Naver Cloud Executive Details Sovereign AI Strategy and 4,000 B200 GPU Investment for Security Model

Kim Jin-hwi, an executive director at South Korean cloud provider Naver Cloud, announced during a special lecture at the Financial Information Security Conference (FISCON 2026) in Seoul on November 12 that the company is developing a security-specialized AI foundation model named 'HyperCLOVA X-Cyber.' This initiative aims to establish autonomous AI capabilities to reduce dependence on overseas frontier models and mitigate risks associated with geopolitical tensions and supply chain disruptions.

Kim described the current threat landscape as a 'cyber pandemic,' noting that traditional defense mechanisms are insufficient against AI-accelerated cyberattacks. He specifically cited instances where US export controls temporarily prevented Naver from accessing Anthropic's Claude model, highlighting the vulnerability of relying on external services for critical security strategies. This experience drove the decision to build an internal, sovereign AI solution.

To support this development, Naver Cloud has committed significant computational resources. According to Kim, beyond the 256 GPUs provided through a government project, the company has proactively deployed 4,000 NVIDIA B200 GPUs for pre-training. He estimated the value of this compute capacity at approximately 200 billion KRW (around $140 million USD), stating that management approved additional internal resources because government-provided assets alone were insufficient for the project's goals.

Technically, the model employs a 'Purple Loop' strategy, integrating offensive testing with defensive validation to continuously improve detection and response capabilities. Naver Cloud is collaborating with institutions like the Korea Financial Security Institute to deploy the model in closed network environments, catering to the strict data sovereignty and security requirements of the financial and public sectors.