AI Data Centers Must Prove Their Viability

AI is driving an unprecedented wave of data center announcements as governments view computing capacity as critical infrastructure. 
 
But these projects are placing growing pressure on electricity grids, water resources and local communities. 
 
Uptime Institute analyzed 479 large data center projects announced between 2021 and 2025. 
 
By July 2026, only 10% had at least one building operational, while 40% were stalled or delayed and 9% cancelled. 
 
The concern is that AI demand could grow substantially while individual developers overestimate the capacity their projects can realistically capture.
 
Authorities therefore need evidence before allocating scarce infrastructure. 
 
Developers should demonstrate committed tenants, contracted power, water availability, realistic construction schedules and access to critical equipment. 
 
Infrastructure design also matters. 
 
AI training and GPU inference can require high-density electrical systems and liquid cooling, while other workloads may have significantly different requirements. 
 
Transparency must improve around projected energy and water consumption, cooling technologies, emissions, noise and grid impacts. 
 
As AI infrastructure expands, the message is clear: data center capacity should be built on evidence, not announcements. 
 
Governments need proof of viability before committing critical public resources.