Fujitsu Limited has announced the upcoming global release of its FUJITSU-MONAKA CPU, a 2nm processor developed in Japan. The company plans to begin sales of the standalone CPU to data center operators and server vendors in November 2026, alongside a new line of dedicated servers. These systems are intended to provide a sovereign AI infrastructure, emphasizing supply chain transparency and domestic manufacturing.
The FUJITSU-MONAKA processor features a 3D-stacked architecture, combining 2nm CPU cores with 5nm cache and I/O components. Each chip contains 144 cores, with base frequencies reaching up to 2.1GHz in air-cooled configurations and 2.9GHz in liquid-cooled setups, while the processor can achieve a maximum operating frequency of 3.8GHz. The design incorporates dedicated matrix instructions and SVE2 vector processing to accelerate AI inference workloads directly on the CPU.
Fujitsu claims the processor delivers twice the AI inference throughput of unspecified competing CPUs, potentially halving the number of servers and power consumption required for equivalent tasks. The hardware supports memory transfer speeds of 8800MT/s and utilizes Arm CCA for hardware-level confidential computing, which encrypts data and applications during memory-resident operations. This security feature is intended to support multi-tenant cloud environments.
The server lineup includes 1U and 2U rackmount models, as well as a 2U multi-node configuration for high-performance computing and academic research. The 1U systems are specifically engineered for air-cooled data centers, supporting ambient temperatures up to 40 degrees Celsius. Fujitsu states that its cooling technology can reduce cooling power consumption by up to 80 percent, though specific test conditions for this figure were not disclosed.
Beyond the hardware, Fujitsu intends to integrate the MONAKA platform with its proprietary software, including the Fujitsu Kozuchi AI platform and Takane generative AI models. The company is also working on future rack-scale systems that incorporate robotic maintenance capabilities. Global availability for the CPU is scheduled for the fourth quarter of fiscal year 2026, with server shipments to select sectors in Japan and Europe beginning in April 2027.