SK Telecom noted that the infrastructure will be housed in its Gasan AI data center
In sum – what to know:
B200 GPU cluster launched in Korea – SK Telecom’s new sovereign GPUaaS platform features over 1,000 NVIDIA B200s, forming one of the country’s most powerful AI infrastructure clusters to date.
Selected for national AI program – The “Haein” system will support Korea’s AI Foundation Model initiative, aimed at building sovereign, globally competitive models and growing the local AI industry.
Built on SKT’s AI stack – Leveraging its Petasus AI Cloud and AI Cloud Manager, SKT offers dynamic GPU provisioning and full AI lifecycle support for training, development, and operations.
Korean carrier SK Telecom has officially launched a sovereign GPU-as-a-Service (GPUaaS) platform built on more than 1,000 Nvidia B200 GPUs — one of the largest clusters in South Korea, the telco said in a release.
This new GPUaaS infrastructure, branded ‘Haein’, offers top-tier computing performance and is a successor to the company’s previous H100-based offering introduced in late 2024.
The system has been selected by Korea’s Ministry of Science and ICT to support its “Proprietary AI Foundation Model” program, which aims to develop globally competitive national AI models and foster the domestic AI ecosystem.
The Asian telco noted that the infrastructure will be housed in SK Telecom’s Gasan AI data center (AIDC). The deployment includes partnerships with Penguin Solutions and Supermicro for system integration and AI server procurement, it added.
By utilizing its in-house virtualization solution, ‘Petasus AI Cloud,’ SK Telecom said it can instantly partition and reconfigure the GPU cluster according to customer needs, maximizing utilization. In addition, the company provides the ‘AI Cloud Manager,’ an AIOps (AI for IT Operations) platform that efficiently manages the entire AI service lifecycle — from development and training to deployment. This platform, backed by extensive expertise in large-scale model development, enhances user work efficiency and development convenience, according to the Asian carrier.
This launch follows the opening of Lambda’s Korea region in December 2024, marking another key milestone in Korea’s AI infrastructure roadmap. SK Telecom noted it will continue expanding its GPUaaS business to provide infrastructure support that enables Korea’s AI industry to compete on the global stage.
South Korea’s Ministry of Science and ICT has recently named five teams to spearhead its sovereign AI foundation model initiative, part of a national strategy to build high-performance, locally developed AI systems, according to local press reports.
The selected consortiums are Naver Cloud, Upstage, SK Telecom, NC AI, and LG AI Research. These teams will be officially designated as “K-AI model” and “K-AI company,” the ministry said.
The ministry also noted that evaluation criteria included technical strength, prior experience, strategic clarity, and commitment to open-source development and ecosystem impact. All five teams proposed ambitious AI models with goals extending beyond large language models to multimodal and omnimodel architectures, according to the reports.
“The selected groups have shown outstanding capabilities and a shared dedication to advancing sovereign AI,” said an ICT ministry official. “Their open-source strategies will enable broader industry use and commercialization.”
The project, which drew proposals from 15 consortiums, aims to create Korean AI models that can match at least 95% of the performance of top global models released within the past six months.
SK Telecom and Naver Cloud were also chosen to provide GPU infrastructure. From late 2025 into early 2026, they will lease GPUs to support the other three teams: Upstage, NC AI, and LG AI Research.
Formal agreements with the five groups are expected in early August. The ministry plans biannual evaluations to narrow the pool: five will become four by year-end, and just two will remain by 2027.
