-Building ultra-high-performance AI-dedicated GPU infrastructure in the Digital Edge Bupyeong data center
-Launching a high-end cloud-based service for high-performance, low-power AI, HPC, and data analysis at a stable and competitive price in line with domestic GPU demand
Korea Quantum Computing (KQC, CEO Kim Jun-young), a quantum computing research and development, quantum security, and AI infrastructure operation company, announced that it will build an ultra-high-performance 'AI GPU Farm' based on NVIDIA H200 GPUs at the Digital Edge Bupyeong Data Center and begin a cloud-based high-performance computing service (GPUaaS) from July 4.
KQC will enable domestic companies and institutions to flexibly utilize computing resources through the cloud in line with the recent market where new application markets such as LLM, AI copilot, and multi-model AI are growing and the need for larger memory and computational demands are exploding. Now, customers can enjoy high performance and stability similar to using GPUs directly in a physical server (Dedicated Bare Metal) environment where they can install and control the operating system themselves.
The AI GPU farm introduced by KQC this time is composed of a parallel cluster based on NVIDIA's latest GPU architecture Hopper, and is an infrastructure optimized for computationally intensive workloads such as ultra-large language models (hereinafter referred to as LLMs) such as GPT-4, Claude, and Gemini, generative AI (GenAI), complex simulations, and high-performance data analysis. The H200 GPU has significantly improved memory capacity and speed compared to the existing H100, and can maximize LLM training and inference efficiency through the built-in Transformer Engine.
KQC CEO Kim Jun-young said, “KQC’s GPU farm will be in great demand from big tech companies, AI startups, industries requiring complex simulations and data analysis, national research institutes, HPC centers, etc.” He added, “By providing infrastructure optimized for AI and HPC at a stable and reasonable price, we plan to introduce advanced services linked with quantum computing infrastructure in the future.”
KQC's AI GPUaaS (GPU as a Service) service It can accelerate the processing of large data sets and the training of complex AI models, and GPU virtualization and sharing functions allow multiple users to efficiently utilize GPU resources simultaneously. In addition, it includes the NVIDIA AI Enterprise software stack to smoothly support development and operation in the latest technology environment.
In addition, computing, storage, and network resources can be flexibly adjusted according to the user's needs, and can be used economically without initial investment burden through a pay-as-you-go billing method. Budget management can be performed more effectively through predictive analysis functions, and operational efficiency can be significantly improved through infrastructure monitoring, automatic deployment, and updates.
“This will be a very competitive service in terms of service stability and cost,” said KQC Managing Director Kim Chang-hoe, who oversaw the launch of this service. “We plan to continuously introduce various systems to meet the needs of the market,” he added.
You must be logged in to post a comment.