Mobilint Deploys Ceva AI IP in REGULUS SoC
The edge AI chip integrates Ceva's SensPro AI DSP and NeuPro-Nano NPU for sensing and inference in Physical AI devices.
Updated on Nov. 2, 2026 in Semiconductors
Mobilint, a Seoul-based AI semiconductor company, has deployed Ceva's SensPro AI DSP and NeuPro-Nano NPU in its REGULUS edge AI System-on-Chip. The deployment is the first production use of Ceva's NeuPro NPU family.
Why it matters
AI workloads are increasingly moving from the cloud onto devices themselves, and Physical AI systems that must sense and respond in real time need intelligence in power-constrained edge hardware. This integration reflects that shift toward on-device processing.
The REGULUS Soc combines Ceva's SensPro AI DSP with the NeuPro-Nano NPU to handle sensing and inference on a single edge chip. Specific throughput, power draw and memory figures were not included in the announcement.
The players
Mobilint
Mobilint is an AI semiconductor company based in Seoul, South Korea, which designs edge AI chips for on-device inference.
Ceva, Inc.
Ceva, Inc. is a Rockville, Maryland-based licensor of semiconductor IP, including DSPs and neural processing units used in AI and connectivity products.
The details
Mobilint integrated Ceva's SensPro AI DSP and NeuPro-Nano NPU into its REGULUS edge AI System-on-Chip, which is aimed at Physical AI devices. The companies frame the deployment as the first production use of Ceva's NeuPro NPU family.
Timeline
November 2, 2026: Mobilint announced the deployment of Ceva's IP in the REGULUS SoC.
The Tech Race
The NeuPro family moving into production through Mobilint's REGULUS places Ceva in a competitive push to supply edge AI inference IP to chip makers. It positions the company against other licensors courting device makers building on-device AI.
Edge AI chips like REGULUS allow devices to run inference locally rather than sending data to the cloud, which can reduce latency and keep sensing data on the device. Consumers may see these capabilities appear in future Physical AI products, though no product launch dates were announced.
The takeaway
Moving AI inference onto power-constrained chips is becoming a central design goal for devices that must act on their surroundings in real time. Watching which products adopt NeuPro-based silicon will show whether this first deployment becomes a broader pattern.
Further reading
Explore more coverage of chip design and licensing in our Semiconductors section.
