Summary
A reference comparing the energy and water consumption of common technologies including AI queries, Google searches, video streaming, and paper printing. Intended to help Bowdoin community members make more environmentally informed choices about technology use.
Body
Questions
- How much energy does a ChatGPT query use compared to a Google search?
- Does using AI use more water than other technologies?
- How environmentally efficient is AI compared to video streaming or printing?
- What is the carbon or energy footprint of generative AI tools?
- How does the water usage of AI compare to everyday activities?
- What technologies should I use to minimize my environmental impact?
Environment
This is a reference article for all Bowdoin faculty, staff, and students. It provides approximate figures for the energy and water consumption of common technologies to support informed, environmentally conscious choices. Figures are estimates based on published research and will vary depending on the specific model, infrastructure, and usage context.
Resolution
Energy consumption comparison
The table below shows approximate energy use per unit of activity for common technologies.
| Technology |
Energy per unit |
Notes |
| AI query (e.g. ChatGPT) |
0.001–0.01 kWh per query |
Varies by model size and number of tokens processed |
| Google search (traditional) |
~0.0003 kWh per query |
Roughly 3–30× more efficient than a generative AI query |
| Database query (web browsing) |
0.0001–0.001 kWh per query |
~10× more efficient than AI, though a single web page may involve many queries |
| Video streaming (4K, 10 min) |
0.017–0.076 kWh |
Includes device, network, and server energy; roughly 7× less efficient than a ChatGPT query at the midpoint |
| AI-assisted search (e.g. Bing AI, Perplexity) |
Combined AI + traditional search cost |
AI-enhanced search results add generative AI energy on top of the traditional search cost |
Water consumption comparison
Data centers that support AI and other cloud services require large amounts of water for cooling. Some water evaporates through cooling towers; many systems recycle or treat water for reuse.
| Activity |
Water used |
Notes |
| AI query (e.g. ChatGPT) |
~0.5 liters per query |
Average estimate across data center infrastructure |
| Video streaming (10 min) |
~0.5 liters |
Comparable to a ChatGPT query |
| Toilet flush |
6–9 liters per flush |
For comparison only |
| Paper production (1 ream, 500 sheets, non-recycled) |
~6,500 liters total (~13 liters per sheet) |
Covers the full production cycle; printing less can meaningfully reduce water use |
What this means in practice
Different technologies consume varying amounts of energy and water depending on their purpose and infrastructure. A few practical takeaways:
- A traditional web search is significantly more energy-efficient than a generative AI query — use search for simple fact lookups where AI isn't necessary.
- Video streaming and AI queries have similar water footprints per unit.
- Reducing paper printing — especially for large documents — has a meaningful impact on water consumption across the full production lifecycle.
- AI-enhanced search (where AI summarizes results) carries the combined cost of both AI and traditional search infrastructure.
Note: All figures are estimates from published research and should be treated as approximations. Actual consumption varies significantly based on the specific AI model, data center efficiency, network infrastructure, and device type.
Additional Help
If you need further assistance, you have several options:
- Bowdoin Bot: Chat with Bowdoin Bot directly from any KB page for instant answers.
- Phone: Call the Bowdoin College Service Desk at (207) 725-3030.
- In person: Visit the Tech Hub in Smith Union during business hours.
- Submit a ticket: Request assistance through the Service Catalog.