The Arm AGI CPU will run Red Hat’s enterprise software stack to support agentic artificial intelligence workloads in data centers. This collaboration aims to deliver a production-ready infrastructure for both cloud and on-premises environments. Meanwhile, the shift from training to inference requires higher compute density and efficient orchestration across distributed systems.
To address these infrastructure demands, Red Hat Enterprise Linux will provide a stable operating system on the architecture. In addition, Red Hat OpenShift will allow enterprises to manage Kubernetes clusters at scale. Consequently, organizations can run virtual machines and containers side-by-side using OpenShift Virtualization.
Technical Specifications of the Arm AGI CPU
Specifically, the Arm AGI CPU features 136 Arm Neoverse V3 cores and 96 lanes of PCIe Gen6. This system-on-chip also includes 12 channels of DDR5 memory running at speeds up to 8800 MT/s. As a result, the hardware provides the high memory bandwidth necessary for complex computers and server tasks.
Software Integration and Virtualization
The software integration allows workloads to move between cloud instances and local servers. Notably, existing enterprise tools and workflows remain fully compatible with this platform. Furthermore, this compatibility helps businesses modernize their infrastructure without experiencing operational disruptions.
Power Efficiency and Rack Density
Consequently, the Arm AGI CPU operates at a 300W thermal design power (TDP). This power level is lower than traditional 500W-class x86 processors. Therefore, a standard 36kW air-cooled Open Compute Project rack can support approximately 8,160 cores. In comparison, traditional systems support only 4,352 cores per rack.
Moreover, liquid-cooled environments can support up to 336 processors per rack. This configuration increases the total compute density up to fivefold, reaching 45,696 cores. Meanwhile, the processor handles critical bottlenecks in data pre-processing and large-scale inference.
Market Availability and Ecosystem Support
Several hardware partners will support the deployment of this integrated platform. Specifically, companies such as Supermicro, Lenovo, and ASRock Rack will provide compatible systems. In addition, cloud platforms like AWS Graviton, Microsoft Azure Cobalt, and Google Axion already use similar architectures.
Finally, the integrated software and hardware stack is expected to be available in the fourth quarter of 2026. This timeline gives enterprises a clear path to plan their next-generation data center upgrades. Consequently, organizations can prepare their environments for these upcoming systems.
According to Red Hat, the expansion of the partner network helps deliver a broader set of customer solutions on a trusted open foundation.
“Red Hat delivers enterprise-grade, open-source platforms that enable innovation across hybrid cloud environments while providing our customers with extensive choice.”
Steven Huels, Vice President of AI Engineering at Red Hat
