Overview

The compute platform for the agentic era

AI Summary

At OCP Global Summit 2026, Arm showed how agentic AI is reshaping data center infrastructure, from persistent agent execution and AI head nodes to cloud-native services and the infrastructure that secures them. With Arm AGI CPU, independent 2x rack-scale performance validation, new open-system designs and a growing ecosystem across silicon, systems and software, Arm demonstrated how open infrastructure can scale the next generation of AI.


Announcements at OCP
Discover the latest at OCP

Up to 2x rack-level performance for agentic AI

Independent Signal65 testing shows that Arm AGI CPU delivers up to 2x rack-level performance across key agentic workloads versus AMD EPYC Turin, highlighting why completed tasks matter more than inference speed alone.

AI agents are redefining CPU performance

As agentic AI scales, CPU performance is increasingly measured by the useful work completed under real data center constraints. See how Arm is introducing a new methodology for agentic AI workloads.

Defining CPU and software stack performance for agentic AI

Agentic AI expands CPU workloads beyond inference. This white paper proposes a use-case-driven methodology that combines component-level and end-to-end benchmarking to measure the infrastructure behind completed tasks.

Exclusive sessions

Agenda at glance

Monday, October 12, 2026
Time Sessions Location

1:15 pm – 1:35 pm

Rethinking CPU performance for the agentic AI era

SJCC - Concourse Level - 210AE

2:30 pm – 2:50 pm

Scaling AI inference infrastructure from silicon to system presented by ASRock Rack

SJCC - Concourse Level - 212

5:10 pm – 5:40 pm

Arm Keynote – The agentic decade starts with open infrastructure

SJCC - Street Level - South Hall

Tuesday, October 13, 2026
Time Sessions Location

1:10 pm – 1:25 pm

Designing infrastructure for agentic AI: From servers to systems presented by Arm

SJCC - Concourse Level - Expo Hall Theater

4:15 pm – 4:30 pm

From AI PC to AI infrastructure: Platform contracts for scalable CPU–accelerator systems

SJCC - Concourse Level - 211

4:25 pm – 4:40 pm

OCP RAS API PLDM Modeling

SJCC – Lower Level LL21BC

4:45 pm – 5:00 pm

SBSA: Advancing Arm server system architecture in the OCP ecosystem

SJCC - Concourse Level - 212

Wednesday, October 14, 2026
Time Sessions Location

8:15 am – 9:10 am

OCP hardware fault management workstreams status and relationship

SJCC - Concourse Level - 211

3:55 pm - 4:10 pm

Near data processing opportunities for context engineering

SJCC - Lower Level LL21DE

Thursday, October 15, 2026
Time Sessions Location

8:00 am – 8:15 am

The open chiplet economy: Progress, challenges, and the road ahead

SJCC - Concourse Level - 211

8:25 am – 8:45 am

Node-to-rack efficiency for real-world AI inference: Arm and FuriosaAI for scalable open infrastructure

SJCC - Concourse Level - 220B

8:45 am – 9:05 am

Endorsements for attestation: A common language for the whole supply chain

SJCC - Concourse Level - 212

9:10 am – 9:30 am

Building interoperable AI systems with chiplets: Arm Neoverse and Cadence in an open ecosystem

SJCC - Concourse Level - 211

9:45 am – 10:10 am

USB to rule them all: One manageability interface for AI platforms

SJCC - Concourse Level - 220B

10:30 am – 11:10 am

AI-native virtual chiplet eco-systems with industry best practices to accelerate chiplet adoption

SJCC - Concourse Level - 211

12:30 pm – 12:50 pm

Scaling AI inference with Arm AGI CPU, Rebellions, RebelCard, and OCP ORv3 rack design

SJCC - Concourse Level -220B

1:10 pm – 1:30 pm

Securing the open chiplet ecosystem: Defining a common security architecture for FCSA

SJCC - Concourse Level - 211

2:05 pm – 2:30 pm

Building bridges for adoption: OCP, standards bodies, and open source firmware

SJCC - Concourse Level - LL20BCD

2:45 pm – 3:05 pm

Introduction to a new OCP FCSA sub-project workstream

SJCC - Concourse Level - 211

3:10 pm – 3:30 pm

Applying OCP ORv3 and ORW design principles to rack-scale Arm AI infrastructure with Supermicro

SJCC - Concourse Level - 230A

Explore Arm AGI CPU

The world’s most efficient agentic CPU

Arm first silicon CPU - AGI CPU

Arm AGI CPU

Arm AGI CPU is Arm's first production silicon, purpose-built for agentic AI. It delivers more than 2x performance per rack versus x86 and is used in servers from Meta, Lenovo, Supermicro, and ASRock.

Learn more

Key features of the Arm AGI CPU

Rack-level performance

Servers and rack-scale systems

Arm AGI CPU servers

 

Arm AGI CPU 1OU Dual Node Reference Server

Arm AGI CPU 1OU Dual Node
Reference Server

Reference design for maximum density deployments of Arm AGI CPU – in a OCP DC-MHS standard form factor 1OU Dual Node server.

Arm AGI CPU 2U2P Reference Server

Arm AGI CPU 2U2P
Reference Server

19” 2U2P reference design for Arm AGI CPU deployment in a traditional form factor.

Lenovo HR650a V3 2U Arm AGI CPU System

Lenovo HR650a V3 2U Arm AGI CPU System

Enterprise-class 2U Arm AGI server optimized for cloud infrastructure, delivering reliable performance and low TCO.

Supermicro 5U Arm AGI CPU PCIe GPU System

Supermicro 5U Arm AGI CPU PCIe GPU System

High-density 5U AI platform combining dual Arm AGI CPUs with extensive PCIe GPU expansion.

Supermicro 2U Hyper Arm AGI CPU Hyper system

Supermicro 2U Hyper Arm AGI CPU Hyper System

Compact 2U dual-socket Arm AGI server designed for efficient cloud and AI infrastructure deployments.

Supermicro 2U4N Arm AGI CPU Liquid Cooled Server

Supermicro 2U4N Arm AGI CPU Liquid-Cooled Server

Liquid-cooled 2U 4-node server for cloud and AI workloads, including agentic AI and AI pre-processing, offering maximum performance, density, and efficiency.

Supermicro 2U Hyper-E Arm AGI CPU Server

Supermicro 2U Hyper-E
Arm AGI CPU Server

Single-socket 2U server with up to 3TB of memory for cloud and edge computing.

ASRock Rack 2OU2N-Arm System

ASRock Rack 2OU2N-Arm
System

High-density dual-node Arm server built to OCP ORv3 standards for scalable, power-efficient cloud deployments.

Arm AGI CPU liquid cooled rack-scale systems

 

Supermicro 1U 4-Node Arm AGI CPU Rack-Scale Solution

Supermicro 1U 4-Node
Arm AGI CPU Rack-Scale Solution

Maximum compute density multi-node liquid cooled ORW rack architecture, featuring 168 nodes with 336 Arm AGI CPUs totaling 45,696 cores.

Supermico 2U4N Arm AGI CPU Rack Scale solution

Supermicro 2U 4-Node
Arm AGI CPU Rack-Scale Solution

Ultra-density multi-node liquid cooled ORv3 rack architecture, featuring 76 nodes with 152 Arm AGI CPUs totaling 20,672 cores.


  1. Based on estimates.