Power Distortions Linked to Rapid Expansion of AI Data Centers
Reports have shown that using additional data centers to solve computation problems of Artificial Intelligence (AI) affects power fluctuations in nearby houses. Data from Bloomberg revealed that over half of homes experiencing severe electrical disturbance are within 20 miles of data centers.
These irregular power fluctuations, known as the ‘bad harmonics,’ pose serious threats as they can burn appliances, increase fire risks, and potentially lead to power outages.
A report by Whisker Labs, based on data collected by 1 million residential sensors, suggests that these centres contribute to power grid pressure. Commonwealth Edison in Chicago has dismissed these attributions, questioning the accuracy of the assumptions used in the research.
Data Centers Growing Faster Than Grid Capacity
Due to the ever-growing development of AI technology, there is a heightened demand for more data centers. This has aggravated a condition where the electrical demand has outpaced the ability of the electric power grids to cope with the situation. According to the Electric Power Research Institute, AI data centers will account for 9% of US electricity consumption in 2030.
Bloomberg cites that data centers are emerging as “city-sized” electricity loads and are usually developed in less than two years, thus outpacing efforts to modernize the grid. This demand has more than doubled the local grid capacity, especially in global data center regions such as Northern Virginia, increasing grid concern.
Sudden energy surges and sags caused by fluctuating demands can exacerbate risks, including sparks and home fires. These issues are more obvious with ageing infrastructure, which explains why the latest grid structures must adapt to support AI facilities.
Environmental Policies and Data Center Operations
The combination of AI and environmental policy development raises more issues. Most Big Tech firms have pledged to transition to renewable energy and power their data centers with wind and nuclear energy. However, Bloomberg discusses that these plans may not work as expected under Trump’s presidency because the administration is highly associated with fossil fuel retailing.
Over 110 climate policies, including pro-climate policies on clean energy, were rolled back during Trump’s previous administration. Analysts also suggest that reducing subsidies on renewable energy may lead to a rise in the costs of clean energy, which poses a challenge for big tech companies in implementing new ways of green AI.
