QNX adds Hailo-8 support for the edge AI developers

QNX has added support for the Hailo-8 AI accelerator to its QNX Software Development Platform 8.0, expanding the hardware options available to developers building edge AI systems.

The support applies to sectors including automotive, industrial automation, robotics and smart infrastructure, where companies are seeking to run AI workloads on devices rather than in the cloud.

Physical AI, or intelligent systems operating in real-world environments, has become a growing focus for suppliers of embedded software and specialist chips. Developers in robotics, medical devices, industrial systems and software-defined vehicles are under pressure to combine AI processing with software that responds predictably.

QNX, a division of BlackBerry, said the new support combines its real-time operating system with Hailo's edge AI processor. Hailo's chips are used in embedded devices across several industrial and transport markets.

Benchmark results

The companies also disclosed benchmark results from tests using a Raspberry Pi 5 connected to the Hailo-8 accelerator. The tests compared QNX SDP 8.0 with a real-time Linux environment while running AI workloads.

According to the companies, the QNX setup delivered up to 14 times greater performance consistency, 2.6 times tighter latency distribution, 4.1% higher throughput and 3.9% lower average latency. They said the figures show the software and chip combination can deliver more predictable behaviour for embedded AI tasks.

That emphasis on predictability reflects a long-standing distinction in the market for operating systems used in critical machines. In sectors such as healthcare, transport and industrial controls, vendors often argue that determinism matters as much as raw speed because delays or inconsistent response times can affect how a system behaves in the field.

For QNX, the partnership adds another AI-related hardware option as it seeks to deepen its role in systems that must meet strict performance and reliability requirements. The business has an established position in embedded and safety-critical software, particularly in automotive and medical applications.

BlackBerry said QNX software is used by nine of the top 10 medical device manufacturers. It added that the technology underpins hundreds of millions of vehicles, giving the unit a broad installed base as developers look to add more AI functions to existing and new products.

Edge push

The announcement comes as chipmakers and software groups try to capture demand for AI processing outside data centres. Running models locally on devices can reduce dependence on cloud connectivity and support applications that need faster response times in changing physical environments.

Hailo has focused on that segment with processors designed for edge devices. Its products are used for tasks including computer vision and other AI workloads in machines constrained by space, power consumption and cost.

Grant Courville, Senior Vice President, Products and Strategy, QNX, described the broader shift in embedded systems.

"Increasingly, advanced embedded devices are looking to AI to unlock new functionality," said Courville. "Together, QNX and Hailo are helping customers turn that potential into reality by bringing advanced AI capabilities to the edge on a QNX trusted and deterministic software foundation."

Hailo described the partnership as part of a broader effort to ease deployment for device makers that want to introduce AI functions without rebuilding their systems around cloud infrastructure.

"As demand for intelligent edge systems continues to grow, developers are looking for technologies that simplify the path from innovation to deployment," said Max Glover, Chief Revenue Officer at Hailo. "Together with QNX, we're enabling customers to bring advanced AI capabilities to a broader range of embedded applications and accelerate the next generation of intelligent products."

The addition of Hailo-8 support broadens QNX's list of AI-enabled hardware options at a time when embedded developers are weighing trade-offs among computing speed, latency, power use and system behaviour. For manufacturers in regulated or safety-sensitive industries, those choices can shape which software and silicon combinations are suitable for products deployed in the physical world.