TOKYO—In a bold move underscoring Japan’s ambitions in the global AI race, Tokyo-based startup Sakana AI has secured $135 million in Series B funding, catapulting its valuation to $2.65 billion. This latest round, announced on November 17, 2025, positions the company as Japan’s most valuable unicorn and highlights a strategic push toward culturally aligned generative AI models tailored for critical sectors like finance and defense.

Founded in 2023 by former Google researchers David Ha, Llion Jones, and Ren Ito, Sakana AI draws its name from the Japanese word for ‘fish,’ symbolizing collective intelligence akin to a school of fish. The company’s approach diverges from the compute-heavy giants dominating the AI landscape, focusing instead on efficient, nature-inspired models that merge existing open-source AI to create new, optimized foundation models.

This funding comes amid Japan’s broader drive for AI sovereignty, with the government supporting initiatives to reduce reliance on U.S. and Chinese tech. Sakana’s models are designed to be culturally attuned to Japanese language and nuances, making them ideal for domestic applications where precision and efficiency matter most.

From Seed to Series B: A Rapid Ascent

Sakana AI’s journey began with a $30 million seed round in January 2024, led by Lux Capital and Khosla Ventures, as reported by the company’s own announcements. This was followed by a substantial Series A round in September 2024, raising approximately $200 million from a consortium including Mitsubishi UFJ, SMBC, Mizuho, Itochu, KDDI, Nomura, and Nvidia, according to Wikipedia.

The Series B, detailed in a TechCrunch article published on November 17, 2025, was led by investors including Mouro Capital and Factorial Funds, with participation from U.S. intelligence-linked In-Q-Tel. This inclusion of In-Q-Tel, the CIA’s venture arm, signals potential defense applications, as noted in posts on X (formerly Twitter) from users like StartupHub.ai.

CEO David Ha emphasized the round’s focus on sustainable AI, stating in a company post on X: ‘From day one, we’ve taken a different path. Our research has always focused on developing efficient AI technology sustainably, driven by the belief that resource constraints—not limitless compute—are key to advancement.’

Nature-Inspired Innovation at the Core

Sakana AI’s technological edge lies in its ‘evolutionary model merge’ method, which uses evolutionary techniques to combine open-source models into new ones with desired capabilities. A March 2024 post on X by AI enthusiast Lior Alexander highlighted this breakthrough: ‘Sakana AI just released a new method to combine the 500,000 open-source models to build new ones… We find that our approach can automatically discover entirely new ways to merge models with different architectures.’

This approach contrasts with the resource-intensive training of massive models by companies like OpenAI or Google. Instead, Sakana leverages collective intelligence, inspired by natural systems, to create efficient AI. The company has access to a government-sponsored GPU cluster through Japan’s NEDO GENIAC program, as announced in a February 2024 X post by Sakana AI, enabling rapid experimentation.

In a September 2024 X update, co-founder David Ha (posting as hardmaru) celebrated receiving an award from Japan’s METI/NEDO for their work, which led to research breakthroughs. This government backing, including grants for foundation model development, underscores Sakana’s role in strengthening Japan’s AI ecosystem.

Culturally Aligned AI for Japan’s Needs

What sets Sakana apart is its emphasis on models optimized for Japanese language and culture. As detailed in a Investing.com report from November 17, 2025, the startup develops ‘affordable generative AI models optimized for Japanese language and culture.’ This localization addresses gaps in global models that often underperform on non-English tasks.

Industry insiders note that this cultural alignment is crucial for sectors requiring high accuracy, such as legal and financial documentation in Japan. Sakana’s models aim to be more efficient, reducing the environmental and computational footprint—a key concern as AI’s energy demands soar globally.

The company’s Wikipedia page elaborates on its research in evolution and collective intelligence, with Ha explaining the fish analogy: ‘a school of fish coming together and forming a coherent entity from simple rules.’

Targeting Finance: Efficiency Meets Precision

With strategic investors like Japan’s three megabanks—Mitsubishi UFJ, SMBC, and Mizuho—Sakana is poised to disrupt finance. A September 2024 X post by hardmaru announced their support, highlighting partnerships across Japanese industries.

In finance, Sakana’s models could power fraud detection, risk assessment, and automated trading systems tailored to Japan’s regulatory environment. The Japan Times reported on November 17, 2025, that the startup is targeting banking markets, noting a ‘tailwind’ from Japan’s new government, which is pushing for tech innovation.

Experts suggest these culturally aligned LLMs could handle nuanced Japanese financial terminology better than Western models, potentially reducing errors in high-stakes transactions. The funding will support R&D for deploying these models in real-world banking scenarios, as per a Pulse 2.0 article from the same day.

Defense Applications: A Sovereign Imperative

Sakana’s foray into defense is particularly intriguing, especially with In-Q-Tel’s involvement. The Nikkei Asia on November 16, 2025, valued the company at $2.63 billion, emphasizing its unicorn status.

Japan’s push for sovereign AI aligns with national security goals, amid tensions in the Asia-Pacific. Sakana’s efficient models could enhance cybersecurity, intelligence analysis, and autonomous systems, all while being attuned to Japanese operational needs.

An X post from The Japan Times on November 17, 2025, stated: ‘With a ¥400 billion valuation after a big funding round, Sakana AI is targeting defense and banking markets.’ This reflects growing sentiment on X, where users like PARAS ARORA discussed the ‘global AI race’ and Japan’s bet on efficiency over scale.

Japan’s Broader AI Strategy and Challenges

Sakana’s rise is part of Japan’s concerted effort to catch up in AI. The company was one of seven institutions selected for NEDO’s supercomputer grant in 2024, as per their X announcement, aimed at developing foundation models.

However, challenges remain, including talent shortages and competition from U.S. and Chinese firms. Co-founder Llion Jones, co-author of the seminal ‘Attention Is All You Need’ paper, brings credibility, but scaling efficiently is key.

In a 2023 NHK interview shared on X, Ha said: ‘I don’t want the future of an important technology like AI to be dominated by a few companies in the Bay Area or by the government in Beijing. Sitting right between the US and China, Japan should play a key role.’

Investor Confidence and Future Horizons

The diverse investor base, including tech giants like Nvidia and Japanese conglomerates, signals strong confidence. A Ventureburn piece from November 17, 2025, noted the funds will expand research, develop new models, and grow teams.

Looking ahead, Sakana aims to deploy AI across manufacturing and other sectors, as mentioned by executive Sota Omura in an X post. The company’s 2024 end-of-year reflection on X by hardmaru celebrated ‘AI creating AI,’ hinting at self-evolving systems.

As global AI investments surge, Sakana’s model-merging approach could inspire a shift toward sustainability. With Japan’s government backing and international interest, the startup is well-positioned to lead in culturally sovereign AI.

The Global Implications of Localized AI

Beyond Japan, Sakana’s success could encourage other nations to pursue tailored AI, reducing dependence on Big Tech. The inclusion of In-Q-Tel suggests cross-border collaborations in defense tech.

Analysts on X, like Ray, shared links to funding news, reflecting buzz in tech circles. As one StartupHub.ai post put it: ‘a $135M bet against giant AI models.’

Ultimately, Sakana AI represents a paradigm where efficiency and cultural relevance trump sheer scale, potentially reshaping the AI landscape for years to come.