Simplifying Safety in Robotics
The next generation of physical AI demands a compute foundation that is as rigorous about safety as it is capable of intelligence. QNX OS for Safety 8.0 integrated with NVIDIA IGX ThorTM delivers exactly that - a single consolidated platform where high-performance AI inference and certified real-time safety run together, partitioned by virtualization. Safety and compliance teams get a pre-certified foundation, enabling faster path to market and a platform that scales from today's collaborative robots to tomorrow's autonomous systems — all without compromise.

NVIDIA IGX Thor

https://delivery-p151589-e1582228.adobeaemcloud.com/adobe/assets/urn:aaid:aem:a4051336-8cce-4ec9-9163-74e34999149d/as/qnx-ndivia-dark-mode.avif

_self
_self

Safety That Starts at the Architecture

Isolation by Design, Not Configuration

QNX is built on a microkernel design, which means only the most essential functions run in the kernel itself. Every other service (e.g. drivers, file systems, networking, applications) runs in isolated user space. If any component fails, the fault is contained. It cannot propagate to safety-critical processes. For physical AI systems operating in close proximity to people and infrastructure, this architectural isolation isn't just a feature, it is the foundation to everything else its built on.
QNX microkernel
_self
_self

Effortless Scalability

QNX OS for Safety 8.0 along with NVIDIA IGX Thor is pre-certified up to IEC 61508 SIL 3 and ISO 26262 ASIL D — functional safety standards most widely recognized in industrial and robotics applications. Rather than constructing an entire safety case from scratch, this foundational platform frees engineers and compliance resources to focus on the product itself rather than the qualification process underneath it.
QNX OS 8.0 Scalability
_self
_self

Real-Time Guarantees

Physical AI systems live and fail by timing. A perception pipeline that delivers results 50 milliseconds late, a control loop that misses its window, a safety interlock that responds unpredictably under load aren't just performance issues, they're safety issues. QNX OS for Safety 8.0 provides hard real-time guarantees with deterministic, bounded response times that hold under any system load. Paired with the compute performance of NVIDIA IGX Thor and Halos Safety Stack, you get an AI platform where the intelligence is fast, and the safety responses are guaranteed.
Real-Time Guarantees
_self
_self

Empower Safety Across Industries

teaser, teaser-background-black, teaser-opacity-10
Humanoids
Real‑time, certified safety control for humanoid robots.
Humanoids
_self
teaser, teaser-background-inherit, teaser-opacity-0
Autonomous Off-Road Vehicles
Reliable autonomous control in dynamic terrain.
Autonomous Off-Road Vehicles
_self
teaser, teaser-background-black, teaser-opacity-10
Autonomous Warehouse Robotics
Unified AI + Safety Stack for AMRs and AGVs
Autonomous Warehouse Robotics
_self
teaser, teaser-background-black, teaser-opacity-20
Surgical Robotics and Medical Imaging
Safety-certified OS + AI compute for medical‑grade precision.
Surgical Robotics and Medical Imaging
_self

Apply For Early Access

NVIDIA IGX Thor early access is now available to qualified developers building production-grade physical AI systems. To ensure the platform is deployed in environments where safety and performance capabilities are fully realized, access to IGX Thor hardware and the Board Support Package requires review and approval by NVIDIA. Submit your application to begin the qualification process and get your program on the path to certified physical AI deployment.

FAQs

NVIDIA IGX Thor is a production-ready platform that handles the hardest parts of building a certified physical AI system — the safety architecture, the real-time OS, and the AI compute — so engineers can focus on the differentiating applications. NVIDIA's hypervisor integrated in Halos Safety Stack enforces strict partitioning between the safety-certified QNX environment and the AI compute partition, meaning the separation between safe and non-safe workloads is managed in software. QNX OS for Safety 8.0's pre-certified software stack shortens the path through functional safety qualification, while NVIDIA's Isaac toolchain accelerates development of AI applications.
Any organization looking to develop products on NVIDIA IGX Thor with QNX is required to submit an application for review. NVIDIA evaluates applicants based on their intended use case, deployment environment, and program readiness. Once approved, organizations will then need to acquire the appropriate QNX GEDP Premium licenses before the hardware and BSP are released.
The Board Support Package includes everything your team needs to begin software development on IGX Thor - OS images for QNX OS for Safety 8.0 / QNX OS 8.0, bootloader, hardware drivers, SDK components, and documentation covering the platform's safety architecture. It is the complete software foundation for building certified physical AI applications on IGX Thor.
QNX OS for Safety 8.0 carries pre-existing certification artifacts for IEC 61508 up to SIL 3 and ISO 26262 up to ASIL D. These cover the primary functional safety frameworks applicable to industrial robotics, autonomous systems, and machinery. Developers will inherit this certification evidence with the purchase of QNX GEDP Premium licenses.
Approved developers get access to NVIDIA's Isaac platform for robotics, including Isaac ROS 2 packages, Isaac Sim for virtual development and testing, Holoscan for streaming data analysis, and the NVIDIA CUDA and TensorRT toolchains optimized for IGX Thor. On the QNX side, developers have access to the QNX Software Development Platform 8.0 and the various profiling and debugging tools built for real-time safety environments. The combined toolchain covers the full development workflow from simulation through to on-hardware validation.