TLDR: Tokyo-based startup Sakana AI, founded by former Google researchers David Ha and Llion Jones, has achieved unicorn status by focusing on a unique ‘collective intelligence’ approach to AI, inspired by natural systems. Instead of building massive, energy-intensive models like industry giants, Sakana AI merges existing and new systems to create more adaptable and resilient AI. The company has secured support from Nvidia and major Japanese financial institutions, and is developing practical applications such as an offline Japanese chatbot and AI tools for scientific discovery and business efficiency, addressing Japan’s unique societal challenges.
Tokyo, Japan – Sakana AI, a burgeoning startup co-founded by former Google Brain research scientist David Ha and ex-Google researcher Llion Jones, has rapidly ascended to unicorn status, surpassing a valuation of $1 billion since its launch in 2023. This achievement marks it as Japan’s fastest startup to reach this milestone, driven by its distinctive ‘nature-inspired’ approach to artificial intelligence.
In a landscape dominated by companies building colossal, energy-intensive AI models, Sakana AI has chosen a different path. CEO David Ha articulated this philosophy, stating, “In a space like AI, everyone’s kind of doing the same thing. They’re just like collecting the world’s data, building a gigantic model, sucking up all this energy.” He emphasized the challenges of such an approach in Japan, where resources for power-hungry data centers are scarce, making the “chance of success actually quite low.”
Instead of direct competition with industry titans like OpenAI or Alibaba, Sakana AI is pioneering ‘collective intelligence.’ This method involves merging diverse existing and novel AI systems, both large and small, to foster more adaptable and robust AI programs. Ha highlighted the strategic advantage of this approach in a volatile global environment, noting that combining multiple systems could allow Sakana’s AI to “continue to operate, even if with slightly lower performance at the beginning,” should a single provider’s models become unavailable.
The company’s innovative strategy has garnered significant backing, including support from U.S. chip giant Nvidia and prominent Japanese financial institutions such as MUFG, with whom Sakana announced a “multi-year partnership” in May to develop “bank-specific AI systems.” Other Japanese businesses are also keen to leverage Sakana AI’s advancements in generative AI.
The name ‘Sakana,’ meaning ‘fish’ in Japanese, reflects the startup’s core inspiration from nature, observing how species from ants to humans collaborate to solve complex problems. The team, comprising over 100 individuals including contractors, operates from unassuming offices in a shared Tokyo startup space.
Sakana AI’s leadership team includes Ren Ito, who brings experience from Japan’s foreign ministry, alongside Ha, a Hong Kong-born Canadian and long-term Japan resident with a background in derivatives trading at Goldman Sachs, and Jones, a co-author of the seminal 2017 machine learning paper ‘Attention Is All You Need.’
Japan is actively investing tens of billions of dollars in AI and semiconductors, aiming to revitalize its technological prowess and enhance the productivity of its aging workforce. However, challenges persist, as indicated by Japan’s 35th ranking in IMD’s 2025 World Competitiveness Ranking, underscoring the need for greater entrepreneurship and solutions to labor shortages. Ha acknowledged these “business challenges in Japan,” noting the distinct working environments compared to Europe, China, or the United States. Nevertheless, he sees “building from the ground up” as an advantage, enabling Sakana AI to leverage “Japan’s soft power” to attract top software engineering talent.
Sakana AI has already developed practical applications, including ‘Tiny Sparrow,’ an offline Japanese-language chatbot designed to protect user privacy. Another project involves an AI trained on historical Japanese literature, capable of responding to text queries in a samurai-like linguistic style. The company is also exploring ‘AI Scientist,’ an initiative to automate scientific discovery, which has successfully passed a peer-review process this year, and other tools aimed at improving AI engineering efficiency. Ha emphasized the importance of experimentation, stating, “We have to experiment… things may not go your way, and we learn from that.”
While Ha intends to maintain a focused and “niche” AI research team, the business arm dedicated to deploying AI solutions for Japanese enterprises and the public sector is expanding. He acknowledged the inherent tension in this dual focus: “It is a challenge to balance between working on cutting-edge… AI development, and on the business side to deploy trusted solutions that are known to work.”
Also Read:
- AI-Powered Chemistry Startup CuspAI Secures $100 Million Series A Funding, Valued at $520 Million
- DataCrunch Secures $64 Million to Advance European AI Sovereignty with Renewable Cloud Infrastructure
Looking ahead, Ha anticipates an eventual cooling of investor enthusiasm for AI, drawing parallels to historical technological revolutions. “Every major technological revolution, from canals to the dot-com boom, has been fueled by a bubble,” he remarked, concluding that “speculative investment, while risky, is what funds the foundational breakthroughs.”
