Ceva Expands Physical AI IP Architecture Integrating Connectivity, Sensing, and NPU Inference
Frontier AI models · Enterprise AI
Semiconductor IP vendor Ceva has integrated wireless connectivity, sensing, and inference architecture to launch its next-generation NeuPro NPU series, addressing Physical AI demands from edge devices to data centers.
As artificial intelligence applications extend into physical environments and edge devices, semiconductor IP provider Ceva is combining wireless connectivity, sensor fusion, and inference processing to support robotics and autonomous systems. According to Ceva executives, Physical AI requires efficient coordination between edge devices and the cloud while managing power and latency. The company has extended its NeuPro NPU lineup across Nano, Prime, Max, and Ultra tiers, optimizing for memory bandwidth and data-movement bottlenecks.
Key takeaways
- Ceva integrates Bluetooth 7, Wi-Fi 7/8, UWB, and satellite communication IP with sensor fusion processors to connect edge sensor data with inference processing.
- The next-generation NeuPro NPU reportedly delivers up to 2x inference data throughput per square millimeter and 50% better energy efficiency compared to the first generation, using a unified SDK.
- The portfolio includes Nano, Prime, Max, and Ultra series, with NeuPro Ultra tailored for cloud and data-center inference alongside KV cache management.
Source：EE Times Taiwan 電子工程專輯 Read the original ↗
Prepared, checked and automatically published by AI; not individually reviewed by a human. Translation checked by AI.