AI Faces Huge Electricity Barrier

By Douglas McIntyre  •  April 10, 2024
Matheus Bertelli Pexels

There is no way to estimate AI’s electricity use compared to the overall US electricity generating capacity. At the same time, it may be possible to estimate US capacity, particularly by region. The amount of electricity used by AI facilities so early in its evolution as perhaps the world’s most influential technology has created a world of anxious debate. For now, all forecasts are a guess. 

Effects Of $5 Gas –The Future Of EVs?

Climate Price Tag –LA’s $12 Billion Challenge

Rene Haas, the CEO of chip company Arm, told The Wall Street Journal, “by the end of the decade, AI data centers could consume as much as 20% to 25% of U.S. power requirements. Today that’s probably 4% or less.” 

Other scientists and tech executives have made similar comments. The issue is that the general population, commercial enterprises, and the government will not allow this level of consumption to occur. Utilities could not build capacity to this level with the massive infracture construction costs required. Even if they could, the increased use of fossil fuels would further damage the environment. 

The Problem Of The Grid

Additionally, even if utilities could handle the construction, renewables need to create more electricity. The level of energy use would also mean that current customers would be deprived of electricity during some periods like hot summer days when there is heavy use of air conditioning. Therefore, the supply of AI applications would be significantly curtailed, hampering the industry. 

One thing is sure. Many electrical utilities in the US operate at or near total capacity. This causes occasional brownouts and loss of service. There have been, and will continue to be, blackouts because, in many parts of America, the grid is old and not likely to be updated.

AI has an electricity problem.

More from ClimateCrisis 247

Similar Posts