Arm and Meta Join Forces for Comprehensive Collaboration Across AI Software and Data Centers

Joint Hardware and Software Co-Design
Combining Arm Computing with Meta Infrastructure
Building and Deploying AI Through Open-Source Projects

British semiconductor design company Arm is strengthening its strategic partnership with U.S. technology firm Meta to expand AI efficiency across the entire computing spectrum, from AI software to data center infrastructure.

On the 17th, the two companies announced that this collaboration aims to deliver an enhanced AI experience to users worldwide. They will optimize AI performance across all computing environments powering Meta's global platforms, ranging from milliwatt-level low-power devices supporting on-device intelligence (proprietary AI computation technology) to megawatt-scale large systems responsible for training cutting-edge AI models.

This enhanced strategic partnership builds on years of joint hardware and software co-design between the two companies. By combining Arm’s power-efficient AI computing technology with Meta’s AI products, infrastructure, and open-source technology innovations, the goal is to significantly improve both performance and efficiency.

Santosh Janardhan, Head of Infrastructure at Meta, stated, “From platform experiences to the devices we build, AI is transforming the way people connect and create. Our partnership with Arm enables us to deliver innovation efficiently to more than three billion users of Meta applications and technologies.”

Rene Haas, CEO of Arm, said, “The next era of AI will be defined by achieving massive efficiency. Through our collaboration with Meta, we will combine Arm’s performance-per-watt advantage with Meta’s AI innovation capabilities to enable smarter and more efficient intelligence across everything from milliwatt-level to megawatt-level systems.”

Meta’s AI-powered ranking and recommendation systems, which support search and personalization across Meta’s suite of applications such as Facebook and Instagram, will utilize Arm Neoverse-based data center platforms to deliver high performance and low power consumption compared to x86 systems. Arm Neoverse helps Meta achieve performance efficiency across its infrastructure and enhances the efficiency and scalability of Arm computing in hyperscale environments.

The two companies have closely collaborated to optimize Meta’s AI infrastructure software stack for the Arm architecture, from compilers and libraries to major AI frameworks. This includes joint tuning of open-source components such as Facebook General Matrix Multiplication (FBGEMM) and PyTorch. By actively leveraging Arm’s vector extension features and performance libraries, they have achieved measurable improvements in inference efficiency and throughput. These optimizations are being provided to the open-source community, expanding their influence across the global AI ecosystem.

This partnership further strengthens AI software optimization across the PyTorch machine learning framework, the ExecuTorch edge inference runtime engine, and the vLLM data center inference engine. In particular, by advancing the ExecuTorch foundation optimized with Arm KlediAI, they plan to enhance efficiency on billions of devices and simplify model deployment, thereby accelerating AI application performance from the edge to the cloud.

These open-source technology projects are central to Meta’s AI strategy, enabling a wide range of AI development and deployment from recommendation systems to conversational intelligence. The two companies plan to continue expanding optimizations for open-source projects, supporting millions of developers worldwide in building and deploying more efficient AI in Arm-based environments.

This content was produced with the assistance of AI translation services.