Ceva and Mobilint Bring Physical AI Closer to Everyday Devices

Physical AI is moving closer to everyday reality, and a new collaboration between Ceva and Mobilint offers an encouraging example of how that progress can reach commercial devices. Mobilint’s REGULUS edge AI System-on-Chip brings together Ceva’s SensPro AI DSP and NeuPro-Nano NPU, marking the first production deployment of Ceva’s NeuPro NPU family and opening new possibilities for intelligent machines.

The announcement reflects a broader shift in how artificial intelligence is deployed. While cloud computing remains valuable for demanding workloads, devices interacting with the physical world increasingly need intelligence close to their sensors. Robots, drones, cameras and industrial equipment must interpret changing conditions and respond quickly. Processing more information locally can help developers deliver responsive capabilities within practical power and space limits.

REGULUS is designed to meet those requirements as a standalone processor for on-device AI. It delivers 10 TOPS of AI performance at under three watts and supports up to eight gigabytes of memory. That combination creates an attractive foundation for manufacturers seeking to bring sophisticated AI functions into compact products, where energy consumption, available space and real-time operation all matter.

Its architecture combines specialized processing engines that can handle different parts of an application. SensPro supplies programmable sensing and signal-processing capabilities, while NeuPro-Nano provides efficient neural processing for always-on inference. These technologies complement Mobilint’s proprietary AI processing architecture, giving developers a flexible platform for workloads that involve both understanding sensor inputs and making decisions based on them.

This approach is particularly promising because Physical AI rarely depends on a single processing task. A machine may need to prepare incoming sensor data, recognize an object, interpret its surroundings and trigger an appropriate response. Matching workloads to suitable engines gives product designers more options as they balance performance and efficiency, while accommodating the changing requirements of emerging applications.

For Mobilint, the integration strengthens its ability to serve customers developing the next generation of intelligent edge products. CEO Dongjoo Shin described REGULUS as bringing together specialized capabilities for sophisticated AI under tight power constraints. Ceva’s contribution adds sensing, signal processing and inference resources that can support a broad range of customer needs across different device categories.

Encouragingly, the collaboration has already progressed into commercial hardware. REGULUS powers Mobilint’s MLD-R1 AI USB accelerator, which was first previewed at COMPUTEX 2026 and is now shipping. By offering a way to add on-device AI to existing systems, the accelerator provides another route for businesses exploring computer vision, video analytics and generative AI applications.

The potential applications span smart factories, robotics, machine vision, security, medical devices and digital signage. Support for widely used frameworks and ecosystems, including PyTorch, ONNX, Hugging Face, TensorFlow, TFLite and Keras, also gives developers familiar starting points. That compatibility can help teams connect their existing AI development work with new opportunities at the edge.

For Ceva, the production milestone provides a concrete demonstration of its AI technology reaching commercial silicon. Yaron Galitzky, Executive Vice President and General Manager of Ceva’s AI Division, highlighted the combination of efficient inference and flexible sensing as an important enabler for devices that must operate locally and respond in real time.

The partnership also brings together complementary strengths. Mobilint contributes its expertise in energy-efficient AI semiconductors and integrated SoC design, while Ceva supplies established silicon and software IP for intelligent edge applications. Their combined work points toward a future in which more manufacturers can build products that understand and engage with their surroundings.

Bottom line: REGULUS offers a tangible reason for optimism about Physical AI’s next chapter. With production silicon, a shipping accelerator and support for diverse development ecosystems, Ceva and Mobilint are helping translate technical ambition into usable tools. The opportunity ahead is to turn that foundation into increasingly capable devices that make workplaces more productive, systems more responsive and intelligent technology more accessible across a wider range of applications and environments around the world.

Contact CEVA

Also Read:

Ceva’s next NeuPro generation opens new possibilities across AI

Ceva’s Next-Generation NeuPro Brings a Unified AI Platform from Sensors to Infrastructure

UWB Without the Integration Headache: Ceva and LG Bring Precision to More Chips