‘The new battle ground for scaling and protecting margins’: AI servers set to consume more power than every conventional data center by 2027 (www.techradar.com)

A recent Gartner forecast reveals a dramatic shift in data center power consumption, predicting that AI-optimized data centers will surpass conventional data centers in energy use by 2027. In 2026, AI centers are expected to consume around 175 terawatt-hours (TWh), a startling increase that could rise to 258 TWh the following year. Currently, AI data centers in the U.S. account for 36% of total data center power consumption, and this demand is projected to soar as AI-driven workloads continue to grow, with Gartner estimating that AI-specific servers will comprise 31% of data center power use by 2026. This surge in consumption poses significant challenges for the AI/ML community, as the increasing demand for compute-intensive tasks strains power availability, potentially stalling future AI expansion. Experts suggest that addressing this energy constraint will require advancements in the efficiency of power grids and data center hardware, particularly in cooling systems. Moreover, transitioning toward edge deployments may offer a solution by reducing reliance on centralized cloud infrastructures and mitigating the risks of outages. As the race for AI capabilities escalates, power security has become a critical issue that needs immediate attention to support sustainable growth in the field.