LG AI Research has adopted FuriosaAI’s RNGD AI accelerator to enhance the performance of its Exaone models after extensive testing on performance, energy efficiency, and software compatibility.

FuriosaAI, the Seoul-based startup seeking to design chips to compete with Nvidia, has sealed its first major contract months after rejecting an US$800 million acquisition offer from Meta Platforms Inc.

RNGD offers 2.25 times better performance per watt for LLM inference compared to GPU-based solutions.

It meets LG AI Research’s latency and throughput needs for real-world applications.

The collaboration will allow both companies to provide RNGD-powered servers to enterprises using Exaone models across industries like electronics, finance, telecommunications, and biotechnology.

This integration supports LG AI Research’s plan to scale its Exaone-powered services, including the enterprise AI agent ChatEXAONE. The company plans to expand service availability to external clients in the future.