Research Finder
Find by Keyword
What If the Future of AI “Factory” Data Centers Wasn’t Proprietary at All?
AMD Delivers on Meta’s Open Rack Wide Specification with Helios, a Double-Wide Rack Designed for Large-Scale AI Infrastructure
Key Highlights:
- Meta introduced the Open Rack Wide (ORW) specification, an open, double-wide rack designed for the power, cooling, and serviceability demands of next generation AI systems
- Bringing this vision to life, AMD announced its Helios reference system, representing a milestone in OCP's evolution extending to rack-scale infrastructure for AI
- Built on the AMD Instinct MI450 Series GPUs, Helios presents a credible alternative for hyperscale AI build-outs
- AMD claims Helios delivers up to 17.9× higher performance compared to previous generations while projecting 50% more memory capacity and bandwidth than NVIDIA’s forthcoming Vera Rubin system.
The News
At the 2025 Open Compute Project (OCP) Global Summit in San Jose, Meta introduced specifications for a new Open Rack for AI featuring a double-wide Open Rack Wide (ORW) form factor. This open, rack-scale standard is designed to meet the mechanical, electrical, and cooling demands of large-scale AI infrastructure.
AMD unveiled Helios, its most advanced rack-scale reference system, built on ORW standards. Helios translates Meta’s open-rack blueprint into a physical, deployable platform for both scale-out and scale-up AI clusters.
Analyst Take
It’s easy to say that open standards such as ORW are essential for sustainable innovation, but openness alone is insufficient. Success depends on broad support from both the vendor community and adoption by real-world customers. As with all open standards, history shows that OCP gains momentum when the ecosystem is aligned from silicon to software to integrator.
The Open Compute Project was launched in 2011 by Facebook (now Meta), Intel, and Rackspace, bringing open-source principles to data-center hardware. Over the past decade, OCP has expanded from servers and networking into complete rack and power systems. Today OCP counts 400+ member companies, including Meta, Microsoft, Google, Intel, AMD, Dell, NVIDIA, HPE, and Supermicro. Growth forecasts reflect expansion of OCP-compliant hardware from hyperscalers into enterprise and edge deployments.
With the Helios announcement, AMD asserts itself as a contender at the rack level, bridging the gap between chip, interconnect, and rack-level infrastructure. We view this as a direct challenge to NVIDIA’s vertically integrated approach, and AMD’s performance claims are aggressive. As we see the industry evolving from data centers to “AI factories,” Helios helps position AMD as a legitimate alternative for hyperscale AI build-outs.
Meta's unveiling of the ORW specification, coupled with AMD's realization of it in their Helios platform, signifies a pivotal moment for the Open Compute Project (OCP), as well. This represents both a technical demonstration and proof that open collaboration can yield rack-scale systems competitive with proprietary designs. In our view, this move will influence considerations when building, operating and scaling high-performance computing and AI data centers.
The performance claims associated with Helios include an impressive 1.4 exaFLOPS in FP8 precision and a substantial 31 TB of HBM4 memory. This should present a compelling choice for data center operators seeking to diversify their AI hardware ecosystems and potentially reduce vendor lock-in. At scale, standardization and resource efficiency are critically important to data center design and ongoing operations.
Vendors–including chip manufacturers, system integrators, and component suppliers–must wholeheartedly commit to developing and delivering interoperable implementations of these open standards. This commitment extends to the timely contribution of reference designs and intellectual property back to the OCP community, enriching the ecosystem for everyone. Simultaneously, hyperscalers and large enterprises actively participate in validating, deploying, and providing constructive feedback on these open solutions. Without this comprehensive alignment, even the most promising open standards see limited uptake in niche applications.
What Was Announced
AMD and Meta jointly rolled out a next-generation open‐rack specification at the OCP 2025 summit: Open Rack Wide (ORW), a double-wide rack form factor optimized for AI-scale power, cooling, and serviceability. AMD’s reference implementation, called Helios, brings the ORW to reality in a deployable rack.
Helios is centered on AMD’s upcoming Instinct MI450 Series GPUs. In a full rack of 72 these accelerators, the system is projected to deliver up to 1.4 exaFLOPS FP8 (and 2.9 exaFLOPS FP4), along with 31 TB of HBM4 memory and 1.4 PB/s aggregate memory bandwidth. It also supports 260 TB/s internal scale-up interconnect and 43 TB/s Ethernet scale-out throughput. In its announcement, AMD claims industry-leading capacity and bandwidth for data-hungry AI models.
From a systems engineering standpoint, Helios is targeted at dense AI deployments: a double-wide layout focused on weight density, liquid cooling with backside quick-disconnect architecture, and standards-based Ethernet for scale-out interoperability. AMD positions it as the first rack-scale design optimized for resource-intensive frontier AI workloads, with emphasis on combining performance, flexibility, and openness.
Helios is being released to OEMs and ODMs as a reference design in 2025, with volume deployments expected in 2026. AMD says its partners can integrate differentiated solutions with AMD Instinct GPUs, EPYC CPUs, and Pensando DPUs. By aligning with open standards like ORW, AMD aims to catalyze ecosystem adoption, reduce fragmentation, and allow partners to build differentiated systems without reinventing foundational infrastructure.
Looking Ahead
Meta’s ORW initiative, in conjunction with AMD’s Helios system, signifies a pivotal transformation in the approach to AI infrastructure strategy. This shift moves beyond the traditional focus on GPU-centric performance scaling, instead emphasizing rack-level system optimization. This broader perspective encompasses crucial elements such as power delivery, advanced cooling mechanisms, and interoperability between components within the rack. The efficiency of the entire system, rather than just individual processing units, dictates overall AI workload performance.
We will be watching several key developments over the next year that will shape the future of this evolving infrastructure, including:
The rate at which major hardware manufacturers and ODMs integrate the standard. Kudos to AMD for being first out of the gate. We will look for announcements from industry leaders such as Dell, HPE, Supermicro, and others, detailing their own plans for ORW-compliant rack solutions. This widespread adoption will be crucial for establishing ORW as a de facto standard.
Robust software support: We expect advancements in orchestration and telemetry standards for efficient management and monitoring of ORW-compliant systems. Effective software integration can automate resource allocation, predict maintenance needs, and help optimize system performance.
Independent validation of claimed performance and benefits: This will involve benchmarking to provide concrete data on power efficiency, scaling across various workloads, and thermal performance under demanding conditions. Such independent data can help build trust among potential adopters, including vendors and end users alike.
OCP helping to formalize and expand the ORW initiative: This may include the formation of dedicated sub-projects within OCP focusing on specialized aspects of rack design. These could include specific guidelines and standards for advanced liquid cooling solutions, optimized power delivery architectures, and standardized rack-management APIs, fostering a more collaborative and open development environment. Meta and AMD might consider seeding grants to drive ORW adoption, and use co-piloted Helios racks at smaller scale to drive use case analysis and benchmarking.
We will be watching for the broader OCP community and marquee early adopters to embrace ORW. All the other players have to do is delay their adoption, while doubling down on proprietary offerings, to threaten the emerging open standard. In an industry where so much is custom (right down to the silicon) openness will only succeed if the cost/benefit is compelling. Really compelling. If other players follow AMD and Meta into the standard, then the companies will have helped write a new chapter in data center planning and design built
for the AI era.
We envision openness, collaboration, and standardized approaches, sharing floor space with proprietary control and isolated systems, as the defining characteristics of future high-performance computing environments. They can coexist. Organizations will need to find performance, security, compliance and cost reasons to justify building and maintaining both environments to meet rapidly evolving
business needs.
Don Gentile | Analyst-in-Residence -- Storage & Data Resiliency
Don Gentile brings three decades of experience turning complex enterprise technologies into clear, differentiated narratives that drive competitive relevance and market leadership. He has helped shape iconic infrastructure platforms including IBM z16 and z17 mainframes, HPE ProLiant servers, and HPE GreenLake — guiding strategies that connect technology innovation with customer needs and fast-moving market dynamics.
His current focus spans flash storage, storage area networking, hyperconverged infrastructure (HCI), software-defined storage (SDS), hybrid cloud storage, Ceph/open source, cyber resiliency, and emerging models for integrating AI workloads across storage and compute. By applying deep knowledge of infrastructure technologies with proven skills in positioning, content strategy, and thought leadership, Don helps vendors sharpen their story, differentiate their offerings, and achieve stronger competitive standing across business, media, and technical audiences.
Stephen Sopko | Analyst-in-Residence – Semiconductors & Deep Tech
Stephen Sopko is an Analyst-in-Residence specializing in semiconductors and the deep technologies powering today’s innovation ecosystem. With decades of executive experience spanning Fortune 100, government, and startups, he provides actionable insights by connecting market trends and cutting-edge technologies to business outcomes.
Stephen’s expertise in analyzing the entire buyer’s journey, from technology acquisition to implementation, was refined during his tenure as co-founder and COO of Palisade Compliance, where he helped Fortune 500 clients optimize technology investments. His ability to identify opportunities at the intersection of semiconductors, emerging technologies, and enterprise needs makes him a sought-after advisor to stakeholders navigating complex decisions.