aims to give manufacturers a ready-made safety architecture for machines that work alongside humans.
What Halos OS offers
Halos OS is positioned as an end-to-end safety operating system for robots, borrowing validation and redundancy methods from the company's autonomous vehicle work. It includes a real-time kernel, a hypervisor, and a safety-certified middleware layer. The system is intended to run on IGX Thor, NVIDIA's industrial-grade computing module designed for high-reliability environments.
NVIDIA says the OS provides deterministic scheduling, hardware-backed isolation, and fault management — features typically required for ASIL-D (Automotive Safety Integrity Level D) certification. By packaging these into a single OS, the company aims to reduce the engineering burden on robot makers who would otherwise have to build such safety mechanisms from scratch.
The role of IGX Thor and Halos Core
IGX Thor is the hardware foundation, a module that integrates NVIDIA's GPU, CPU, and networking capabilities in a form factor validated for industrial and medical applications. Halos Core is the software layer that includes the safety kernel, communication stacks, and application interfaces. Together, they form a platform that NVIDIA believes can handle both real-time control and high-performance AI inferencing without compromising safety.
The company notes that Halos OS is not just for traditional industrial arms — it is also aimed at physical AI, a category that includes mobile robots, collaborative robots, and autonomous mobile manipulators. These systems often need to react to unpredictable human movement in factories or warehouses, making safety certification a critical requirement.
Industrial robots have historically relied on proprietary safety controllers and separate logic systems. That approach can be slow to update and hard to integrate with modern AI workloads. Halos OS promises a unified environment where safety and AI coexist on the same hardware, potentially speeding up development cycles for new robot models.
NVIDIA's move also signals a push beyond the automotive sector. The company has long supplied chips for self-driving cars, but industrial robotics is a larger market by unit volume, and safety requirements are similarly strict in many jurisdictions. By offering a certified OS off the shelf, NVIDIA lowers the barrier for smaller robotics startups that cannot afford a lengthy certification process.
What comes next
NVIDIA has not announced which robot manufacturers will first adopt Halos OS, nor a specific release date for the software. The company typically reveals customer partnerships at its annual GTC conference, which is scheduled for March 2025. Until then, the robotics industry will be watching for early integration tests and certification results from safety assessors.
