Fujitsu is preparing to commercially sell its custom Monaka Arm-based server chip, targeting sovereign-sensitive cloud providers and inference-focused workloads.

Fujitsu will begin selling its Monaka processors and associated servers in November. The Japanese technology giant first introduced Monaka in 2023, teasing that the processor was derived from the custom silicon designed for the Fugaku supercomputer but adapted for datacenter workloads. Fugaku held the title of the world’s most powerful supercomputer from 2020 to 2022 and remains ninth on the Top 500 list six years after its debut. Given that legacy, demand for infrastructure powered by such technology is high. However, as The Register reported earlier this year, Monaka diverges significantly from the silicon inside Fugaku.

Built on Broadcom’s 3D-chip stacking technology, Monaka utilizes the Armv9 instruction set architecture. The chip integrates four 2nm compute dies, each containing 36 cores, alongside four 5nm SRAM chiplets. These components are linked by a central I/O and memory die featuring 12 channels of DDR5 and PCIe 6.0 connectivity. While Fujitsu previously indicated that Monaka would not reach the market until 2027, the company confirmed this week that commercial sales will commence in November.

Fujitsu will sell the chips directly to cloud operators and third-party server manufacturers, who will inevitably compete with Fujitsu’s own hardware offerings. The company has developed three distinct server configurations, all built around the 144-core Monaka processor. The first is a 1U chassis supporting one or two processors running at either 2.1GHz or 2.9GHz, accommodating eight E3.S SSDs and two M.2 drives. It operates in air-cooled environments up to 40°C and can handle temperatures up to 45°C with liquid cooling. The second is a 2U dual-CPU model running at 2.1GHz, equipped with four E3.S disks and two M.2 drives, designed exclusively for air cooling. The third is a 2U dual-CPU system optimized for multi-node deployments of four servers per rack, operating at 2.9GHz with two E1.S SSDs and two M.2 drives, requiring liquid cooling. Fujitsu identifies this third configuration as best suited for high-performance computing and academic users.

Fujitsu has not publicly disclosed pricing. The company is positioning Monaka as a strong candidate for AI inference workloads, claiming it delivers twice the inference throughput of competing CPUs. This performance claim, combined with ongoing fragility in global silicon supply chains, is likely to attract attention from major AI companies, hyperscalers, and emerging cloud providers. Additionally, the servers’ ability to operate reliably in high-temperature air-cooled datacenters adds practical appeal. Nevertheless, adoption beyond hyperscale suppliers remains uncertain. Despite nearly a decade of Arm CPU vendors emphasizing low power consumption as a critical advantage, server manufacturers have historically made only minimal efforts to deploy Arm-based systems.

Fujitsu is marketing its compact servers as a sovereign computing solution, partly because they are manufactured in Japan. Japan and Europe constitute the initial target markets for these systems, with European enterprises increasingly prioritizing domestic and sovereign technology stacks.