ソフトバンク株式会社. announced that it will begin offering the “AI Data Center GPU Server” cloud service on the 8th, which allows customers to flexibly use the GPU system it uses as its AI computing platform, starting from a single unit, as a dedicated resource. The AI Data Center GPU Server is a cloud service that allows companies and research institutions developing and training generative AI to use the NVIDIA DGX System, an AI computing platform composed of NVIDIA Ampere GPUs or NVIDIA Hopper GPUs, as a dedicated resource.
こちらもお読みください: NTT DATA Launches Slopebase for Expense Management
It allows flexible use of one or multiple units, with short-term contracts available for as little as seven days. It meets a variety of user needs, including training large-scale language models (LLMs) and data analysis and simulations using generative AI. A GPU cluster connected via a high-speed InfiniBand line can be used exclusively with a single GPU system. For multiple units, the cluster is provided pre-configured, allowing immediate use for large-scale AI development and scientific computing. In addition to a wide range of usage periods, from as little as seven days to annual contracts, customers can flexibly choose the number of GPU systems and storage capacity to suit their needs. In addition to the AI software platform ” NVIDIA AI Enterprise ” optimized for businesses, it also comes with the job scheduler “Slurm” that handles calculation processing and resource management, so it can be used immediately without complex environment configuration.
ソース ヤフー