SoftBank launches beta of AI data centre service in Japan

TL;DR

• SoftBank Corp launched a beta of its AI Data Centre GPU Cloud in Japan ahead of a planned commercial launch in October.
• The platform combines SoftBank hardware with Infrinia AI Cloud OS for large language model development and multi-tenant workloads.
• SoftBank said the hardware uses Nvidia GB200 NVL72 systems with 72 Blackwell GPUs and 36 Grace CPUs in a single rack.

SoftBank Corp launched a beta of its AI Data Centre GPU Cloud in Japan, with plans to launch the service commercially in October. The Japanese operator said the beta covers an infrastructure and software setup that will form the basis of a domestic AI data centre service.

The AI Data Centre GPU Cloud uses SoftBank hardware and Infrinia AI Cloud OS to provide compute-intensive services. SoftBank said the platform supports large language model development and multi-tenant capabilities. SoftBank president and CEO Junichi Miyakawa said AI competitiveness increasingly requires both computing power and operational software.

SoftBank said its hardware setup uses Nvidia's GB200 NVL72, a single-rack system with 72 Blackwell GPUs and 36 Grace CPUs for large AI workloads. The company also highlighted two as-a-service elements in the offering, one for Kubernetes and one for inference. Miyakawa said Infrinia AI Cloud OS and the data centre programme are intended to support secure use within Japan.

Related questions

• What is SoftBank's planned launch date for its AI Data Centre GPU Cloud commercial service?
• What hardware does SoftBank use in its AI data centre platform?
• What workloads does SoftBank say its AI Data Centre GPU Cloud supports?