The recent partnership between Mitsubishi Heavy Industries (MHI) and Preferred Networks (PFN) signals a pivotal shift towards developing a sovereign AI capability for Japan's defense sector.
This collaboration is the direct result of a clear national strategy. For years, Japan has been laying the groundwork through policy. The Economic Security Promotion Act of 2022 mandated the protection of critical technologies, and the Ministry of Defense's 2025 "Responsible AI" guidelines specifically called for AI systems to be developed and operated domestically, setting the stage for a homegrown solution.
Two major external forces are also accelerating this move. First, the persistent weakness of the yen makes importing foreign high-tech components, like advanced GPUs, increasingly expensive. This creates a strong economic incentive to adopt domestic hardware, such as PFN's own MN-Core AI chip.
Second, the global technological arms race is heating up. The U.S. Department of Defense is aggressively deploying AI from major tech firms onto its classified networks. For Japan, a key U.S. ally, this creates a dual pressure: the need to ensure its systems are interoperable with American forces, while also protecting its own sensitive data and maintaining technological independence.
This is where the MHI-PFN partnership becomes so powerful. It's a perfect synergy of capabilities. PFN brings a complete, vertically integrated AI stack—from the MN-Core chip and the PLaMo large language model to the PFCI AI cloud platform. They provide the core intelligence.
MHI, on the other hand, provides the robust industrial and defense framework. They have the platforms—ships, aircraft, and unmanned vehicles—and the critical infrastructure for secure, on-premise deployment. MHI's DIAVAULT edge data center, for example, allows AI models to run in a closed, air-gapped environment, which is essential for handling classified military data. This collaboration effectively connects PFN's advanced AI brain to MHI's powerful industrial body.
- On-premise (On-prem): A computing environment where an organization manages its own hardware and software in its own physical facilities, rather than using a public cloud. This offers maximum security and control.
- C2 (Command and Control): The set of organizational and technical attributes and processes that a commander uses to plan, direct, coordinate, and control forces and operations.
- Sensor Fusion: The process of combining data from multiple different sensors to produce more accurate and complete information than what could be derived from any single sensor.