SoftBank has ambitious plans to construct a cutting-edge supercomputer explicitly designed for AI applications, using NVIDIA’s Blackwell GPU architecture

This powerful collaboration will allow SoftBank to leverage NVIDIA’s GB200 Superchips, including NVIDIA’s Grace CPUs and Blackwell GPUs, tailored for high-performance computing and demanding AI applications. This development aligns with SoftBank’s broader AI ambitions, especially in Japan’s evolving tech landscape. AI infrastructure and 5G/6G advancements will transform the telecommunications and healthcare industries.

At the core of this supercomputer project are NVIDIA’s GB200 NVL72 Superchips, which bring significant upgrades over previous generations. These Blackwell chips support new AI-centric capabilities, including second-generation transformer engines optimized for large language models (LLMs) and other generative AI workloads. This architecture enables precise model scaling for improved performance, allowing SoftBank to run complex AI models, like trillion-parameter systems, with greater efficiency and lower power consumption—a critical advantage for both cost and sustainability​

The system’s architecture also incorporates NVIDIA’s Confidential Computing features, which enhance data security by encrypting sensitive AI workloads and protecting intellectual property within the AI ecosystem. This security focus makes it suitable for federated learning, where models are trained across decentralized data without compromising privacy. Combined with the high-throughput NVLink connections, the Blackwell architecture provides fast, secure data exchange across GPUs, ensuring that SoftBank’s supercomputer will meet the demands of large-scale AI operations and low-latency performance essential for future AI applications.​

This project is part of a larger vision by SoftBank’s CEO Masayoshi Son to position the company as a leader in AI technology. Given SoftBank’s investments in AI and telecommunications, this supercomputer may eventually support SoftBank’s 5G/6G network rollout, delivering AI-driven insights and operational efficiency to telecom systems across Japan. NVIDIA and SoftBank’s collaboration highlights an exciting era for AI infrastructure, where innovations in GPU architecture and data center design are reshaping what’s possible in both research and commercial AI applications​

Focusing on groundbreaking AI and energy-efficient computing solutions, this move is expected to propel Japan’s AI capabilities forward. It is a pivotal development in the global race for AI infrastructure dominance.

• Mail