Mitsubishi Electric invests in Sakana AI, bringing AI Agents into factories

Top News

Sakana AI has secured an investment from Mitsubishi Electric, and the two sides are collaborating to apply AI to factories and infrastructure.

Overview

On March 25, 2026, Sakana AI announced that it received an investment from Mitsubishi Electric (the amount was not disclosed), and integrated its own AI Agent technology into Mitsubishi’s Serendie digital platform. These technologies include AI Scientist for conducting automated research and the Continuous Thought Machine for reasoning, with the goal of making manufacturing and infrastructure operations more automated.

Sakana has already had hands-on experience in the financial sector (working with Citigroup, Daiwa Securities, and Mitsubishi UFJ Bank) and in the defense sector (winning the U.S.-Japan Biological Defense Innovation Prize). This time, it has set manufacturing as the company’s third key focus area. This also aligns with Japan’s broader push for domestic AI: Mitsubishi has years of industrial know-how and the Maisart AI toolset, while Sakana focuses on a small-but-efficient model approach.

Analysis

AI Agents are expanding from the software domain into physical industries. With Japan’s severe population aging, the manufacturing sector’s need for automation is urgent.

• Route difference: Sakana follows an efficiency-first approach. It doesn’t aim to simply build larger models; instead, it uses less compute to handle multi-step tasks and research-oriented workflows.
• Industry enablement: Mitsubishi’s years of accumulated factory data and real test environments can provide engineering-level validation and conditions for large-scale deployment.
• Task types: What this collaboration aims to address are complex tasks that require human judgment, such as equipment fault prediction, supply chain scheduling, and cross-system operational optimization.

Based on the industry context, after Sakana completed its $135 million Series B round, it has continued to emphasize doing more with fewer resources, which contrasts with the U.S. trend of throwing compute at the problem. Whether this path can truly work remains to be seen based on real-world results, but for Japanese industrial customers, it is clearly compelling.

Impact

For the market: Japan’s manufacturing industry is roughly $1.5 trillion in size. If this collaboration performs well, it could spur more manufacturers to adopt similar Agent-driven automation solutions. Since the investment amount has not been disclosed, it’s hard to gauge the scale, but Sakana is reportedly already profitable from its projects in finance and defense, which provides a commercialization path that can be referenced.

For both sides:

• Sakana AI: Gains proprietary industrial data and top-tier customer use-case scenarios, enabling it to validate how far efficient models can go in environments with hard constraints.
• Mitsubishi Electric: Doesn’t have to develop Agent capabilities and research-oriented workflows from scratch; it can obtain them directly through the collaboration.

Assessment: This round is still in the early stage of industrial validation. It has more value for Builders deeply engaged in real scenarios and for mid- to long-term strategy funds; short-term traders can temporarily ignore it.**