Schneider Electric and AMD have released a jointly developed reference design intended to serve as a blueprint for data center operators deploying high-density AI infrastructure.

The design is built around AMD's Helios rackscale platform and is capable of supporting AI racks with power densities of up to 246kW, as well as modular, multi-cluster environments with up to 10.4MW of IT capacity. To put that density figure in perspective, traditional enterprise data center racks typically operate in the 5–15kW range, meaning the blueprint targets an order-of-magnitude increase in per-rack power draw driven by GPU-heavy AI training and inference workloads. According to the companies, the reference design can be used both by operators building new AI factories and by those retrofitting high-density AI infrastructure into existing data centers.

Rob Bunger, global director of data center solution architecture at Schneider Electric, said the design arrives as intensifying AI workloads push data center infrastructure to what he described as "unprecedented limits."

"Operators must manage significantly higher power densities, thermal requirements and operational complexity while reducing deployment risk," Bunger told AI Business.

Simulation Tools and Infrastructure Coverage

To address these challenges, the design incorporates Schneider Electric's ETAP and EcoStruxure IT Design CFD simulation tools, enabling operators to test, predict, and manage infrastructure performance before deployment.

The reference design covers four infrastructure domains: facility power, facility cooling, IT space, and lifecycle software. By modeling the physical infrastructure performance of a data center, the companies say the blueprint shortens planning processes by clearly defining how power, cooling, and IT infrastructure should be organized.

"These capabilities help organizations build reliable, scalable AI factories faster with greater confidence, reduced risk and less complexity," Bunger added.

The design also features real-time monitoring and analytics, AI-driven predictive maintenance, and system-level optimization spanning power, cooling, and IT infrastructure.

AI Infrastructure Deployment as an Engineering Challenge

The blueprint's release comes at a time when AI infrastructure deployment is shifting in focus — from securing compute capacity toward integrating complex systems across an entire operation, including power, cooling, networking, and ongoing management. This shift has prompted a wave of vendor collaborations producing pre-validated reference architectures, as individual operators increasingly struggle to integrate rackscale GPU systems, liquid cooling, and high-voltage power distribution on their own.

"The AMD Helios reference design bridges the gap between advanced AI compute platforms, energy tech and real-world data center implementation through a jointly developed and validated blueprint," Bunger said. "While access to power remains a major industry constraint, engineering-backed reference designs enable organizations to move from planning to deployment faster with greater confidence, efficiency and reliability."

Bunger's reference to power as a constraint reflects a widely reported industry challenge: utility interconnection timelines, grid capacity limits, and local permitting have become significant bottlenecks for hyperscale and enterprise AI buildouts across multiple regions.

The design is intended to support modular scaling, accommodating AI clusters of up to 10.4MW of IT capacity for greenfield deployments while also providing infrastructure capable of handling high-density AI workloads in existing facilities.

Standards Validation

The reference design has been validated to ANSI standards for U.S. deployments, with plans to extend the framework to support International Electrotechnical Commission (IEC) standards for global implementations.

Source: AI Business