Oracle Cloud Infrastructure joins Arm AGI CPU ecosystem as agent AI accelerates

news

June 2, 2026

This move strengthens the role of the Arm computing platform as agent AI drives new demand for efficient rack-scale cloud infrastructure.

Written by Mohamed Awad Executive Vice President, Cloud AI Business Unit arm

In Arm Everywhere on March 24th, we said that agent AI will dramatically increase CPU demand in data centers. Just two months later, the market is moving even faster than we expected.

Today at COMPUTEX, we announced that Oracle Cloud Infrastructure (OCI) is joining the Arm AGI CPU ecosystem, building on the momentum we shared with Arm Everywhere with customers such as Cerebras, Cloudflare, F5, Meta, OpenAI, Positron, Rebellions, SAP, SK Telecom, and Verda.

“OCI is demonstrating strong momentum for Arm-based infrastructure across large-scale cloud-native workloads, including with customers like Uber, and we are excited to explore how Arm AGI CPUs can extend those benefits to next-generation agent AI systems,” said Mahesh Thiagarajan, EVP, Oracle Cloud Infrastructure. “By combining high-performance Arm compute with OCI’s scalable cloud infrastructure, we believe Arm AGI CPUs have the potential to help customers deploy more efficient and tuned AI environments at scale.”

These deployments span hyperscalers, AI model providers, enterprises, and cloud infrastructure leaders and highlight the growing demand for dedicated infrastructure for agent AI.

The age of agents is coming sooner than expected.

The rapid growth of agent AI confirms what we outlined on March 24th. Anthropic reported a revenue run rate of nearly $50 billion in May, up from about $9 billion at the end of 2025. Meanwhile, companies across industries are racing to deploy increasingly sophisticated AI agents. What was predicted just 70 days ago is rapidly expanding faster than expected.

Agent workloads continuously coordinate across tools, services, and data sources, routing tasks, executing code, and maintaining context across an increasingly distributed system. As a result, more work occurs outside of the model itself. SemiAnalysis recently estimated that 42% of the execution time of modern agent coding workloads is spent using CPU-driven tools, highlighting the growing importance of CPUs in AI infrastructure.

Agent CPU

Designed specifically for the agent era, Arm AGI CPUs deliver more than double the performance per rack compared to traditional x86 CPU deployments, enabling cloud providers and AI infrastructure operators to significantly increase compute density while staying within power and thermal constraints.

Arm Everywhere estimated that Arm AGI CPUs can save operators up to $10 billion in capital expenditures for every gigawatt of AI infrastructure capacity deployed. Given the accelerating pace of agent AI adoption, we currently believe the economic impact could be even greater.

At COMPUTEX, Supermicro joined systems from ASRock Rack, Lenovo, and other ecosystem partners to announce the new Arm AGI CPU platform, which spans both air-cooled and water-cooled rack-scale deployments.

The momentum of Arm Neoverse continues to grow

The same forces driving adoption of Arm AGI CPUs are accelerating Arm Neoverse momentum across the cloud.

Google recently announced Axion as the head node for its latest TPU system to replace x86 CPUs.
AWS continues to expand adoption of Graviton, and major customers such as Snowflake are seeing increased demand for Arm-based infrastructure to support agent workloads. AWS also noted that customers are seeking additional Graviton capacity to support next-generation AI services.
NVIDIA’s Arm-based Vera platform is currently being deployed by multiple customers, including OpenAI, Anthropic, and SpaceX, reinforcing the value of dedicated CPUs working in conjunction with AI accelerators in tightly integrated rack-scale systems.

These deployments show that Arm-based CPUs are becoming the control plane of modern AI infrastructure.

Defining the architecture of the agent era

The past two months have highlighted the simple reality that Agentic AI is increasing the strategic importance of CPUs.

But agent AI doesn’t just exist in data centers. Agents span cloud AI infrastructure, edge AI devices, and physical AI systems. With Arm’s pervasive computing platform, we are uniquely positioned to provide the foundation for the age of agents.