AMD and Schneider Electric have produced a reference design for data centres built around AMD’s Helios AI racks, with configurations reaching 246kW per rack. The plans give operators a detailed starting point for the power and liquid-cooling demands of dense AI clusters, rather than leaving every project to begin with a blank sheet.
AMD Watch analysis
What happened
The designs cover new-build AI facilities and retrofits, with equipment lists, physical layouts and guidance on electrical systems and cooling. Data Center Dynamics reports that the largest rack configuration reaches 246kW, and that the designs include liquid-cooled AI pods alongside air-cooled racks.
The examples specify redundant power arrangements for IT loads and cooling equipment. The companies describe the designs as validated and say they can help shorten planning cycles and reduce deployment risks. Those are the partners’ claims; the practical value will depend on how well the blueprints translate to individual sites.
Why it matters
At these power densities, fitting servers into a room is only the beginning. Operators also need to plan electrical capacity, cooling and the physical infrastructure that keeps a cluster running. A shared reference design could make early planning less bespoke, particularly for organisations adapting existing facilities to workloads they were not built to serve.
It also makes the demands visible: 246kW in a rack is a serious power and heat-management problem, not a modest equipment refresh. The Data Center Dynamics report details the designs and example configurations.
Our read
This is the less glamorous side of AI progress, where cooling pipes and power trains get a starring role. The useful promise is not that a blueprint removes the hard work, but that operators can start from a concrete design rather than reinventing the basics. The real test is whether it saves time and delivers reliable, efficient deployments outside the drawing board.
What to watch
- Whether operators adopt the designs for real Helios deployments.
- How the layouts adapt to the constraints of existing data centres.
- What measured power use and cooling performance look like in operation.
- Whether the plans are updated as rack configurations and site requirements change.
Discussion spark: Should AI rack designs be standardised around shared reference blueprints, or do the differences between data centres make bespoke engineering the safer bet?
Sources and evidence
Independent WittyWires tracker for public updates about AMD. Not affiliated with or endorsed by AMD; this is not an official account.