Selected as GPU supplier for government's AI research computing project

Academics, researchers to receive GPU access; service launched last month

Nvidia B300-based GPU cluster to be provided

Samsung SDS headquarters [Samsung SDS]
Samsung SDS headquarters [Samsung SDS]

Samsung SDS has launched a GPU computing service for national AI research.

The company announced Tuesday that it has been selected as the GPU resource supplier for the "2026 AI Research Computing Support Project," backed by the Ministry of Science and ICT and the Institute of Information and Communications Technology Planning and Evaluation. The service began July 20.

The project's core aim is to provide GPU resources for AI research to academics and researchers at domestic universities and government-funded research institutes — institutions that typically struggle to secure large-scale computing infrastructure on their own. The initiative is designed to give those researchers the computing power needed to conduct work on large-scale AI models and large language models.

The project is valued at about 15.3 billion won ($10.8 million) and will run through March next year.

According to Samsung SDS, the supplier selection process evaluated bidders on four criteria: GPU supply plans and available resource scale, research environment quality, operational support systems, and security capabilities. Multiple vendors were selected, with Samsung SDS ranked first.

Samsung SDS plans to provide not only Nvidia H100 GPUs but also clusters based on the latest B300 architecture. The B300 exceeds the Hopper-class specifications required in the request for proposals and offers significantly improved performance and larger memory capacity compared with the previous generation for training and inference of large-parameter models.

Samsung SDS said domestic research institutions have faced an urgent need to close the gap between available resources and global infrastructure standards, driven by intense competition for the latest GPUs. By deploying cutting-edge GPUs directly into the national AI research computing project, the company said it has laid the groundwork for Korean researchers to work in a world-class computing environment.

In addition, Samsung SDS will provide large-scale cluster configurations that pool multiple GPUs into a single resource pool, along with a high-speed interconnect network. The setup is intended to give researchers the flexibility to handle everything from long-term, large-scale projects to short-term experimental tasks.

Beyond GPU supply, the company will also deliver pre-configured environments covering CPUs, memory chips, storage, networking and development tools, allowing researchers to focus on their work rather than building infrastructure from scratch.

Meanwhile, Samsung SDS became the first company in South Korea to launch an Nvidia B300-based GPU-as-a-service offering through its Samsung Cloud Platform in March. In July, it added a neural processing unit service to the platform based on the second-generation Renegade NPU from Furiosa AI, a domestic chip developer specializing in AI inference.

"Through this project, we will provide stable service so that domestic researchers can achieve results based on world-class AI infrastructure," said Lee Ho-jun, head of Samsung SDS's cloud services division. "We will continue to broaden the foundation of the national AI research infrastructure and contribute to strengthening South Korea's AI competitiveness."


chami@heraldcorp.com