AI Footprint Lab · everyday comparisons

How much does your AI use add up to?

Enter your AI usage, compare it with everyday life, then try changing one assumption.

Published estimates and assumptions. No account connection needed.

Electricity3.1 kWhText requests

Based on a published statement

◐ Published values + assumptionsscenarios.chatgpt.wh lengths.options.1.mult

Sam Altman’s June 2025 reported average; model mix, task mix and accounting method were not disclosed. Not a current per-model measurement.

A sourced input does not make this a measurement of your account.

scenarios.chatgpt.wh: Reported source value. First-party statement; no published measurement method 0.34 Wh per reported average ChatGPT query. 0.34 Wh × 1

lengths.options.1.mult: AI Footprint Lab-authored teaching input. Background context only; does not substantiate this exact value No exact source extraction asserted. Author-selected value: 1; no empirical conversion asserted.

Open original source ↗The Gentle Singularity — ChatGPT resource-use statementaltman-chatgpt-2025 · Sam Altman · Sam Altman’s blog · 2025Locator: Paragraph beginning ‘As datacenter production gets automated’ (June 2025)Boundary: Reported average query; energy and water boundaries unspecifiedGeography: Not disclosedReview status: not peer reviewed; methodology undisclosedChecked: 2026-09-05 · review by 2027-03-05

Open original source ↗Power Hungry Processing: Watts Driving the Cost of AI Deployment?luccioni-facct-2024 · Luccioni, Jernite & Strubell · ACM FAccT · 2024Locator: Table 2 image generation: mean 2.907 kWh/1,000 inferences, standard deviation 3.31; section 3 methodologyBoundary: CodeCarbon inference hardware energy for selected open models; excludes facility overhead, training and waterGeography: Test hardware and grid in paperReview status: peer-reviewed publication; AI Footprint Lab extraction not independently reviewedChecked: 2026-09-05 · review by 2027-03-05

Open original source ↗How Hungry is AI? Benchmarking Energy, Water, and Carbon Footprint of LLM Inferencejegham-2025 · Jegham et al. · arXiv preprint · 2025Locator: v6 abstract: most energy-intensive models exceed 29 Wh per long prompt, over 65× the most efficient; 0.42 Wh short-query exampleBoundary: Inference estimates across selected models and tasksGeography: Varies by modeled infrastructureReview status: preprint (v6; headline values changed between v1–v6); closed-model results are modeledChecked: 2026-10-12 · review by 2027-04-12

Water2.94 LText requests · chosen water example

Based on a published statement

○ Study or modeled estimatewaterCases.chatgpt.ml

0.000085 US gallons per query, reported by Sam Altman in June 2025. Accounting boundary and method were not disclosed; do not assume direct-only or life-cycle coverage.

A sourced input does not make this a measurement of your account.

waterCases.chatgpt.ml: Reported source value. First-party statement; water boundary unspecified 0.000085 US gallons per reported average ChatGPT query. 0.000085 × 3785.411784 = 0.32176000164 mL

CO₂ emissions1.21 kg CO₂eFrom electricity · CO₂ equivalent

Electricity × selected grid factor

◐ Published values + assumptionsgrid.value scenarios.chatgpt.wh lengths.options.1.mult

Uses 390 g CO₂e/kWh. Serving location is not established by this choice.

A sourced input does not make this a measurement of your account.

grid.value: User-selected teaching input. Background only; EPA does not substantiate the selected grid for this request Selected grid intensity: 390 gCO₂e/kWh. User input 390; not an inferred serving location.

scenarios.chatgpt.wh: Reported source value. First-party statement; no published measurement method 0.34 Wh per reported average ChatGPT query. 0.34 Wh × 1

Open original source ↗eGRID 2023 Summary Data, Revision 2epa-egrid-2023 · US Environmental Protection Agency · US EPA · 2025Locator: US and subregion output emission ratesBoundary: Electricity-generation output rates; not the serving data centre's measured gridGeography: United StatesReview status: government statistical datasetChecked: 2026-09-05 · review by 2027-03-05

ChatGPT estimate. 25 requests/day × 365 days. Assumes this daily use continues all year.

Your AI use

Who is this estimate for?

One request means one message you send to AI and its response.

Text requests per day

requests / day

Choose an AI example

Published examples, not measurements of your account. Actual requests vary.

Fine-tune this estimate

ChatGPT’s published figures have no disclosed calculation method. They are a starting estimate, not a reading of your account. Source ↗

0.34 Wh/request · 0.322 mL/request

Electricity estimateReported / illustrative waterResponse-length sensitivity

Response multipliers are authored assumptions. Reported water applies to ordinary text only. Changing a starting estimate resets response length to reference.

Advanced: images, video & vibe coding Optional · off

Add only activity not counted in text requests above. Include retries and discarded generations. Use a daily average; zero excludes an activity.

Images per day 2.907 Wh/image · 2024 benchmark mean

Image source & limits

○ Study or modeled estimateimage.wh

Luccioni et al., FAccT 2024, Table 2: mean 2.907 kWh per 1,000 image generations = 2.907 Wh/image across tested open models.

Historical benchmark mean, not a current product average. Model, image size and sampling settings vary; the reported standard deviation is 3.31 Wh, not a confidence interval. Inference hardware energy only; no facility overhead, training or water added. Newer GPU-only ML.ENERGY v3 measurements of 1024 × 1024 images on one H100 run from 0.116 Wh (SANA 1.5 1.6B, 20 steps) to 2.208 Wh (FLUX.1-dev, 50 steps); they are separate cases and are not averaged into this value.

image.wh: Reported source value. Supports this numerical value

Open original source ↗ML.ENERGY v3: text-to-image, text-to-video, chat and GPQA measurementsmlenergy-v3-tasks-2026 · ML.ENERGY research team · ML.ENERGY · 2026Locator: text-to-image.json, text-to-video.json, lm-arena-chat.json and gpqa.json at commit 5d899ea (unchanged through 2bdd007, 23 Sep 2026); H100 rows; energy_per_image_joules, energy_per_video_joules, energy_per_request_joulesBoundary: Steady-state GPU energy measured with Zeus; no CPU, memory, facility overhead or water. Reasoning comparison uses each model’s lowest-energy configuration, so prompt sets and batch settings differ.Geography: Laboratory hardware; grid selected separatelyReview status: Benchmark framework published at NeurIPS 2025 Datasets & Benchmarks; v3 dataset extraction not independently reviewedChecked: 2026-10-12 · review by 2027-04-12

Video clips per day 25.3 Wh/clip · CogVideoX-5b

Reference clip: about 6 seconds at 720 × 480, without audio. Other models and clip sizes can differ greatly.

Video source & limits

○ Study or modeled estimatevideo.wh

Delavande, Pierrard & Luccioni (NeurIPS 2025 workshop), Table 4, CogVideoX-5B: GPU 21.6 + CPU 2.4 + RAM 1.3 = 25.3 Wh/clip. One H100; 49 frames at 8 fps, 480 × 720, 50 steps; mean of 50 prompts × 5 runs after 2 warmups.

One open model at its default settings, not a product average. Energy grows roughly quadratically with resolution and frame count, so this value does not scale to other clips: the same table reports 0.14 Wh for AnimateDiff (16 frames, 512 × 512) and over 415 Wh for WAN2.1-T2V-14B (81 frames, 720 × 1280), and GPU-only ML.ENERGY v3 runs range from 7.95 Wh (CogVideoX-2b) to 323 Wh (HunyuanVideo). Excludes facility overhead, training and water. Does not estimate Sora, Veo, audio, upscaling or arbitrary clip lengths.

video.wh: Reported source value. Supports this CogVideoX-5B clip specification only; NeurIPS 2025 workshop paper, not a product average

Open original source ↗Video Killed the Energy Budget: Characterizing the Latency and Power Regimes of Open Text-to-Video Modelsdelavande-video-neurips-2025 · Delavande, Pierrard & Luccioni · NeurIPS 2025 workshop paper (arXiv) · 2025Locator: Section 4.3 protocol (50 prompts, 2 warmups, 5 runs); Table 2 CogVideoX-5b defaults; Table 4 (PDF p. 9) CogVideoX-5B rowBoundary: CodeCarbon GPU (NVML) + CPU (pyRAPL) + RAM heuristic on one H100 SXM; 49 frames, 8 fps, 480 × 720, 50 steps; warmups excluded from the mean. No facility overhead, training or water.Geography: Laboratory hardware; selected grid applied separatelyReview status: Accepted as an oral at the NeurIPS 2025 NextVid workshop; not a journal article; AI Footprint Lab extraction not independently reviewedChecked: 2026-10-12 · review by 2027-04-12

Vibe coding sessions per day Approximation from generated output

Code-completion proxy, not a measured agent session. Excludes your computer, builds, tests and hosting.

Coding sources & calculation

○ Study or modeled estimatecodingCases.large.joulesPerToken

Sessions/day × output tokens/session × 19.601410817113816 J/token ÷ 3,600 = GPU Wh/day. Qwen3-Coder 480B FP8 · 4 B200s. Code completion, maximum batch 8, mean output about 261 tokens; GPU energy only.

Measured short code completions are used as a proxy for a larger session. Input processing is implicit in the benchmark workload; long-context and agent-tool overhead are not separately modeled. No facility overhead or water is added. No provider-specific agent estimate is claimed.

codingCases.large.joulesPerToken: Reported source value. Supports this code-completion configuration only; not a measured vibe coding session

Open original source ↗ML.ENERGY v3: Qwen3-Coder code-completion measurementsmlenergy-coding-2026 · ML.ENERGY research team · ML.ENERGY · 2026Locator: Qwen__Qwen3-Coder-30B-A3B-Instruct__sourcegraph-fim.json and Qwen__Qwen3-Coder-480B-A35B-Instruct-FP8__sourcegraph-fim.json; configurations with B200, max_num_seqs=8, num_gpus=1 and 4 respectively; energy_per_token_joulesBoundary: Steady-state GPU energy on Sourcegraph fill-in-the-middle code completion. GPU joules divided by generated tokens includes the benchmark input-processing workload; no separate input-token term. Session scaling is an authored extrapolation, not a measured coding agent.Geography: Laboratory hardware; grid selected separatelyReview status: Benchmark framework published at NeurIPS 2025; v3 dataset extraction not independently reviewedChecked: 2026-09-05 · review by 2027-03-05

Adds electricity and estimated emissions. Reported water remains text-only. These benchmarks cover different hardware boundaries; their sum is a scenario, not a complete footprint.

Explore an alternative

What changes my estimate?

Try one change to see its effect. Your current estimate will be saved alongside the alternative.

Picture this estimate

Put the number in perspective

Picture your AI estimate

2.94 L per year · selected AI water case

500 mL water bottles

5.87

Each full symbol = 1 bottle.

A chosen container size to picture the water volume. Bottle production and transport are excluded.

Toilet flushes

0.606

Each full symbol = 1 flush. Highlighted share: 60.6% of one.

4.85 L per flush (1.28 US gal). Uses your toilet reference setting.

Same water volume does not mean the same local impact. Highlighted fractions are proportional symbols, not liquid levels. Flush reference ↗

Compare with everyday life

Choose a reference. Both sides show the same time period.

Everyday reference

Refrigerator

Efficient top-freezer · 360 kWh/year

Electricity

Your AI / year: 3.1 kWh
Refrigerator / year: 360 kWh

0.862% of refrigerator electricity

Like 3.15 days of refrigerator use.

Water

Your AI / year: 2.94 L
Refrigerator / year: Not estimated

Refrigeration uses no running water. Ice-making and electricity-supply water are excluded.

CO₂ emissions

Your AI / year: 1.21 kg CO₂e
Refrigerator / year: 140 kg CO₂e

0.862% of refrigerator electricity emissions

Like 3.15 days of refrigerator use on the same grid.

Colored bar = your AI estimate. Gray bar = the reference. “Not estimated” means no comparable figure is available; it does not mean zero.

Adjust electricity & water assumptions

Electricity emissions factor

g CO₂e / kWh

Used for AI, appliances and EV charging. Gasoline tailpipe CO₂ uses a separate fuel factor. Your AI’s location is unknown. Grid context ↗

Water comparison includes

AI uses the selected per-request case. Golf and showers show water applied or delivered.

Estimates cover the selected uses. Manufacturing, AI training, refrigerant leaks and wider life-cycle impacts are excluded.

The numbers behind this comparison

Sources and assumptions for the settings above. All conversions are available in the source ledger; independent scientific review is pending.

Your AI estimate

Electricity: Sam Altman’s June 2025 reported average; model mix, task mix and accounting method were not disclosed. Not a current per-model measurement.

Water: 0.000085 US gallons per query, reported by Sam Altman in June 2025. Accounting boundary and method were not disclosed; do not assume direct-only or life-cycle coverage.

Sam Altman’s blog ↗

Refrigerator

ENERGY STAR’s 2024 guide: about 360 kWh/year for a certified top-freezer refrigerator, divided by 365. A representative efficient type, not an average of all refrigerators.

US EPA ↗

360 × 1000 ÷ 365 Wh/day

Full source details and correction IDs

○ Study or modeled estimatescenarios.chatgpt.wh

Sam Altman’s June 2025 reported average; model mix, task mix and accounting method were not disclosed. Not a current per-model measurement.

scenarios.chatgpt.wh: Reported source value. First-party statement; no published measurement method

○ Study or modeled estimatewaterCases.chatgpt.ml

0.000085 US gallons per query, reported by Sam Altman in June 2025. Accounting boundary and method were not disclosed; do not assume direct-only or life-cycle coverage.

waterCases.chatgpt.ml: Reported source value. First-party statement; water boundary unspecified

○ Study or modeled estimatefridge.whDay

ENERGY STAR’s 2024 guide: about 360 kWh/year for a certified top-freezer refrigerator, divided by 365. A representative efficient type, not an average of all refrigerators.

fridge.whDay: Reported source value. Supports this numerical value

Open original source ↗2024 Quick Reference Guide — Refrigeratorsenergy-star-fridge-2024 · ENERGY STAR · US EPA · 2024Locator: PDF page 14, Refrigerator & Freezer ConfigurationsBoundary: Certified refrigerator operational electricity; excludes manufacturing and refrigerant leakageGeography: United StatesReview status: government product guidanceChecked: 2026-09-05 · review by 2027-03-05

A refrigerator’s size and age, AC operating hours, and a golf course’s climate can change these comparisons substantially. Water volume does not measure local water stress. CO₂e expresses greenhouse gases as an equivalent amount of CO₂.

Explore the data and tools

Built to make the assumptions visible.

Created by Chris Meehan, an AP Environmental Science educator and academic technology director. Independent energy and water review, and classroom pilots, are pending.

Method 1.7.0 · dataset 2026.10.12.2

Your estimate per year

What changes my estimate?

Picture your AI estimate

Compare with everyday life

Refrigerator

↯Electricity

≈Water

CO₂CO₂ emissions

Your AI estimate

Refrigerator

Explore the data and tools

Built to make the assumptions visible.

Electricity

Water

CO₂ emissions