By 2030, AI data centers could take a bigger share of carbon emissions than flights do currently
Running AI compute on phones could offer a glimmer of hope for reducing the tech’s mounting power use.
The AI boom has come with a slew of side effects, but beyond social implications, geopolitical complications, and economic transformations, the growing number of data centers used to power the tech could have a devastating environmental impact — and it might come sooner than we initially imagined.
New forecasts from Accenture, reported by Axios on Wednesday, show that carbon emissions from AI data centers could increase 11-fold through the decade, accounting for a 3.4% share of total global CO2 emissions by 2030 in the “base case” scenario.
That’s a considerably greater share of total emissions than the entire aviation industry — a sector often admonished for its carbon footprint — notched at the last count, when the flying business took 2.5% of global CO2 emissions. It would also exceed the emissions contributed by both Germany (1.75%) and Saudi Arabia (1.58%) combined in 2022, per estimates from the IEA.
Dialing down
Accenture’s analysis offered some suggestions for making AI less resource-demanding, including utilizing “smart silicon,” a term used to describe running AI workloads on “smarter, compute-efficient hardware and models,” as well as integrating hardware and software to reduce the movement of data between memory and processors.
Just a day later, Axios reported new academic research also supporting the move toward AI power efficiency. In a Qualcomm-sponsored study, researchers at the University of California, Riverside, found that running AI directly on smaller devices like phones rather than the cloud could reduce the power consumption of queries by about 90%.
Moving AI compute to specialized small devices, thus bypassing energy-guzzling data centers for simple tasks, has been hailed by the industry for some time now. But, even if tiny chips in smartphones are the solution to a tech-driven energy crisis, it’ll be difficult to slow down the data center building spree of Amazon, Microsoft, OpenAI, and co. anytime soon.
