LOGO

As generative AI accelerates worldwide, demand is surging for specialized computing capacity capable of handling massive AI workloads. Emerging at the center of this transition is a new category of infrastructure provider: the neocloud.

Unlike traditional hyperscale cloud platforms built to host general-purpose applications like corporate databases and web services, a neocloud is a purpose-built platform designed specifically for high-performance AI tasks, such as large-scale model training and inference. By integrating dense clusters of top-tier GPUs with tailored AI software, neoclouds deliver the extreme compute power required by modern AI architectures.

SoftBank Accelerates Its Neocloud Strategy

In Japan, SoftBank is aggressively stepping up its neocloud investments. In May 2026, the company unveiled its “AI Data Center GPU Cloud”—a unified platform combining proprietary management software, Infrinia AI Cloud OS, with high-density GPU infrastructure.

Scheduled for launch in October 2026, the platform leverages state-of-the-art hardware such as NVIDIA’s GB200 NVL72 to cover the full AI lifecycle, from data processing and training to real-time inference. By centralizing GPU cluster management and workload orchestration, SoftBank aims to drastically reduce the operational friction companies face when building AI development environments.

The Convergence of Neoclouds and Next-Gen Data Centers

Scaling a neocloud business requires much more than simply racking GPUs; it depends on a tight integration of underlying facilities, high-capacity power, and advanced cooling technologies.

A prime example is SoftBank’s flagship AI data center currently under development in Tomakomai, Hokkaido. Slated to open in FY2026, the campus is projected to expand to roughly 700,000 square meters with a total power capacity exceeding 300 MW. Designed to run on 100% locally produced renewable energy, this facility will support commercial generative AI development while also providing computing resources to universities, research institutes, and enterprises.

Further expanding this vision internationally, SoftBank Group and SoftBank Corp. announced in July 2026 the establishment of SB Neo in the United States. Aiming to roll out neocloud services for US enterprises starting in FY2027, SB Neo plans to scale incrementally on top of a massive 10 GW energy and AI infrastructure portfolio developed across the group.

As these moves show, the neocloud is evolving beyond a basic GPU-as-a-service model into a fully integrated stack spanning facilities, power, liquid cooling, high-speed networking, and orchestration software. The strategic choices made by neocloud operators will heavily influence the growth trajectory and site selection for future data centers globally.

Pure-Play Neoclouds Gain Global Traction—and Eye Japan

Outside Japan, dedicated neocloud providers are already asserting immense influence. US-based CoreWeave, the leading pure-play AI GPU cloud provider, successfully went public in 2025 following a string of landmark deals with major AI developers. Other fast-growing players include Europe’s Nebius, alongside US-based Lambda and Crusoe, all of which are actively securing power capacities that rival established hyperscalers.

These global neocloud giants are increasingly turning their attention toward Japan. In March 2026, GPU cloud provider GMI Cloud partnered with Wistron to announce a massive $12 billion AI infrastructure plan in Kagoshima Prefecture with a target capacity of 1 GW. If global neocloud players accelerate their market entry into Japan, it will significantly reshape local data center demand and site development trends.

The neocloud landscape is set to scale rapidly and diversify both in Japan and internationally. We will continue to track these global operators and their expansion into the Japanese market—stay tuned for further updates.