Featured image: Electricity pylons illustrate the wider power system serving data centres. Contextual photograph; not Schneider equipment or an installation identified in its results. Photo: Matthew Henry / Unsplash. Unsplash licence.
AI’s construction boom reaches well beyond the companies selling processors. Schneider Electric’s latest update puts electrical distribution, backup power and cooling near the centre of that expansion.
In its 16 October 2026 third-quarter revenue release, Schneider reported sales of €11.46 billion, up 16.9% on its organic measure of comparable growth. It also reported triple-digit growth in data-centre demand. Those are supplier-reported business figures, not a count of completed AI facilities.
The important infrastructure story is what customers are buying around the chips. Computing capacity becomes useful only when the surrounding physical systems can supply it, cool it and keep it operating together.
The supporting equipment has become a major system
Schneider’s release points to data-centre growth in prefabricated solutions, three-phase uninterruptible power supplies and cooling. An uninterruptible power supply, or UPS, provides immediate electrical support during disturbances while the wider backup system responds.
These systems have different jobs from the AI accelerator itself. Distribution equipment routes and controls electricity. Cooling removes heat. Monitoring helps operators understand whether the electrical and mechanical equipment is behaving as intended.
A rise in demand across these categories is therefore evidence of a broader infrastructure supply chain serving AI. It does not show that each ordered component has been installed, commissioned or used at its intended capacity.
Schneider also described a record backlog. Backlog is future work awaiting delivery, not a separate measure of functioning compute. Reading it alongside revenue helps avoid treating every commercial commitment as capacity that is already available.
Cooling architecture depends on the building
Schneider’s Reference Design 100, released in its current version in March 2026, illustrates why the surrounding building matters. It describes air-cooled and liquid-cooled AI configurations, including retrofit scenarios beside existing conventional IT equipment.
In one arrangement, a coolant distribution unit transfers heat from a liquid-cooled computing system to air. Another transfers heat into a facility water circuit. The design distinguishes sites where suitable water systems are available from those where they are not.
That does not mean either arrangement is universally better. It means an equipment decision depends on the facility receiving it. A liquid-cooled cluster is not specified completely by saying that liquid reaches the processors.
The reference design covers facility power, cooling, IT space and lifecycle software. It is background engineering material, rather than a new product announcement in the quarterly release or independent proof that every installation achieves a particular saving.
Efficiency needs a measured boundary
The US Department of Energy’s July 2024 design guide treats IT efficiency, environmental conditions, air management, cooling, electrical systems and heat recovery as connected decisions. Improving one area can reduce the work required from others.
For example, reducing the heat that must be removed can change the demands on supporting equipment. But a component’s efficiency rating is not the same as a measured reduction for an entire operating site. The load, configuration and comparison point matter.
This is why the most revealing follow-up to an infrastructure announcement is often an operating record. Useful evidence includes how much computing the site delivers, its total electricity use and how the result varies at different loads.
Supplier growth and delivered compute answer different questions
Schneider’s update is a concrete commercial signal from the companies building AI’s physical surroundings. It is more substantial than a speculative campus rendering, because the group is reporting recognised quarterly revenue.
It still covers a business serving several end markets. The total sales figure is not all AI revenue, and growth in one supplier’s demand cannot be converted directly into global AI energy consumption.
The distinction makes the story more useful. Revenue shows business delivered under the company’s reporting framework; backlog indicates further work; commissioned facilities show physical delivery; operating measurements show what those facilities actually do.
AI’s expansion needs all of them. The new results underline the scale of activity around the chips, while the test for the wider buildout remains whether those electrical and cooling systems become reliable, well-used computing infrastructure.
Reporting and source checks completed on 16 October 2026. By FutureTechDose Editor.
Leave a Reply