The Scale Math: How 0.24 Wh Becomes 347 TWh

Small numbers, enormous impact. This is the defining paradox of AI economics. A single prompt consumes just 0.24 watt-hours—barely enough to light an LED for a minute. Multiply by billions of daily queries, and you get an energy footprint that rivals nation-states.

The Multiplication Effect framework reveals how microscopic per-query costs aggregate into macroeconomic forces:

ChatGPT: A Scale Study

• 2.5 billion queries per day
• 850 MWh daily electricity consumption
• 311 GWh annually (~30,000 US homes)
• 100,000+ tons CO2/year (equivalent to 20,000 cars)
• Electricity cost: $30-40M annually

Water: The Hidden Cost

• Each query: ~0.26 mL of water for cooling
• Daily total: 450 million gallons by 2030
• Equivalent to 8 million people’s daily water use
• Half of US data centers sit in water-stressed regions

2030 Projection: All Generative AI

• 347 TWh annual consumption (Schneider Electric estimate)
• 4.5% of global electricity generation (IMF projection)
• 38 Google-class (1 GW) data centers required

The insight for business leaders: at planetary scale, microscopic costs aggregate into trillions of dollars and terawatts of demand. Companies that ignore the multiplication effect will find their unit economics destroyed by infrastructure — as explored in the economics of AI compute infrastructure — costs they never modeled.

The prompt is cheap. A billion prompts reshape the global energy grid.

This analysis applies The Business Engineer’s Scale Math framework to understand how AI’s unit economics translate to planetary impact. Read the full analysis: The Economics of an AI Prompt →