Samsung and Naver are collaborating on the development of an AI chip that is said to be 8 times more power efficient than NVIDIA H100.

Naver, the company behind the HyperCLOVA X large language model (LLM), has partnered with Samsung Electronics to develop power-efficient AI accelerators. This collaboration combines Naver's expertise with Samsung's extensive systems IP in silicon design, complex SoC building, semiconductor fabrication, and DRAM technologies. Recently, the two companies created a proof of concept for an upcoming AI chip, which they tested on an FPGA. Naver claims that the AI chip being co-developed with Samsung will be 8 times more energy efficient than NVIDIA's H100 AI accelerator, although they did not provide specific details about its actual throughput. The solution also utilizes energy-efficient LPDDR memory from Samsung. The collaboration between the two companies on this project began in December 2022.