SoftBank to Invest a Billion Dollars in GenAI

Japanese multinational investment holding company SoftBank is reportedly boosting its investment in generative AI with a strategic allocation of 150 billion yen (approximately $960 million). This investment, spread over the next two years, aims to enhance its computing facilities to develop and refine a Japanese-language-specific large language model (LLM).

To make this happen, the company is sourcing advanced graphics processing units (GPUs) from chip giant NVIDIA, which also includes lending some of these GPUs to other companies.

This move comes on the heels of a prior 20 billion yen investment in computing infrastructure. With this added expenditure, SoftBank is setting a record for the largest investment in computing power by any Japanese company to date.

The company is not only focusing on developing a highly competent model with 390 billion parameters but also plans to initiate a more ambitious project to create a model with one trillion parameters in the coming year. This scale is comparable to OpenAI’s GPT-4, which is known for its high performance, including passing a mock bar exam.

Japan’s generative AI market is expected to grow to about US$ 13 billion by 2030. This forecast represents a growth of 17 times from its size in 2023. The Japanese market's focus on "data sovereignty" emphasizes the importance of managing and controlling data within national borders to ensure economic security and adaptability in the face of international tech dependencies.