3.6 Times the Capacity and 2.4 Times the Bandwidth Compared to the H100
Efficient Processing of Large-Scale AI Models

Samsung SDS has launched the country's first GPU as a Service (GPUaaS) subscription model based on Nvidia's latest graphics processing unit (GPU). With this service, corporate clients pay only for what they use, allowing them to reduce costs while operating artificial intelligence (AI) at a hyperscaler level.

Samsung SDS Launches Korea's First Subscription-Based GPU Service Powered by NVIDIA B300 View original image

On March 23, Samsung SDS announced the release of the GPUaaS service based on Nvidia's latest GPU, the 'B300', via the Samsung Cloud Platform (SCP). This service enables companies to utilize required GPU resources from the cloud without the need to purchase or operate GPUs themselves. The launch reflects the surging demand for high-performance computing among companies expanding their AI operations.


Nvidia's B300 GPU is equipped with 12-stack HBM3E (High Bandwidth Memory), offering 288GB of memory per GPU and a bandwidth of 8TB per second. This results in a 3.6-fold increase in capacity and a 2.4-fold increase in bandwidth compared to the H100, significantly enhancing memory performance for AI inference workloads that require complex computations. As a result, companies using SCP B300 GPUaaS can dramatically improve the data bottleneck issues that occur when running large language models (LLMs).


Samsung SDS explained that with SCP B300 GPUaaS, companies can efficiently process massive AI models. In particular, the service reduces AI service latency for applications such as AI agents, image, video, and code generation and analysis. The subscription-based model allows companies to lower initial investment risks by paying only for what they use, and also enables immediate access to the latest GPUs even when GPU supply is tight.

Samsung SDS Launches Korea's First Subscription-Based GPU Service Powered by NVIDIA B300 View original image

The GPUaaS market is expected to continue its robust growth. According to Fortune Business Insights, the global GPUaaS market size is projected to grow at a compound annual rate of up to 35.8% through 2032.


Samsung SDS plans to launch a 'serverless inference service' in the third quarter of this year, which charges only for the amount of tokens used without separate infrastructure fees, as well as an 'AI training service' that automatically distributes AI training when users input code and data.



Hojoon Lee, Head of Cloud Services at Samsung SDS, stated, "Leveraging the resource optimization and energy-saving capabilities of the Samsung Cloud Platform, we will provide B300 GPU services to large enterprises, as well as mid-sized, small, and public sector clients seeking to apply AI to their work, thereby supporting the AI transformation (AX) of our customers."


This content was produced with the assistance of AI translation services.

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing