Skip to content
AI.info

The Pulse

Eole starts reselling SoftBank GPU services under partnership

Eole has begun selling usage rights to two SoftBank AI infrastructure services to businesses and resellers. The companies also plan to pursue sales of Eole's computing resources to SoftBank customers.

Eole starts reselling SoftBank GPU services under partnership

Eole has begun selling usage rights to two SoftBank AI infrastructure services to businesses and resellers in Japan and abroad. The move is the first phase of an agreement to sell each company's AI data center and GPU computing resources to the other's customers, Eole said in a November 20 release.

Eole and SoftBank signed a basic transaction agreement covering the resale of SoftBank's AI Data Center GPU Server and Cloud PF Type A services. The intended customers include companies that need computing infrastructure for large language model training and inference or AI agent development. The GPU resources are located in Japan, but their specific sites are not disclosed.

A two-way sales plan

The agreement also covers proposed sales of data centers and GPU resources owned, operated or managed by Eole to SoftBank customers. Eole said discussions on that part of the arrangement will continue.

The partners also intend to make joint proposals for large GPU clusters and dedicated AI computing environments. They plan to build a system for providing computing resources across multiple sites and GPU environments by combining their infrastructure, and to consider future cooperation on AI data centers, GPU cloud and related services.

Eole has been developing AI data centers across Japan. It has worked with Digital Dynamic on activities ranging from GPU server procurement and sales to deployment in Japanese data centers and the provision of AI computing capacity.

GPU server terms

SoftBank's AI Data Center GPU Server offering uses NVIDIA DGX A100 and NVIDIA DGX H100 systems. Customers can reserve servers for exclusive use, from a single machine to multiple machines in a high-bandwidth cluster, for periods as short as seven days.

Storage is offered in 100GB units using the Lustre file system and NVMe, with capacities exceeding 1PB also supported. The service includes NVIDIA AI Enterprise, the Slurm job scheduler and NGC Private Container Registry. SoftBank also offers an optional private connection through its SmartVPN service.

Cloud service

Cloud PF Type A offers GPU servers using NVIDIA HGX B200. The cloud service operates across two Japanese regions, in eastern and western Japan, where systems and data can be distributed.

It uses Oracle Alloy to provide infrastructure and platform services based on Oracle Cloud Infrastructure technology. Eole said users can also access SoftBank AI services, including the Japanese generative AI model Sarashina.

Sources

  1. Original source · Press release Eole prtimes.jp

Related stories

Explore

More articles