AI and its 10 million increase in energy consumption

Between 2010 and 2018, a worldwide increase of 550 % in the number of data center workloads and compute instances is estimated. Data centers and transmission networks together participate up to 1,5 % on global energy consumption. They emit an amount of carbon dioxide comparable to the annual production of Brazil.

Hyperscalers like Google , Microsoft and Amazon have committed to ambitious climate goals that aim to decarbonize their operations. Hyperscalers are large, highly optimized and efficient devices.

However, the spread of AI poses a huge challenge to these goals. The power-intensive nature of graphics processing units (GPUs), which are necessary for training AI models, increases the pressure on power resources. (Jennifer L, more at carboncredits.com)