Fujitsu Shows Production-Ready 2nm MONAKA Processor at SC26, a 144-Core Bet on Sovereign AI

Fujitsu showcased its production-intent 2-nanometer Arm processor, FUJITSU-MONAKA, and its server lineup at the SC26 supercomputing conference in Chicago: 144 Armv9.3-A cores, TSMC 2nm compute dies stacked over 5nm cache and I/O dies, aimed at sovereign AI and high-performance scientific computing. The processor goes on sale globally this month, with servers ramping into 2027.
Fujitsu showcased its production-intent next-generation 2-nanometer Arm processor, FUJITSU-MONAKA, and the accompanying MONAKA Server lineup at the SC26 supercomputing conference in Chicago (November 15-20), with live demonstrations of the hardware. The MONAKA chip packs 144 Armv9.3-A cores and targets sovereign artificial intelligence and high-performance scientific computing.
144 cores in a 3D stack, with 2nm reserved for compute
MONAKA uses an advanced 3D disaggregated System-in-Package design that places compute cores, cache memory and I/O controllers on separate silicon dies. The compute layer holds 144 Armv9.3-A cores fabricated on TSMC's 2-nanometer N2P process, each core with dual 256-bit SVE2 vector pipelines; dual-socket server configurations deliver 288 physical cores per node. Fujitsu restricts 2nm silicon to the compute logic: the entire last-level SRAM cache and peripheral I/O subsystems are built on TSMC's mature 5-nanometer process, with the 2nm compute dies bonded face-to-face directly atop the 5nm SRAM dies. The processor also has 12 DDR5 memory channels and supports FP8/INT8 matrix operations.
Two models; on sale this month, volume production in 2027
Fujitsu is preparing two models: a 350W air-cooled version with a 2.1 GHz base clock, and a 500W liquid-cooled variant at 2.9 GHz. Global sales of the processor begin in November 2026, commercial server deployments ramp through late 2026, and widespread enterprise and university deliveries are slated for 2027. Sample silicon and early validation systems have gone to strategic partners, including national laboratories and supercomputing research consortia, for testing. Fujitsu claims twice the inference throughput of competing CPUs and cooling-power reductions of up to 80 percent — company claims that still await independent production benchmarks.
Successor to the Fugaku chip, aimed at AI compute free of foreign hardware
MONAKA succeeds Fujitsu's A64FX processor, whose 48 cores on a 7nm process power Japan's Fugaku supercomputer. Fujitsu pitches MONAKA as 'sovereign AI' infrastructure: universities, research agencies and regional computing centres can build their own compute capacity, free of proprietary GPU licensing and supply bottlenecks. Its AI-HPC Platform, under the banner 'Powering AI for Science', unifies numerical simulation with generative AI acceleration, and Fujitsu also laid out a hybrid quantum computing roadmap in which MONAKA host nodes interface with superconducting quantum processors. The processor is designed and developed by Fujitsu in Japan and the servers are assembled at its Kasashima Plant, while the advanced compute dies are fabricated by TSMC in Taiwan; the planned successor, Monaka-X, is to use Japan's Rapidus 1.4nm process and power the FugakuNEXT system.
Sources: CASRAI; Fujitsu; ServeTheHome / Hot Chips 2026 coverage.
N環球 
💬 Comments
No comments yet. Be the first!