Sakana AI: Towards Fully Automated Open-Ended Scientific Discovery

Company & Team Introduction

Sakana AI is a Tokyo-based artificial intelligence startup founded in 2023 by former Google researchers. The company focuses on developing AI models inspired by natural systems, such as evolution and collective intelligence. Their approach emphasizes creating smaller, collaborative models that work together to perform complex tasks efficiently.

Founders:

• David Ha: Former head of research at Stability AI and a researcher at Google Brain. He has a background in developing innovative AI models and has contributed significantly to the field of machine learning.

• Llion Jones: Co-author of the seminal 2017 research paper "Attention Is All You Need," which introduced the transformer architecture foundational to many modern AI models. He previously worked at Google Research.

• Ren Ito: Graduated from the University of Tokyo in 2001 with a law degree. He joined the Ministry of Foreign Affairs of Japan the same year, later working at the Japanese Embassy in the U.S. and engaging in international negotiations. He also served as an interpreter to the Prime Minister and worked at the World Bank for 11 years. In 2015, he became an executive officer at Mercari, overseeing global business, and later served as COO of Stability AI in the U.K. before co-founding Sakana AI.

Company Size and Growth Potential:

As of May 2025, Sakana AI has approximately 28 employees and is valued at $1.5 billion. The company's unique approach to AI development, combined with its experienced leadership and strategic partnerships, positions it for significant growth in the AI industry.

Product Overview

Sakana AI’s innovative approach emphasizes creating smaller, collaborative AI models that work together efficiently, aiming to revolutionize AI development with more adaptable and resource-efficient systems.

Key Products and Technologies:

• Evolutionary Model Merge: A technique that combines multiple existing AI models using evolutionary algorithms to create new, optimized models. This method enables the development of specialized models tailored to specific tasks or domains.

• The AI Scientist: An autonomous AI system designed to conduct end-to-end scientific research. It can generate research ideas, design experiments, analyze data, and draft scientific papers without human intervention. This tool aims to accelerate discoveries in fields like drug development and materials science.

• CycleQD: A population-based model merging approach that evolves a swarm of specialized agents to complement one another, laying the groundwork for lifelong learning by enabling diverse, adaptable skill acquisition at the population level.

• Experimental Models: Sakana AI has developed various experimental models, such as Llama-3-EvoVLM-JP-v2, a Japanese Visual Language Model, and Evo-Ukiyoe-v1, an image generation model specializing in traditional Japanese Ukiyoe woodblock print style. These models are built using advanced techniques like Evolutionary Model Merge and ControlNet.

Market Opportunity

Sakana AI is strategically positioned at the convergence of generative AI, automated scientific research, and region-specific language modelling. By focusing on developing nature-inspired AI systems, such as their flagship product, the AI Scientist, Sakana AI aims to revolutionize the way scientific research is conducted. Their approach not only targets the global scientific community but also addresses the unique linguistic and technological needs of the Asia-Pacific region, particularly Japan:

• Generative AI: The global generative AI market is projected to reach $109.37 billion by 2030, growing at a CAGR of 34.3% from 2023 to 2030.

• AI in Scientific Research: The AI in drug discovery market alone is expected to grow from $1.01 billion in 2021 to $5.71 billion by 2028.

• AI in Asia-Pacific: The Asia-Pacific AI market is anticipated to grow at a CAGR of 42.2% from 2021 to 2028, with Japan playing a significant role in this expansion.

Sakana AI's focus on developing efficient, collaborative AI models positions it well to capitalize on these growing markets, particularly in sectors requiring advanced AI solutions like scientific research and development.

Business Model & Traction

Sakana AI employs a multifaceted business model that combines open-source innovation with strategic enterprise partnerships. By developing nature-inspired AI models and tools, the company generates revenue through licensing, collaborations, and customized solutions for various industries.

Key Components of Sakana AI's Business Model:

• Enterprise Licensing and Custom Solutions: Sakana AI offers tailored AI solutions to enterprise clients, particularly in sectors like finance, telecommunications, and manufacturing. These solutions leverage their proprietary models to address specific business challenges.

• Open-Source Community Engagement: The company releases certain models and tools as open-source, fostering a community of developers and researchers. This approach not only accelerates innovation but also positions Sakana AI as a thought leader in the AI space.

• Research and Development Collaborations: Sakana AI partners with academic institutions and research organizations to co-develop AI technologies, sharing resources and expertise to advance the field collectively.

• Strategic Partnerships and Investments: The company has secured investments from major corporations and venture capital firms, including NVIDIA, Khosla Ventures, and Lux Capital. These partnerships provide both financial backing and opportunities for collaborative projects.

This diversified business model enables Sakana AI to sustain innovation while delivering value to a broad spectrum of stakeholders.

Competitors

Sakana AI operates in the rapidly evolving landscape of artificial intelligence, where several companies are pioneering advancements in AI research, foundation models, and generative technologies. While Sakana AI distinguishes itself with a nature-inspired approach to AI development, it shares the space with other notable entities:

Key Competitors:

• OpenAI: A leading AI research organization known for developing advanced language models like GPT-4, focusing on ensuring that artificial general intelligence benefits all of humanity.

• Anthropic: An AI safety and research company that emphasizes the development of AI systems aligned with human values, known for its AI assistant, Claude.

• Stability AI: Specializes in developing generative AI models across various domains, including imagery, video, audio, and language, contributing significantly to multimodal AI and creative industries.

• Hugging Face: An open-source machine learning platform that provides tools and models to build, train, and deploy transformers at scale, fostering collaboration within the AI community.

• Mistral AI: Focuses on the development of open-source artificial intelligence models, offering efficient and adaptable AI solutions.

These companies, alongside Sakana AI, are at the forefront of AI innovation, each bringing unique approaches and solutions to the challenges and opportunities within the field.

Funding

Sakana AI has raised a total of $244 million across two rounds in 2024, marking one of the most substantial early-stage funding efforts in the AI sector. The company's innovative approach to collaborative, nature-inspired AI has attracted top-tier investors including NVIDIA, Khosla Ventures, New Enterprise Associates, 500 Global, and Global Brain Corporation.

In January 2024, Sakana AI closed a $30 million seed round led by Lux Capital with participation from 16 investors. Just eight months later, in September 2024, the company announced a $214 million Series A round led by Global Brain Corporation, JAFCO, Khosla Ventures, Lux Capital, Miyako Capital, and NEA, bringing the total number of investors across both rounds to 35.

This rapid and sizable fundraising underscores investor confidence in Sakana AI’s potential to reshape the future of AI research and applications, especially in emerging domains like automated scientific discovery and region-specific language models.