Aetina edge AI systems deliver 100 TOPS for multi-camera video perception

The launch of these compact systems addresses the growing demand for localized, high-bandwidth video processing in environments where cloud connectivity is unreliable. By integrating four GMSL2 ports into a palm-sized form factor, Aetina enables sophisticated multi-camera perception—such as 360-degree surround view and automated inspection—without the latency overhead of traditional networking. This move strengthens the NVIDIA Jetson ecosystem's footprint in the autonomous mobile robot and industrial sectors, providing a ruggedized path to mass production for vision-based AI agents. Watch for the scheduled update to NVIDIA JetPack 7.2 to see how it enhances the multi-vision capabilities of these edge platforms.

NVIDIA's Jetson Orin platform continues to attract hardware partners building ruggedized edge systems for autonomous mobility and industrial vision. In March 2026, NVIDIA announced at GTC that the Jetson Orin NX Super module delivers 157 TOPS of AI performance while maintaining a 15-watt power envelope, positioning it as the highest-performance option in the compact edge AI segment. Aetina's DeviceEdge AIE-VN34 and AIE-VN44 units build on this trajectory by pairing Jetson Orin NX and Orin Nano modules with GMSL2 camera interfaces, a combination that targets autonomous mobile robots and factory inspection lines where space and thermal budgets are constrained.

The business case for edge AI in mobility applications is strengthening as enterprises seek to reduce cloud dependency for latency-sensitive workloads. NVIDIA reported in its fiscal Q2 2026 earnings that embedded and edge segment revenue grew 38% year over year, driven by demand from robotics, logistics, and manufacturing customers deploying vision-based AI at the point of capture. This revenue momentum signals that OEMs like Aetina are entering a market with proven buyer intent, particularly in sectors where multi-camera perception must operate without reliable network connectivity. The GMSL2 interface standard, developed by Analog Devices, supports up to 15 meters of cable length with deserializers that aggregate multiple camera feeds into a single processing node, making it the de facto interconnect for automotive and industrial vision systems.

On the technical side, independent benchmarking of Jetson Orin platforms shows meaningful gains over prior-generation hardware. MLPerf Inference v5.0 results published in February 2026 showed the Jetson Orin NX achieving 2.4x higher throughput on YOLOv8 object detection compared to the Jetson Xavier NX, while consuming only marginally more power. For Aetina's target use cases, such as 360-degree surround view and automated visual inspection, this throughput improvement translates directly into the ability to process four simultaneous camera streams at higher frame rates without dropping frames. The DeviceEdge AIE-VO series, which uses the lower-power Jetson Orin Nano, serves applications where thermal dissipation is the primary constraint, such as enclosed robotic housings or outdoor enclosures without active cooling.