REGULUS Edge AI SoC: 10 TOPS under 3W for robotics and IoT

On November 3, 2026, IoT Now reported that Ceva and Mobilint integrated Ceva's SensPro AI DSP and NeuPro-Nano NPU into Mobilint's REGULUS Edge AI system-on-chip.

According to the companies' announcement, REGULUS delivers 10 TOPS of AI compute performance at under 3W, with support for up to 8GB of memory. It targets AI inference on devices rather than sending every sensor reading or video frame to the cloud.

Why does on-device AI matter?

Continuously transmitting sensor data and video to remote servers consumes bandwidth, introduces latency and can raise data-privacy concerns.

Edge AI enables certain tasks—such as object detection, video analysis and sensor-driven responses—to run locally. Potential applications include smart cameras, industrial machine vision, robots, drones and power-constrained IoT devices.

What's inside REGULUS?

SensPro AI DSP: programmable sensing and signal-processing capabilities.
NeuPro-Nano NPU: energy-efficient neural-network inference acceleration.
Multi-engine architecture: assigns different sensing and AI workloads to suitable processing engines.

REGULUS also powers Mobilint's commercially available MLD-R1 AI USB accelerator, intended to add local AI processing to existing platforms.

Potential applications in Vietnamese factories

Potential uses include conveyor-based defect detection, product counting, safety-zone monitoring and local video analytics at the edge. Real deployment depends on model accuracy, supported frameworks, memory needs and inference optimization.

Important: 10 TOPS is a manufacturer-reported compute metric. It does not guarantee a particular inference speed or accuracy for every model or workload.

Original sources

Summarized and edited by Kỹ Sư Thập Cẩm from the cited sources. Cover image: REGULUS chip image from Mobilint's official website, linked to the manufacturer's image host. Image rights remain with Mobilint.