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

• Electricity demand is large and growing fast: the IEA estimates data centres consumed about 415 TWh in 2024 (~1.5% of global electricity) and projects roughly a doubling to ~945 TWh by 2030, with AI the main driver. In the U.S., Lawrence Berkeley National Laboratory/DOE estimate data centres used ~176 TWh in 2023 (~4.4% of U.S. electricity) and could reach 6.7–12% by 2028.
• Fast load growth is being met partly with fossil generation and is delaying coal/gas retirements in some regions; several utilities have proposed new gas plants explicitly to serve data-centre load, and grid emissions per unit of new demand can be high.
• Water use is real and locally significant: evaporative cooling consumes potable water, and studies (e.g., Mytton, npj Clean Water 2021; Li et al., 'Making AI Less Thirsty') document facility-level withdrawals of hundreds of thousands of gallons per day and note siting in water-stressed regions. Google's own reporting shows billions of gallons of water withdrawal annually.
• Hyperscaler emissions have risen despite clean-energy pledges: Google reported 2023 GHG emissions ~48% above its 2019 baseline, and Microsoft reported ~29–30% above its 2020 baseline, largely attributed to data-centre construction and AI infrastructure.
• Local externalities affect residents: diesel backup generators (criteria pollutants), constant low-frequency cooling-fan noise (documented complaints/litigation in Chandler AZ, Northern Virginia, and elsewhere), land conversion, and transmission-line buildout.
• Cost-shifting risk to households: researchers at Harvard's Electricity Law Initiative and multiple state regulators have raised concerns that special utility contracts and rate structures can shift grid-upgrade costs onto ordinary ratepayers; PJM's 2025/26 capacity auction cleared at record prices, with data-centre load growth a cited contributor.
• Job creation per dollar and per megawatt is low: Virginia's JLARC found the industry generates substantial tax revenue but relatively few permanent jobs relative to capital investment and energy footprint.
• Embodied impacts: server manufacturing, rare-earth/critical-mineral demand, and short hardware refresh cycles generate e-waste and upstream mining impacts.

• The word 'terrible' overstates the aggregate scale: data centres are still only ~1.5% of global electricity use, smaller than, for example, global road-transport, steel, or cement emissions, and far smaller than agriculture's water footprint (agriculture is ~70% of global freshwater withdrawals).
• Efficiency gains have been dramatic: Masanet et al. (Science, 2020) found global data-centre compute grew ~550% from 2010–2018 while energy use rose only ~6%, because of improvements in server efficiency, virtualisation, and the shift to hyperscale facilities. Average hyperscale PUE is ~1.1–1.2 versus ~2.0 for legacy on-premise rooms.
• Migrating workloads from small enterprise server rooms to cloud data centres typically reduces total energy and carbon per unit of compute; the 'data centre' category absorbing computing is not the same as net new environmental harm.
• Data centres are among the largest corporate purchasers of clean energy in the world; Amazon, Microsoft, Google, and Meta have collectively contracted tens of gigawatts of renewables and are financing new nuclear and geothermal capacity that would not otherwise exist. Several operators match 100% of annual consumption with carbon-free purchases and are pursuing 24/7 hourly matching.
• Cooling water intensity varies enormously by design; air-cooled, closed-loop, and reclaimed/non-potable-water designs exist and are increasingly mandated, and some operators (e.g., Microsoft's zero-water evaporation designs, Meta and Google water-replenishment programmes) are cutting potable water draw.
• 'Terrible for people' ignores benefits: data centres underpin hospitals, banking, emergency services, remote work, education, scientific computing, and weather/climate modelling; digital substitution (telework, videoconferencing, dematerialised media) offsets some emissions.
• Economic benefits are documented: JLARC found Virginia's data-centre industry supports tens of thousands of direct and indirect jobs and generates roughly $1 billion/year in local tax revenue that funds schools and services, and can lower residents' property-tax burden.
• The claim is a universal statement about a heterogeneous category: impacts depend heavily on grid mix, cooling design, siting, and local governance — a Nordic hydro-powered, heat-recovering facility (district-heating reuse in Denmark, Finland, Sweden) is not comparable to a gas-turbine-powered one in a water-stressed region.

78/100 — the lab’s own confidence in the score above, not a second reading of the claim.

International Energy Agencygov · 93 · International Energy Agencygov · 92 · U.S. Department of Energygov · 90 · Lawrence Berkeley National Laboratory (Energy Technologies Area)gov · 92 · Science (Masanet et al., 'Recalibrating global data center energy-use estimates')journal · 94 · npj Clean Water (Mytton, 'Data centre water consumption')journal · 88 · arXiv (Li et al., 'Making AI Less Thirsty')journal · 70 · Joule (de Vries, 'The growing energy footprint of artificial intelligence')journal · 85 · Google (2024 Environmental Report)other · 62 · Microsoft Environmental Sustainability Reportother · 60 · Virginia Joint Legislative Audit and Review Commission, 'Data Centers in Virginia' (2024)gov · 90 · Monitoring Analytics (PJM Independent Market Monitor)other · 82

• Data centers do have real environmental burdens: the IEA says data centers and data-transmission networks use roughly 1–1.3% of global electricity, and IEA’s Electricity 2024 report says electricity use from data centers, AI and cryptocurrency could roughly double from 2022 to 2026 if trends continue.
• A peer-reviewed Environmental Research Letters study found U.S. data centers have sizable electricity, carbon and water footprints, including indirect water use from electricity generation and direct cooling water use; these impacts can matter more in water-stressed regions.
• When data centers are built on fossil-heavy grids, their 24/7 power demand can increase greenhouse-gas emissions and local air-pollution burdens. They can also impose local costs such as land use, transmission infrastructure, cooling-water demand and backup-generator/noise concerns.
• The claim is strongest for poorly sited or poorly powered data centers: in communities with constrained water supplies, fossil-fuel electricity, weak grid capacity, or nearby residents exposed to construction/noise/diesel backup generators, the harms can be substantial.

• The statement is too sweeping. A Science study found global data-center energy use rose only modestly from 2010 to 2018 despite huge increases in workloads, because efficiency improvements were large; data centers were estimated at about 1% of global electricity use, not a dominant share of global energy use.
• Impacts vary greatly by facility, grid mix, cooling design, utilization, and procurement of low-carbon electricity. A renewable-powered, efficient data center in a cool climate has a very different footprint from one on a stressed fossil grid in a water-scarce area.
• Data centers also provide major social and economic benefits: cloud computing, hospitals’ records, scientific research, communications, banking, emergency services, and business productivity. The claim weighs harms but ignores these benefits to people.
• The best-supported conclusion is not that data centers are categorically 'terrible,' but that they are energy- and resource-intensive infrastructure whose impacts can be serious and are growing, especially without clean power, efficient cooling, transparency, and careful siting.

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

International Energy Agencyother · 92 · International Energy Agencyother · 92 · Sciencejournal · 95 · Environmental Research Lettersjournal · 93 · Lawrence Berkeley National Laboratory / U.S. Department of Energygov · 92 · International Energy Agencyother · 90

• Data centers are highly resource-intensive, accounting for approximately 1% to 1.5% of global electricity demand. With the rapid expansion of Artificial Intelligence (AI) and cryptocurrency, the International Energy Agency (IEA) projects this energy consumption could double by 2026.
• Cooling data center servers requires massive amounts of fresh water. A single large data center can consume millions of gallons of water per day, often straining local water supplies in drought-prone regions like the American Southwest.
• Local communities frequently oppose new data centers due to immediate negative impacts on people, including strain on local power grids (which can drive up utility rates for residents), massive land use, and severe, constant noise pollution generated by industrial cooling fans.
• The rapid turnover of servers and hardware to keep up with computing demands generates significant amounts of electronic waste (e-waste), which poses environmental and human health risks if not properly recycled.

• The claim that they are 'terrible for people' ignores the foundational societal benefits of data centers; they form the backbone of the modern internet, enabling global communication, advanced medical and scientific research, digital banking, and remote work.
• Data centers enable remote work and the virtualization of services, which can actually reduce broader societal greenhouse gas emissions by drastically cutting down on physical commuting and travel.
• The industry has achieved massive efficiency gains. A landmark 2020 study in the journal Science found that while global data center computing workloads increased by 550% between 2010 and 2018, their actual energy use grew by only 6%.
• Major data center operators (such as Google, Microsoft, Amazon, and Meta) are the largest corporate purchasers of renewable energy in the world. The EPA frequently lists tech giants at the very top of their Green Power Partnership list, as they often match 100% of their electricity consumption with renewable energy purchases and invest heavily in carbon-free grid initiatives.

90/100 — the lab’s own confidence in the score above, not a second reading of the claim.

International Energy Agency (IEA)other · 95 · Massachusetts Institute of Technology (MIT) Climate Portaledu · 95 · Sciencejournal · 98 · U.S. Environmental Protection Agency (EPA)gov · 98

• Data centers consume substantial and rapidly increasing amounts of electricity: globally they used about 415 TWh in 2024, or 1.5% of world electricity consumption, and the IEA projects demand could roughly double by 2030. This increases associated greenhouse-gas emissions, especially where electricity comes from fossil fuels.[https://www.iea.org/reports/energy-and-ai/executive-summary]
• U.S. data centers consumed approximately 176 TWh of electricity in 2023. Their electricity-related indirect water footprint was estimated at nearly 800 billion liters, while direct on-site water consumption was estimated at about 66 billion liters.[https://eta-publications.lbl.gov/sites/default/files/2024-12/us_data_center_energy_usage_report_lbnl-2001637_0.pdf]
• The environmental burden is locally concentrated: large facilities can intensify electricity-grid demand, require new generation and transmission, and compete for water in water-stressed regions. The Berkeley Lab report documents substantial energy, greenhouse-gas, and water impacts, although the severity varies by facility design, location, cooling system, and electricity mix.[https://eta-publications.lbl.gov/sites/default/files/2024-12/us_data_center_energy_usage_report_lbnl-2001637_0.pdf]
• The IEA estimates data centers currently cause about 180 million tonnes of indirect CO2 emissions from electricity use, with emissions projected to rise substantially through the decade in its central scenario.[https://www.iea.org/reports/energy-and-ai/ai-and-climate-change]

• The claim is overly absolute: data centers currently account for a relatively small share of global emissions—about 0.5% of combustion emissions according to the IEA—even though their electricity demand is growing quickly.[https://www.iea.org/reports/energy-and-ai/ai-and-climate-change]
• Data centers provide the infrastructure for communications, digital public services, commerce, research, banking, and other activities; the environmental assessment therefore depends partly on what services they enable and what activities they replace. The available evidence does not support treating every data center as uniformly harmful to people or the environment.
• The IEA projects data-center electricity use to reach about 3% of global electricity demand by 2030 in its central outlook, meaning the sector is significant but not the dominant source of global energy use or emissions.[https://www.iea.org/reports/energy-and-ai/energy-supply-for-ai]
• Impacts vary materially with efficiency, cooling technology, water source, local climate, grid mix, and use of low-carbon electricity. The Berkeley Lab analysis reports both rising aggregate demand and major efficiency changes, including a shift toward more efficient hyperscale and colocation facilities; this undermines the blanket claim that all facilities are equally damaging.[https://eta-publications.lbl.gov/sites/default/files/2024-12/lbnl-2024-united-states-data-center-energy-usage-report_1.pdf]

88/100 — the lab’s own confidence in the score above, not a second reading of the claim.

International Energy Agencyother · 96 · International Energy Agencyother · 96 · International Energy Agencyother · 96 · Lawrence Berkeley National Laboratory, U.S. Department of Energygov · 98 · Lawrence Berkeley National Laboratory, U.S. Department of Energygov · 98

• Data centers use a large and growing share of electricity (about 1–1.5% globally, higher in some countries), and growth from AI/cloud is projected to raise demand sharply; when grids are fossil-heavy this increases GHG emissions.
• Many facilities use large volumes of water for cooling, which can stress local supplies in drought-prone or water-scarce areas.
• Construction, land use, heat rejection, grid strain, noise, and e-waste create local environmental and quality-of-life impacts; some communities report higher costs or opposition.
• Rapid hyperscale/AI buildout has outpaced some grids and water systems, producing documented local conflicts and emissions from new generation.

• Efficiency gains (servers, cooling, PUE) have historically kept energy-use growth far below growth in compute/data traffic.
• Major operators buy large volumes of renewables/PPAs and are shifting toward carbon-free energy and better cooling; impact depends heavily on location and design, not an inherent property of all data centers.
• Centralized cloud is often more energy-efficient than equivalent on-premises IT; digital services also enable remote work and other substitutions that can cut other emissions.
• Facilities create jobs, tax base, and infrastructure investment; ‘for people’ impacts are mixed rather than uniformly terrible.
• Absolute wording (‘are terrible’) ignores variation, ongoing mitigation, and net benefits of the services they host.

78/100 — the lab’s own confidence in the score above, not a second reading of the claim.

International Energy Agencyother · 92 · International Energy Agencyother · 92 · Sciencejournal · 95 · U.S. Department of Energygov · 90