A processor uses 3D stacking and AI instructions to run inference while reducing power use and the number of servers required.
Fujitsu’s FUJITSU-MONAKA processor is designed for AI inference and other workloads where computing performance, power consumption, and data security are key considerations. It operates at up to 3.8GHz and supports memory transfer speeds of up to 8800MT/s. Fujitsu says it can deliver twice the AI inference throughput of other CPUs while improving power efficiency, which could reduce the number of processors and servers required for the same AI workload.
The processor uses a 3D-stacked architecture that combines a 2nm process for the CPU cores with a 5nm process for the cache and I/O sections. These process technologies are combined with Fujitsu’s ultra-low-voltage operation technology to improve both performance and power efficiency.
For AI workloads, the CPU includes dedicated instructions for matrix operations used in AI inference. These work alongside Arm SVE2 vector operations and software optimisation to increase inference throughput. According to Fujitsu, this approach allows the CPU to provide practical AI inference performance without requiring a separate accelerator.
The processor is designed for environments where power and cooling capacity are limited. By increasing AI inference performance at the CPU level, it can give organisations more flexibility when scaling AI workloads while reducing infrastructure requirements.
The chip also supports confidential computing through hardware-based encryption. This protects data and applications while they are stored in memory during operation and is intended to support secure AI workloads, including those running in multi-tenant cloud environments.
Global availability is planned from the fourth quarter of fiscal year 2026, including markets such as the US and Asia-Pacific. Fujitsu will initially offer the processor as a standalone product to cloud and data centre operators and server vendors.
The processor is being developed in response to growing demand for AI computing and the resulting increase in power consumption. Fujitsu is positioning it for organisations that need higher AI inference performance while keeping power consumption and infrastructure requirements under control.
Click here for the original announcement.
