The AI infrastructure industry is entering a fundamentally different phase—one where the biggest challenges are no longer simply securing land, power, or capital, but making infrastructure decisions that will shape operations for decades.

A new report from the International Data Center Authority (IDCA) argues that gigawatt-scale AI campuses have transformed data center development from a cycle of reversible technology refreshes into long-term infrastructure bets with 20- to 40-year consequences. Unlike previous generations of facilities, today’s AI campuses cannot be easily relocated, repowered, or redesigned once construction begins.

The scale of the challenge is reflected in the industry’s rapidly rising energy demand. According to the report, data centers now consume 67.7 GW of electricity globally, equivalent to 1.9% of worldwide power generation, after increasing 17% year over year. The United States alone accounts for 43% of global data center electricity consumption, with data centers representing 6% of the country’s grid demand. Meanwhile, electricity use by AI-focused facilities surged 50% in 2025, and global data center consumption is projected to reach approximately 945 TWh by 2030, more than double current levels.

The report argues that these trends are reshaping investment priorities. Grid constraints, water availability, financing models, workload orchestration, and community acceptance are emerging as strategic bottlenecks alongside power generation itself. As AI infrastructure scales toward gigawatt campuses, success will increasingly depend on integrating energy strategy, infrastructure financing, and long-term operational control rather than simply deploying more compute.