According to the International Energy Agency, global data center electricity demand is set to more than double by 2026, surpassing 1,000 terawatt-hours –roughly the annual consumption of Japan. The surge is being driven by generative AI and other high-density compute platforms that require enormous amounts of power.
Transmission bottlenecks, regulatory delays, and workforce shortages, along with lack of generation capacity, are turning power into the defining constraint for digital infrastructure. Without new strategies, the mismatch between demand and supply could become the defining constraint on the next decade of innovation.
Several interlocking challenges are shaping the power equation for today’s data centers.
The first is utility interconnection delays. In regions like Northern Virginia, substations are already tapped out, leaving developers waiting years for new capacity. Meanwhile, areas such as West Texas or Pennsylvania have surplus generation but lack the transmission infrastructure to deliver it where demand is highest. The result is a paradox: stranded supply in some areas and stalled projects in others.
Another challenge is the shortage of skilled labour. Even when power can be generated, infrastructure buildouts depend on people, and there simply aren’t enough of them. According to the National Electrical Contractors Association and the U.S. Bureau of Labor Statistics, there are a projected 81,000 openings for electricians every year through 2034 – and the shortfall extends across mechanical, electrical, and controls disciplines. Prefabrication and modular construction are helping by shifting more work into factory environments, which tend to be more appealing to younger workers. Still, labor capacity remains one of the biggest factors limiting how quickly projects can scale.
Finally, the rapid technology cycles are driving compute demand. AI platforms are evolving at a pace unlike anything the industry has seen before, with new, denser systems released every 12 to 18 months. Traditional project timelines – from permitting through commissioning – are often longer than a single hardware cycle. As a result, facilities designed for today’s density risk being outdated before they even become fully operational.
Given these pressures, it is no longer possible to rely on a single source of energy. Each option brings advantages and limitations.
Historically, grid power has been the backbone – reliable and cost-efficient; however, rising electricity costs around the U.S. are creating backlash on the impact that data center buildouts are having on the everyday energy consumer.
Renewables are critical to long-term sustainability, yet often intermittent and geographically constrained. Texas, for instance, has 1.2 GW of renewable energy per hour annually sitting unused because of congestion in the transmission system.
Onsite generation offers speed and autonomy through gas turbines or reciprocating engines, but it effectively turns data centers into “mini utilities.” This adds complexity in emissions management, staffing, and continuous operations.