South Korea Accelerates AI Infrastructure Ambitions, Taps NAVER and NVIDIA

South Korea is accelerating its ambitions to build artificial intelligence (AI) infrastructure through a partnership involving NAVER, NVIDIA, and Brookfield. The three companies announced an increase in the capacity of a sovereign AI factory to 200 megawatts (MW), more than triple the initial plan of 55 MW unveiled last month. In the long term, NAVER is targeting a capacity of 1 gigawatt (GW). Less than a month after announcing the 55 MW project with NVIDIA, NAVER immediately raised its target to 200 MW. This move indicates that demand for AI computing in South Korea has exceeded initial estimates. For context, a 200 MW capacity is sufficient to operate tens of thousands of NVIDIA’s latest GPUs, used to train advanced AI models and serve millions of AI requests. The involvement of Brookfield suggests that investors view AI infrastructure as a strategic sector with significant growth prospects, particularly for countries seeking to reduce reliance on US-based cloud services. According to NVIDIA, the facility will utilise the NVIDIA DSX AI Factory platform, a system that integrates GPUs, networking, software, and AI computing management to train and run large language models (LLMs) and other AI applications. As South Korea’s largest internet company, NAVER aims to bolster its AI capabilities after US technology firms surged ahead with models like ChatGPT. This initiative also mirrors a regional trend; Japan has announced domestic AI computing capacity projects, whilst Singapore is striving to become a regional AI hub. Many nations are beginning to view AI infrastructure as a strategic asset, akin to the semiconductor and energy industries. For NVIDIA, the project represents a significant business opportunity. Although the contract value was not disclosed, a 200 MW AI facility is estimated to require GPUs worth hundreds of millions to over a billion US dollars. A detailed timeline for reaching the 200 MW capacity or the 1 GW target has not yet been announced, with power supply, licensing processes, and GPU availability remaining key challenges. Nevertheless, the rapid increase in targets within a short period demonstrates that the global race to build AI infrastructure is accelerating and is expected to intensify further in the coming years.