Fujitsu will begin selling its Japan-developed FUJITSU-MONAKA CPU in November in Japan and overseas markets, alongside the Japan-made Fujitsu MONAKA Server in Japan and Europe. Designed for AI inference, the chip uses a 3D-stacked architecture with 2nm compute technology and 5nm cache and input/output components. Fujitsu says matrix-operation acceleration delivers twice the throughput of competing CPUs, while power-saving technology derived from the Fugaku supercomputer can halve server requirements and electricity use for equivalent workloads. The CPU reaches a maximum 3.8GHz clock speed and memory transfer speeds of 8800MT/s. Fujitsu is positioning the package as a CPU-based alternative to GPU-dependent AI infrastructure for security-conscious customers seeking domestic or geographically diversified supply chains. The server is available in 2U dual-CPU and 1U one- or two-CPU configurations, with proprietary cooling supporting air-cooled operation at ambient temperatures up to 40°C and liquid cooling at water temperatures up to 45°C. TSMC is manufacturing the chip, while Fujitsu is considering future production outsourcing to Rapidus. The company previously targeted external CPU sales in the U.S. and Asia-Pacific from 2027, with Super Micro Computer confirmed as a customer, 24 companies testing units and more than 40 discussing adoption. Fujitsu aims for 60,000 AI-server sales and ¥250 billion in cumulative external CPU revenue by the end of fiscal 2030.