The rapid expansion of artificial intelligence has shifted attention away from software alone towards the infrastructure required to deploy it at scale. As AI workloads become larger and more computationally demanding, organisations are increasingly confronting practical constraints around power density, cooling capacity and operational complexity.
Against that backdrop, Schneider Electric and AMD have released a jointly developed reference design for the AMD Helios rackscale solution, intended to provide data centre operators with a validated blueprint for deploying high-density AI environments more quickly and with reduced implementation risk.
Rather than introducing a new AI platform, the companies are focusing on one of the industry’s emerging priorities: reducing the uncertainty involved in building what are increasingly being described as AI factories. These facilities combine high-performance computing, networking, power and cooling infrastructure to support large-scale artificial intelligence workloads.
The reference design represents the first milestone in the companies’ collaboration and supports the AMD Helios rackscale solution, powered by AMD Instinct MI455X GPUs, sixth-generation AMD EPYC CPUs, AMD Pensando Vulcano network interface cards and the ROCm software ecosystem.
Building AI infrastructure as an integrated system
One of the clearest messages emerging from the announcement is that AI infrastructure can no longer be designed in isolation.
As AI clusters become denser, the interaction between computing hardware, electrical systems and cooling technology has become increasingly important. The reference design provides guidance across four technical areas: facility power, facility cooling, IT space and lifecycle software, with the aim of shortening planning cycles by defining how these elements should be organised into a scalable deployment.
The design supports AI racks with power densities of up to 246kW and modular AI clusters delivering up to 10.4MW of IT capacity for large-scale deployments. It also incorporates advanced liquid cooling using Motivair by Schneider Electric cooling distribution units alongside hybrid air and liquid cooling approaches capable of removing up to 84 per cent of generated heat.
The announcement also reflects the growing role of digital modelling in infrastructure planning. Electrical and thermal designs are validated using ETAP and EcoStruxure IT Design computational fluid dynamics simulation tools, while integrated Electrical Digital Twin capabilities enable infrastructure performance to be modelled, analysed and managed. The design also includes support from AVEVA’s Unified Operations Center to provide operational visibility and real-time monitoring.
Collectively, these technologies are intended to support AI-driven predictive maintenance and system-level optimisation across power, cooling and IT infrastructure rather than treating each discipline independently.
From planning to deployment
Reference designs have become increasingly important as organisations move beyond AI experimentation towards production-scale deployments.
Instead of developing every project from first principles, operators can begin with engineering models that have already been validated against expected power densities, thermal loads and operational requirements. Schneider Electric argues that this reduces both planning time and deployment risk.
Manish Kumar, Executive Vice President, Secure Power & Data Centers at Schneider Electric, said organisations now require “comprehensive, AI-ready reference designs” that support the journey from planning through deployment.
He said the collaboration with AMD delivers “an engineering-backed reference design that bridges the gap between advanced AI compute platforms, energy tech, and real-world data centre implementation”, enabling customers to deploy scalable, high-density AI environments with greater confidence, efficiency and speed.
AMD also framed the announcement as evidence that AI infrastructure is evolving beyond individual servers towards complete AI factories.
Forrest Norrod, executive vice president and general manager of AMD’s Data Center Solutions Business Group, said compute, networking, power and cooling increasingly need to be designed together from the outset rather than integrated later. He said the validated reference design provides customers with “a practical blueprint to accelerate high-density AI deployments, reduce integration risk and scale with greater confidence and efficiency.”
The reference design has been validated to ANSI standards for US deployments, with plans to extend the framework to support IEC standards for global implementations.
Perhaps the broader significance lies less in the individual technologies than in the industry’s changing priorities. As artificial intelligence moves from research projects to operational infrastructure, competitive advantage is likely to depend as much on the engineering of the physical environment as the capability of the AI hardware it contains. In that sense, the AI factory is becoming an infrastructure challenge first and a computing challenge second.
