The artificial-intelligence boom is usually narrated through chips, models and cloud companies. Schneider Electric’s record quarter shows the boom from the power room. Before an accelerator can train a model, electricity has to be transformed, protected, distributed and cooled; racks must be assembled; control systems must keep thousands of components within safe limits. That physical chain has become one of the clearest ways to see AI capital spending outside the semiconductor industry.
Schneider reported third-quarter revenue of €11.46 billion, up 16.9% organically from a year earlier and above the company-provided analyst consensus. Data-centre demand grew at a triple-digit rate in the quarter, according to management, and the company raised its 2026 organic revenue-growth target to 12–14% and adjusted EBITA-growth target to 18–21%.[1]
A record quarter
| Measure | Value | Context |
|---|---|---|
| Group revenue | €11.46 billion | Up 17.9% as reported and 16.9% organically.Source 1 |
| Energy Management revenue | €9.62 billion | The segment supplying much of the power and data-centre infrastructure.Source 1 |
| Industrial Automation revenue | €1.85 billion | A smaller but growing part of the quarter.Source 1 |
| Energy Management share | 83.9% | Our calculation: €9.62 billion ÷ €11.46 billion × 100. Rounding means segment totals need not reconcile perfectly.Source 1 |
| 2026 organic revenue target | 12–14% growth | Raised from 10–13%.Source 1 |
AI demand reaches Schneider through several physical layers
A data centre does not buy “power” as one undifferentiated object. Utility supply must enter a site, be switched and protected, converted to voltages the computing equipment can use, distributed to racks and monitored continuously. High-density accelerators produce heat that has to be removed close to the chips. A delay in any one layer can leave expensive processors waiting for a building that is not ready.
From grid connection to a working AI rack
| Layer | What it does | Why AI density changes the requirement |
|---|---|---|
| Site power | Switchgear, transformers, protection and backup systems bring electricity safely into the facility | Large clusters require more capacity and resilience before the first server is installed.Source 2 |
| Distribution | Power modules, busways, shelves and high-power PDUs move electricity to rows and racks | Higher rack density increases current, cable and conversion challenges.Source 2Source 3 |
| Rack and compute enclosure | Standardised or custom racks hold processors, networking and power equipment | New accelerator platforms may require different mechanical and electrical designs.Source 2 |
| Cooling | Liquid and air systems remove heat from chips, racks and rooms | Power consumed by processors becomes heat that must be carried away continuously.Source 2 |
| Management software | Monitors capacity, temperature, faults and energy use | Operators need to coordinate infrastructure whose loads can change rapidly.Source 2 |
Higher voltage is one response to higher rack power
Schneider’s 800-volt direct-current architecture is aimed at AI racks in roughly the 400-kilowatt to 1-megawatt range. For the same power, raising voltage lowers current. Lower current can reduce conductor size and resistive losses, which matters when electricity must be delivered through dense equipment. It does not remove the need for conversion, protection or safety engineering; it changes where those tasks happen and how the system is designed.[3]
Prefabrication attacks a different bottleneck. Instead of assembling every power room, cooling loop and row from scratch on site, manufacturers can deliver tested modules and pods as repeatable building blocks. Schneider describes configurations spanning racks, power, cooling and management. Standardisation can shorten deployment, but it also makes supplier capacity and interface choices more consequential.[2]
Demand is not the same as durable profit
The quarter is strong evidence that AI construction is reaching electrical and thermal suppliers. It is not proof that every announced data centre will be completed, fully utilised or economically successful. Orders can be delayed; hyperscalers can redesign facilities; power connections can become the limiting factor; and a record backlog can create execution pressure as well as revenue visibility.
The business also extends beyond data centres. Energy Management supplies buildings, factories and infrastructure, while Industrial Automation serves manufacturing and process industries. That diversification matters because it means the €9.62 billion segment cannot be read as one pure AI revenue line. The company’s result shows exposure to the build-out, not a clean measurement of global AI spending.
The AI boom becomes tangible when the question changes from “How many models can be trained?” to “How many megawatts can be delivered, cooled and kept reliable?” Schneider’s results suggest that this second question is no longer a supporting detail. It is one of the central industrial constraints—and one of the places where the spending is already visible.
Sources & further reading
- Schneider Electric raises 2026 targets again on booming data centre demand Reuters · 16 Oct 2026 · Accessed 16 Oct 2026
- EcoStruxure Data Centre Solutions Schneider Electric · Accessed 16 Oct 2026
- 800 VDC power architecture for AI factories Schneider Electric · Accessed 16 Oct 2026