SKT (Haein Cluster)
Korea's largest NVIDIA B200 dedicated GPU cluster
Last reviewed Apr 5, 2026
SK Telecom's GPUaaS infrastructure featuring over 1,000 NVIDIA Blackwell B200 GPUs in a single cluster, powered by proprietary Petasus virtualization for isolated, flexible AI training environments.
SKT (Haein Cluster) does not publish public hourly pricing.
Contact the provider directly for a quote.
Pros & Cons
Advantages
- Access to latest Blackwell B200 architecture in a large-scale cluster configuration
- 24/7 operations in partnership with Penguin Solutions
- Fine-grained GPU partitioning supports workloads from mid-sized models to large LLM training
- Near-bare-metal interconnect performance through virtualized NVLink and InfiniBand
Limitations
- Optimized for enterprise-level and large parameter training rather than small individual instances
- Primarily operated from Seoul (Gasan) AIDC with limited global multi-region options
- Contact-based pricing — no self-service or published rates
Key Features
Large-Scale B200 Cluster
Over 1,000 NVIDIA Blackwell B200 GPUs grouped into a single logical cluster for large-scale AI training
Petasus Virtualization
Proprietary virtualization layer enabling flexible partitioning and reconfiguration of GPU resources
Near-Bare-Metal Performance
Virtualized NVLink and InfiniBand support providing near-physical interconnect speeds within isolated tenants
AI Cloud Manager
Intelligent scheduling and multi-tenancy management for large training workloads
Sovereign AI Infrastructure
Located in South Korea (Gasan AIDC), providing data sovereignty and enterprise-grade compliance
Rapid Provisioning
Fully isolated, dedicated training environments provisioned in under an hour
Compute Services
GPUaaS
Virtualized, high-performance B200 dedicated GPU resources
Dedicated Clusters
Isolated, bare-metal-grade performance clusters for enterprise clients
Pricing Options
| Option | Details |
|---|---|
| Reserved Commitments | 1-year or long-term dedicated resource reservations for enterprise use |
| Subscription-Based | Pricing based on scale of reserved resources and duration of use |
Availability & Support
Regions
South Korea (Seoul/Gasan AI Data Center)
Support
24/7 dedicated enterprise technical support in collaboration with Penguin Solutions
Getting Started
- 1
Consultation & Sizing
Contact the SKT GPUaaS team to determine required B200 GPU count and commitment duration
- 2
Tenant Provisioning
Deploy an isolated GPU and networking environment in under an hour via Petasus virtualization
- 3
Environment Setup
Configure containers and library environments using the AI Cloud Manager
- 4
Launch Training
Schedule and execute large-scale model training jobs using the intelligent job scheduler