Arm launches AGI CPU, its first data center chip
Arm has announced the AGI CPU, its first production silicon product and its first Arm-designed data center CPU. The company says this marks a shift beyond licensing IP and Compute Subsystems, with Arm now offering complete silicon for AI infrastructure.
The AGI CPU is aimed at what Arm calls agentic AI infrastructure, where CPUs handle orchestration, scheduling, memory, storage, and data movement around large accelerator clusters. Arm says the chip offers up to 136 Neoverse V3 cores , 6GB/s of memory bandwidth per core at sub-100ns latency, and a 300W TDP with one dedicated core per program thread.
Source: ARM
The Arm AGI CPU delivers:
- Performance: Up to 136 Arm Neoverse V3 cores per CPU, delivering leading performance per core, SoC, blade and rack*, with 6GB/s memory bandwidth per core at sub-100ns latency.
- Scale: 300-watt TDP with a dedicated core per program thread enables deterministic performance under sustained load, eliminating throttling and idle threads.
- Efficiency: Supports high-density 1U server chassis that supports air-cooled deployments with up to 8,160 cores per rack, and liquid-cooled systems delivering 45,000+ cores per rack.
Arm is also putting heavy focus on rack density. Its reference design uses a 1OU dual-node server with 272 cores per blade, scaling to 8,160 cores in a 36kW air-cooled rack. Arm says a liquid-cooled Supermicro design can scale past 45,000 cores per rack. The company claims more than 2x performance per rack versus current x86 systems, although that figure comes from Arm’s own materials.
Source: ARM
Meta is the lead partner and co-developer for the launch, and Arm says the AGI CPU will work alongside Meta’s MTIA accelerators. Arm also named Cerebras, Cloudflare, F5, OpenAI, Positron, Rebellions, SAP, and SK Telecom as deployment partners for AI control plane, orchestration, API hosting, and enterprise workloads.
Commercial systems are already available to order from ASRock Rack, Lenovo, and Supermicro, while Arm says broader availability is expected in the second half of 2026. The launch is also backed by a long list of ecosystem partners, including AWS, Google, Microsoft, NVIDIA, Samsung, SK hynix, and TSMC, which Arm says will manufacture the AGI CPU on its 3nm process.
Source: ARM
