Fujitsu Backs AI Startup to Create Next-Gen Enterprise AI Solutions

Fujitsu has entered into an exciting partnership with AI startup Cohere to create advanced large language models tailored specifically for Japanese enterprises. Together, they are developing a language model provisionally named Takane, which will be available for deployment in secure private cloud environments. Takane will leverage Cohere’s state-of-the-art Command R+ model, known for its ability to manage long text inputs and its effectiveness in retrieval augmented generation (RAG) systems that support AI-powered applications.

This innovative model will allow users to address their unique industry and business challenges while ensuring data security and confidentiality. Takane will be accessible through Fujitsu’s cloud-based AI platform, Kozuchi, with an expected official launch in September.

In addition to Takane, Fujitsu and Cohere plan to integrate a knowledge graph-enhanced RAG tool into the Kozuchi platform, along with a generative AI-driven auditing tool. "Fujitsu has pioneered a knowledge graph extended RAG technology designed for logical inferences, along with generative AI amalgamation technology that facilitates the automatic creation of tailored generative AI models. This enables us to meet the diverse needs of enterprises," commented Vivek Mahajan, Fujitsu’s Chief Technology Officer and Chief Product Officer. “By combining our advancements with Cohere's robust enterprise-grade language models, we aspire to offer powerful and customizable AI solutions that effectively cater to the specific needs of businesses while accelerating the global adoption of generative AI.”

In a significant move, Fujitsu has also made a substantial investment in Cohere. As part of this collaboration, Fujitsu will be the exclusive provider of their jointly developed services, planning to make the co-developed models available through its Data Intelligence Platform as a Service (PaaS) and Uvance platform. Fujitsu joins previous investors including Nvidia and Oracle, further solidifying the startup’s standing in the AI landscape. Earlier this year, Cohere was actively seeking to raise up to $1 billion to expand its innovative capabilities.

Cohere distinguishes itself from competitors like OpenAI by focusing exclusively on enterprise needs, creating customized applications utilizing clients' data. Founded by former Google researchers Aidan Gomez, Nick Frosst, and Ivan Zhang, the team behind Cohere brings a wealth of experience, particularly Gomez, who is known for co-authoring the transformative paper on Transformers, a framework that has profoundly influenced the development of large language models.

"We believe that our strategic partnership with Fujitsu is a critical milestone in delivering premier large language model capabilities to one of the most influential enterprise markets globally,” said Gomez. “To realize the full potential of AI technologies, we must cater to enterprises in ways that resonate with their operational environment, whether that’s within their own cloud systems or in the languages relevant to their business practices. We are thrilled that our collaboration with Fujitsu will harness Cohere’s technology to empower the next generation of Japanese businesses.”