QNX and Hailo Integrate Edge AI for Critical Systems

QNX, a division of BlackBerry Limited, has announced support for the Hailo-8 AI accelerator on its QNX^® Software Development Platform (SDP) 8.0. By combining Hailo’s edge AI processors with QNX’s trusted real-time software, developers gain a robust hardware and software foundation for safety- and mission-critical applications. This partnership aims to advance the next generation of "Physical AI" across sectors such as automotive, industrial automation, robotics, and medical technology, where performance and predictability are essential.

Benchmark-Proven Deterministic Performance
To demonstrate integration efficiency, the companies conducted benchmark tests utilizing a Raspberry Pi 5 paired with the Hailo-8 accelerator. Compared to a real-time Linux environment, AI workloads running on the QNX operating system achieved up to 14 times higher performance consistency, a 2.6 times tighter latency distribution, and higher throughput with lower average latency. Leaders from both companies emphasized that this deterministic real-time performance gives developers the flexibility to design intelligent edge systems that operate safely, swiftly, and reliably in dynamic real-world environments.

Additional Context: Technological Background and Market Dynamics
This section provides technological and market-specific background information not explicitly detailed in the original press release.

In safety-critical applications such as autonomous vehicles, industrial controllers, or surgical robots, a Real-Time Operating System (RTOS) like QNX is mandatory because it guarantees deterministic response times—meaning tasks are always executed within a strictly defined time window. Traditional operating systems can experience unpredictable latency spikes, which can have fatal consequences in physical AI systems. The native integration of powerful Neural Processing Units (NPUs) like the Hailo-8 allows complex machine learning models (for example, for image recognition or sensor fusion) to run directly at the edge. This eliminates reliance on error-prone cloud connections and ensures immediate, fail-safe decision-making right at the point of operation.

Edited by Lekshman Ramdas, Induportals editor – adapted by AI.

www.qnx.com