SoftBank Group Corp. has announced its ambitious plans to build Japan's most powerful AI supercomputer, marking the first-ever use of Nvidia Corp.’s latest Blackwell design chips. This partnership highlights SoftBank's commitment to advancing AI capabilities across Japan, leveraging Nvidia’s state-of-the-art technology to support a wide range of AI-powered services in the country.
SoftBank’s telecom division is set to develop the AI supercomputer in partnership with Nvidia, utilizing Nvidia’s DGX B200 platform, which combines general-purpose processors and AI accelerator chips. This cutting-edge system will be instrumental in supporting diverse applications, including remote robotics and autonomous vehicles, creating a strong foundation for Japan’s AI ecosystem.
SoftBank’s supercomputer will run on the DGX B200, Nvidia’s high-performance AI computing system that integrates both processors and AI accelerator chips. This combination enables efficient data processing, crucial for developing and operating large-scale AI models. In the next phase, the company plans to incorporate the Grace Blackwell platform, an advanced configuration to further enhance processing power and efficiency.
Grace Blackwell, an evolved version of Nvidia’s Blackwell architecture, will boost computing power, ensuring SoftBank’s AI supercomputer can handle more complex and large-scale AI applications. This advancement will place Japan at the forefront of AI-driven technology, enhancing the country’s AI development capabilities.
Nvidia’s chips have become indispensable for global tech companies that rely on accelerated data processing capabilities to run AI models effectively. These chips, particularly Nvidia’s Blackwell design, are tailored to process large datasets at higher speeds, ideal for machine learning applications requiring rapid data analysis and integration.
Initially announced by Nvidia CEO Jensen Huang in 2024, the Blackwell lineup faced production delays. However, Huang has assured partners that supply will soon ramp up to meet the significant demand. This partnership provides SoftBank with early access to Nvidia’s highly anticipated chips, giving it a strategic edge in AI initiatives.
Nvidia is actively promoting its Blackwell chip lineup worldwide. Events in India and Japan underscore its dedication to spreading AI technology beyond a few U.S.-based tech giants, building a comprehensive ecosystem to foster national AI capabilities across various regions.
Beyond building the AI supercomputer, SoftBank’s telecom division plans to leverage Nvidia’s advanced hardware to transform its cellular network into an “AI grid.” This AI network, known as an AI-RAN (AI Radio Access Network), will support AI-based applications with more efficiency and lower power consumption, bringing smarter services to Japan’s wireless infrastructure.
An AI-RAN utilizes AI-optimized hardware rather than traditional telecom systems. This approach allows telecom companies to deliver next-generation services like remote-controlled robotics and autonomous driving support, all while consuming less electricity.
SoftBank is partnering with Fujitsu Ltd. and IBM’s Red Hat to initiate AI-RAN trials. This collaboration will help to transform Japan’s telecom sector, moving towards AI-driven networks that can handle the rising demand for AI-enhanced mobile services.
Masayoshi Son, SoftBank's founder, has long championed the development of advanced AI technologies. He joined Nvidia CEO Jensen Huang at the AI Summit in Tokyo, emphasizing the importance of AI to Japan’s future. Reflecting on SoftBank’s past investments in Nvidia, Son noted that Japan’s regulatory environment is currently conducive to AI advancements.
Son remarked that this moment represents Japan’s opportunity to “catch up” in the AI revolution, urging the country to continue advancing its tech landscape. This development reflects a concerted effort to establish Japan as a leader in AI, with a robust ecosystem capable of supporting diverse applications in industry and society.
As Japan embraces AI, SoftBank’s project aims to unify resources across industries, working with Nvidia, Fujitsu, and IBM’s Red Hat to build a comprehensive AI network that can support both economic growth and technological innovation.
SoftBank’s collaboration with Nvidia to build Japan’s first AI supercomputer exemplifies its commitment to elevating AI capabilities nationwide. By deploying Nvidia’s Blackwell chip technology and creating an AI-optimized telecommunications grid, SoftBank aims to enable advanced AI services across Japan, including autonomous robotics and smart communication networks.
Partnering with major tech firms like Fujitsu and IBM, this initiative reinforces Japan’s standing in the global AI arena, supporting economic progress and technological leadership. With these advancements, SoftBank is not only enhancing Japan’s AI infrastructure but is also paving the way for the country’s strategic role in the AI-driven future.