Arm Sets the Standard for Open, Converged AI Data Centers

By Mohamed Awad, SVP and GM of the Infrastructure Business, Arm

News highlights: 

• OCP appoints Arm to Board of Directors alongside AMD and NVIDIA, underscoring leadership role in defining open standards for converged AI data centers 
• Arm contributes vendor-neutral Foundation Chiplet System Architecture specification to the Open Compute Project, with plans to drive further innovation up the compute stack 
• Arm Total Design ecosystem adds new capabilities to advance chiplet supply chain for AI innovation, tripling in size since its 2023 launch

October 14, 2025 -- The AI economy is reshaping compute infrastructure, requiring unprecedented performance, efficiency, and scale from cloud to edge. Data centers are undergoing their biggest shift yet, moving from commodity servers to rack-level systems and large-scale clusters designed specifically for AI. But we face a massive power challenge: A single AI rack in 2025 is using the same amount of power as ~100 U.S. households, and delivering as much compute as a top supercomputer from 2020. 

Arm sees the converged AI data center as the next phase of infrastructure, maximizing AI compute per unit of area to reduce the volume of power and related costs associated with powering AI. To build this next phase of infrastructure requires co-designed capabilities across compute, acceleration, memory, storage, networking and beyond. Arm Neoverse is now the core technology at each layer of the AI stack, enabling frontier AI providers to optimize how data becomes tokens, how tokens fuel advanced AI models and agents, and how AI delivers real impact through applications in science, medicine, and commerce. 

While significant progress has been made, there is still much to be done as the industry reconfigures for this challenge. Open collaboration across a fast-moving ecosystem is vital to building the next phase of infrastructure, and that’s why I’m thrilled that OCP has appointed Arm to its Board of Directors, along with leaders from AMD and NVIDIA. This appointment underscores Arm’s unique ability to drive openness and industry standards that shape the future of AI data centers. With nearly 50% of compute shipped to top hyperscalers this year powered by Arm, this is a fantastic opportunity to share what we’ve learned as the most pervasive compute platform spanning everything from the smallest sensor to the most powerful data centers.   

As part of the OCP board, we’ll work on advancing open, interoperable design for the AI data center alongside key influential companies in computing like Meta, Google, Intel, and Microsoft. Power demand is rising fast and we believe Arm’s efficiency leadership and central role in the compute ecosystem will be key to making AI infrastructure investments more sustainable and impactful. 

Scaling the silicon supply chain for AI  

The converged AI data center won’t be powered by individual commodity chips. To increase density of these systems, advanced purpose-built silicon is required. Chiplets provide a path through in-package integration, 2.5D and 3D technologies, while opening new opportunities for multi-vendor co-design of critical capabilities. Today we’re taking a step toward deeper industry collaboration around chiplets by contributing the Foundation Chiplet System Architecture (FCSA) specification to the Open Compute Project (OCP). 

FCSA leverages our ongoing work with the Arm Chiplet System Architecture (CSA) but addresses an industry demand for a framework that aligns to vendor- and CPU architecture-neutral requirements. It delivers a common set of standards for chiplet system and interface definitions, accelerating chiplet design and integration while enabling reuse and interoperability at scale – regardless of CPU architecture. 

Today’s contribution of FCSA is a testament to the progress made by the Arm Total Design ecosystem since it was introduced in 2023. In that time, partners have begun delivering the first chiplets to market aligned with the Arm CSA, with many more projects in development, laying the groundwork for the next phase of open chiplet standards. Building on that success, we’re expanding the Arm Total Design ecosystem with 10 new partners: Alchip, ASE, Astera Labs, CoAsia, Credo, Eliyan, Insyde Software, Marvell, Rebellions, and VIA NEXT. Their combined expertise in advanced packaging, interconnect, and system integration will help drive the next phase of standards and specification work, accelerating chiplet innovation across the silicon design lifecycle from IP and EDA tools to manufacturing, packaging, and validation.  

Charting the path to the converged AI data center 

Moving forward, I am most excited about the opportunity that comes with joining the OCP Board. We’re already active in OCP workstreams on firmware, manageability, and server hardware design. In fact, just this week we joined the collaborative effort to advance Ethernet technologies for large-scale AI called ESUN under the OCP Networking Project. We’re committed to helping solve how we scale AI efficiently in every deployment, from milliwatts to gigawatts in the data center. The contribution of FCSA to OCP is just the beginning.  

Today is another step toward an AI-ready foundation that is powerful, open, and accessible to all. The value of open, shared standards extends well beyond the data center, driving innovation in markets where full system co-design is essential such as the AI-defined vehicle in automotive. We’ll have more to share on that specific example soon!

Supporting partner quotes:

“Our customers are embracing open, collaborative approaches to AI infrastructure. By joining the Arm Total Design ecosystem and aligning our purpose-built silicon and software with open standards like OCP’s Foundation Chiplet System Architecture, we’re helping partners accelerate time-to-market while unlocking higher bandwidth, lower latency connectivity—scaling XPUs seamlessly from systems to racks and beyond.” – Chris Petersen, Fellow of Technology and Solution Architecture, Astera Labs

“The industry is increasingly adopting chiplet-based architectures to accelerate time to market, reduce engineering effort, and scale efficiently amid unprecedented design complexity. In addition to EDA and IP, Cadence has a long history of collaboration with Arm on the CSA specification and chiplets, as demonstrated by our silicon-proven Physical AI System Chiplet and Cadence Chiplet Framework. We fully support Arm’s contribution to OCP, which represents a significant step forward in unifying chiplet architectures around a common, open standard. We are committed to the evolution of the new Foundation Chiplet Architecture as we continue to drive the transition to chiplet-based designs and an open chiplet marketplace.” – David Glasco, vice president of the Compute Solutions Group at Cadence

“Cisco welcomes Arm’s contribution of the Foundation Chiplet System Architecture (FCSA) specification to the Open Compute Project. Collaborative initiatives like this help accelerate the pace of industry-wide innovation in the AI era.” – Mohammad Issa, SVP of Engineering for Cisco Silicon One

“As the AI infrastructure build-out continues to ramp at unprecedented velocity, we see intensifying demand for custom silicon solutions across the data center. By joining Arm Total Design, with Marvell’s industry leadership position in chiplets and custom solutions supporting Arm’s Foundation Chiplet System Architecture, we can advance towards our shared vision of providing high-performance, efficient XPU and XPU attach accelerated computing solutions wherever needed.” – Will Chu, senior vice president and general manager, Custom Cloud Solutions at Marvell