Lee Gyu-lee is a business writer at The Korea Times, focusing primarily on IT & telecommunications, the Ministry of Trade, Industry and Energy and KOTRA. Prior to this, she has covered a wide range of cultural news, from film, television and K-pop to lifestyle and fashion.
Samsung SDS debuts Korea’s 1st Nvidia B300 GPU service subscription
A poster for Samsung SDS' new graphics processing unit-as-a-service (GPUaaS) powered by Nvidia B300 GPUs / Courtesy of Samsung SDS
By Lee Gyu-lee
Samsung SDS has rolled out Korea’s first Nvidia B300-based graphics processing unit-as-a-service (GPUaaS) on its Samsung Cloud Platform (SCP).
The launch comes as enterprises move beyond developing artificial intelligence (AI) models into applying trained models to real world services, driving surging demand for high-performance computing.
The B300 GPU is equipped with 12-layer high-bandwidth memory 3E, delivering 288 gigabytes of memory per unit and bandwidth of up to 8 terabytes per second. This represents a 3.6-fold increase in memory capacity and a 2.4-fold improvement in bandwidth over the previous generation H100, significantly boosting performance for computing-intensive AI inference workloads.
The company has been building out its GPUaaS ecosystem since introducing A100-based services in 2021 and H100-based offerings in 2023, supporting cloud-based infrastructure and AI adoption across enterprise operations.
With the latest launch, it aims to minimize latency and unlock greater efficiency for large-scale AI applications such as agents; image, video and code generation; and analysis.
The subscription-based model will allow customers to reduce upfront investment risks and improve cost efficiency, as they pay only for what they use.
Samsung SDS explained that enterprise customers can immediately access Nvidia’s latest architecture via SCP even as the global GPU supply tightens, while processing sensitive enterprise data within its secure cloud environment.
“Leveraging SCP’s capabilities in resource optimization and energy efficiency, we plan to offer Korea’s first B300 GPU service to large companies, small and medium-sized enterprises (SMEs) and public sector clients seeking to integrate AI into their operations, supporting their transition to AI-driven business,” said Lee Ho-joon, Samsung SDS’ head of cloud service division.
The company will further expand its AI cloud portfolio in the third quarter of the year, launching a serverless inference service that charges only based on tokens used, eliminating separate infrastructure fees. It will also introduce an AI training service that automatically distributes and scales training jobs when developers upload code and data.
Meanwhile, as a public-private partnership cloud provider, Samsung SDS was the first to deploy H100-based GPUaaS at its Daegu data center in North Gyeongsang Province and has been involved in multiple government-led AI initiatives, including building a large-scale AI foundation and developing intelligent workflow platforms.
The company also supplies GPUs to more than 60 AI-focused startups, SMEs, research institutes and universities under the Ministry of Science and ICT-led program.