Fujitsu

Semiconductors · AI Infrastructure · Data Centers

Launch

Fujitsu unveils MONAKA 2nm AI chip, cutting server count 50% for same workload

The launch bets that Japan and Europe will pay a premium for domestically built AI hardware to escape Nvidia-centric supply chains.

• Fujitsu announced on September 14, 2026 that it will begin global sales of the FUJITSU-MONAKA processor and Fujitsu MONAKA Server starting November 2026, positioned as sovereign AI infrastructure.
• Fujitsu is a Japanese IT and electronics conglomerate; MONAKA draws on chip technology it developed for Japan's Fugaku supercomputer.
• The 144-core Arm-based chip uses a 2nm process for its compute cores and 5nm for cache and I/O in a 3D-stacked design, hitting 3.8GHz with 8,800MT/s memory transfer speed.
• Fujitsu says MONAKA delivers twice the AI inference throughput of comparable CPUs at half the power, letting customers run the same workloads on 50% fewer servers.
• The 1U air-cooled server can run in ambient temperatures up to 40C, which Fujitsu says can cut cooling energy use by up to 80% without requiring liquid-cooling infrastructure.
• Chip production is outsourced to TSMC, though Fujitsu is also considering Japan's Rapidus as a future manufacturer, and the server itself is built at a Fujitsu plant in Ishikawa Prefecture.
• The launch targets a market currently dominated by Nvidia GPUs, as European and Japanese buyers increasingly seek domestically traceable AI hardware amid rising geopolitical and supply-chain security concerns.
• MONAKA signals a broader shift toward AI hardware sovereignty, where governments and enterprises weigh supply-chain traceability and power efficiency as heavily as raw compute performance.