Korean telecom giant SK Telecom (SKT) is building a sovereign artificial intelligence (AI) infrastructure powered by Supermicro's Nvidia Blackwell server and an AI operating system (OS) of vast amounts of data.
This new infrastructure, known as Haein Cluster, is at the heart of SKT's transformation from traditional telcos to AI-first companies. Designed to support Telco's Petasus AI cloud service, it meets the needs of AI model development and ensures that all data and processing remains in Korea. The infrastructure already selected for the Ministry of Science and ICT's AI Computer Rental Program will be used to develop the National AI Foundation model.
Traditional AI infrastructure built on bare metal environments often take days or weeks to deliver new workloads, a major bottleneck for developers and businesses. SKT has been dramatically reduced this time by deploying a virtualized AI infrastructure. This allows you to spin up your GPU (graphics processing unit) environment in just 10 minutes, maintaining a performance level comparable to bare metal systems.
Dk Lee, Vice President and Head of SKT's AI DataCentre Lab, said the vast amount of data is helping Telco move from the legacy bare metal deployment to a fully virtualized, production-grade AI cloud. “The vast AI OS gives you the confidence to scale quickly and safely, with the added performance, simplicity and flexibility needed to support the next generation of sovereign AI workloads,” he added.
Essentially a GPU-As-a-Service, the Petasus AI Cloud is built on a vast, vastly disassembled shared (Dase) architecture running on Supermicro's Nvidia HGX servers. This creates a high-throughput and secure multi-tenant environment that can support both large-scale AI training and real-time inference.
The service is designed for resilience with enterprise-grade uptime and minimal operational overhead, allowing dynamic allocation of GPU, storage and networking resources for each tenant. AI application developers can also expect a unified pipeline for full lifecycle AI development and deployment. It also provides protocol flexibility without the need for a client-side gateway, simplifying access to government, research and corporate AI customers across Korea.
“SK Telecom defines the future of a national AI infrastructure, and Vast is proud to be able to support their vision,” said Sunil Chavan, vice president of Asia-Pacific. “By eliminating traditional bottlenecks in data movement, provisioning and security, SKT is now able to launch sovereign and secure AI infrastructure that offers speed and flexibility at scale.”
SKT drives the use of AI across the business, from providing services powered by AI to improving customer experiences and expanding robotics capabilities through AI's AI.
This initiative is based on many of the AI capabilities that SKT has deployed over the years. This includes AI chatbots that use OpenAI's GPT model to handle customer requests. Telco has also developed its own Sapeon AI chips used in its own neural processing unit farms, as well as its own Sapeon AI chips used by OpenStack-based cloud service providers and by the smart factory of semiconductor company SK Hynix.
