Google's Gemini AI has made significant strides in reducing its environmental footprint. According to Google's recent transparency report, a median Gemini query now requires 0.24 watt-hours of electricity, equivalent to watching 9 seconds of TV. This translates to 0.03 grams of CO₂ emissions and about 5 drops of water usage per query.
This achievement is attributed to streamlined AI architecture, optimized hardware, and improved data center operations, resulting in a 97% energy efficiency improvement in just one year. Google's Gemini AI is powered by its sixth-generation TPU (Tensor Processing Unit) hardware, dubbed Trillium, which has made its way to customers.
However, experts warn that these per-query figures only tell part of the story. The actual environmental impact depends on the scale at which these queries are deployed. Billions of queries daily can transform even "tiny" resource use into a substantial environmental load.
To mitigate this impact, Google is investing heavily in clean energy, securing large-scale deals to power its data centers with nuclear and other low-carbon energy sources. The company's Trillium processors are designed to improve efficiency and reduce environmental impact. Google aims to operate on carbon-free energy around the clock by the end of the decade.
As AI technology continues to evolve, Google's efforts to reduce its environmental footprint serve as a model for the industry. By prioritizing sustainability and innovation, companies can minimize the environmental impact of AI while maximizing its benefits.
