KT is embarking on building a “full AI stack” for the South Korean government’s nationwide AI service “AI for All,” connecting everything from homegrown AI semiconductors to models, platforms, cloud, and network into a single integrated system. The core strategy goes beyond simply combining individual companies’ technologies—it centers on “AI orchestration” that makes them work together in actual large-scale service environments.
According to KT on the 23rd, the consortium includes more than ten South Korean AI companies, including Rebellions, Lablup, Upstage, and Motif Technologies. As the lead contractor, KT is responsible for linking each partner’s AI semiconductors, computing infrastructure, and AI models and platforms with its own cloud and network to deliver a large-scale public-facing service.
South Korean Companies Converge Across AI Infrastructure and Model Domains
The consortium is broadly divided into the AI infrastructure domain and the AI model and platform domain. In the AI infrastructure domain, Rebellions, Lablup, and Vessl AI Korea are participating. The structure combines Rebellions’ homegrown NPU (neural processing unit) with Lablup’s AI computing resource management technology and Vessl AI Korea’s AI development and operations platform. The goal is to efficiently allocate computing resources and deliver stable service even when AI service users surge at massive scale.
The model and platform camp is also built around South Korean AI companies. Upstage, Motif Technologies, NC AI, Saltlux, and ActionPower are participating, each contributing their AI models and specialized technologies. Rather than relying on a single AI model, the structure is designed to leverage multiple models tailored to different usage environments and service characteristics.
KT will connect these technologies to its nationwide telecommunications network and cloud and data center infrastructure. The company will also draw on its experience operating large-scale user-facing services accumulated through its telecom offerings. KT’s assessment is that moving beyond AI model performance competition to scaling into a service used simultaneously by the general public requires the capability to reliably integrate semiconductors, computing resources, and cloud and network infrastructure.
KT plans to evolve the consortium’s technology integration into “AI orchestration.” The approach organically connects different NPUs, computing infrastructure, and AI models and platforms according to service demand, enabling South Korean AI companies’ technologies to be applied in actual nationwide services. This means “AI for All” could serve as a large-scale validation ground for homegrown AI technology.
Lee Jin-hyung, Senior Vice President and Head of the AX Business Division at KT, said: “For AI for All to become a service that citizens use reliably in their daily lives, it requires not only AI models but also the semiconductors and computing infrastructure that support them, cloud and network capabilities, and large-scale operational expertise.” He added: “We will combine the technological capabilities of South Korean AI specialists with KT’s operational experience to establish a stable and trustworthy AI service foundation.”
Beta in September, Full Launch in December
The Ministry of Science and ICT plans to select the service provider for AI for All by the end of this month, followed by a beta service in late September and a full service launch in December. The selected company must provide a general-purpose AI chatbot service and public AI agents to all citizens nationwide. The government plans to allow participating companies to charge for advanced features beyond basic functionality or personalized services.
KT’s full-stack AI strategy is expected to give homegrown AI semiconductors, models, and cloud infrastructure an opportunity to be validated in actual large-scale service environments. In particular, if Rebellions’ NPU is deployed in a nationwide public service, it could become a significant case demonstrating the commercial viability of South Korean AI semiconductors. However, ensuring interoperability and stable operations while integrating technologies from different companies into a single service remains a key challenge.