Hitachi Energy, one of the world’s leading power infrastructure companies, has issued a stark warning: the explosive and unpredictable power demands driven by artificial intelligence (AI) and data centres are threatening to destabilise global electricity supply. As AI adoption accelerates, particularly in generative AI applications, the resulting power spikes have introduced unprecedented volatility into energy systems, challenging utilities and grid operators worldwide.
The rapid expansion of AI-powered data centres is fundamentally reshaping the global energy landscape. According to Hitachi Energy’s CEO, Andreas Schierenbeck, demand volatility during the generative AI learning process has introduced a whole new factor into managing energy systems. Unlike traditional industrial or residential loads, AI data centres can produce sudden, sharp surges in electricity consumption, making it difficult for utilities to forecast and balance supply and demand in real time.
Data centres are projected to consume up to 12% of U.S. electricity by 2028, up from just 4.4% in 2023, with similar trends emerging globally. This surge is primarily attributed to the immense computational requirements of training and running advanced AI models, which can draw vast amounts of power in unpredictable bursts.
The wild fluctuations in demand caused by AI workloads risk overwhelming existing grid infrastructure, particularly in regions where planning and investment have not kept pace with digital expansion.
The company is already partnering with major grid operators, such as the Southwest Power Pool in the U.S., to deploy AI-driven solutions that accelerate grid planning and improve reliability in the face of rising data centre demand.
While the warning is global, the impact will be felt most acutely in regions with high concentrations of data centres and slower infrastructure upgrades. Countries like the U.S., India, and those in Europe are racing to adapt, investing in advanced transformers, voltage regulators, and grid optimisation technologies to stay ahead of the AI-driven demand curve.
