Fujitsu sets a 2nm CPU for sovereign AI

Fujitsu announced a 2nm MONAKA CPU and server line for sovereign AI, with sales starting in November 2026. The launch details say MONAKA reaches 3.8 GHz and claims twice the AI-inference throughput with half the servers and power for equivalent loads; these are Fujitsu’s claims, not independent benchmarks. Operators needing inference in air-cooled or jurisdiction-bound facilities should treat this as a procurement option, while the CPU-first data-center thesis shows why compatibility and rack-level tests matter more than a process-node headline.