November 13, 2025 — Leads & Copy —
VeriSilicon (688521.SH) and Google have jointly launched the Coral NPU IP, designed for always-on, ultra-low-energy edge Large Language Model (LLM) applications. The IP leverages Google’s research in open machine learning compilers and incorporates AI security features.
The Coral NPU is based on the open RISC-V instruction set architecture and features native tensor processing capabilities. It supports mainstream machine learning frameworks, including JAX, PyTorch, and TensorFlow Lite (TFLite). The IP uses open-standard tools like Multi-Level Intermediate Representation (MLIR) from the Low Level Virtual Machine (LLVM) project for compiler infrastructure. It is designed for wearable devices and ambient sensing systems.
The Coral NPU IP is open-sourced on the Google Developers website and is accessible to developers worldwide. VeriSilicon will offer a commercial-ready, enterprise-grade version of the IP. The company plans to use its IP portfolio and chip design capabilities to provide custom silicon services. VeriSilicon is developing a validation chip based on the Coral NPU IP, targeting AI/AR glasses and smart home applications, to accelerate LLM deployment at the edge.
Wiseway Wang, Executive Vice President and General Manager of Custom Silicon Platform Division at VeriSilicon, stated that the launch builds on the Open Se Cura open-source project and integrates Google’s open-source technology with VeriSilicon’s chip design and commercialization capabilities. Wang added that the company will continue to leverage its strengths in chip design, verification, and system-level optimization to advance the edge AI ecosystem and support the deployment of open-source technologies in real-world products.
VeriSilicon provides platform-based, all-around, one-stop custom silicon services and semiconductor IP licensing services using its in-house semiconductor IP.
Source: VeriSilicon
