Sustainability Tools

AI Consumption Calculator

Understand the water, electricity, and carbon footprint of AI usage

AI Usage Input

Enter your AI usage to calculate environmental impact

Large language model, highest capability

Type of task affects energy consumption

Total number of AI requests/prompts

Time frame for your usage

Optional: Training Impact

Add your share of model training costs

For calculating your training share

Quick Scenarios

Real-World Comparisons

Compare your AI usage to everyday activities

⚠️ No data to compare
Calculate your AI consumption in the Calculator tab first

Visualizations

Charts showing AI consumption breakdown

⚠️ No charts to display
Calculate your AI consumption in the Calculator tab first

AI Model Reference Data

Consumption data for various AI models and tasks

Reference

Understanding AI environmental impact

Key Insight
1 AI query ≈ 0.1-3.5 Wh + 2-55 mL water
Energy varies widely by model: small models (0.1 Wh) to reasoning models like o3 (3.5 Wh). Image/video generation uses significantly more.
Why Water?

Data centers require massive cooling systems.

  • GPUs generate significant heat
  • Evaporative cooling uses water
  • ~500mL per 10-50 AI queries
  • Varies by location and climate
Energy Breakdown

Where the electricity goes.

  • ~60% GPU computation
  • ~25% cooling systems
  • ~10% networking/storage
  • ~5% other infrastructure
Training vs Inference

One-time vs ongoing costs.

  • GPT-5.2 Pro training: ~120,000 MWh
  • Gemini 3 Pro training: ~70,000 MWh
  • Single query: ~0.3-3.0 Wh
  • Reasoning models use 3-10x more per query
Calculation Sources
Data Sources:
• Research papers from UC Riverside, Google, Microsoft
• Published carbon reports from AI companies
• Data center efficiency metrics (PUE 1.1-1.5)
• EPA carbon intensity factors

Assumptions:
• US average grid carbon intensity (~0.4 kg CO₂/kWh)
• Modern data center efficiency (PUE ~1.2)
• Water usage varies significantly by location
• Estimates are approximate and for educational purposes

Understanding AI's Environmental Footprint

The Hidden Costs

Every AI query has an environmental impact that's often invisible to users.

  • Data centers use 1-2% of global electricity
  • AI is the fastest-growing segment
  • Water scarcity concerns in some regions
  • Carbon footprint varies by energy source
Reducing Your Impact

Simple ways to use AI more sustainably.

  • Use smaller models when possible
  • Batch similar queries together
  • Be specific to reduce back-and-forth
  • Consider if AI is necessary for the task
Industry Progress

Companies are working to reduce AI's footprint.

  • Renewable energy commitments
  • More efficient model architectures
  • Improved data center cooling
  • Carbon offset programs