Student tool

Everyday AI Impact

Choose a common AI task and see its estimated environmental impact in units that are easier to picture.

1. Choose a task

Pick the option closest to what you did.

2. How many times?

Enter the total number of times the selected AI task was completed.

Number of times Example: Enter 30 if a class completed the task 30 times.

Estimated impact

1 quick AI question

This is an estimate, not a direct measurement.

Research-based scenario

Electricity

About 1.9 minutes of LED light

Equivalent to 0.31 watt-hours.

Phone battery

About 1.4% of one charge

Compared with an assumed 22 Wh smartphone charge.

Estimated U.S. water footprint

About 30 drops of water

An educational estimate combining average data-center cooling water and electricity-related water. It is not a measurement of a specific AI provider or data center.

What this means One quick AI question has a small estimated impact, but repeated use across a class or school adds up.

Important limitation This classroom tool estimates scale using published reference values. Actual environmental impact varies by AI model, output length, hardware, data-center location, cooling system, and electricity source. Results should not be presented as exact measurements of a particular AI service.

How infrastructure changes the picture

Your choices affect how often AI is used, but the data center's design can also make a big difference. The same AI task can have very different environmental impacts depending on how the data center is cooled and powered.

Two types of water use Using less water at the data center does not always mean an AI system has zero water impact. Electricity generation can also require water.

Direct water is used at the data center itself — mainly for cooling servers.
Indirect water is used off-site to generate the electricity the data center needs (for example, to cool power-plant condensers or produce steam).

What makes the difference?

AI's environmental impact depends on more than how many prompts you send. These five factors shape the total footprint:

Number and type of AI requests — more complex tasks take more computing power.
Computing hardware and efficiency — newer chips often do more work per watt.
Cooling technology — closed-loop liquid cooling uses far less water than evaporative cooling towers.
Source of electricity — renewable energy produces less CO₂ and can require less water than fossil fuels.
Data center location and climate — cooler climates and regions with cleaner grids have lower footprints.

What you can do — and what companies can do Your choices can reduce unnecessary AI use, but infrastructure matters too. More efficient cooling and cleaner electricity can greatly reduce the impact of the same AI task. Using AI intentionally — and avoiding unnecessary retries — is still a good habit, but the biggest gains come from better data center design.

Actual impact varies by model, response length, hardware, location, cooling system, and electricity source. The rewrite, summary, and document-analysis presets use classroom comparison values anchored to the published typical text-query estimate.

Data sources

These studies provide the published reference values used in this calculator.

Microsoft Research — Energy Use of AI Inference Supports the 0.31 Wh typical text-query estimate, its 0.16–0.60 Wh range, and the 3.91 Wh extended-reasoning scenario with a 2.15–7.05 Wh range. Luccioni, Jernite & Strubell — Power Hungry Processing Supports the 2.907 Wh mean image-generation benchmark. The study found substantial variation among models. Berkeley Lab — 2024 United States Data Center Energy Usage Report Supports the national-average estimates of just over 0.36 L/kWh of direct data-center water and 4.52 L/kWh of electricity-related water. EPA eGRID 2022 — Emissions & Generation Resource Integrated Database Supports the ~0.386 kg CO₂e/kWh U.S. national average grid emissions factor used in the Conventional Infrastructure scenario. IPCC Sixth Assessment Report (AR6), Working Group III ⚠️ Used as a reference for illustrative low-carbon electricity CO₂e ranges (~0.007–0.048 kg CO₂e/kWh for wind and solar). The blended 0.02 kg CO₂e/kWh value used in Scenario 3 is an estimate requiring expert validation.

The calculator combines estimates from sources with different methods and system boundaries. Summary and document-analysis values are classroom comparison estimates, not measured averages. Infrastructure scenario values for Scenarios 2 and 3 are illustrative estimates; see ⚠️ notes above.