Fujitsu has announced its investment in Cohere, a North American generative AI startup, aiming to develop a large language model (LLM) tailored for the Japanese market. The model, tentatively named Takane, will integrate Cohere’s technology with Japanese business data provided by Fujitsu.
The specific investment amount remains undisclosed, but the collaboration underscores Fujitsu’s commitment to advancing AI capabilities in Japan. Takane will be trained on business-specific terminology, enhancing its application in various industries, from manufacturing to digital services.
Projected to feature between 50 billion and 100 billion parameters, Takane aims to bridge the gap in Japan’s AI infrastructure, offering a powerful tool for complex tasks. In comparison, OpenAI’s GPT-4 boasts 1.76 trillion parameters, highlighting the ambitious scale of Fujitsu’s project.
Fujitsu plans to launch Takane commercially by September, integrating it into its suite of AI products designed to meet specific customer needs. This initiative aligns with Fujitsu’s broader strategy, including collaborations with Tokyo Institute of Technology, to diversify its AI solutions portfolio.
The Takane model is expected to support Japan’s ongoing digital transformation, particularly in modernizing aging systems. Fujitsu aims to generate significant revenue from this venture, targeting sales of NT$700 billion ($4.43 billion) in digital transformation support by March 2026, a 90% increase from 2024.
