SoftBank unveils sovereign GPU cloud on NVIDIA GB200 to challenge hyperscalers

SoftBank Corp. will launch an “AI Data Center GPU Cloud” built on NVIDIA GB200 NVL72 systems hosted in Japan, targeting enterprises that need data residency.

SoftBank Corp. announced on May 25, 2026 that it will commercially launch its “AI Data Center GPU Cloud” in October 2026, with a beta version live immediately, according to the company’s press release. The service runs on NVIDIA GB200 NVL72 Blackwell systems deployed in SoftBank’s Japan-based data centers.

The offering is positioned as sovereign, distributed AI infrastructure, letting customers train and run models within Japan to meet data-residency and security requirements. SoftBank frames it as part of its broader “Telco AI Cloud” vision.

The initial rollout will be restricted to SoftBank group companies before opening to wider commercial customers. NVIDIA’s Charlie Boyle provided supporting commentary in the announcement.

The launch reflects a wider Japanese push against dependence on US hyperscalers. For investors, it marks SoftBank’s move into the neocloud market and a domestic alternative for regulated industries seeking sovereign AI capacity.

Sources: Source 1 · Source 2

About the author

Related stories

関連記事More stories →