Samsung SDS targets Asia-first Nvidia Vera Rubin cloud launch

Samsung SDS plans to offer Nvidia’s next-generation Vera Rubin GPUs through its cloud business. It is targeting availability in March 2027. The move would make Samsung SDS Asia’s first cloud provider to offer these chips, according to the company’s plan. It reflects the rapid rise in demand for computing power for artificial intelligence.

The company will use part of the west wing at its Dongtan data center in Hwaseong, Gyeonggi Province. It plans to start with Nvidia B300 GPU-as-a-service later this year. In April 2027, it also expects to provide interim computing services from the same facility before the Vera Rubin offering and the national center’s opening.

Samsung SDS is preparing for a temporary capacity gap before South Korea’s national AI computing center opens in Haenam in 2028. Its wider plan is to exceed 800 megawatts of AI infrastructure capacity by 2031, using several investment and construction models.

Nvidia Vera Rubin GPUs are a next-generation family of specialized processors from Nvidia. GPUs handle many calculations at the same time, making them useful for demanding AI tasks. The article does not provide Vera Rubin’s technical specifications, so its exact performance is not stated. It identifies the chips as Samsung SDS’s planned next major GPU offering.

AI companies and other customers can use these processors through a cloud provider instead of buying and operating them themselves. Samsung SDS would place the GPUs in its data center and sell access as a service. Customers could then obtain computing capacity for AI work when needed, while Samsung SDS manages the underlying infrastructure.

The planned launch is important because demand for AI computing is surging. Samsung SDS expects to offer Vera Rubin GPUs in March 2027. Before then, it plans to start B300 GPU services and provide interim capacity while South Korea’s national AI computing center is still under construction.

Samsung SDS aims to expand its artificial intelligence infrastructure capacity to more than 800 megawatts by 2031. This is a major capacity goal for supporting growing demand for AI computing. The article presents the target as part of the company’s broader push into AI infrastructure and cloud-based GPU services.

The company plans to use three business models. It can invest directly in infrastructure, take an equity stake in projects, or design, build, and operate data centers for customers. Together, these approaches give Samsung SDS several ways to add capacity rather than relying on only one ownership structure.

The target sits alongside near-term projects at the Dongtan data center. B300 GPU-as-a-service is planned by the end of this year, followed by interim services in April 2027. Samsung SDS also hopes to support demand before the national AI computing center opens in 2028 and is targeting more than 1 trillion won in operating profit next year.

GPU-as-a-service is a cloud model in which customers access graphics processing units remotely instead of owning them. The provider supplies the chips, data-center space, power, networking, and operations. Customers pay for or otherwise use computing capacity without building a specialized GPU facility themselves. The article identifies this as a growing Samsung SDS business.

For example, Samsung SDS plans to offer Nvidia B300 GPU-as-a-service from the west wing of its Dongtan data center by the end of this year. Customers would connect to the hosted infrastructure and run AI workloads on the available GPUs. Samsung SDS would manage the physical equipment and supporting systems.

This model matters because demand for AI computing is surging, while building dedicated infrastructure can require substantial investment and expertise. Samsung SDS plans interim services in April 2027 and Vera Rubin GPU services in March 2027. It also wants more than 800 megawatts of AI capacity by 2031.

Samsung SDS is offering interim GPU services because South Korea’s national AI computing center in Haenam, South Jeolla Province, is not expected to begin operations until 2028. The company anticipates that AI computing demand will continue rising before that facility is ready. Its interim service is intended to provide capacity during this transition.

The service is planned for April 2027, using part of the west wing of Samsung SDS’s Dongtan data center in Hwaseong, Gyeonggi Province. The company also plans B300 GPU-as-a-service at Dongtan by the end of this year. These services would give customers access to hosted GPU resources before the national center opens.

The plan positions Samsung SDS as a bridge between immediate demand and future public infrastructure. The company separately aims to offer Nvidia Vera Rubin GPUs in March 2027. Its longer-term goal is more than 800 megawatts of AI infrastructure capacity by 2031, supported by three business models.

The three models differ mainly in ownership, financial exposure, and responsibility. With direct investment, Samsung SDS puts its own capital into AI infrastructure and directly supports the project. With equity participation, it takes a financial stake in a project, sharing ownership or economic interest without necessarily controlling every activity.

The design-build-operate model is more service-focused. Samsung SDS designs a customer data center, builds it, and then operates it. The customer receives a complete infrastructure service, while Samsung SDS applies its technical and operating expertise. The article names this model as one way to expand capacity.

Using all three approaches gives Samsung SDS flexibility as AI demand grows. It can own infrastructure, participate in projects, or build and run facilities for customers. These models support its goal of exceeding 800 megawatts of AI infrastructure capacity by 2031, alongside GPU services from its Dongtan data center.

AI systems need specialized computing because training and running advanced models involves huge numbers of calculations. Those calculations must often be repeated across large datasets. GPUs are well suited to this work because they can perform many similar operations in parallel. The article connects this need to surging demand for AI computing, though it does not explain the technical process in detail.

Data centers turn GPU hardware into usable computing capacity. They provide physical space, electricity, cooling, networking, and operational support. In Samsung SDS’s case, the west wing of its Dongtan data center will host B300 GPU-as-a-service. Customers can use that capacity remotely instead of owning GPU systems themselves.

This infrastructure is becoming strategically important as more organizations seek AI resources. Samsung SDS plans interim services in April 2027 and Vera Rubin GPU services in March 2027. By 2031, it wants more than 800 megawatts of AI infrastructure capacity through several expansion models.