Contextualising AI Consumption

A comparative analysis of the environmental footprint of Artificial Intelligence against other less visible costs, such as social media or the tangible impacts of leisure activities.

The AI Baseline: Training vs. Leisure

To understand AI's impact, we must look at the massive singular event of Training, which can take months of continuous energy consumption. Training GPT-3 generated approximately 552 tonnes of CO₂e (Patterson et al., 2021).

To evaluate fairness, we compare this against sustained leisure activities. A footprint analysis for an "average" US golf course estimates approximately 796.6 tCO₂e per year; therefore, maintaining it for 6 months equals ~398 tCO₂e (CleanFi, 2021).

The "PL Football Weekend" reference is calculated from an aggregated estimate of 80,651 tCO₂e for the 2018/19 Premier League season (fan travel, club travel/accommodation, and stadium energy), implying ~212 tCO₂e per match and ~2,120 tCO₂e for a typical 10-match matchday (Murray, 2021).

Figure 1: Comparative Emissions (tCO₂e). Sources: CleanFi (2021); Murray (2021); Patterson et al. (2021).

The Digital Aggregate

Here we compare 1 hour of continuous active use. AI queries are energy-intensive per unit, but social media's video bandwidth creates a massive aggregate footprint.

  • • AI Interaction (1 Hour): ~30 queries totals ~90g CO₂e (≈3 gCO₂e/query; high estimate) (Ritchie, 2025).
  • • TikTok (1 Hour): ~158g CO₂e (≈2.63 gCO₂e/min) (Compare the Market, n.d.; The Independent, 2024).
  • • Netflix (1 Hour): ~56g CO₂e per hour of streaming (European average cited by Carbon Trust) (Carbon Brief, 2020; The Guardian, 2021).

Figure 2: Approximate CO₂e impact of 1 hour activity (gCO₂e). Values sensitive to device, resolution, network, and electricity mix (IEA, 2020). TikTok estimated from 2.63 gCO₂e/min. Netflix approximated at 55–56 gCO₂e/h. AI assumes 3 gCO₂e/query (high end) and 30 queries/h.

Daily Footprint Calculator

Compare your daily habits. Toggle metrics to see hidden costs.

💻 Digital Habits (Daily)

20 queries
2 hours
1 hours

🏠 Home & Lifestyle (Daily)

1 meals
10 mins
0.5 avg/day

⛳ Leisure (Amortised Daily)

*Enter monthly frequency. We calculate the daily share.

15 miles
0 rounds
0 matches

Daily Breakdown

Showing Carbon Emissions

Total Daily CO₂e
0 g

Modify the sliders to see how AI compares to your physical habits.

Methodology & Data Sources

This section explicitly lists the constants used in the calculator above and the scientific references validating them.

Activity Carbon Factor Carbon Justification Water Factor Water Justification
AI Query (Avg) 3g / query Conservative estimate (high end) of ~2–3 gCO₂e per query including training amortization (Ritchie, 2025; Tomlinson et al., 2024). Estimates vary 0.4–4.3g depending on model, length, and datacenter energy source. 5ml / query Updated cooling water consumption estimate based on recent analyses by Li et al. (2023). Original figures of ~500ml/20-50 queries have been refined to ~5ml per query as a likely average.
Streaming (HD) 56g / hour European average per Carbon Trust (2021) and Netflix/DIMPACT. IEA estimates ~36g/hr global average (2020 update). 56g is used as a conservative European reference. 2 L / hour Conservative estimate of water footprint associated with electricity supporting streaming (cooling/evaporation in datacenters and networks). Literature reports ranges of ~2–12 L per hour; lower end (2 L/h) used to avoid overestimation (Obringer et al., 2021; Purdue University, 2021).
Video Call 150g / hour Based on Purdue University Study (2021) analyzing symmetric upload/download data intensity compared to streaming. 2 L / hour Conservative estimate based on studies quantifying water footprints per hour of videoconferencing. Ranges depend heavily on camera usage, resolution, and device; 2 L/h used as reported minimum (Obringer et al., 2021; Purdue University, 2021).
Meal (Beef) 9,000g / meal Based on Poore & Nemecek (2018). Producing 1 kg of beef emits ~60 kgCO₂e global average. For a ~150g portion, this equals ~9,000g CO₂e. 2,200 L / meal Based on Water Footprint Network. Includes "Virtual Water" (Green+Blue+Grey) required to raise feed/livestock.
Shower 30g / min Assumption of ~10 kW electric shower: ~1 kWh every 6 mins (≈0.167 kWh/min). Converted to CO₂e using official UK grid factor (0.177 kgCO₂e/kWh; 2025 conversion). Result ≈ 0.0295 kg/min ≈ 30 g/min (Viessmann, 2022; DESNZ, 2025). 9 L / min Approximate direct consumption of an efficient shower (typical range 5–8 L/min for electric/mains-fed; 9 L/min used as prudent reference). Equals ~90 L in 10 minutes (CCW, n.d.; Waterwise, 2007).
Car Travel 404g / mile Based on EPA (2023) average for a typical passenger gasoline vehicle. 5 L / day Indirect usage for fuel processing and manufacturing amortised daily (Source: Argonne National Lab).
Golf (1 Round) 9,000g / round Derived from Turtureanu et al. analyzing maintenance equipment, fertiliser production, and player transport share. 7,800 L / round Based on USGA data. Average course uses 300k-1M gallons/day. Figure represents per-player share assuming 150 players/day.

Bibliography

  • Consumer Council for Water (CCW). (n.d.). How much water do you use?
  • Department for Energy Security and Net Zero (DESNZ). (2025). Greenhouse gas reporting: conversion factors 2025.
  • Obringer, R., Rachunok, B., Maia-Silva, D., Arbabzadeh, M., Nateghi, R., & Madani, K. (2021). The overlooked environmental footprint of increasing Internet use. Resources, Conservation and Recycling, 167, 105389.
  • Purdue University. (2021, January 14). Turn off that camera during virtual meetings, environmental study says.
  • Waterwise. (2007). Evidence base for large-scale water efficiency in homes.
  • Viessmann. (2022, May 31). Which appliances use the most electricity at home?
  • Carbon Trust. (2021). Carbon impact of video streaming. White paper.
  • FIFA. (2022). FIFA World Cup Qatar 2022: Greenhouse gas accounting report.
  • International Energy Agency. (2020). The carbon footprint of streaming video: Fact-checking the headlines. IEA.
  • Li, P., Yang, J., Islam, M. A., & Ren, S. (2023). Making AI Less "Thirsty": Uncovering and Addressing the Secret Water Footprint of AI Models. arXiv preprint arXiv:2304.03271.
  • Carbon Brief. (2020). Factcheck: What is the carbon footprint of streaming video on Netflix? Carbon Brief.
  • The Guardian. (2021, October 29). Streaming’s dirty secret: How viewing Netflix creates vast quantities of CO₂. The Guardian.
  • Compare the Market. (n.d.). Social carbon footprint calculator. comparethemarket.com.au.
  • The Independent. (2024, October 29). Every email, TikTok and text we send is killing the planet. The Independent.
  • Ritchie, H. (2025). What’s the carbon footprint of using ChatGPT? Substack / Our World in Data.
  • Tomlinson, B., et al. (2024). The carbon emissions of writing and illustrating are lower for AI than for humans. Scientific Reports, 14, 3732.
  • CleanFi. (2021). Carbon footprint of an average U.S. golf course. Report.
  • Murray, A. (2021). An estimation of emissions from the English Premier League. University of Strathclyde (thesis).
  • Patterson, D., et al. (2021). Carbon emissions and large neural network training. arXiv preprint arXiv:2104.10350.
  • Poore, J., & Nemecek, T. (2018). Reducing food’s environmental impacts through producers and consumers. Science, 360(6392).
  • USGA. (n.d.). Water Resource Management. United States Golf Association.
  • Water Footprint Network. Product Water Footprints: Beef.
Copied to clipboard!