SoftBank Corp.'s Japan Top-level Generative AI Computing Platform Now Operational, Subsidiary Begins Full-fledged Development of Homegrown LLMs

SoftBank Corp. (“SoftBank”) today announced it began operating its computing platform for the development of generative AI. Furthermore, SoftBank's subsidiary, SB Intuitions Corp. (“SB Intuitions”), started utilizing this computing platform to begin the full-fledged development of homegrown large language models (LLMs) specialized for the Japanese language. SoftBank will aim to complete construction of these homegrown LLMs with 350 billion parameters in 2024.

To realize its corporate philosophy, “Information Revolution — Happiness for everyone,” SoftBank announced its long-term vision to provide next-generation social infrastructure essential for the development of a digital society in May 2023. To advance this long-term vision, SoftBank is constructing a computing platform for the development of generative AI. In July 2023, the computing platform was certified for supply assurance by Japan's Ministry of Economy, Trade and Industry (METI) as a “Cloud Program,” which designates important materials based on its Economic Security Promotion Act.

This Japan top-level^* computing platform comprises an NVIDIA DGX SuperPOD™ AI supercomputer with over 2,000 NVIDIA Tensor Core GPUs, NVIDIA Networking and NVIDIA AI Enterprise software. With the support of ITOCHU Techno-Solutions Corporation, equipment installation and construction were quickly completed to make the computing platform operational. While using the computing platform in phases, SoftBank and SB Intuitions plan to complete all investments and construction within the 2023 fiscal year, which ends on March 31, 2024, so they can provide the computing platform to universities, research institutions, companies and other entities at an early date.

With the computing platform now operational, SoftBank's subsidiary SB Intuitions has completed pre-verification and commenced with the full-fledged development of homegrown LLMs specialized for the Japanese language. With its computing platform capable of high-speed processing, an abundance of engineers and an overwhelming number of customer touchpoints, SoftBank will develop homegrown LLMs with Japanese language datasets to provide generative AI services tailored to Japanese business practices and culture. SoftBank will aim to complete construction of these homegrown LLMs with 350 billion parameters in 2024.

Junichi Miyakawa, President & CEO of SoftBank Corp., commented, “I am delighted that our Japan top-level generative AI computing platform is now operational so we can commence with the full-scale development of homegrown LLMs. Generative AI is a ground-breaking technology that has the potential to fundamentally transform society. However, significant investments in computing platforms and LLMs are now primarily being made by companies based outside of Japan, and AI development based on Japanese datasets is lagging behind development for other languages. Given this situation, SoftBank is aiming to create its own advanced and large-scale computing platform and develop homegrown LLMs that are tailored to Japanese culture and business practices. By doing this we will support the implementation of generative AI solutions across various industries and work toward realizing the implementation of digitalization into society.”

Hironobu Tamba, President of SB Intuitions Corp., commented, “I am tremendously pleased that we can now begin the full-scale development of homegrown LLMs. Using this computing platform, SB Intuitions will develop high-performance homegrown LLMs that are reliable and secure by utilizing our proprietary Japanese language datasets. SB Intuitions will work to drive innovation in Japan's new digital industries by using Japanese language LLMs and, in collaboration with various partners, we will build an ecosystem to accelerate the implementation of generative AI in Japan.”

Jensen Huang, founder and CEO of NVIDIA, commented, “Computing is undergoing a massive transition, driven by GPU-accelerated computing and generative AI. DGX SuperPOD is the ultimate AI supercomputer. We are delighted to partner with SoftBank to build a state-of-the-art AI infrastructure for the industry and researchers of Japan.”

Ichiro Tsuge, President & CEO of ITOCHU Techno-Solutions Corporation, commented, “We, ITOCHU Techno-Solutions, are excited to co-work with SoftBank's operating a computing platform for the development of generative AI, and SB Intuitions' utilizing this platform to begin the full-fledged development of homegrown LLMs specialized for the Japanese language in order to provide generative AI services tailored to Japanese business customs and culture. We also provide various services, including services with respect to the products of NVIDIA, support for the deployment of generative AI environments, and support for the development of customer-owned LLMs. We will continue to support SoftBank's business and contribute to the prosperity of society through the utilization of AI.”

[Notes]

1. *
   One of the largest computing platforms in Japan for LLMs learning. Based on publicly available information as of October 31, 2023, and SoftBank Corp. research.

• SoftBank, the SoftBank name and logo are registered trademarks or trademarks of SoftBank Group Corp. in Japan and other countries.
• Other company, product and service names in this press release are registered trademarks or trademarks of the respective companies.