The Ministry of Science and ICT in South Korea selected a consortium led by Naver Cloud Corp. to develop a sovereign AI foundation model specialized in cybersecurity. ^[1] The consortium plans to deploy 4,000 B200 GPUs from Naver Cloud and 256 H200 GPUs from LG to train two cybersecurity AI models based on Naver's HyperCLOVA X and LG's EXAONE foundation models. ^[2] The government plans to provide the consortium with up to 256 Nvidia B200 graphics processing units (GPUs) over the next 10 months to support the development of the model. ^[3] The project aims to develop a sovereign cybersecurity AI model tailored to South Korea's security environment amid growing cyber risks stemming from rapidly evolving AI technologies. ^[4] The consortium consists of 33 organizations, including LG CNS, LG U+, Sands Lab, and East Security, alongside universities such as KAIST, Seoul National University, Korea University, and POSTECH. ^[5] The consortium plans to utilize a total of 4,256 GPUs for the project, comprising 4,000 NVIDIA B200 chips and 256 LG H200 chips, separate from the government-provided equipment. ^[6] The project aims to build a 700B-parameter Mixture of Experts (MoE) model based on Naver's HyperCLOVA X and LG's EXAONE AI models. ^[7]