Written from 61 named sources · Aug 27 · first result Water Consumption Ranking: AI Data Centers vs. Everyday Goods and Services Prepared as of August 27, 2026 The Accounting Problem First: Why Every Popular Comparison Breaks Before ranking anything, we must fix the unit of account. The viral claim that "golf uses 100 times more water than data centers" and the rebuttal that "data centers are close behind" are both technically defensible — because they are measuring different things. Three distinctions do the work: Withdrawal vs. consumption. Withdrawal is water taken from a river, aquifer, or municipal main; consumption is the fraction evaporated, transpired, or otherwise not returned. Data centers using evaporative cooling consume 70–90% of what they withdraw [58]. Golf courses also consume 70–90% of what they apply to turf [36], but their headline 1.63 million acre-feet figure is applied water — a large share soaks back into the ground or runs off. Comparing golf's applied number (531 billion gallons) to data centers' direct consumed number (17 billion gallons) stacks the deck before the analysis begins. Direct vs. indirect/embedded. A data center's operational water is the cooling tower and humidification. But the electricity powering it carries an embedded water cost at the thermoelectric power plant. Lawrence Berkeley National Laboratory estimated that embedded cost at roughly 211 billion gallons in 2023 — about 12 times the direct on-site figure [36][59]. Nearly every agricultural and livestock number you have encountered is, by contrast, already an embedded life-cycle total. Beef "uses" 1,675 gallons per pound only if you count the rainfall on pasture and feed crops; the blue-and-gray portion (water actually drawn from surface or groundwater that could serve humans) is 158 gallons per pound [56]. Mixing scopes is how the gap between two sectors swings by orders of magnitude. Green vs. blue vs. gray. Green water is rainfall or soil moisture that would fall on the landscape regardless of whether a crop or golf course occupied it. Blue water is diverted surface or groundwater — the societally contested kind. Gray water is the assimilation volume needed to dilute pollution. More than 90% of beef's water footprint is green; nearly all of a data center's direct footprint is blue [56]. Environmentally, a gallon of blue water withdrawn from a stressed municipal system is not interchangeable with a gallon of rain on pasture. Every number in the ranking below is labeled by scope. If a comparison leaves the scope unstated, treat it as unreliable. The Ranking: 2020 vs. 2026 vs. 2029–2031 All figures are U.S.-focused unless noted; volumes are billion gallons per year unless marked otherwise. "Applied" means water delivered to turf or fields; "consumed" means not returned. One billion gallons = 3.785 gigaliters (GL) = 0.003785 km³. Sector Type / scope 2020 (B gal/yr) 2026 (B gal/yr) 2029–2031 (proj.) Metric (GL/yr) WUE / intensity notes Environmental significance U.S. irrigated agriculture (all crops) Food & fiber; consumed, blue + green ~27,000 ~27,000 ~27,000 ~102,200 GL (~102 km³) ≈73 B gal/day irrigation; 1–4 acre-ft/acre by crop [36] Dominant national user; one to two orders of magnitude above everything else California almonds (subset of the agriculture row) Irrigated specialty crop; nearly all blue, consumed 1,600–1,800 1,500–1,800 1,400–1,700 ~5,800–6,800 GL 3–3.5 acre-ft/acre on ~1.56M acres; ≈1 gal per almond [6] ~15–17% of California's developed agricultural water; concentrated in one stressed state, 70% exported Golf courses (U.S.) Recreational; applied irrigation, 70–90% consumed ~550 (applied) ~531 (applied; 2024 survey) ~500–530 (flat to declining) ~2,010 GL applied; ~1,400–1,800 GL consumed Median 1.1 acre-ft/acre; down 31% since 2005, 3.2% since 2020 [11][46] Largest discretionary use; only ~7–9% from municipal potable supply [38] Data centers (U.S.) — direct on-site only Industrial cooling; blue, 70–90% consumed ~8–10 ~22–27 (2025 Bluefield: 62 MGD consumed) ~30–45 ~80–170 GL WUE 0.5–1.5 L/kWh across operators [33]; 97% municipal-sourced [55] Fastest-growing category; hyper-local peak-day impacts Data centers (U.S.) — total incl. power-plant indirect Industrial + embedded electricity water ~140–160 ~200–240 (228 in 2023 per LBNL) ~250–350 (Bluefield: indirect 54→91 B gal, 2025→2030) ~750–1,300 GL Indirect = 4–12× direct; Bluefield projects 72% of DC water off-site by 2030 [42] Comparable to golf's consumed total; water risk migrating upstream to power generation U.S. public water systems (reference row, not a comparator) Municipal supply baseline ~12,900 withdrawal / ~1,540 consumed ~13,000 / ~1,550 ~13,000 / ~1,550 ~48,900 GL withdrawal 35,400 MGD total public withdrawal; 4,219 MGD consumptive [38] The reservoir data centers and households compete within Livestock / meat (embedded in the agriculture row above) Because livestock water use is overwhelmingly the green water used to grow feed — already counted in the agriculture total — presenting a separate national "billion gallons" figure would double-count. The honest comparison is per-pound intensity [56]: Product Total water footprint Blue + gray (contested) water Per kg (blue + gray) Beef ~1,675 gal/lb 158 gal/lb ~1,320 L/kg Pork ~545 gal/lb 146 gal/lb ~1,220 L/kg Poultry ~257 gal/lb 55 gal/lb ~460 L/kg More than 85% of livestock water use is feed production; more than 90% of beef's total footprint is green rainfall [56]. On the blue-and-gray basis that matters for competition with households and industry, beef is roughly comparable to pork, not 10 times thirstier. Where AI Data Centers Actually Rank Direct on-site: near the bottom. In 2026, data center direct consumption is on the order of 25 billion gallons — about 1/20th of golf's applied water, roughly 1/70th of California almonds alone, and well under 0.1% of irrigated agriculture. A single soft-drink supply chain (169–310 liters per 0.5-liter bottle, 99.7% of it in agricultural ingredients) illustrates the same structural point: industrial operations are almost always a rounding error next to the agricultural inputs behind them [28]. Total footprint including electricity: roughly the size of golf's consumed volume. Adding the indirect power-plant water brings U.S. data centers to roughly 200–240 billion gallons today — the same order as golf courses' consumptive use of ~400–480 billion gallons. The two sectors are closing but have not crossed, except under the most aggressive grid-average assumptions [36]. The comparison depends entirely on whether the marginal new electricity comes from gas-and-renewables (~0.8 L/kWh) or averages in hydropower evaporation (~4.5 L/kWh). Bluefield, using a different intensity basis than LBNL, projects total U.S. data center water (direct + indirect) at roughly 125 billion gallons by 2030 [42] — below the LBNL-derived 228 billion gallon 2023 figure. The spread is a measurement dispute, not a rhetorical one. The trajectory is the headline. Data center water use is compounding at 15–25% annually while golf declines and agriculture is flat. By 2029–2031, even the moderate scenario has U.S. data center total water approaching 300 billion gallons, and the water risk shifts decisively off-site: 72% of data-center-related water consumption is projected to occur at power plants by 2030 [42]. Direct on-site consumption grows slowly as operators pivot to closed-loop and direct-to-chip liquid cooling, which cut water 20–90% [2]. Why Localized Stress Matters More Than National Totals Data centers use water differently from agriculture in two ways that national aggregates hide. First, the destructive mismatch: a 100 MW campus consumes more than it withdraws. A typical residential user returns 85–95% of the water they take from the pipe. A data center using evaporative cooling returns 10–30%; the rest goes up as vapor. On a withdrawal basis, data centers are trivial. On a consumptive basis relative to the municipal system's total consumptive use, they jump to 4–7% by 2030 [38]. Municipal systems are not volume-limited; they are replenishment-limited. Second, the peaking factor. Data centers' water demand is not steady. A facility operating with dry cooling 90% of the year can spike its evaporative use on the hottest days, producing peaking factors of 6 to over 30 relative to average daily demand [38][43]. Planned campuses request 0.7–8 million gallons per day of capacity. Water utilities plan around peak-day flow. When a single hyperscale campus requests capacity equal to 30 times the previous largest municipal user, one building can exhaust the uncommitted surplus of a small or medium system. Agriculture and golf are spread across 14,000 facilities and entire watersheds; data centers concentrate industrial-scale demand at a single address. The $10–58 billion range for U.S. public-water-system upgrades needed through 2030 [38] reflects this geometry. It is small compared to the AI industry's capital spending, but concentrated in communities with 12–19 million households already facing water affordability stress. The 2020–2031 Trajectory in One View Agriculture's dominance is not forecast to change. Golf's decline is real but slowing. Almonds are declining modestly from their 2021 acreage peak, but remain a multi-trillion-gallon California issue — roughly triple golf's national total on a consumptive basis, concentrated in a single drought-prone state, with 70% exported [6]. The only category with an unmistakable acceleration is data centers, and the nature of their water risk is changing faster than the magnitude. The next three to five years will see less argument over whether data centers are "thirsty" in absolute terms — they will remain small against agriculture — and more argument over who pays for the peak-day pipe capacity and whether off-site power-plant water gets counted in corporate water-positive claims. The latter is uncertain: Bluefield's 2030 projection (91 billion gallons indirect) and LBNL's 2023-derived figure (211 billion gallons indirect) disagree by more than a factor of two because they disagree about whether new marginal load draws on the grid-average mix or on marginal gas and renewables. That single methodological choice, not the cooling technology inside the data hall, is now the largest variable in the AI water debate. Sources [2] How thirsty is the AI Industry? Bordier — https://www.bordier.com/how-thirsty-is-the-ai-industry/ [6] California Almond Water Usage: Updated — https://www.c-win.org/cwin-water-blog/2024/9/23/california-almond-water-usage-updated [11] Golf Course Environmental Profile — https://www.gcsaa.org/docs/default-source/what-we-do/gcep-phase-4-water-report.pdf?sfvrsn=c829dd3e_0 [28] Corporate Water Footprint Accounting and Impact Assessment: The Case of the Water Footprint of a Sugar-Containing Carbonated Beverage Water Resources Management Springer Nature Link — https://link.springer.com/article/10.1007/s11269-010-9723-8 [33] Water usage effectiveness — https://en.wikipedia.org/wiki/Water_usage_effectiveness [36] Golf uses more water than data centers. By 2x, or by 100x, depending how you count. • Buttondown — https://buttondown.com/napkinquest/archive/golf-uses-more-water-than-data-centers-by-2x-or [38] 1Introduction — https://arxiv.org/html/2603.02705v2 [42] Data Center Electricity Demand Nearly Doubles to 8.9% by 2030, Shifting Water Risk from Cooling to Power Generation - Bluefield Research — https://www.bluefieldresearch.com/ns/data-center-electricity-demand-nearly-doubles-to-8-9-by-2030/ [43] How Much Water Does AI Use? The $58 Billion Risk — https://www.forbes.com/sites/robertszczerba/2026/07/21/how-much-water-does-ai-use-the-58-billion-risk/ [46] Golf courses reduce water usage by 31 percent according ... — https://www.gcsaa.org/who-we-are/media/news-release/2025-news-releases/2025/12/30/golf-courses-reduce-water-usage-by-31-percent-according-to-national-survey [55] U.S. Water-Related Expenditures for Data Centers to Exceed US$4.1 Billion Through 2030 - Bluefield Research — https://www.bluefieldresearch.com/ns/u-s-water-related-expenditures-for-data-centers-to-exceed-us4-1-billion-through-2030 [56] Does beef production really use that much water? — https://ksubci.org/2020/11/16/does-beef-production-really-use-that-much-water [58] Data Center Water Use — https://mostpolicyinitiative.org/science-note/data-center-water-use [59] Hyperscale Data Center Water Report - Bill Carley — https://billcarley.substack.com/p/hyperscale-data-center-water-report AI's environmental costs threaten water, land and climate — https://news.un.org/en/story/2026/06/1167658 Multiscale Assessment of Agricultural Consumptive Water Use in California's Central Valley — https://pmc.ncbi.nlm.nih.gov/articles/PMC8519118/ Assessment of the water footprint of the brewery industry — https://www.sciencedirect.com/science/article/pii/S2590123024011836 Assessment of the water footprint of the brewery industry: The case of Heineken, Addis Ababa, Ethiopia — https://doi.org/10.1016/j.rineng.2024.102928 Carbon and water footprints assessment of cotton jeans using the method based on modularity: A full life cycle perspective — https://doi.org/10.1016/j.jclepro.2021.130042 How Much Water Does Golf Use and Where Does It Come From? — https://usgatero.msu.edu/v11/216335.pdf Journal list menu — https://acsess.onlinelibrary.wiley.com/doi/10.1002/cft2.20182 Livestock and Water Resources: A Comparative Study of Water Footprint in Different Farming Systems — https://www.mdpi.com/2071-1050/17/5/2251 Rising Emissions, Depleting Water and Vanishing Land— ... — https://unu.edu/ Rising Emissions, Depleting Water and Vanishing Land—UN Scientists: AI Is Threatening Natural Resources for Billions United Nations University — https://unu.edu/inweh/news/environmental-cost-of-AIs-Enrgy-use-carbon-water-and-land-footprints The water footprint of poultry, pork and beef: A comparative ... — https://www.researchgate.net/publication/259172043_The_water_footprint_of_poultry_pork_and_beef_A_comparative_study_in_different_countries_and_production_systems The water footprint of poultry, pork and beef: A comparative study in different countries and production systems — https://www.sciencedirect.com/science/article/pii/S2212371713000024 Virtual carbon and water flows embodied in global fashion trade - a case study of denim products — https://discovery.ucl.ac.uk/id/eprint/10128720/3/Meng_Revised%20manuscript%20-%20clean%20version.pdf Water use and management practices on U.S. golf courses — https://doi.org/10.1002/cft2.20182 Water-indexed benefits and impacts of California almonds — https://doi.org/10.1016/j.ecolind.2017.12.063 Beverage Industry Works to Cap Its Water Use — https://www.nytimes.com/2011/03/22/business/energy-environment/22iht-rbog-beverage-22.html What uses more water: golf courses or data centers? — https://golfweek.usatoday.com/story/sports/golf/courses/2026/07/09/what-uses-more-water-golf-courses-or-data-centers/90855401007/ What uses more water: golf courses or data centers? — https://golfweek.usatoday.com/story/sports/golf/courses/2026/07/09/what-uses-more-water-golf-courses-or-data-centers/90855401007 California Nut Crop Expansion — https://www.foodandwaterwatch.org/wp-content/uploads/2022/07/California-Nut-Crop-Expansion-Research-For-Media.docx.pdf doi:10.1016/j.ecolecon.2005.11.027 — https://www.waterfootprint.org/resources/multimediahub/Chapagain_et_al_2006_cotton_2.pdf Golf Course Environmental Profile — https://www.gcsaa.org/docs/default-source/environment/22_waterreport_web.pdf?sfvrsn=862fcd3e_0 Golf Course Environmental Profile — https://www.gcsaa.org/docs/default-source/what-we-do/gcep-phase-4-water-report.pdf Golf Course Environmental Profile — https://www.gcsaa.org/docs/default-source/environment/phase-3-water-survey-full-report.pdf No, Almonds Don't Use 10 Percent of California's Water — https://www.almonds.org/why-almonds/almond-living-magazine/no-almonds-dont-use-10-percent-californias-water Survey of Water Use and Management Practices on US ... — http://journals.ashs.org/view/journals/horttech/35/5/article-p848.xml The countries with a relatively high rate of evapotranspiration and a high gross national income per capita (which often results in large consumption of meat and industrial goods) have large water footprints, such as: Portugal (2260 m3/yr/cap), Italy (23 — https://www.waterfootprint.org/resources/Report39-WaterFootprintCarbonatedBeverage.pdf Volume 1: Main Report — https://www.waterfootprint.org/resources/Report-48-WaterFootprint-AnimalProducts-Vol1.pdf Water - Beef Research — https://www.beefresearch.org/resources/beef-sustainability/fact-sheets/water waterfootprint.org — https://waterfootprint.org/resources/A_guide_to_reduce_water_footprint_of_cotton_cultivation.pdf Do water footprint data keep in mind that the animal still ... — https://www.reddit.com/r/DebateAVegan/comments/ta3d0v/do_water_footprint_data_keep_in_mind_that_the A Guide to Data Center Water Usage Effectiveness (WUE) and Best Practices — https://www.datacenterknowledge.com/cooling/a-guide-to-data-center-water-usage-effectiveness-wue-and-best-practices AI is set to consume up to 600 billion gallons of water by 2030 — rising energy consumption primarily to blame as data center power demands rise Tom's Hardware — https://www.tomshardware.com/tech-industry/ai-is-set-to-consume-up-to-600-billion-gallons-of-water-by-2030-rising-energy-consumption-primarily-to-blame-as-data-center-power-demands-rise AI Used More Water Than Entire Countries. Here's the Data. · WaterFreeChat — https://waterfreechat.com/blog/ai-used-more-water-than AI's Real Infrastructure Limit Isn't Power. It's Water, and It's Hyperlocal — Codexical — https://www.codexical.com/posts/2026-07-06-ai-data-center-water-wall Data Center Electricity Demand Nearly Doubles to 8.9% by 2030, Shifting Water Risk from Cooling to Power Generation - Bluefield Research — https://www.bluefieldresearch.com/ns/data-center-electricity-demand-nearly-doubles-to-8-9-by-2030 Data Center Water Secrecy Hurts Communities (and the Industry Itself) - Bluefield Research — https://www.bluefieldresearch.com/data-center-water-secrecy-hurts-communities-and-the-industry-itself Data Centers in the AI Era: — https://www.globalefficiencyintel.com/s/GEI-data-centers-report-1152025-clean-E4.pdf Do golf courses use more water than data centers? Alabama golf course — https://www.montgomeryadvertiser.com/story/news/2026/07/07/do-golf-courses-use-more-water-than-data-centers-alabama-golf-course/90833406007 GCSAA reveals water survey results - Golf Course Industry — https://www.golfcourseindustry.com/news/gcsaa-reveals-water-survey-results Golf Course Irrigation Management: Save Water & Money — https://fairwaycontrol.com/blog/golf-course-irrigation-management-saving-water-and-money Measuring energy and water efficiency for Microsoft ... — https://datacenters.microsoft.com/ WATER USAGE EFFECTIVENESS (WUE™): A GREEN GRID DATA — https://img1.wsimg.com/blobby/go/6992f185-d35b-4204-bd47-2dea1008c968/downloads/The-Green-Grid-White-Paper-35-WUE-Usage-Guidel.pdf?ver=1724446188262 What is WUE in Data Centers? Trane Commercial HVAC — https://www.trane.com/commercial/north-america/us/en/about-us/newsroom/glossary/water-usage-effectiveness.html BOARD-INVITED REVIEW: — https://ayhoekstra.nl/pubs/Legesse-et-al-2017.pdf Fashion's environmental price tag — https://publications.parliament.uk/ Water footprint of crop and animal products: a comparison — https://web.archive.org/web/20230113183241/https:/waterfootprint.org/en/water-footprint/product-water-footprint/water-footprint-crop-and-animal-products/ Water Usage Reduction at US Courses - Asian Golf Industry Federation — https://agif.asia/2026/01/water-usage-reduction-at-us-courses