Arm Joins OCP Board, Leading Open Standards for Integrated AI Data Centers

Standing Shoulder to Shoulder with AMD and Nvidia
Leading Open Standards for Integrated AI Data Centers
FCSA Specifications Provided to OCP... Driving Innovation in the Computing Stack

Arm, the UK-based semiconductor design specialist (fabless), announced on October 21 that it will join the board of the Open Compute Project (OCP) alongside global companies such as AMD and Nvidia.

Arm Logo

As a result, Arm will be granted the qualifications and responsibilities of an OCP board member and will work with companies such as Meta, Google, Intel, and Microsoft to advance open and interoperable designs for artificial intelligence (AI) data centers.

Mohamed Awad, Senior Vice President and General Manager of Arm’s Infrastructure Line of Business, emphasized, "The AI economy is reshaping computing infrastructure, demanding unprecedented performance, efficiency, and scalability from cloud to edge."

He continued, "Data centers are undergoing an unprecedented transformation from general-purpose servers to AI-dedicated rack systems and large-scale clusters. At the same time, there is a significant power challenge. As of 2025, a single AI rack will consume as much electricity as 100 American households while delivering the performance level of the top supercomputers in 2020. To address these challenges, open collaboration across the rapidly evolving ecosystem, along with next-generation infrastructure, is essential."

Currently, Arm views integrated AI data centers as next-generation infrastructure and is pursuing the necessary business and technology development to maximize AI computing per unit area, thereby reducing power consumption and costs. Building next-generation infrastructure requires joint design across computing, accelerators, memory, storage, and networking. Arm has also introduced its Arm® Neoverse™ technology, which serves as a core technology at every layer of the AI stack, enabling leading AI companies to optimize everything from data tokenization to running AI models and agents, and to generate real-world impact through applications in scientific, medical, and commercial fields.

It is known that integrated AI data centers are not powered by general-purpose chips. To increase the density of such systems, advanced dedicated silicon is required. The 'chiplet' approach, which combines multiple chips into one, addresses this need through in-package integration and 2.5D/3D technologies, providing new opportunities for multiple vendors to jointly design core functions.

Accordingly, Arm also announced that it will contribute the FCSA specifications to the OCP and strengthen industry collaboration related to chiplets. Based on Arm’s Chiplet System Architecture (CSA) work, FCSA meets industry demand for an open framework that is not tied to any specific company or CPU architecture. By providing common standards for chiplet systems and interface definitions, it accelerates chiplet design and integration regardless of CPU architecture, enabling large-scale reuse and interoperability.

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