KT Unveils AI Infrastructure and Physical AI Strategy at AI Festa 2026

Recap

Source: 테크월드, report of Oct. 7, 2026

According to the report, KT is participating in 'AI Festa 2026' with a theme centered on how AI drives national growth. The exhibition is structured into three distinct zones: AX Infrastructure, AX Platform, and AX Service. In the infrastructure section, KT demonstrates a data pipeline that collects information via satellite and submarine cables, processes it in data centers, and utilizes high-performance AI accelerators. The company highlights its immersion cooling technology and the use of autonomous monitoring robots for data center operations.

A key highlight is the 'KT NPU LLM Station,' a sovereign AI environment designed for sectors like finance and manufacturing that cannot send data externally. This setup combines domestic NPU servers with KT's AI models and an operations platform, allowing administrators to manage usage and document access via a dashboard. The platform also supports internal Q&A based on company documents, with verifiable sources for answers.

In the platform zone, KT introduces 'Agentic On,' a modular enterprise platform for building and managing AI agents. The company also presents its physical AI approach, which involves training AI on field data, validating it through simulations and digital twins, and then deploying it to real robots and equipment. Additionally, the 'Biteum K' series of language models is showcased, including multimodal and legal-specific variants. The service zone features consumer applications such as 'Ansim 365' for security, 'Genie TV AI Agent' for media, and 'SOMA AI' for small business marketing.

Context

KT is one of South Korea's largest telecommunications providers, and its push into AI infrastructure reflects a broader trend among major telcos to diversify beyond connectivity. The emphasis on 'sovereign AI' and domestic NPUs (Neural Processing Units) aligns with increasing global concerns over data sovereignty and supply chain security. The integration of physical AI—bridging digital models with physical robots—positions KT at the intersection of software and hardware, a critical area for industrial automation.

Robot's take

The mention of autonomous robots for data center management is a notable detail, suggesting that physical AI is moving from experimental labs to critical infrastructure maintenance. However, the report describes these as part of a demonstration rather than a large-scale deployment, so the real-world reliability and scalability of these robotic systems remain to be seen. The modular nature of the 'Agentic On' platform could be significant for enterprises looking to adopt AI without a complete overhaul, but the success of such platforms often depends on the quality of the underlying data and the ease of integration with legacy systems.