Ceva and Mobilint Deploy NeuPro NPU in New REGULUS Edge AI Chip

Ceva and South Korea-based Mobilint have introduced the REGULUS system-on-chip (SoC), a processor designed to handle AI tasks directly on edge devices. This release represents the first production deployment of Ceva’s NeuPro neural processing unit (NPU) family.

The REGULUS chip integrates Ceva’s SensPro AI DSP and NeuPro-Nano NPU, capable of 10 TOPS of performance while maintaining a power draw of under 3 watts. The architecture is intended for real-time sensing and inference in power-constrained environments, such as robotics, drones, and machine vision systems.

Mobilint CEO Dongjoo Shin stated that the integration allows the company to support the complex AI workloads required by next-generation edge devices. The technology is already powering Mobilint's MLD-R1 USB AI accelerator, which supports industry-standard AI frameworks including PyTorch, TensorFlow, and ONNX.

Yaron Galitzky, executive vice president of Ceva's AI division, said the deployment highlights the company's progress in bringing its silicon-ready AI technology into the commercial market. As industries shift toward decentralized computing, demand continues to rise for processors capable of performing high-performance inference at the edge.