Data Center Operators Sound Alarm Over Power Constraints Amid AI Boom

Data center operators are increasingly worried about their ability to secure sufficient power as demand for electricity surges due to the rapid expansion of artificial intelligence. Amazon CEO Andy Jassy has identified power as the single biggest challenge for the company as it scales its AI initiatives. A recent survey by the Uptime Institute reveals that concerns over forecasting data center capacity and planning for power needs have risen by 9% since 2023—a jump the industry advisory called "significant." While rising costs remain the top concern, with 38% of respondents saying they are "very concerned," power constraints are quickly catching up, with 36% expressing similar anxiety. The same percentage also cited demand forecasting as a major worry. The International Energy Agency projects that global data center electricity consumption could double by 2030. In the United States, major tech companies are investing heavily in infrastructure to support AI, with Amazon, Microsoft, Meta, Google, and Apple collectively expected to spend more than $350 billion on data centers in 2025. Recent quarterly earnings reports from four of the five companies showed they plan to spend more than initially projected, driven by the need for additional computing capacity and infrastructure. While inflation, labor shortages, and higher energy prices contribute to rising costs, data center operators cannot control the broader electric grid, making power supply a fundamental and growing constraint. Despite ongoing efforts to improve energy efficiency, progress has stalled. The industry’s Power Usage Effectiveness (PUE) metric, which measures how efficiently a data center uses energy, has remained flat for the past six years. Although new cooling technologies are emerging, older facilities and regional limitations have hindered widespread adoption. Moreover, data center operators are still hesitant to fully integrate AI into facility operations. While most are comfortable using AI for tasks like analyzing sensor data and predictive maintenance, they are reluctant to let AI make critical decisions such as configuring systems, controlling equipment, or managing staff. Trust in AI remains highly dependent on the specific use case.