Green credentials come at a cost
Silicon Valley’s tech giants are fond of publicizing their green credentials by installing everything from waterless urinals to solar power. But, according to a new report from the International Energy Agency (IEA), tech giants’ latest offerings, primarily artificial intelligence (AI), are driving energy consumption to unprecedented levels.
The report, Electricity 2024 Analysis and Forecast to 2026, predicts that, if current trends continue, AI and cryptocurrency power consumption could more than double from 460 TWh in 2022 to up to 1,050 TWh in 2026, roughly equivalent to adding another Germany to global electricity consumption. According to the IEA, there are currently over 8,000 data centers globally, with about 33% of these located in the United States, with the largest data center hubs located in California, Texas, and Virginia.
In environmental terms, the US could end up paying a very high price indeed for the as yet largely uncertain benefits of AI. According to Rene Haas, CEO of giant chip designer Arm, if AI adoption continues unchecked, its consumption of electricity could soon mushroom. While US power consumption by AI data centers currently accounts for only four percent of US power consumption, Haas predicts that it will account for between 20-25 percent by 2030.
AI has an “insatiable thirst” for power
Haas blames popular large language models {LLMs) such as ChatGPT, which he says are “insatiable in terms of their thirst.” The IEA also estimates that one ChatGPT request consumes almost ten times as much power as a Google search. For example, were Google to switch its search engine entirely to AI software and hardware in order to keep pace with other popular services such as Microsoft-backed ChatGPT, it would increase its power consumption roughly tenfold.
Some countries, such as Ireland, a popular location for international data centers, could see a third of total energy consumption gobbled up as early as 2026. The Emerald Isle is already starting to experience power shortages with Amazon Web Service servers there reportedly being hindered by power limitations.
But given the drive to drastically reduce the use of fossil fuels to generate electricity, AI’s power-hungry consumption may not go unchecked for long. The IEA is already calling for government legislation to remedy the rapidly growing problem.
