NVIDIA Halos for Robotics launch: IGX Thor and Halos OS

NVIDIA launched Halos for Robotics on Jun 22, 2026, pairing the IGX Thor AI compute module with Halos OS to deliver a full-stack functional safety system for industrial robots, humanoids, and autonomous mobile robots. The platform combines built-in hardware safety, a certified safety operating system, and sensor-edge safety integration to help teams build standards-aligned autonomous machines.

What is NVIDIA Halos for Robotics?

NVIDIA Halos for Robotics is a unified hardware and software safety platform built around the IGX Thor compute module and Halos OS, aimed at extending NVIDIA’s autonomous vehicle safety work into robotics. The stack includes Halos Core (the safety OS), reference Halos Applications such as the Outside-In Safety Blueprint, and infrastructure components like the Holoscan Sensor Bridge (HSB). The Halos AI Systems Inspection Lab supports ecosystem certification with 43 members: 16 AV, 23 Robotics, and 4 spanning both.

Halos OS provides two Halos Core configurations today: Halos Core Linux, and Halos Core Linux plus QNX with an NV Hypervisor that partitions IGX into a Linux VM for AI workloads and a QNX VM for safety-critical functions. Both configurations are available now for early access.

How does IGX Thor and Holoscan Sensor Bridge provide safety?

IGX Thor delivers platform safety through a dedicated IEC 61508 SIL 3 capable Safety Island (FSI) and extensive on-chip diagnostics, while Halos Core’s Safety Extension Package (SEP) collects and dispatches hardware errors to the FSI and the Safety MCU. IGX Thor advertises up to 2,070 FP4 TFLOPs of AI performance, 14x Neoverse ARM CPU cores, and 128 GB of memory at 273 GB/s bandwidth. The chip includes more than 22,000 safety mechanisms and supports systematics per IEC 61508 SC 3.

Holoscan Sensor Bridge extends the safety chain to sensors and actuators over Ethernet, with low-latency ConnectX RDMA and RTX GPU Direct streaming, scalability to hundreds of sensors and hundreds of Gbit/s, and domain-agnostic multimodal support. HSB also offers MACsec for device authentication, an end-to-end IEC 61508 SIL 2 safety protocol, watermarking, and camera testing services integrated into Halos Core.

SEP within Halos Core bundles FSI and SMCU reference firmware, an Error Propagation Layer, and the Edge Safety Link safety protocol. An application note describing IGX and Halos OS for safety is available under NDA, and the NVIDIA IGX Safety Product Brief is available for registered developers.

How does this build on NVIDIA’s prior safety work?

NVIDIA is extending a decade-plus AV safety investment into robotics rather than starting from scratch. The company cites over 18,000 engineering years on vehicle safety, assessment of more than 21 billion safety transistors, and production of over 7 million lines of safety-assessed code. NVIDIA also notes development of more than 22,000 platform safety monitors, publication of 330+ AV safety research papers, and issuance of 30+ certificates and assessment reports. Those assets and processes are being reused to align robotics safety with standards such as IEC 61508 and ISO 13849, and NVIDIA holds convenorship and leadership roles in multiple standards efforts including IEC 61508, IEC TC 65 AhG 30, and contribution to ISO 25785-1.

Why it matters

Halos packages a tested safety pedigree into a robotics-specific stack, which lowers the barrier for teams that must meet functional safety standards. By offering an SoC with a physically isolated Safety Island, an OS layer with SEP and Edge Safety Link, and a sensor bridge that carries safety protocols to the edge, the platform addresses both diagnostic coverage and end-to-end safety communication—two perennial pain points for certifying robots that operate near people. Early partner adoption, such as Agility integrating IGX Thor and Halos OS into its Digit humanoid safety system, signals industry interest in a shared, standards-aligned foundation.

What to watch

Monitor partner certifications and Inspection Lab outputs: Halos AI Systems Inspection Lab membership and third-party certification artifacts will show whether NVIDIA’s approach reduces the time and cost of compliance. Also watch for broader ecosystem uptake of Halos Core Linux plus QNX configurations and the availability of the IGX Safety Product Brief and the SEP application note from NVIDIA for developers.