Data Centre Construction Excitement Needs a Dash of Caution

Why It Matters

If construction firms fail to adapt to modular methods and the shift toward distributed data centres, they risk being sidelined in a market where AI‑driven compute demand will soon prioritize speed, locality, and sustainability over sheer size.

Key Takeaways

• Hyper‑scale data centres now average 224 acres, up 144% since 2022
• MEP systems can exceed 70% of construction costs in demanding facilities
• Prefabricated pods and digital workflows cut on‑site time and errors
• Utah’s Stratos campus would consume >2× the state’s average electricity
• Gray Wolf proposes 5‑20 MW franchise hubs to serve local AI inference

Pulse Analysis

The data‑centre construction boom is being driven by relentless AI growth, yet the industry is hitting practical limits. Large‑scale facilities consume massive power and water, prompting state‑level moratoria and community pushback. The Stratos project in Utah, backed by Kevin O’Leary, illustrates the scale problem: its projected electricity use would exceed twice the entire state’s average consumption, raising alarms about grid strain and environmental impact. Simultaneously, land scarcity forces developers into rural zones, displacing agricultural acreage and sparking zoning reforms aimed at protecting prime farmland.

In response, builders are embracing prefabrication and digital integration to stay relevant. Companies like Schneider Electric promote factory‑assembled pods that bundle power, cooling, and cabling, dramatically shortening on‑site timelines and reducing error rates. This modular approach redefines the contractor’s role from a pure executor of concrete and steel to a system integrator that must coordinate complex IT loads, redundancy strategies, and MEP engineering. As Jacobs’ CEO notes, the investment cycle remains early, but success now hinges on mastering these integrated delivery platforms rather than merely meeting construction schedules.

A counter‑trend is emerging with the rise of small, distributed data centres. Gray Wolf’s franchise model envisions 5‑20 MW hubs placed near end users to meet the millisecond latency required for AI inference, which is projected to account for 55% of AI compute demand by 2027. By fragmenting capacity into dozens of 10‑MW sites, the model reduces land footprints, eases regulatory hurdles, and aligns with the industry’s push toward rapid, repeatable builds. For construction firms, this signals a strategic pivot: invest in modular, scalable solutions and develop capabilities to support a network of micro‑facilities, or risk obsolescence as the market fragments.