Sakana AI raises $135M Series B at a $2.65B valuation to continue building AI models for Japan (techcrunch.com)

Tokyo-based Sakana AI has closed a ¥20 billion (~$135 million) Series B at a $2.65 billion post-money valuation to accelerate development of generative AI models optimized for the Japanese language, culture and enterprise use. Founded in 2023 by former Google and Stability AI researchers Llion Jones, Ren Ito and David Ha, Sakana focuses on data- and cost-efficient models that perform well with small datasets—an approach that favors targeted adaptation and productization of LLM capabilities over competing with hyperscalers on model scale. The round included strategic domestic and global investors such as Mitsubishi UFJ Financial Group (MUFG), Daiwa partnerships, Khosla Ventures, NEA, Lux Capital, Macquarie Capital and In-Q-Tel, bringing Sakana’s total funding to roughly $379 million. For the AI/ML community, Sakana’s raise underscores a maturing market for “sovereign” and regionally specialized models that encode language, culture and regulatory needs rather than one-size-fits-all global LLMs. The company will deploy the capital into R&D, engineering and commercial expansion across finance, industrial, manufacturing, government—and potentially defense/intelligence—verticals, signaling growing enterprise demand for fine-tuned, efficient models and localized AI stacks. Sakana’s strategy highlights a viable alternative path in the ecosystem: building competitive, production-ready capabilities via targeted adaptation and partnerships instead of matching scale at the foundation-model level.