{"id":2693,"date":"2026-02-04T18:37:25","date_gmt":"2026-02-04T09:37:25","guid":{"rendered":"https:\/\/news.sktelecom.com\/en\/?p=2693"},"modified":"2026-02-04T18:37:25","modified_gmt":"2026-02-04T09:37:25","slug":"skts-haein-gains-momentum-as-a-core-driver-of-koreas-ai-strategy","status":"publish","type":"post","link":"https:\/\/news.sktelecom.com\/en\/2693","title":{"rendered":"SKT\u2019s Haein Gains Momentum as a Core Driver of Korea\u2019s AI Strategy"},"content":{"rendered":"

\u2022 The company celebrates six months since the launch of Haein, its GPU cluster designed to support large-scale AI workloads and foundation model development.<\/strong>
\n\u2022 Haein is differentiated by an integrated full stack software platform \u2014 including Petasus AI Cloud, AI Cloud Manager, and GPUaaS Service Orchestrator \u2014 streamlining AI infrastructure deployment and operations.<\/strong>
\n\u2022 Haein\u2019s GPUaaS is being utilized for the Korean Government\u2019s Sovereign AI Foundation Model Project, supporting the development of globally competitive and nationally strategic AI models.<\/strong><\/p>\n

SK Telecom (NYSE: SKM) today marked the six-month anniversary of Haein, its GPU cluster built for GPUaaS, launched on August 1, 2025, designed to support large-scale AI workloads and the development of advanced AI models.<\/p>\n

The name Haein is inspired by Haeinsa Temple, home to the Tripitaka Koreana (Palman Daejanggyeong) \u2014 one of the world\u2019s most complete and enduring collections of Buddhist scriptures. SK Telecom chose the name to reflect its vision of building a foundational AI infrastructure that enables the creation, accumulation, and responsible use of knowledge over time.<\/p>\n

Since its launch, Haein has become a core GPU computing platform supporting Korea\u2019s AI initiatives, including foundation model development and large-scale AI training. Over the past six months, the cluster has progressed from initial deployment to stable operations, enabling faster experimentation, improved training efficiency, and scalable collaboration across teams and partners.<\/p>\n

A key differentiator of SKT Haein Cluster is its fully integrated full stack software, comprising Petasus AI Cloud, AI Cloud Manager, GPUaaS Service Orchestrator. Together, these components form an end-to-end platform optimized for complex AI workloads, allowing customers to move seamlessly from infrastructure provisioning to model development and deployment.<\/p>\n

At the infrastructure layer, Petasus AI Cloud provides a high-performance, AI-optimized data center virtualization platform that supports heterogeneous xPU environments, including GPUs and CPUs. Its Dynamic Allocation technology enables flexible partitioning and assignment of large GPU clusters by virtualizing various transport protocols such as NVLink, InfiniBand, and RoCEv2, thereby enhancing security, performance, and resource efficiency.<\/p>\n

Through Petasus AI Cloud\u2019s virtualization features, Haein provides each customer with a fully segregated environment by logically isolating both the GPU and network fabric at the tenancy level. This process takes less than an hour to prepare an isolated environment, compared to physical isolation, which can take several days to weeks depending on the size of the cluster.<\/p>\n

Building on this foundation, AI Cloud Manager simplifies large-scale AI training through intelligent job scheduling and multi-tenant user environment, enabling multiple teams to train massive models concurrently on shared GPU clusters while maintaining predictable performance and optimal utilization.<\/p>\n

The GPUaaS Service Orchestrator provides unified, real-time visibility into GPU, network, and storage resources, delivering actionable insights that help customers run their AI jobs in a more stable and efficient manner. This enables customers to develop models and bring them to market faster, while allowing SK Telecom to minimize operational overhead and total cost of ownership (TCO).<\/p>\n

By fully integrating these software layers, Haein streamlines the traditionally complex process of configuring and validating multiple infrastructure tools. This integration significantly reduces deployment lead times and mitigates economic losses associated with delayed fault detection. Furthermore, Haein\u2019s comprehensive monitoring and AI-driven anomaly detection capabilities\u2014delivered through the GPUaaS Observability solution\u2014enable immediate identification of failures and rapid service restoration, minimizing customer downtime and ensuring reliable service continuity.<\/p>\n

In August 2025, Haein was selected for a national initiative aimed at strengthening the foundation for developing globally competitive AI foundation models. Through this initiative, Haein has been actively utilized as a core AI computing infrastructure supporting large-scale AI model training and experimentation under Korea\u2019s Sovereign AI Foundation Model Project.<\/p>\n

\u00a0<\/p>\n

\n