Japanese multinational holding company SoftBank Group will invest $960m by 2025 to create state-of-the-art GenAI, according to a Nikkei report.
SoftBank‘s computing network will likely be the most powerful in Japan once the investment is completed, according to the report, as the company looks to create powerful specialised Japanese GenAI.
The graphics processing units used to develop Softbank’s GenAI technology will be bought from NVIDIA, the largest chip designer in the US.
In 2023, SoftBank invested $120m into building its computing infrastructure. The added $960m will be spent, in addition, throughout 2024 and 2025.
SoftBank is currently building a large language model with 390 billion parameters, which it expects to be completed in 2024. The company said it will begin developing a higher-powered model with one trillion parameters as early as next year.
The largest version of Meta’s Llama 3 has 70 billion parameters, while OpenAI’s GPT 4 has 1.76 trillion parameters.
