 |
LG Twin Towers headquarters building. (Photo: Yonhap News Agency) |
[Alpha Biz= Kim Jisun] FuriosaAI, a Korean artificial intelligence (AI) semiconductor company, announced on July 22 that its second-generation AI inference accelerator Renegade (RNGD) has been adopted by LG for its large language model (LLM), EXAONE.
Building on this collaboration, FuriosaAI plans to launch an enterprise-grade EXAONE solution powered by Renegade, aimed at advancing generative AI infrastructure efficiency in the corporate sector.
FuriosaAI and LG AI Research jointly tested Renegade in a pilot environment for the EXAONE 3.5 model over the past eight months. The results confirmed that Renegade met LG’s technical requirements and achieved 2.25 times higher performance-per-watt compared to conventional GPUs, while maintaining performance specifications necessary for large-scale generative AI services.
This marks a significant step forward in solving the high energy consumption problem of GPUs without compromising performance.
Jeon Gi-jeong, Head of the Product Unit at LG AI Research, stated:
“After evaluating various GPUs and NPUs, we determined that Renegade was the most suitable for our needs. It not only delivers outstanding absolute performance, but also drastically reduces total cost of ownership (TCO) and provides a streamlined model support process.”
FuriosaAI is also preparing support for EXAONE 4.0, the successor to the recently released EXAONE 3.5, with ambitions to establish global competitiveness. The company plans to continually enhance its inference-optimized technologies and software capabilities to transition enterprise AI ecosystems from GPU-centric architectures to its proprietary NPU-based platforms.
As part of this initiative, FuriosaAI and LG are collaborating on an on-premises turnkey EXAONE AI Solution built around Renegade.
Alphabiz Reporter Kim Jisun(stockmk2020@alphabiz.co.kr)
