ALBAWABA – While Artificial intelligence grows in popularity, use and power, with data centers being set up all over the globe in a exceedingly fast speed, a new threat looms over the horizon; as experts warn of the massive energy these AI services are hogging.
A report by Goldman Sachs Research uncovers that data centers’ power need could rise 160 percent by 2030, reaching 3 to 4 percent of overall global energy, in contrast to the current 1 to 2 percent.
In a practical example, according to Goldman Sachs, a single ChatGPT inquiry consumes approximately ten times the amount of energy as a regular Google search. On the other hand, the generation of an artificial intelligence picture may use as much power as charging a smartphone.
As per the findings of a recent research conducted by Dr. Luccioni and his colleagues, posted on arXiv, a Generative AI system has the potential to use around 33 times more energy compared to typical machines which execute task-specific software. This development is already disturbing the tech industry, with Google disclosing this month a massive 48 percent increase in its carbon impact over the preceding five years.
More and more power is being used by data centers throughout the globe. According to the International Energy Agency (IEA), they used 460 terawatt hours of power in 2022, and in only four years, this amount is expected to quadruple. As BBC reports, by 2026, data centers may use 1,000 terawatt-hours of energy yearly.
Dipti Vachani, head of automotive at Arm, which energy-efficient small chips to giants like Google, Microsoft and Amazon, says “if we don’t start thinking about this power problem differently now, we’re never going to see this dream we have,” CNBC reports, adding that “saving every last bit of power is going to be a fundamentally different design than when you’re trying to maximize the performance.”
