AI PMDC
AI-dedicated modular data center delivered in 12 weeks
From hardware to software, a full-stack AI solution
Asia's first liquid-cooled B200 data center, Elice AI PMDC
Ready in 12 Weeks
From design to installation in just 12 weeks Start AI training immediately without GPU procurement delays
Up to 640kW+ Scalability
Large-scale GPU cluster operation Handle training and inference for large models at once
PUE 1.1 Ultra-High Efficiency
35% power savings compared to standard data centers (PUE 1.7)
Up to 50% Cost Savings
No upfront CAPEX required Pay only for the AI infrastructure you need
Flexible data centers without compromising performance
General Data Center
36-48 month construction period
Long timelines for design, bidding, and construction
High upfront construction costs
Requires large-scale CAPEX investment
Average PUE 1.7
High energy waste structure
GPU scaling constraints
General cooling limits high-density GPU heat, requiring redesign for next-gen GPUs
InfiniBand-only configuration
Vendor lock-in and increased cost burden
Certification and security burden
Difficult to meet AI-specialized infrastructure and security certification simultaneously
Elice AI PMDC
Ready in 12 weeks
Portable modular design for rapid infrastructure deployment
Up to 50% cost savings
Flexible infrastructure without upfront investment
PUE below 1.1
High-density power design, up to 1.2MW IT power
High-density AI compute optimization
Cooling design capable of handling ultra-high-power GPUs like NVL72
InfiniBand + RoCE dual support
Choose the right network for your workload and budget
CSAP + ISO certified
Ready for public and financial sector workloads
Data center architecture that determines AI performance
Optimal infrastructure for high-performance AI, from space to power to cooling
Space
Integrated configuration from servers, GPU/NPU, network, power, to cooling
Up to 8x GPU density per unit area
High-density structure for more AI servers
AI workload optimized deployment
Server configuration considering training and inference efficiency
Modular expansion structure
Flexible capacity scaling when needed
Maximized space efficiency
Reduced installation footprint vs. traditional data centers
Power
Up to 80kW high-density power per rack
Stable operation of latest GPU/NPU
Up to 1.2MW IT power supply
*Based on GB300, stable operation of large-scale GPU clusters
Over 20% power efficiency savings with liquid cooling (15kWh → 12kWh)
Up to 8,192 GPU scaling support
High-performance AI training and inference environment
Cooling
Liquid/air-cooled selectable cooling system
Flexible design tailored to environment and budget
PUE 1.1 level high-efficiency cooling
Improved energy efficiency compared to conventional systems
Optimal GPU temperature stabilization
No performance degradation even under heavy loads
High-density environment cooling system
Stable operation of high-power GPU clusters
Choose the AI infrastructure that fits your goals
From standalone infrastructure to integrated AI operations, flexibly configure to match your needs.
AI PMDC standalone
High-performance AI infrastructure chosen to match your operational goals
Model training
PMDC-R6-480-DLC
liquid-cooling
Max IT Load 480kW
Cluster-scale
4 x PMDC-R6-480-DLC
liquid-cooling
Max IT Load 1.92MW
Inference or mixed use
PMDC-R6-240-AC
air-cooling
Max IT Load 240kW
PMDC-HYBRID
Container options combination and customization available
AI PMDC + ECI integrated solution
An integrated solution for flexibly scaling large-scale AI resources
PMDC Hardware
AI workload-optimized containerized modular data center AI
ECI Software
Cloud IaaS platform for AI infrastructure operations
GPU/NPU optimized VMs
Virtualize and utilize high-performance resources without performance degradation
Multi-tenant security isolation
Minimize security concerns with complete data separation
IaC tool integration
Support for automation environments including Terraform, OpenAPI
Turnkey solution
Integrated environment ready for operation immediately after PMDC installation
Support for various scale clustering
Stable scaling support in Infiniband, RoCEv2 environments
PMDC solution built and proven in-house
AI PMDC, the foundation of Elice's full-stack AI solution
Elice internal deployment case
GPUaaS solution built in 4 months
Launched GPUaaS services immediately after building an H100-based AI PMDC. Provides a structure for rapid transition from design to operations in a single environment.
Construction period: 4 months (July-October 2024)
Configuration : H100-based PMDC container
Results : GPU cloud service operation immediately after deployment
"From deployment to operation in just 4 months. Completed AI infrastructure at 1/3 the time and cost compared to traditional DC."
Elice Infrastructure TeamElice AI infrastructure built with global tech leaders
MOU signed for AI PMDC construction
NVIDIA Deep Learning Institute (DLI) official education service partnership
Strategic cooperation MOU signed for the activation of the domestic AI semiconductor ecosystem
Joint development partnership established for next-generation AI PMDC and cloud solutions
From public agencies to enterprises, cloud infrastructure you can trust
CSAP certified
Meets mandatory requirements for public sector adoption Ready for immediate deployment without additional security review
*For CSAP details, visit gov.elice.cloud or contact our sales teamGlobal standard security management
ISO 27001/27701 certified - International verification of security and privacy management systems
Proven operational expertise
Proven through government-led AI foundation model projects, NPU farm construction, and large-scale service operations
