How Worth the Compute? works

1. 1. Normalizes daily, weekly, workday, and monthly activity into one typical month.
2. 2. Multiplies each activity by low, typical, and high impact factors.
3. 3. Adds activity totals together.
4. 4. Converts the results into water, carbon, and energy units.
5. 5. Produces separate real-world comparisons.

Real-world comparisons are alternatives. They help interpret one metric at a time and are not added together.

The calculator can accept activity frequencies such as queries per day, tasks per week, coding sessions per workday, and videos per month. All of those inputs are normalized into one typical month before environmental impact is calculated.

Days per month

30.44

Weeks per month

4.345

Workdays per month

21.75

Months per year

12, used internally for yearly methodology comparisons

Users can estimate usage with an AI assistant or enter it manually. Either way, the displayed environmental result is monthly.

Low, typical, and high values reflect uncertainty from model size, architecture, prompt length, response length, generated tokens, reasoning duration, generation count, hardware, utilization, data-center efficiency, cooling, geography, electricity source, and provider implementation.

The range is not a guarantee that the actual result falls between these exact values.

A scenario can be uncertain because the user may not know exactly how much AI they use, and because the environmental cost of each activity varies. Imported low and high usage estimates are shown for context but are not yet included in the final environmental range.

Each comparison stands alone. They are alternative ways to understand the same monthly result, not values to add together. Water can be compared with bottles, showers, laundry, or bathtubs; carbon with driving or a flight percentage; and energy with charging devices, television use, or electric-vehicle miles. The calculator presents those metrics in the order Water, Carbon, Energy.

• Operational AI inference or generation
• User-selected activity volume
• Estimated operational water
• Estimated operational carbon
• Estimated electricity
• The selected number of users
• One typical month

• Manufacturing AI chips and servers
• Constructing data centers
• Training a model, unless a modeled factor implicitly approximates it
• The user's laptop or phone electricity
• Network infrastructure
• Storage of conversations and generated media
• Repeated hidden provider-side operations
• AI activity the user did not report
• Provider-specific optimizations not represented by public data

1. [1] modeled assumption

   Worth the Compute working AI activity factors

   Worth the Compute, 2026. Reviewed 2026-07-31.

   Internal MVP working factors used when public per-request measurements are not standardized enough to cite as exact measurements.

   Used for: Low, typical, and high energy factors for AI activity categories; Proxy mapping for agent and automation runs.

   Modeled assumption documented in this methodology.

2. [2] international organization

   Energy and AI

   International Energy Agency, 2025. Reviewed 2026-07-31.

   IEA analysis describes AI and data centers as a growing electricity demand category while emphasizing variation by geography and deployment.

   Used for: Context on AI and data-center electricity demand; Uncertainty framing for rapidly changing AI workloads.

   Open source

3. [3] government

   Emissions & Generation Resource Integrated Database (eGRID)

   U.S. Environmental Protection Agency, 2025. Reviewed 2026-07-31.

   EPA eGRID provides U.S. electricity generation, emissions, and output emission-rate data.

   Used for: Explaining grid emissions factors; Basis for the calculator's broad grid-average carbon factor.

   Open source

4. [4] modeled assumption

   Data Centers and Water Consumption

   Environmental and Energy Study Institute, 2025. Reviewed 2026-07-31.

   Explains Water Usage Effectiveness as liters of water per kWh and discusses why data-center water use varies by cooling system and location.

   Used for: Explaining water usage effectiveness in liters per kWh; Context for the calculator's operational water factor.

   Open source

5. [5] government

   Greenhouse Gas Emissions from a Typical Passenger Vehicle

   U.S. Environmental Protection Agency, 2016. Reviewed 2026-07-31.

   EPA explains typical passenger-vehicle CO2 emissions and the assumptions behind average emissions per mile.

   Used for: Gasoline vehicle carbon equivalency.

   Open source

6. [6] government

   How much electricity does an American home use?

   U.S. Energy Information Administration, 2024. Reviewed 2026-07-31.

   EIA publishes average annual and monthly electricity use for U.S. residential electric-utility customers.

   Used for: Household electricity equivalency.

   Open source

7. [7] modeled assumption

   Showers

   Home Water Works / Alliance for Water Efficiency, 2016. Reviewed 2026-07-31.

   Provides average shower duration, flow rate, and total gallons for a typical American shower.

   Used for: Average shower water equivalency.

   Open source

8. [8] modeled assumption

   Worth the Compute equivalency assumptions

   Worth the Compute, 2026. Reviewed 2026-07-31.

   Internal, clearly labeled conversion factors used to turn calculator outputs into familiar comparisons.

   Used for: Smartphone, laptop, LED bulb, TV, dishwasher, laundry, EV, flight, beef meal, drinking bottle, and bathtub comparisons.

   Modeled assumption documented in this methodology.

Methodology version

1.0

Environmental methodology version

1.0

Last reviewed

July 31, 2026

Import schema version

1.1

Usage-estimation prompt version

1.1