While Arm primarily licenses CPU and GPU designs to other manufacturers, they also produce their own silicon for specific datacenter and AI infrastructure applications.
ARM AGI CPU#
Announced at the Arm Everywhere Keynote on March 24, 2026, the ARM AGI CPU is Arm’s first production silicon, designed for AI infrastructure at scale. It delivers a new class of CPU with high performance and extreme rack-level density support for agentic AI operations across modern data centers.
Specifications#
- Cores: Up to 136 Neoverse V3 cores (2x 128 SVE, 2MB/core L2)
- Architecture: Armv9.2 with bfloat16 and INT8 AI instructions
- Clock Speed: Up to 3.7GHz boost
- PCIe: 96 Lanes of PCIe Gen6, CXL 3.0 Type 3
- Process: 3nm lithography process
- TDP: Up to 420W TDP
- Memory: Up to 6 Terabyte of DDR5-8800 (12x DDR5 channels)
- Design: Dual-Chiplet Design
SKUs#
The ARM AGI CPU is available in three distinct SKUs:
- SP113012: 136-core flagship model for maximum core count.
- SP113012S: 128-core model optimized for Total Cost of Ownership (TCO).
- SP113012A: 64-core model optimized for maximum memory bandwidth per core.
Server Configurations#
Arm’s reference server configuration is a 10U, 2-node design – packing in two chips with dedicated memory and I/O for a total of 272 cores per blade. These blades are designed to fully populate a standard air-cooled 36kW rack – 30 blades delivering a total of 8160 cores.
Arm has additionally partnered with Supermicro on a liquid-cooled 200kW design capable of housing 336 Arm AGI CPUs for over 45,000 cores.
For more information, visit the official Arm AGI CPU page or read the Product Brief.