본문 바로가기
bar_progress

Text Size

Close

Japan's SoftBank Directly Enters Enterprise Generative AI Development... "Equipped with NVIDIA GPUs"

200 Billion Yen Invested in Supercomputer Maintenance
Targeting AI Production for Medical and Financial Sector Enterprises

Japanese SoftBank has announced that it will independently develop generative artificial intelligence (AI), drawing attention both inside and outside Japan. Following the return of Chairman Masayoshi Son to the public stage last month, where he emphasized aggressive investment in the AI sector, all eyes are on how far the company can secure technology by directly engaging in development.


Japan's SoftBank Directly Enters Enterprise Generative AI Development... "Equipped with NVIDIA GPUs" [Image source=Yonhap News]

On the 29th, Nihon Keizai Shimbun (Nikkei) reported that SoftBank will invest 20 billion yen (approximately 191.8 billion KRW) to develop enterprise generative AI. The investment will be used for development costs, including maintenance based on supercomputer calculations. Additionally, the plan is to enhance competitiveness by equipping the AI with graphics processing units (GPUs) from the U.S. company NVIDIA.


Nikkei stated, "The AI developed by SoftBank is expected to have the highest level of capability among domestic companies by incorporating NVIDIA GPUs," adding, "High-performance computing infrastructure is essential for cutting-edge AI development due to the need for massive data learning."


SoftBank’s generative AI is designed for enterprise use. Initially, it plans to develop generative AI for call center operations, followed by specialized AI tailored for fields requiring expert knowledge such as finance and healthcare.


SoftBank has also begun developing large-scale language models that form the basis of generative AI. The number of parameters, which determine performance, is currently about 1 billion, with a goal to increase this to 60 billion. For comparison, GPT-3, which underpins ChatGPT, has 175 billion parameters. However, since ChatGPT is a general-purpose generative AI that incurs high development and operational costs, SoftBank appears to have judged that focusing first on specialized AI development is preferable.


SoftBank plans to generate revenue by renting out the supercomputer it is building for this development to other companies.


Within the industry, SoftBank’s direct AI development is seen as a virtually anticipated scenario. Chairman Son appeared publicly for the first time in seven months at the shareholders’ meeting on the 20th. At this event, he expressed deep interest, stating, "The AI revolution is growing explosively," and announced that SoftBank would make aggressive investments to support this.


Japan's SoftBank Directly Enters Enterprise Generative AI Development... "Equipped with NVIDIA GPUs" [Image source=Yonhap News]

In fact, SoftBank’s medium-term management plan announced last month emphasized business expansion targeting external companies, signaling a shift from the identity of a mobile carrier toward entering non-mobile businesses. The company aims to explore new opportunities by providing generative AI to enterprises and renting out computing infrastructure.


The industry is also watching closely to see if SoftBank can succeed in its revival through this expanded AI investment. Previously, SoftBank recorded massive losses due to Vision Fund investment failures. In response, Chairman Son did not attend the Q3 conference held in February for the first time since the company’s founding, citing a focus on pushing forward the Nasdaq listing of the UK semiconductor design firm ARM. However, some speculated at the time that this was actually to avoid investor dissatisfaction and media attention.


© The Asia Business Daily(www.asiae.co.kr). All rights reserved.

Special Coverage


Join us on social!

Top