Industry news. A report on AI data centre growth concludes that expansion now hinges on solving two constraints at once: available power supply and community concerns about new facilities.
Why power is the binding constraint
As rack densities rise into the hundreds of kilowatts, the electricity a site needs grows faster than the site footprint. In many regions the limiting factor for a new campus is not land or cooling equipment but the grid connection.
Efficiency moves up the agenda
When power is scarce, every kilowatt spent on cooling is a kilowatt not available for compute. That shifts attention onto power usage effectiveness (PUE) and onto control strategies that avoid unnecessary energy use – not just more capacity.
What reduces cooling energy
- Modulating pumps and fans continuously instead of switching them, which avoids repeated start-stop losses.
- Raising coolant supply temperature where the design allows, reducing chiller load.
- Measuring rather than assuming: airflow, differential pressure and temperature data at the points that matter, so control responds to actual conditions.
Together these make control and instrumentation a direct contributor to how much compute a given power allocation can support.
Source: industry report coverage on AI data centre growth constraints, 2026.