Key Points:
- Samsung SDS launched Korea’s first Nvidia B300-based GPU-as-a-Service (GPUaaS).
- The new B300 GPU offers huge jumps in memory and speed for AI tasks.
- Businesses can now access top-tier AI power without big upfront costs.
- Samsung SDS plans to expand its AI cloud services further this year.
Samsung SDS just took a big leap in artificial intelligence (AI) services. They’ve rolled out Korea’s very first Nvidia B300-based graphics processing unit-as-a-service (GPUaaS) on their Samsung Cloud Platform (SCP). This is a game-changer as more businesses are moving beyond just building AI models and are now putting those trained models to work in real-world services, creating a massive demand for powerful computing.
The new B300 GPU is seriously impressive. It comes with 12 layers of super-fast memory, packing 288 gigabytes per unit and a lightning-quick bandwidth of up to 8 terabytes per second. Compared to the older H100, this means 3.6 times more memory and 2.4 times faster bandwidth. This huge boost makes it incredibly powerful for intense AI tasks like inference, which is when AI models make predictions or decisions.
Samsung SDS has been building up its GPUaaS offerings for a while, starting with A100-based services in 2021 and H100 offerings in 2023. They’ve been helping businesses adopt cloud-based AI across their operations. With this latest launch, they aim to cut down delays and make large-scale AI applications much more efficient. Think AI agents, generating images, videos, and code, or complex analysis – all running smoother and faster.
This new service works on a subscription model, meaning customers only pay for what they use. This helps businesses avoid big upfront investments and saves them money. Samsung SDS pointed out that even with the global shortage of GPUs, their enterprise customers can immediately get their hands on Nvidia’s newest tech through SCP. Plus, sensitive business data stays safe within their secure cloud.
Lee Ho-joon, Samsung SDS’s head of cloud service division, stated, “We plan to offer Korea’s first B300 GPU service to large companies, small and medium businesses, and government clients looking to use AI, helping them move towards AI-driven operations.”
Samsung SDS isn’t stopping there. In the third quarter of this year, they’ll launch a serverless inference service where you only pay based on the tokens you use, ditching separate infrastructure fees. They’ll also introduce an AI training service that automatically scales and distributes training jobs when developers upload their code and data.
As a cloud provider working with both public and private sectors, Samsung SDS was also the first to deploy H100-based GPUaaS in its Daegu data center. They’ve been involved in many government AI projects, like building a huge AI foundation and creating smart workflow platforms. They even supply GPUs to over 60 AI startups, research groups, and universities through a government program.
