Can Generative AI get the power it demands? And without reversing Grid Decarbonization?

AI and Cloud computing’s power growth will increase US electric power consumption by 2% (2026) and to 8% (2030) (EIA,GS). Jevon’s paradox has overcome 50 years of energy efficiency improvements, driving computing power consumption to TerawattHours (BIG!). This growth both power grid stability and decarbonization, datacenters are slowing the retirement of and even adding new fossil fuel generation. In short, the climate change impact of computing is poised to grow rapidly. We discuss how computing flexibility (power, reliability) can help the grid decarbonize (Zero-carbon cloud), and how generative AI applications such as ChatGPT can be greened. The key is cooperation to manage reliability and to balance power load with variable generation. It depends on redesigning the service model for cooperation between datacenter loads and power grids. This approach can support both corporate goals (efficient computing) and societal goals (power grid decarbonization), and is a key to preventing AI from accelerating climate change. More information is available at http://zccloud.cs.uchicago.edu/.

Speaker Andrew Chien, University of Chicago

Attend in person or online (see weblink for connection information)