AI's Rising Electricity Demand Raises Concerns, But Improvements in Efficiency Could Help

Recent interest in AI models like ChatGPT is raising worries about surging electricity use for AI processing.

Operating trained AI models (inference) may contribute significantly to lifecycle costs, not just model training.

Google adding AI to search could consume as much electricity as Ireland (29 TWh/year) in a worst case scenario.

More realistic estimates put AI server electricity use at 5.7 - 8.9 TWh in 2023, still minor compared to total datacenter use.

Improvements in model efficiency may be offset by greater AI adoption, limiting reductions in electricity consumption.