On October 15 Eastern Time, Arm Holdings announced on its official website that it had reached a strategic partnership with Meta Platforms to comprehensively enhance AI computing efficiency. Below is the full text of the press release.
Key News: The strategic partnership combines Arm's leadership in energy-efficient computing with Meta's innovation capabilities in infrastructure, AI products, and open technologies, aiming to deliver richer and more accessible AI experiences to billions of users worldwide. Meta’s foundational AI software technologies, including PyTorch, have now been optimized for the Arm architecture. Specifically, PyTorch’s Executorch runtime leverages Arm KleidiAI technology to achieve maximum energy efficiency (performance per watt) for Meta and the global open-source community.
$Arm Holdings (ARM.US)$ and $Meta Platforms (META.US)$ Announced the establishment of a strategic partnership aimed at enhancing AI efficiency across all layers of the computing stack, covering AI software and data center infrastructure, with the goal of delivering richer experiences to billions of users globally. The collaboration spans from milliwatt devices that enable on-device intelligence to megawatt systems used to train some of the world’s most advanced AI models, empowering various types of computing, workloads, and experiences to support Meta Platforms' global operations.
This multi-year collaboration builds on the ongoing co-design efforts of both companies in hardware and software, combining Arm Holdings' leadership in energy-efficient AI computing with Meta Platforms' innovative capabilities in AI-driven products, infrastructure, and open technologies to achieve significant improvements in performance and efficiency.
Santosh Janardhan, Head of Infrastructure at Meta Platforms, stated: "From platform experiences to the devices we build, AI is transforming how people connect and create. Partnering with Arm Holdings enables us to efficiently scale these innovations to over three billion users of Meta Platforms’ applications and technologies."
Rene Haas, CEO of Arm Holdings, noted: "The next phase of AI will be defined by scalable efficiency. By partnering with Meta Platforms, we are combining Arm Holdings' leadership in performance-per-watt with Meta Platforms' AI innovations to deliver smarter, more efficient artificial intelligence—from milliwatt devices to megawatt systems—everywhere."
AI Scaling in the Cloud
Meta's AI ranking and recommendation system, which powers the discovery and personalization features of Meta's suite of applications, including Facebook and Instagram, will adopt an Arm Neoverse-based data center platform. This move aims to achieve higher performance and lower power consumption compared to x86 systems. Arm Neoverse will also help Meta meet its per-watt performance targets across its overall infrastructure, highlighting the efficiency and scalability of Arm-based computing in hyperscale scenarios.
Through close collaboration, Meta’s AI infrastructure software stack has been optimized for the Arm architecture, spanning compilers, libraries, and major AI frameworks. This includes joint tuning of open-source components such as Facebook General Matrix Multiplication (FBGEMM) and PyTorch, leveraging Arm's vector extensions and performance libraries. These efforts have yielded measurable improvements in inference efficiency and throughput. The optimizations are being contributed back to the open-source community to expand their impact within the global AI ecosystem.
Accelerating AI Software from Cloud to Edge
The collaboration further deepens both parties' efforts in AI software optimization, covering the PyTorch machine learning framework, the Executorch edge inference runtime engine, and the vLLM data center inference engine. Plans include further enhancements to Executorch, which is already optimized for Arm KleidiAI, to improve efficiency across billions of devices. This partnership will jointly accelerate the ease of deploying AI models and enhance AI application performance from edge to cloud.
These open-source technology initiatives are central to Meta’s AI strategy, supporting the development and deployment of various AI capabilities ranging from recommendation systems to conversational intelligence. Both parties intend to continue extending these optimizations into future open-source projects, enabling millions of developers worldwide to build and deploy efficient AI on Arm architectures anywhere.
Driving the Future of AI Together
From megawatt-scale data centers to foundational AI software, the collaboration between Arm and Meta represents a full-stack partnership aimed at scaling AI efficiency across every level of the compute stack, ushering in a new era of intelligent, efficient, and connected experiences for billions of users globally.
Editor/Joryn
