“Data centers are terrible for the environment and for people.”

• Electricity demand is large and growing fast: the IEA estimates data centres consumed roughly 415 TWh in 2024 (~1.5% of global electricity) and projects that to roughly double by 2030, with AI the main driver. Lawrence Berkeley National Laboratory's 2024 report estimates US data centres used ~176 TWh in 2023 (~4.4% of US electricity) and could reach 6.7–12% of US electricity by 2028.
• Much of that electricity still comes from fossil generation, so growth translates into real emissions. Microsoft reported its total (Scope 1–3) emissions rose ~29% from its 2020 baseline, and Google reported emissions up ~48% since 2019, both attributing much of the increase to data centre construction and AI compute.
• Water consumption is material and locally concentrated: peer-reviewed work (Siddik, Shehabi & Marston, Environmental Research Letters 2021) estimated US data centres directly consume ~0.66 billion m³ of water per year plus large indirect consumption via electricity generation, and a substantial share is drawn from watersheds already under stress.
• Local health and nuisance impacts are documented in specific cases: backup and bridging generators burning diesel or gas emit NOx and PM2.5; xAI's Memphis facility drew formal complaints and litigation threats from the NAACP and the Southern Environmental Law Center over unpermitted gas turbines in a majority-Black neighborhood already burdened by industrial pollution. Noise complaints from cooling equipment have prompted zoning fights in Virginia, Georgia and Arizona.
• A 2024 preprint by Ren and colleagues (UC Riverside/Caltech) attempted to quantify public-health costs of AI data centre air pollution in the US, estimating billions of dollars in annual health damages by 2030 if growth continues on trend.
• Grid and ratepayer stress: Virginia's legislative watchdog JLARC (2024) concluded data centre demand could roughly double the state's electricity needs within a decade and that, absent rate design changes, non-data-centre customers could bear some transmission and capacity costs. Utilities in PJM and elsewhere have delayed coal retirements or approved new gas plants citing data centre load.
• Job benefits are modest relative to footprint and subsidies: large data centres typically employ dozens to a few hundred permanent staff while receiving substantial tax abatements, a pattern documented in multiple state audits and investigative reporting.
• Land, construction, and embodied-carbon impacts (concrete, steel, chips, backup batteries) are significant and are often excluded from operational 'carbon neutral' claims that rely on unbundled renewable energy certificates rather than hourly clean power.

• The word 'terrible' is an absolute; the measurable aggregate impact is real but bounded. At ~1.5% of global electricity, data centres consume less than, for example, global aluminium production or residential air conditioning, and far less than transport or agriculture.
• Efficiency gains have been dramatic: Masanet et al. (Science, 2020) found global data centre compute output rose ~550% from 2010–2018 while energy use rose only ~6%, due to server efficiency, virtualisation and the shift to hyperscale facilities. Average PUE fell from ~2.5 in 2007 to ~1.5–1.6 today, and hyperscale campuses commonly run 1.1–1.2.
• Hyperscale operators are among the world's largest corporate buyers of clean energy; Google and Microsoft match (or aim to match) 100% of annual consumption with renewables and are funding new nuclear, geothermal and storage capacity that can benefit the wider grid. Data centre offtake agreements have financed a meaningful share of new US renewable build.
• Water context matters: many facilities now use closed-loop or air-cooled designs (most large Northern Virginia campuses are predominantly air-cooled), and some operators report water-positive targets. Total US data centre water consumption is a small fraction of that used by thermoelectric power generation, agriculture, or golf courses; the problem is local siting, not aggregate volume.
• Cloud consolidation generally reduces emissions per unit of computing versus the on-premises servers it replaces, and digital services substitute for physically intensive activities (travel, paper, physical media, in-person commerce), producing offsetting benefits that the blanket claim ignores.
• 'Terrible for people' ignores substantial benefits: data centres underpin hospitals, banking, emergency services, scientific computing, weather forecasting and climate modelling, and generate large local property-tax revenues — Loudoun County, Virginia derives roughly a third of its general fund from data centre taxes, funding schools and services.
• Harms are highly site- and design-dependent. A facility on a low-carbon grid with air cooling and no on-site fossil turbines has a very different profile from a gas-turbine-powered site in a polluted neighbourhood. A categorical claim about all data centres is not supportable.
• Regulatory tools are being applied: EPA permitting, state rate-design reforms (special data centre tariffs in Ohio, Georgia, Virginia proposals), and EU Energy Efficiency Directive reporting requirements are already changing the cost and impact allocation.

76/100 — the lab’s own confidence in the score above, not a second reading of the claim.

International Energy Agencygov · 92 · Lawrence Berkeley National Laboratory (US DOE)gov · 93 · Science (Masanet et al., 2020)journal · 95 · Environmental Research Letters (Siddik, Shehabi & Marston, 2021)journal · 90 · arXiv preprint, 'Making AI Less Thirsty' (Li et al.)other · 68 · arXiv preprint, 'The Unpaid Toll' (Ren et al.)other · 65 · Virginia Joint Legislative Audit and Review Commissiongov · 90 · Microsoft Environmental Sustainability Reportother · 58 · Google Environmental Reportother · 58 · Uptime Institute Global Data Center Surveyother · 78 · US Energy Information Administrationgov · 93 · Southern Environmental Law Center (xAI Memphis turbine complaints)other · 65

• Data centers consume large and growing amounts of electricity. IEA reported that data centers, AI and cryptocurrency could roughly double electricity demand from 2022 to 2026, and LBNL estimated U.S. data centers used about 4.4% of U.S. electricity in 2023 with a possible rise to 6.7%-12% by 2028.
• If powered by fossil-heavy grids, that electricity demand contributes to greenhouse-gas emissions and conventional air pollution; EPA documents the environmental impacts of electricity generation, especially from fossil fuels.
• Data centers can use substantial water for cooling. AP reported large water-use increases by Microsoft and Google during the AI buildout and highlighted community concerns where data centers draw from local water systems.
• Local impacts can include noise from cooling equipment and backup generators; WHO finds environmental noise can harm health, so poorly sited or poorly mitigated data centers can negatively affect nearby residents.
• Data centers require frequent hardware replacement and large amounts of servers, networking equipment and batteries, contributing to material demand and e-waste pressures, though this is part of the broader digital-technology footprint rather than unique to data centers.

• The blanket claim is too broad: data centers are not uniformly 'terrible.' Their impact depends heavily on grid mix, cooling design, water source, siting, utilization rates and whether they displace less efficient computing.
• Historically, efficiency gains have greatly moderated energy growth. A 2020 Science paper found global data-center workloads rose sharply from 2010-2018 while data-center energy use rose much more slowly, reaching about 1% of global electricity use.
• IEA’s tracking similarly shows data centers and data transmission networks have represented a relatively small share of global electricity use compared with sectors such as transport, industry and building heating/cooling, though growth pressure is increasing.
• Large cloud data centers can be more energy-efficient than scattered on-premise servers because they run at higher utilization and invest in advanced cooling and power management; consolidation can reduce computing energy per unit of work.
• Data centers provide major social benefits: cloud services, hospitals and research computing, emergency communications, financial systems, remote work, education, AI services and government services. These benefits do not erase environmental costs, but they make the categorical 'for people' claim misleading.

84/100 — the lab’s own confidence in the score above, not a second reading of the claim.

International Energy Agencygov · 95 · International Energy Agencygov · 94 · Lawrence Berkeley National Laboratorygov · 95 · Sciencejournal · 96 · U.S. Environmental Protection Agencygov · 93 · Associated Pressnews · 88 · World Health Organizationgov · 92

• Data centers are highly energy-intensive. The International Energy Agency (IEA) estimates they accounted for 1% to 1.5% of global electricity use in 2022, leading to significant greenhouse gas emissions when regional grids rely on fossil fuels.
• Evaporative cooling systems used in many data centers consume millions of gallons of fresh water, directly competing with local community needs and exacerbating water stress in drought-prone regions.
• Local communities near hyperscale data centers frequently report negative quality-of-life impacts, including persistent low-frequency noise pollution from massive cooling fans and industrial HVAC systems.
• The rapid expansion of AI and cloud computing puts immense strain on local electrical grids, which can necessitate grid upgrades, delay the retirement of fossil fuel plants, and potentially increase utility costs for local residents.
• The constant need to upgrade server hardware (typically every 3 to 5 years) generates significant electronic waste, which poses environmental and human health hazards if not properly recycled.

• Major data center operators (such as Amazon, Google, and Microsoft) are the world's largest corporate purchasers of renewable energy, aggressively funding new wind and solar projects to match their electricity consumption.
• Data center energy efficiency has improved dramatically over the last decade. A landmark study published in Science found that while global data center computing instances increased by 550% between 2010 and 2018, their total energy use grew by only 6%.
• Data centers provide the foundational infrastructure for the digital economy, enabling remote work, virtual meetings, and smart logistics, all of which substantially reduce global carbon emissions associated with commuting and physical transportation.
• Many large tech companies are implementing circular economy practices, refurbishing components, and achieving 'zero waste to landfill' certifications to drastically reduce e-waste.
• Operators are increasingly exploring advanced, low-impact cooling technologies, such as closed-loop liquid cooling systems that consume zero water after their initial charge.

95/100 — the lab’s own confidence in the score above, not a second reading of the claim.

International Energy Agency (IEA)gov · 95 · Sciencejournal · 95 · Cornell University (arXiv)edu · 85

• Global data centers consume a large and rapidly growing share of electricity, making them a significant and fast‑increasing source of greenhouse gas emissions. The International Energy Agency (IEA) estimates data centers used around 415 TWh in 2024 (about 1.5% of global electricity), with emissions from electricity use around 180–300 Mt CO₂ and potentially reaching 500 Mt CO₂ in high‑growth scenarios.[2][6][11][10][14]
• Recent analyses suggest that some official estimates may understate the true carbon footprint of data centers, once manufacturing, construction, and other lifecycle impacts are included. One economic analysis projects that overall data‑center emissions could reach around 286 Mt CO₂ by 2025, about 57% higher than IEA’s estimate, and could more than double to 643 Mt CO₂ by 2030 without grid decarbonization, implying large climate damages.[1][8][15]
• The scale of electricity and water use by data centers already rivals large countries. A United Nations University report summarized by AP found that global data centers used about 448 TWh of electricity in the previous year—more than all but about 10 countries—and produced ~208 million tons of CO₂, similar to Argentina’s emissions; it estimated water use at ~1.2 trillion gallons for that energy and projected data centers could consume nearly 3% of global electricity by 2030 and emit ~440 Mt CO₂.[12][2][11]
• Data centers often rely heavily on fossil‑fuel‑based grids and backup diesel generators, leading to localized air pollution and health risks. Studies of U.S. and global hyperscale data centers report that a majority of their electricity comes from fossil fuels, generating tens of millions of tons of CO₂ per year and elevated levels of nitrogen oxides and fine particulate matter, which increase respiratory, cardiovascular, and cancer risks in nearby communities.[7][9][13][6]
• Water use for cooling and for upstream electricity generation can severely strain local water resources. A public health analysis reports that a hyperscale data center can use 3–7 million gallons of water per day, and in some regions data centers account for up to 57% of potable cooling water withdrawals, exacerbating water scarcity and increasing risks of dehydration, poor hygiene, and waterborne disease in stressed communities.[13][4][10]
• Waste heat from large urban data centers can noticeably raise local temperatures, worsening urban heat‑island effects that disproportionately affect nearby residents. Research summarized by Yale’s Environment School indicates that a large data center can emit waste heat comparable to tens or hundreds of thousands of households, increasing land surface temperatures by up to about 16°F and air temperatures in adjacent neighborhoods by several degrees.[4]
• Lifecycle impacts of data center infrastructure—servers, chips, cooling systems, buildings, backup power, and grid upgrades—carry additional environmental burdens, including embodied carbon, resource extraction, and pollution from manufacturing and construction, beyond operational electricity and water use.[4][8][15]
• Rapid expansion driven by AI workloads is expected to accelerate energy demand, emissions, and resource use. IEA and other analyses project that electricity demand for data centers could approximately double or more by 2030, rising to roughly 1,000 TWh or ~3% of global electricity, making data centers among the fastest growing sources of energy‑sector emissions.[2][11][12][14]
• In some locations, community impacts are already contentious: data centers can compete with residents and agriculture for limited water and grid capacity, potentially driving up electricity prices or forcing new fossil plants, thereby imposing environmental and economic burdens on nearby people.[10][12][13][4]

• Data centers currently account for a relatively small share of total global greenhouse gas emissions compared with major sectors like transportation, buildings, and heavy industry. IEA estimates that data centers contribute around 0.5–1% of global energy‑sector combustion emissions, even under growth scenarios, and remain well below the shares from sectors like road transport or power generation overall.[6][2][11]
• Efficiency improvements in hardware, cooling, and software have significantly constrained the growth of data‑center energy use relative to exploding digital demand. Studies of the ICT sector show that between 2010 and 2018, global data‑center energy use rose only modestly (around 6%) despite much higher increases in workloads, due to efficiency gains and optimization.[15][2]
• Many large operators are investing heavily in renewable energy, low‑carbon procurement, and more efficient designs, which can substantially reduce the carbon intensity of data centers. Analyses from IEA and EU authorities note that data centers on low‑carbon grids have much lower emissions than those using coal‑ or gas‑heavy electricity, and policy pressure in regions like the EU is pushing operators toward climate‑neutral and energy‑efficient practices.[2][10][11][4][15]
• Digital services enabled by data centers can offset environmental impacts by improving efficiency in other sectors (e.g., telecommuting, smart logistics, virtual meetings, digitalization of services) that reduce travel, material use, and other high‑emission activities. While difficult to quantify precisely, ICT and cloud computing are often cited as enabling emissions reductions that could partially counterbalance data‑center footprints.[8][15]
• Compared with many industrial facilities, data centers generally do not emit large quantities of direct toxic substances on‑site; their primary direct local impacts are noise, land use, and waste heat, while most harmful pollutants originate upstream at power plants. This means that health harms are often mediated by the broader energy system rather than inherent to data‑center technology itself.[4][6][13]
• Policy frameworks and regulations, especially in the EU and some U.S. jurisdictions, are starting to require reporting, efficiency standards, and limits on water and energy use, which can mitigate the worst environmental and community impacts over time if effectively implemented.[10][4][15]
• Claims that data centers are uniformly “terrible for people” can be overstated because they also underpin critical services (healthcare data, emergency communications, education platforms, financial infrastructure) that benefit society, including in ways that can improve health, safety, and economic opportunity. The net human impact is mixed and context‑dependent rather than universally negative.[4][8][15]

92/100 — the lab’s own confidence in the score above, not a second reading of the claim.

International Energy Agencygov · 95 · International Energy Agencygov · 95 · International Energy Agencygov · 95 · European Commission / EU Energygov · 93 · National Institutes of Health / PMCjournal · 95 · National Institutes of Health / PMCjournal · 95 · Deutsche Wellenews · 88 · Deutsche Wellenews · 88 · Associated Pressnews · 92 · Yale School of the Environmentedu · 92 · Harvard Universityedu · 93 · SSRNjournal · 85

• Data centers use large amounts of electricity (on the order of 1–2% of global demand, with faster growth from AI and cloud), adding emissions wherever the grid is fossil-heavy.
• Cooling often consumes large volumes of water, which can compete with households and farms in drought-prone regions.
• On-site diesel generators for backup power emit NOx, particulates, and other pollutants that affect nearby air quality.
• Hardware production, mineral supply chains, and e-waste create additional environmental and occupational-health burdens.
• Some host communities report noise from cooling plant, grid strain, higher local power prices, and land/water conflicts.

• Efficiency gains (lower PUE, better servers, virtualization) have historically kept energy growth far below growth in data traffic.
• Many large operators match a high share of electricity with renewables or PPAs, lowering carbon intensity versus the average grid.
• Centralized facilities are typically more efficient than equivalent on-premises or highly distributed computing.
• The same infrastructure enables remote work, logistics and energy optimization, telemedicine, education, and climate research that can cut other emissions and improve living standards.
• Jobs, taxes, and the economic value of digital services are large; net effects on people are not uniformly negative.

72/100 — the lab’s own confidence in the score above, not a second reading of the claim.

International Energy Agencyother · 95 · Sciencejournal · 97 · U.S. Department of Energygov · 92