A case study has been presented in which Yaskawa Electric identifies companies developing “robot brains” through AI-driven M&A and considers acquisition and investment strategies targeting 4 companies. The targets are advanced physical AI companies in the United States, Europe and China, with their integration roadmaps and risks also examined in detail.
This article presents a case study on a defense technology M&A strategy against the backdrop of the situation in Ukraine, narrowing down to 4 global physical AI companies that Yaskawa Electric should consider acquiring or investing in. The targets are companies with technologies and business opportunities in heavy industry, electrical equipment and drones. The lecturer is Daisuke Imaizumi (今泉大輔), and the event is hosted by SSK新社会システム総合研究所. It proposes a roadmap for implementing autonomous operation without teaching and adaptive control through the acquisition of a robot brain. The article compares each candidate in detail, covering its corporate profile, technology architecture, synergies with Yaskawa Electric, estimated acquisition price range and key risks.
In “Screening and Selection of Candidate Companies,” Physical Intelligence, FieldAI, Micropsi Industries and Spirit AI are identified as the 4 leading candidates, with their headquarters locations, fundraising, core AI model characteristics and application areas organized for comparison. Particular emphasis is placed on the feasibility of applying the technologies to the MOTOMAN family and achieving autonomy in unstructured environments.
“Deep Dive into Each Candidate Company” summarizes each company’s technology, capital relationships, specific synergies with Yaskawa Electric, estimated acquisition price range and key risks. Physical Intelligence is characterized by its substantial fundraising and Flow Matching-based VLA model, FieldAI by its FFM, which learns the laws of physics, Micropsi Industries by MIRAI, which specializes in real-time correction, and Spirit AI by its VLA model, which prioritizes real-world data.
“Technology Integration and the Real-Time Control Platform” explains a four-layer hierarchical architecture, latency coordination between inference and control, and integration with the NVIDIA Cosmos/Isaac Lab/GR00T pipeline. The assumed operating frequencies are approximately 20 Hz for inference and 1 kHz for the servo systems. The section outlines the technical challenges and proposed solutions for synchronizing the robot brain with on-site control systems.