Data drought: AI data centers face backlash over water secrecy and soaring thirst

AI datacenter water consumption is highly variable and largely dictated by the cooling technologies and infrastructure design choices employed. For instance, Google's Oregon datacenter relies heavily on evaporative cooling to manage chip temperatures, resulting in substantial water use that accounts for roughly a quarter of the entire city of The Dalles' water consumption. This localized impact became a flashpoint when Google's water sourcing expansion plans into the Mount Hood National Forest sparked environmental concerns, highlighting the tangible strain AI datacenters can place on regional water resources. Importantly, water use in these facilities primarily stems from cooling operations and power generation, with chip manufacturing water consumption excluded from this analysis as it is a broader semiconductor industry issue.

Simplistic comparisons of AI datacenter water use to everyday items like fast food restaurants risk obscuring the nuanced tradeoffs between water and carbon footprints inherent in these operations. A notable example is the SemiAnalysis post equating xAI’s Colossus 2 datacenter water use to that of 2.5 In-and-Out burger restaurants, which David Sacks criticized as a 'narrative violation' for downplaying the significance of water consumption in AI infrastructure. Such analogies fail to capture the complex interplay of environmental factors and can mislead public discourse about the true scale and implications of AI datacenter water demands.

By mid-2026, Texas exemplified the regulatory and transparency challenges surrounding data center water use, with only 17% of data centers responding to the Texas Water Development Board's survey and a mere 8% to the Public Utility Commission's optional questionnaire. Lawmakers like State Rep. Cody Harris criticized operators for avoiding accountability, noting that companies such as Meta have been unresponsive to information requests, while the Public Utility Commission’s reliance on outdated communication methods like physical mail further suppressed participation, leaving officials to make 'huge decisions on bad data.'

Despite some companies like Amazon Web Services and Google highlighting their commitments to water stewardship—AWS emphasizing recycled water use and Google reporting full survey compliance and community water pledges—the overall industry remains opaque. This opacity is compounded by the concentration of data centers in unincorporated Texas areas where counties lack zoning authority, and by operators’ claims that water usage data is proprietary, resulting in a regulatory environment with few legal incentives or requirements for disclosure, as noted by Rachel Hanes of the Greater Edwards Aquifer Alliance.

California and Oklahoma illustrate broader regulatory gaps beyond Texas, with California lacking state laws to control data center water impacts and relying on pending bills like AB2619 and AB2469 that mandate disclosure and water assessments but fall short without coordinated state support. Similarly, Oklahoma data centers largely evade state water permitting by purchasing municipal water, leaving consumption untracked and communities frustrated by the absence of transparency, as highlighted by local officials and the Oklahoma Water Resources Board.

In the absence of federal regulation, states are experimenting with diverse legislative approaches to manage AI data center water use, including mandatory reporting, conservation mandates, incentives for efficient technologies, and water rights frameworks, as detailed in a University of Colorado Boulder report. However, many states face legislative hurdles—Colorado’s failed bills and municipal moratoriums in Longmont, Boulder County, and Broomfield underscore the contentious and evolving nature of this policy area, with officials like Anderson emphasizing the need for ongoing adaptation to balance water conservation with the booming AI data center demand.

By mid-2026, local communities have made it clear that transparent, localized disclosure of data center water use is non-negotiable for trust and project approval. As emphasized on June 22, projects that cannot clearly communicate their water sources, consumption levels, and mitigation strategies in accessible terms face outright rejection, as communities intuitively distrust off-site water replenishment claims and demand alignment with local water realities. Transparency is not a mere PR exercise but a prerequisite that compels better project design and helps prevent opposition fueled by worst-case assumptions.

In Texas, the mounting community backlash against data centers’ water consumption culminated in a state House committee hearing on June 25, where citizens called for moratoriums until environmental impacts are fully understood. Data center operators like Diode Ventures and Calypso LLC faced criticism for declining to testify, undermining trust, while proactive companies such as Amazon Web Services and Vantage Data Centers showcased water stewardship through recycled water use and closed-loop cooling. Yet, local officials like State Reps. Jeff Barry and Mary Gonzalez lamented being sidelined in decision-making, highlighting a persistent disconnect between developers and the communities they affect.

The challenge of managing data center water demand is particularly acute in smaller or resource-limited communities, where lack of transparent data and technical support leaves local leaders overwhelmed and reliant on blunt instruments like moratoriums. As noted in late June, states could better support municipalities by assembling dedicated response teams to help interpret proposals, while legislative efforts such as California’s AB2619 and AB2469 aim to improve water use data collection. However, fragmented data without coordinated analysis risks perpetuating confusion, a tension underscored by Governor Newsom’s veto of prior disclosure bills to protect the tech sector at the expense of local water security.

Across multiple states, including Texas, Oklahoma, and Colorado, local governments and communities are increasingly resorting to moratoriums, zoning restrictions, and calls for mandatory water use disclosures to counteract the environmental strain of data centers. In Texas, nearly half of new centers are planned in unincorporated areas lacking robust regulatory oversight, exacerbating mistrust fueled by industry non-disclosure practices. Oklahoma faces a regulatory loophole allowing data centers to bypass water permits by purchasing municipal water, intensifying political debates over scarce resources amid drought. Meanwhile, Colorado municipalities like Longmont and Boulder have enacted moratoriums while grappling with legislative tensions between economic incentives and environmental safeguards, underscoring the critical role of transparent, proactive community engagement in shaping sustainable data center growth.

By early 2026, AI data centers had emerged as significant contributors to global water scarcity, with their water demand projected to more than double from 23 to over 54 cubic kilometers annually by 2050. This surge intensifies competition for limited freshwater resources, as exemplified by the staggering 5.4 million liters used to train GPT-3 alone, much of which involves clean freshwater evaporated during cooling. Compounding the challenge, the opacity surrounding AI systems' water consumption hinders efforts to fully grasp and mitigate their environmental footprint.

The rapid escalation of water use by U.S. data centers—from 5.6 billion gallons in 2014 to 17.4 billion gallons in 2023, with projections nearing 32.8 billion gallons by 2028—spotlights the urgent environmental challenges posed by AI infrastructure expansion. While states increasingly mandate water-efficient cooling technologies to curb consumption, this well-intentioned shift often backfires by driving up non-renewable energy use, thereby escalating fossil fuel demand and its associated water footprint. This paradox underscores the complex trade-offs between conserving water and managing energy sustainability within the AI sector.

In the absence of comprehensive federal regulation, states and localities have adopted a patchwork of strategies—ranging from mandatory reporting and conservation incentives to moratoria on hyperscale data centers—to address the localized risks that AI data centers pose to water supplies. Notably, municipalities like Longmont and Boulder County have enacted bans reflecting acute water stress concerns. Meanwhile, the indirect water consumption linked to power generation for these centers dwarfs direct cooling needs, with an estimated 211 billion gallons consumed in 2023 alone, revealing the broader, often overlooked water footprint embedded in AI’s energy demands.

By mid-2026, Texas found itself at a crossroads where the economic promise of a data center boom—likened by Governor Greg Abbott to the historic railroad expansion—clashed with mounting infrastructure strains and community concerns. With over 440 gigawatts of data center applications queued against just 110 gigawatts of current peak capacity, political leaders like Rep. John McQueeney recognized the urgent need for a regulatory framework, culminating in the drafting of the Data Center Responsibility Act aimed at balancing growth with accountability. Abbott emphasized protecting everyday Texans by addressing electricity, water, noise, and site selection, while also vowing to flush out 'bad actors' who bypass local governance, underscoring a political will to keep Texas open for business but not at any cost.

Despite the economic incentives, transparency issues have severely hampered policy-making, as fewer than one-third of data center companies responded to state surveys on water and electricity usage, with only 28 companies reporting out of more than 248 proposed facilities. Lawmakers like State Rep. Brad Buckley decried the data as 'bad' and 'pretty pathetic,' warning that decisions cannot be based on 'half-baked information.' This opacity fuels political frustration and intensifies calls for mandatory reporting, especially as Governor Abbott directs utilities to shield consumers from rising infrastructure costs and encourages adoption of efficient technologies like Google's 100% air-cooled, closed-loop cooling systems, reflecting a partial industry response to regulatory pressures.

The political landscape is further complicated by the tech industry's reliance on tax breaks and preference for rural, unincorporated areas with limited zoning oversight, enabling rapid expansion but complicating infrastructure management. Critics highlight billions lost in tax revenue, while proponents argue these incentives are vital to sustain investment and economic benefits. This tension illustrates the broader challenge Texas faces: fostering a competitive environment for data centers to avoid economic stagnation, as warned by Rep. McQueeney's railroad analogy, while instituting regulatory guardrails to manage the environmental and community impacts of this unprecedented growth.

Adding a layer of urgency, data centers have emerged as a national security concern, elevating political pressure for oversight amid industry resistance. The complex approval process involving Transmission Service Providers, ERCOT studies, and regulatory bodies reflects the intricate balancing act between enabling economic development and safeguarding critical infrastructure. The regulatory delay until 2027 underscores the tension between rapid expansion and the necessity for structured policies to address rising energy and water demands, community impacts, and strategic vulnerabilities, signaling that Texas is grappling with how to govern a sector that is both an economic engine and a potential systemic risk.

Generous incentives and rural siting let tech giants sidestep local oversight, sparking debate over lost revenue versus economic growth as Texas struggles to rein in the environmental costs of unchecked data center proliferation.