Elice Cloud Releases Open-Source AI Ecosystem Optimization Solution

Easily Develop Open-Source AI Models in Customized AI Infrastructure Utilizing Latest GPU and NPU

Elice Cloud, the AI-specialized cloud service by Elice, Inc. (CEO Kim Jae-won), has unveiled a solution that allows various general-purpose AI models, including open-source models, to be used safely and conveniently in a cloud environment.

Elice Cloud has upgraded its existing ML API features, introducing an ‘ML API Library’ function that enables the use of open-source AI models in customized AI infrastructure. The key feature of this upgrade is providing an optimized GPU and NPU environment that allows easy use of various open-source AI models without additional installation or configuration.

Unlike API-based solutions for commercial AI models, open-source AI models require environment configuration depending on the specific model and often need fine-tuning for desired tasks. To address these challenges, Elice Cloud provides GPU and NPU environments optimized for each open-source model, along with customized APIs. Users can easily develop, manage, and deploy AI chatbots and other applications in high-performance AI infrastructure tailored to various models, including open-source models like Deep Seek R1 and Meta’s Llama. The service also offers high-performance computing services utilizing the latest NPUs from Furiosa AI and Rebellion.

Executed in a private cloud environment based on the Elice AI PMDC (Portable Modular Data Center), the solution prevents external leakage of sensitive data. The Elice AI PMDC can be configured in modular units, allowing physical isolation, agile response to AI demand changes, and achieving an impressive Power Usage Effectiveness (PUE) of 1.27.

Will Strengthen AI Infrastructure Competitiveness with Largest Domestic High-Performance GPU AI PMDC

Elice Cloud aims to design and construct 1MW-class AI PMDCs this year. To stably manage high-density power and heat, the infrastructure incorporates space optimization design and Direct Liquid Cooling (DLC) technology. It is designed to perfectly support the latest high-performance GPUs like NVIDIA B100, with a rack power capacity of 160kW, which is industry leading.

CEO Kim Jae-won of Elice, Inc. stated, “We have designed and built an AI PMDC that can accommodate up to 5,000 high-performance H100 GPUs, which is expected to be the largest scale among domestic cloud companies. To fully launch the Elice Cloud business, we will focus on GPU acquisition and technological innovation through R&D.”