Overview

Rethinking the Datacenter for the AI Era

Driven by the rapid rise of AI, computing infrastructure is shifting from general-purpose platforms to more specialized, power-efficient, workload-optimized solutions with an increased emphasis on performance, efficiency, and scalability. Arm provides a flexible, power-efficient compute foundation to meet these demands and create opportunities for innovation and disruption—a foundation for an AI infrastructure that scales seamlessly from hyperscale to edge.

Benefits

More Compute, Higher Efficiency, Better Price-Performance

Arm delivers energy-efficient compute that pairs seamlessly with a broad range of AI accelerators—helping you achieve strong performance and efficiency while lowering total cost of ownership.

Up to 8x Faster ML Model Training and 4.5x LLM inference Performance

Delivered by the NVIDIA Grace Hopper Superchip when training a DLRM model and inferencing GPT-65B model, compared to x86+Hopper systems.1
Up to 3x Better Recommender Performance

Delivered by Google Axion processor in MLPerf DLRMv2 benchmark compared to x86 alternatives.2
Up to 2.5x Higher AI Inference Performance

Delivered by Google Axion processor, with 64% cost savings and faster RAG for real-time AI compared to x86 alternatives.3
Up to 2x Better Performance with LLM and ML Inference Tasks

Delivered by Axion-based VMs compared to current-generation x86 instances.4
Partners

Enabling Industry Leaders Though Infrastructure Optimized for Real-World Performance

Arm empowers industry leaders to build a new wave of scalable, efficient data centers with computing solutions optimized for real-world performance. Designed for performance, power efficiency, and seamless scalability, Arm CPUs are perfectly suited to pair with accelerators for the most demanding AI and cloud workloads.

Arm and AWS

Discover how Arm-based AWS Graviton processors are transforming cloud computing with leading price performance and efficiency for AI and cloud-native workloads, now powering over 50% of AWS recent CPU capacity.

Explore Arm and AWS

Explore how Axion, the first Google Cloud custom Arm-based CPU, is advancing performance and efficiency for AI and cloud workloads, with up to 2x better performance than current x86 instances.

Explore Arm and Google Cloud
Arm and NVIDIA

Discover how Arm’s power-efficient compute platform has become a key element in NVIDIA accelerated computing platforms, including the Grace CPU family, delivering up to a 10x performance leap in AI tasks.

Explore Arm and NVIDIA
Compute Platform

Powerful AI/ML Performance with Arm Neoverse

Designed to handle demanding AI workloads efficiently, Arm Neoverse CPUs deliver high throughput, power efficiency, and low TCO—making them ideal when CPUs are the practical choice. From recommendation engines and language model inference to retrieval-augmented generation (RAG), Neoverse scales across a broad range of AI applications.

Performance

Up to 3x better recommendation model performance on Google Axion vs. x86 2.

cost saving

Up to 2.5x higher AI inference throughput with 64% cost savings compared to x86 alternatives 5.

Ecosystem

Broad hyperscaler adoption and multi-cloud availability.

Explore Arm Neoverse for AI WorkloadsLearn About AI/ML on CPU

Arm Compute Platform for Every AI Workload

As AI progresses from classic machine learning to generative AI and now agentic models, workloads are becoming increasingly compute and power intensive. Meeting these demands requires a shift to heterogeneous infrastructure which enables systems to dynamically match each workload with the right processor, optimizing for performance, power efficiency, and cost.

 

Arm Neoverse CPUs provide a power-efficient, scalable compute platform that integrates seamlessly with GPUs, NPUs and custom accelerators and delivers increased performance, flexibility, efficiency, and scalability.

Explore Heterogeneous Computing Solutions
Software and Developer Tools

Optimize AI Workloads with Arm Software and Tools

Developers need optimized tools to deploy AI quickly and efficiently with little effort. The Arm software ecosystem—including Arm Kleidi libraries and broad framework support—helps accelerate time to deployment and boost AI workload performance across cloud and edge.

Resources

Latest News and Resources

  • NEWS and BLOGS
  • Report
  • Podcasts
AI in Datacenters

The Dawn of a New Era for Arm in the Datacenter

Industry analyst Ben Bajarin explores how AI is redefining datacenter architecture and why Arm is emerging as a key player in powering scalable, efficient infrastructure for the AI era.

Podcast icon
AI in Datacenters

Arm and NVIDIA Redefine AI in Datacenters

Listen to our podcast with NVIDIA to explore how our partnership is transforming enterprise computing.

Podcast icon
AI in Datacenters

The Future of AI Infrastructure with Arm and Industry Expert Matt Griffin

Hear Arm and Matt Griffin, founder of the 311 Institute, discuss emerging AI infrastructure trends, challenges in scaling compute, and how Arm is enabling efficient, sustainable AI from cloud to edge.

Frequently Asked Questions: AI in the Datacenter

1. What makes Arm ideal for AI in datacenters?
  • Power-efficient performance: Arm Neoverse CPUs deliver industry-leading performance-per-watt, reducing energy costs and improving operational efficiency.
  • Lower total cost of ownership (TCO): Scalable architectures optimized for modern AI workloads help businesses reduce infrastructure spend.
  • Flexible, workload-optimized systems: Arm-based platforms seamlessly integrate with GPUs, NPUs, and custom accelerators to deliver the right compute for every AI task.
  • Trusted by hyperscalers: —underscoring growing confidence in Arm for large-scale AI deployment.
  • Unified AI infrastructure: A mature software ecosystem and broad adoption support seamless integration across diverse compute engines in cloud and datacenter environments
 2. How do Arm-based platforms enhance AI performance and reduce cloud costs across industry partners like NVIDIA, Google Cloud, and AWS?

Arm-based platforms boost AI performance and efficiency at scale:


  • NVIDIA: Up to (GPT-65B) with Arm CPUs + Grace Hopper compared to x86-based systems.
  • Google Cloud: When compared to x86-based alternatives, .
  • AWS: Graviton CPUs, built on Arm, power over , offering industry-leading price-performance and energy efficiency.

Together, these innovations enable faster, more cost-effective AI across cloud and hyperscale platforms.

 3. What tools does Arm offer to developers for AI workloads?

Developers can accelerate workloads using:


  • Arm Kleidi Libraries
  • Optimized frameworks and toolchains
  • Migration tutorials and learning paths for cloud/server development

保持联繫

註册帳號，以接收有關 Arm Neoverse 與生態系的最新消息。

1.
2.
3.
4.
5.