Water, Power, and the Race for AI:
Solving the Resource Equation for Data Centers in the Western United States
Authored by: RJ Colwell, Brian Annes, Jeff Simonetti, and Patrick Giardina
Editor’s Note: Artificial intelligence is driving one of the largest infrastructure expansions in modern history, but the race to build hyperscale data centers is increasingly constrained by two essential resources: electricity and water. This invited article examines how these competing demands are reshaping investment decisions, regulatory policy, and water resource management across the Western United States. It argues that long-term success will depend not only on securing reliable power, but also on developing innovative, sustainable water supply solutions that balance economic growth with the needs of agriculture, municipalities, and the environment. For policymakers, investors, developers, and water managers alike, the emerging intersection of AI infrastructure and water security represents one of the defining resource challenges of the future.
I. The Collision
The United States is building the physical infrastructure of artificial intelligence at a pace not seen since the interstate highway system. Hyperscale data centers, the warehouse-scale computing facilities that anchor the AI economy, are under construction or in advanced planning across every major power market in the country.
The scale of demand is difficult to overstate. In Texas, ERCOT’s grid currently manages approximately 86 gigawatts of peak demand. As of October 2025, ERCOT’s interconnection queue held more than 200 GW of large-load requests, with 70% attributed to data centers.[1] Thomas Gleeson, Chairman of the Public Utility Commission of Texas (PUCT), has stated publicly that Texas wants to accommodate large-load growth, declaring that the state intends to “hook ‘em and bring them in and have a place for them to go.”[2]
But data centers do not run on electricity alone. They require water, large quantities of it, primarily to cool the servers that process AI workloads. A single gigawatt of data center capacity, together with its co-located power generation, can require 10 million to 21 million gallons of water per day under current evaporative cooling configurations.[3] Emerging technologies including direct-to-chip liquid cooling, immersion cooling, and dry-cooler hybrid systems may significantly reduce water intensity in new facilities, though adoption timelines and costs vary. Multi-gigawatt campuses multiply that demand regardless of cooling method. Data center end users, including hyperscaler developers, sometimes plan these campuses in areas where water supplies are limited.
Water and energy constraints are increasingly likely to shape data center development across the Western United States. The central question may be which states, developers, and investors will solve for those constraints, and how water for data center developments will be prioritized in water-constrained markets.
II. The Colorado River: Test Case for a National Problem
No river system in the United States better illustrates the stakes than the Colorado River.
The Colorado River Compact of 1922 (the Compact) apportions 7.5 million acre-feet of water annually to each of two basins: the Upper Basin (Colorado, Wyoming, Utah, New Mexico) and the Lower Basin (Arizona, Nevada, California).[4] The Compact does not allocate specific shares to individual states. The Boulder Canyon Project Act of 1928 provided a formula for allocating water among the Lower Basin states, and the Secretary of the Interior manages reservoir operations, including Lake Powell and Lake Mead, under the Colorado River Storage Project Act.[5] The United States and Mexico also entered into a treaty in 1944 (1944 Treaty) that guarantees Mexico an annual delivery of 1.5 million acre-feet of Colorado River water, bringing the total allocation across the basin to approximately 16.5 million acre-feet per year.[6] In 1948, the Upper Basin states entered into the Upper Colorado River Basin Compact,[7] which apportions the Upper Basin’s 7.5 million acre-feet among the four Upper Basin states: Colorado receives 51.75%, Utah 23%, Wyoming 14%, and New Mexico 11.25%. The Compact, the 1944 Treaty, and the Upper Colorado River Basin Compact must be read together to understand the full scope of the Colorado River’s over-allocation problem.
The Compact was negotiated during one of the wettest periods in the river’s recorded and reconstructed history. Many hydrologists and water policy analysts have concluded that the river has been over-allocated since the Compact’s inception. A century of population growth, agricultural expansion, and industrial development have compounded the structural deficit. The current operating agreements, including the 2007 Interim Guidelines for Lower Basin Shortages and Coordinated Operations, the 2019 Drought Contingency Plans, and related operational arrangements implementing the 1944 Treaty, are set to expire at the end of 2026.
The federal government is not waiting for the states to reach consensus. On January 9, 2026, the Bureau of Reclamation released a Draft Environmental Impact Statement evaluating five operational alternatives for managing Colorado River reservoirs beginning in 2027. The Draft EIS does not designate a preferred alternative. The 45-day public comment period closed on March 2, 2026, drawing over 18,000 submissions. In February 2026, Interior Secretary Doug Burgum confirmed that the Department is moving forward to finalize operating guidelines by October 1, 2026, the start of the 2027 water year, despite the seven basin states not having reached full consensus. Secretary Burgum described a “fair compromise with shared responsibility” as still “within reach,” but the Department has made clear it will not delay action.[8] Mexico, as a party to the 1944 Treaty, has also been engaged in these discussions, and any restructuring of basin operations may affect the timing and volume of treaty deliveries, adding a cross-border dimension to the supply uncertainty facing Lower Basin water users.
The Upper Basin states seek to protect their allocations and ensure equitable distribution of drought mitigation responsibilities. The Lower Basin states focus on maintaining existing uses and addressing shortages, particularly in light of their higher historical consumption. If the states fail to agree, the Secretary of the Interior may act unilaterally, modifying operating criteria based on actual operating experience or unforeseen circumstances after consulting the basin states.[9] Any basin state may bring an action in the U.S. Supreme Court to enforce the Compact’s provisions, with the United States consenting to be joined as a party.[10] Federal reserved water rights, which vest upon the creation of a federal reservation, are senior to state-law claims made after the reservation’s creation, adding further uncertainty.[11]
These dynamics are not abstractions for the data center industry. The implications are particularly acute for Arizona, which depends on the Colorado River for roughly 36% of its water supply and which is simultaneously one of the fastest-growing data center markets in the country. Phoenix is consistently ranked among the largest U.S. data center markets, with approximately 804 MW of existing inventory and more than 1,000 MW under construction as of year-end 2024, and the development pipeline continuing to grow.[12] But community tolerance for new industrial water demand appears to have reached a breaking point.
In Tucson, a proposed $5 billion, 290-acre data center campus known as Project Blue, developed by Beale Infrastructure with Amazon as the intended tenant, became a political lightning rod. The project’s initial water consumption estimate dominated local media and became a defining issue in municipal elections. The city denied land-use changes and water utility access. Amazon withdrew participation in the project in December 2025, though the project may still be seeking alternative tenants and cooling methods to move forward.[13] In the Phoenix suburb of Chandler, the community debate extended beyond water consumption to include the impact of large-load data centers on local power rates. The city adopted zoning and water-use restrictions, including limiting data center permitting to planned area development districts and imposing mandatory conservation measures during drought conditions, as well as provisions addressing the cost allocation of grid infrastructure upgrades to ensure residential ratepayers are not required to subsidize the transmission and distribution costs attributable to industrial-scale data center load. Industry participants have described these measures as effectively halting new data center development.[14]
The regulatory tools behind these outcomes are not limited to ad hoc political decisions. Arizona’s Groundwater Management Act (A.R.S. §§ 45-561 et seq.) requires groundwater management plans for Active Management Areas to include conservation programs for non-irrigation uses, which regulators have interpreted to require deployment of commercially available conservation technology assessed against reasonable economic return. Municipalities and sanitary districts have statutory authority to impose moratoriums on land development.[15] While a few states have enacted data-center-specific water legislation, existing general-purpose regulatory tools may be sufficient to impede or halt projects, as recent experience in Arizona suggests.
And the pattern is not limited to Arizona. Residents in El Paso, Texas have organized against proposed data centers over water and energy concerns. Communities across Utah, Nevada, and Colorado are asking similar questions. The instinct in many jurisdictions is to slow demand rather than solve for supply. The question of “highest and best use,” whether data center cooling is a higher-value use of scarce water than irrigating farmland or supplying a growing city, has no established legal framework in any Western state, but it is increasingly the question that communities are asking.
III. What Is at Stake
The economic consequences of failing to solve the water-energy equation are substantial and immediate.
AI data centers represent concentrated capital investment (billions of dollars per campus for large data centers), long-duration operational commitments (20-30 year asset lives), high-wage construction and operations employment, and significant tax base contributions. They anchor supply chains in power generation, fiber, construction, and professional services. They are the physical infrastructure underlying a technology expected to reshape many major industries in the coming decade.
The competition for this investment is real and accelerating. States that offer reliable power, secure water supply, predictable regulatory frameworks, and community support are likely to attract investments. States that impose moratoriums, layer on new approval requirements, or allow water and power uncertainty to fester may not.
But competition among states tells only part of the story. There is a harder question underneath the competition. In a basin where water is over-allocated, every gallon consumed by a data center is a gallon not available for agriculture, municipal supply, or environmental flows. Western prior-appropriation states have long operated on a “first in time, first in right” principle.[16] But priority dates do not answer the policy question of whether a data center’s cooling demand should take precedence over a farmer’s irrigation right or a city’s drinking water supply.
Some states address this implicitly through their drought management frameworks. During drought, Texas law mandates pro rata curtailment among customers of a water system, based on entitlement or adjusted entitlement under the system’s conservation plan.[17] The Texas Commission on Environmental Quality (TCEQ) may temporarily suspend or adjust water rights to maximize beneficial use and minimize impacts on rights holders, though courts have held that TCEQ cannot exempt junior preferred rights from suspension.[18] In Arizona, the prior appropriation doctrine protects senior surface-water rights holders, but groundwater is governed by the doctrine of reasonable use, which may limit the enforceability of long-term industrial supply contracts during shortages.[19] No state has adopted a statutory framework for evaluating the relative priority of data center water use against agricultural or municipal use.
The practical implication for developers and investors is clear. Those who do not secure durable water supply, whether through senior water rights, alternative sources, or contractual structures that survive curtailment, may face operational risk that investors are only beginning to price. Hyperscaler site selection typically sequences power first (power purchase agreement (PPA) execution or behind-the-meter generation commitment), then water (firm offtake with contractual curtailment protections), then permitting and community engagement. Water supply contracts at data center scale require duration matching asset life (20-30 years), inflation-adjusted pricing, and step-in rights, features not yet market-tested in produced water or alternative-source deals. The questions, then, are how those water deals can be structured and what jurisdictions and water rights holders have the ability and interest to structure them.
IV. The Texas Model, and What Other Western States Can Learn
Texas is not part of the Colorado River Basin. It faces its own water challenges: a projected 290 billion gallon annual deficit by 2050, groundwater depletion in key aquifers, and rapidly growing industrial demand. But Texas is approaching those challenges with a posture of building supply rather than restricting demand, and several recent legislative and regulatory developments are creating a framework that may be uniquely favorable for integrated water-and-power solutions.
Regulatory pragmatism. The PUCT is surveying data center and cryptocurrency mining facility water usage this spring, gathering data before legislating. The survey, authorized through a budget rider by State Representative Armando Walle, will collect information on direct water use, cooling technology, and indirect water consumption through power generation. Results will be shared with the Texas Water Development Board and TCEQ to inform future planning. That posture – pro-growth, data-driven, and solutions-oriented – contrasts with the moratorium-first instincts emerging in other parts of the United States.
Legislative reform. The 89th Texas Legislature enacted the most significant produced water legislative package in the state’s history. SB 1145 vested TCEQ with land-application permitting authority for treated produced water. TCEQ is finalizing its proposed rule package implementing SB 1145. Developers, investors, and their counsel should monitor the rulemaking closely, including any public comment opportunities, and evaluate the final rules’ treatment of transition provisions, grandfathering, and effluent limits applicable to industrial cooling use. HB 49, meanwhile, limits tort liability for entities involved in the treatment, transfer, or beneficial use of treated produced water, with exceptions for gross negligence, intentional wrongful conduct, or regulatory noncompliance, and bars exemplary damages for negligence claims premised solely on regulatory noncompliance.[20] The Texas Supreme Court’s decision in Cactus Water Services, LLC v. COG Operating, LLC, 718 S.W.3d 214 (2025), resolved a threshold ownership question by holding that produced water belongs to the mineral lessee absent an express contrary conveyance. The court did not address ownership of brine minerals dissolved in produced water, a gap that SB 1763 sought to fill but that stalled in the Senate Natural Resources Committee. Until the legislature or courts resolve that question, investment in lithium and other mineral extraction from produced water may face title uncertainty. Together, these developments may clear significant legal obstacles that have impeded produced water beneficial reuse in Texas, though important questions, including brine mineral ownership and the content of TCEQ’s implementing rules, remain unresolved.
Alternative water sources. Texas sits atop one of the largest alternative water supplies in the world. The Permian Basin alone generates over 20 million barrels of produced water daily, a volume that dwarfs the cooling demand of even the most ambitious data center campuses. Treatment technologies have demonstrated the ability to convert that water from salinity levels of 130,000 to 150,000 mg/L (milligrams per liter) down to below 200 mg/L, a specification clean enough for data center cooling, power generation, and agricultural irrigation. Achieving that specification at commercial scale requires multi-stage treatment trains: pretreatment to remove oils, greases, suspended solids, and organics; membrane-based or thermal desalination; and post-treatment polishing for ammonia (typically 500-700 mg/L in Permian produced water), boron, and residual contaminants. Hybrid membrane-thermal systems demonstrate feasibility at 10,000 to 150,000 barrels per day scale in current pilots; deployment at gigawatt-scale data center demand will require parallel trains and firm offtake commitments to underwrite capital expenditure.
Developers, operators, and their advisors should also consider that evaporative cooling towers concentrate contaminants through repeated cycling, typically three to six cycles, making blowdown water management and discharge permitting a separate compliance consideration beyond initial treatment specifications. Hyperscaler sustainability commitments, including published water-quality specifications and non-potable sourcing requirements in recent sustainability reports, are converging toward standards that treated produced water and other alternative sources can meet, provided treatment trains are designed to the specific chemistry required.
Any beneficial reuse pathway must address the full contaminant profile of produced water, not only TDS but also PFAS (per- and polyfluoroalkyl substances), naturally occurring radioactive materials (NORM), heavy metals, and residual hydrocarbons. Rice University’s WaTER Institute has established a dedicated PFAS Alternatives and Remediation Center that is focused on destruction and removal technologies for these persistent contaminants. Independent third-party monitoring, enforceable effluent limits, and public reporting are likely to be expected, and may be required, as baseline conditions for any commercial-scale reuse project, both to protect public health and to build the social license that large-scale beneficial reuse will require.
The economics of treatment and disposal are converging as injection capacity tightens and disposal costs rise. For oil and gas operators, this convergence could transform a disposal liability into a potential revenue source, depending on basin-specific economics, treatment costs, and regulatory outcomes. Operators may wish to evaluate whether existing mineral leases could be interpreted to require revenue-sharing or royalty payments on proceeds from treated produced water sales, a question Cactus Water did not reach and that existing lease language may not clearly address. For data center developers, treated produced water provides a non-freshwater supply that does not compete with municipal or agricultural users, directly addressing the community opposition dynamic that has impeded projects in Arizona.
Institutional capital for water. Texas voters have approved constitutional amendments dedicating substantial funding to water infrastructure through the Texas Water Fund, administered by the Texas Water Development Board.[21] Eligible projects include water and wastewater infrastructure, water conservation, desalination, aquifer storage and recovery, and, notably, produced water treatment projects other than those solely for oil and gas exploration.[22] This creates a durable, scaled funding mechanism that contrasts with the inconsistent funding history of comparable programs in other Western states.
Power supply solutions. At the federal level, the Trump Administration has reinforced the policy expectation that data center operators should supply their own power rather than rely on grid capacity that increases costs for residential ratepayers. Following President Trump’s February 2026 State of the Union address, Amazon, Google, Meta, Microsoft, OpenAI, Oracle, and xAI signed the Ratepayer Protection Pledge (Pledge)[23] at a White House event on March 4, 2026, committing to build, bring, or buy new generation resources for their data centers and to cover the cost of all required power delivery infrastructure upgrades without passing those costs to American households. Microsoft’s 20-year power purchase agreement to offtake the entire output of the restarted Three Mile Island Unit 1, rebranded the Crane Clean Energy Center, reflects a related strategy: a hyperscaler contracting for dedicated, long-duration generating capacity to match its data center load rather than relying on uncommitted grid supply.[24] While that arrangement delivers power through the PJM grid rather than behind the meter, it illustrates the same underlying logic of securing firm, purpose-built generation rather than drawing on shared system capacity. The Pledge is a voluntary commitment rather than an enforceable federal regulation, but it reinforces the commercial logic of behind-the-meter generation structures and may accelerate regulatory support for co-located and self-supply configurations in states that adopt the Pledge’s framework as policy.
Texas law permits data center operators to self-generate electricity behind the meter without being classified as a utility or power generation company, provided the electricity is not resold.[25] For co-located generation and load, Tex. Utilities Code § 39.169 requires generators to notify ERCOT before implementing net metering arrangements with large-load customers, with exemptions for resources that were registered with co-located load at energization or are majority-owned by the customer’s parent company. ERCOT’s independent operation outside FERC jurisdiction for most intrastate transactions[26] provides additional regulatory flexibility. Texas’s behind-the-meter generation framework and produced water pathway may compress the typical site-selection sequence (power, then water, then permitting), but only if TCEQ rules deliver predictable permitting timelines and groundwater conservation districts do not delay, deny or unduly condition production permits.
Data center developers in Texas are pairing these self-generation structures with long-term gas supply contracts that fix fuel costs for terms of 20 years or more, eliminating commodity price variability for behind-the-meter generation.[27] Parties structuring these arrangements should consider the risk of recharacterization as disguised loans[28] and should ensure compliance with CFTC and SEC requirements. When properly structured, such arrangements may offer data center operators predictable, long-duration fuel supply at a potentially lower fixed price, a compelling proposition for multi-gigawatt behind-the-meter generation.
SB 6, enacted June 20, 2025, establishes interconnection and emergency curtailment requirements for large-load customers above 75 MW and ensures that those customers contribute to the recovery of interconnection costs incurred by utilities.[29] SB 6 suggests that a pro-growth regulatory posture need not come at the expense of ratepayer protection. The bill’s financial security requirements and cost-allocation mechanisms were enacted specifically to protect residential ratepayers from bearing infrastructure costs driven by large-load users, a principle that strengthens, rather than undermines, Texas’s competitive position by demonstrating that growth and ratepayer protection can coexist. Residential electricity prices in Texas have increased approximately 30% since 2021, with forward projections suggesting another 29% increase driven by transmission and distribution investment.[30] The PUCT is implementing SB 6 through active rulemaking dockets that will determine interconnection standards, financial security requirements, and cost recovery mechanisms. Developers evaluating Texas siting should track these dockets in real time; the final rules could significantly affect the viability of behind-the-meter structures without triggering utility classification or cost-socialization obligations.
What other Western states are doing. Arizona’s Water Infrastructure Finance Authority (WIFA), established under A.R.S. §§ 49-1203 through 49-1271, has statutory authority to investigate and facilitate water importation projects, issue bonds, negotiate agreements for water infrastructure, and administer the Long-Term Water Augmentation Fund and the Water Supply Development Revolving Fund.[31] WIFA’s mandate is broad enough to encompass industrial water supply arrangements.[32] But state funding for WIFA has been inconsistent, and its statutory tools do not appear to have been deployed for data center-scale industrial supply to date. Arizona policymakers face the difficult task of balancing economic development with competing funding priorities, environmental justice concerns and agricultural priority in a basin where communities are already experiencing the consequences of over-allocation.
New Mexico’s Strategic Water Supply Act (HB 137, signed April 2025) created the Strategic Water Supply Program and Fund to support brackish water treatment projects.[33] The statute explicitly excludes produced water from the definition of brackish water,[34] reflecting persistent concerns about treatment adequacy and public health risk. The legislature removed produced water reuse provisions during the committee process. New Mexico also prohibits surface discharge of treated produced water under two independent regulatory provisions.[35] New Mexico’s approach reflects the environmental and scientific concerns that any responsible produced water reuse framework will likely need to address.
Colorado adopted the most prescriptive framework. The Energy and Carbon Management Commission is required by statute to adopt rules mandating statewide reductions in fresh water usage and corresponding increases in recycled produced water use at oil and gas locations, with iterative increases over time.[36] Colorado’s nontributary groundwater framework[37] permits produced water from nontributary formations to be used for oil-and-gas-related purposes without a Water Court decree, but that authorization does not extend to beneficial reuse for non-oilfield purposes (such as data center cooling) without landowner consent or additional water rights authorization. The rules do not apply to designated groundwater basins or aquifers within the Southern Ute Indian Reservation.[38] Against this varied and still-developing Western regulatory landscape, the commercial model for delivering water to data centers is itself an open question.
V. Water as a Service: From Concept to Commercial Reality
A data center developer evaluating a West Texas site today faces a sequence of water-related tasks that most developers are not staffed or equipped to handle: identifying whether the site has access to groundwater, surface water, produced water, or some combination; determining the applicable groundwater conservation district rules and whether a production permit can be obtained at the volumes required; evaluating whether existing water rights can be acquired or transferred and at what priority; selecting and contracting with a treatment technology provider if the available source requires desalination or polishing; permitting the treatment and delivery infrastructure across TCEQ, the Texas Railroad Commission, and potentially federal agencies; negotiating a water supply agreement that survives drought curtailment, regulatory change, and production decline over a 20-to-30-year asset life; and managing ongoing compliance, monitoring, and community relations for the duration of operations. Each of these steps involves a different set of counterparties, agencies, and legal frameworks. Most developers would prefer not to manage any of them.
That is a lot to ask of a company whose core competence is computing. The data center industry does not want to be in the water business. What developers need is a counterparty who can deliver water, sourced, permitted, treated, and transported, under a long-term supply agreement at a predictable cost and specification. The same way a developer signs a power purchase agreement and someone else handles generation, a “Water as a Service” (WaaS) model allows the developer to sign a water supply agreement while a dedicated counterparty handles everything upstream of the fence line: water rights acquisition, regulatory approvals, treatment technology selection, infrastructure development, and ongoing compliance.
That model does not exist at scale today. But the building blocks are in place, and assembling them requires coordination across disciplines that have traditionally operated in silos. The WaaS counterparty may take various forms, from a single integrated platform company to a consortium of specialists in water rights, infrastructure development, legal structuring, and land management, but in every case it must bring all four disciplines to bear as a coordinated offering.
Water sourcing and rights. In the Western United States, securing water for a new industrial use, whether through acquisition of existing rights, transfer from agricultural or municipal allocations, or development of alternative sources such as produced water or brackish groundwater, requires deep knowledge of state-specific appropriation systems, groundwater conservation district rules, and the regulatory process for each transfer. In prior appropriation states, the priority date, the point of diversion, the place of use, and the purpose of use are all legally significant, and a transfer that changes any of these elements may require administrative approval or judicial decree. In Texas, the bifurcated system (prior appropriation for surface water; Rule of Capture[39] modified by groundwater conservation districts for groundwater) adds further complexity.
Infrastructure development and capital formation. Moving water from source to end user at data center scale requires physical infrastructure (pipelines, treatment facilities, storage, delivery systems) and the capital to build it. The development pathway, from siting through permitting, financing, construction, and operation, parallels the energy infrastructure development cycle that has been proven in military installations, renewable energy projects, public-private partnerships, and agricultural land platforms. Structuring the capital stack for a WaaS platform, whether as a project-financed special purpose vehicle, a midstream-style infrastructure fund, or a joint venture among a water rights aggregator, treatment technology provider, and infrastructure developer, draws on experience across asset-backed securitization, long-term commodity contracts, and institutional infrastructure investment.
Legal structuring and regulatory compliance. The commercial agreements required to operationalize WaaS are approaching the complexity of hydrocarbon midstream contracts. Water supply agreements with 20-to-30-year terms, volume commitments, treatment specifications, curtailment allocation provisions, force majeure protections, and regulatory risk allocation require careful drafting and deep familiarity with the agencies that administer them. The regulatory landscape spans multiple agencies (including, as applicable, TCEQ, the Texas Railroad Commission, groundwater conservation districts, state water agencies, and federal agencies) with overlapping and sometimes conflicting requirements. Behind-the-meter power supply structuring, FERC and ERCOT regulatory analysis, and co-location arrangements add the energy dimension. On the produced water side, ownership (post-Cactus Water), liability allocation (under HB 49), treatment permitting (under the pending TCEQ rulemaking), and brine mineral rights all require experienced legal and regulatory engagement. Those evaluating a WaaS structure should model TCEQ’s final rule text against the City of Carrollton line of cases before committing capital.
Agricultural and land management context. Many of the most promising WaaS configurations involve land with both water resources and proximity to power, characteristics common in agricultural regions of West Texas, California’s Central Valley, and other Western basins. These regions also vary materially in regulatory characteristics, community receptivity to data center siting, and the adequacy of grid and water delivery infrastructure. These varying characteristics may make each more or less amenable to data center construction. Evaluating these sites requires understanding of agricultural water use patterns, land management economics, and the practical dynamics of transitioning water from agricultural to industrial use in communities where farming is not just an economic activity but an identity. This perspective is essential to building the community relationships and development strategies that distinguish viable projects from paper concepts.
The midstream analogy. The closest precedent is the produced water midstream sector. Five years ago, every oil and gas operator managed its own water gathering and disposal. Today, dedicated infrastructure companies, including WaterBridge, Western Midstream (following its acquisition of Aris Water Solutions), and others, have built institutional-scale platforms under long-term, take-or-pay contracts (under which the buyer commits to pay for a minimum volume whether or not it takes delivery). Two landmark 2025 capital markets transactions confirmed the model: WaterBridge’s upsized IPO (approximately $634 million at $20 per share, top of range, full overallotment exercise) and Western Midstream’s approximately $2 billion Aris acquisition. Water as a Service for data centers is the next iteration: a dedicated water infrastructure counterparty serving data center load instead of oil and gas exploration and production (E&P) demand. The commercial structures are analogous, including acreage dedications (i.e., commitments to deliver all water from a defined geographic area to a single midstream counterparty), volume commitments, take-or-pay terms, treatment specifications, and regulatory compliance obligations.
Competing approaches. WaaS is not the only model. Hyperscalers may choose to self-fund treatment infrastructure through joint ventures with E&P operators, or to contract for municipal reclaimed water where available. Municipal reclaimed water PPAs are already commercially proven in some markets and may be less expensive where supply exists at sufficient volume. The WaaS model is most compelling where the developer lacks in-house water expertise, where freshwater alternatives are politically or physically constrained, and where the developer prefers a single-counterparty solution that allows it to focus on computing operations rather than water management. The service may also fit when the jurisdiction requires developers to provide their own water supplies as part of the conditions of approval. For projects where those conditions are present, particularly in water-constrained Western basins with produced water or brackish groundwater availability, the WaaS model may offer the most efficient path from site control to operations.
Structuring considerations. A WaaS provider must be carefully designed to avoid unintended regulatory classification. Under Texas Water Code Chapter 13, a “retail public utility” is defined as any entity providing potable water service for compensation (§ 13.002(19)), and requires a Certificate of Convenience and Necessity (CCN).[40] Entities that do not serve “the public,” that operate solely within private bilateral agreements for non-potable industrial supply, may fall outside that classification, though this conclusion would depend on the specific facts and the applicable regulatory posture of TCEQ and any affected municipality. Structuring options to consider include private bilateral industrial supply contracts for non-potable water, conveyance through an existing municipal utility district, or formation of a special district under Texas Water Code Chapter 49 with a limited service area and no CCN requirement. Developers and their legal and regulatory advisors should also consider whether TCEQ might assert jurisdiction through the new produced water land-application rules, depending on how those rules define the scope of regulated activity.
Indicative economics. At indicative scale, a 1 GW data center campus requiring 250,000 to 350,000 barrels per day of treated water could, depending on offtake terms and regulatory certainty, potentially support a dedicated water infrastructure investment with structural parallels to recent produced water midstream transactions. The cost spread is converging: produced water disposal currently runs roughly $0.75 to $1.25 per barrel; treatment to industrial cooling quality costs $2 to $4 per barrel; and the gap is narrowing as disposal capacity tightens and treatment technology matures.[41] If the market clears at a delivered price that covers treatment, transport, and a risk-adjusted return on infrastructure capital, a single 1 GW campus could generate annual water supply revenues in the range of $150 million to $300 million, depending on pricing, volume, and contract structure. These figures are directional and will vary with geography, source water chemistry, and contract structure, but they illustrate the scale of the opportunity. The value proposition is straightforward: the developer avoids the community opposition, regulatory risk, permitting delay, and reputational cost of competing for freshwater, and receives a contractually assured supply with specifications and a delivery schedule. The indicative economics for specific geographies and configurations will depend on variables, including offtake tenor, basin-specific treatment costs, and the regulatory certainty delivered by TCEQ’s pending rulemaking, that are best evaluated on a project-specific basis.
Risks to acknowledge. The WaaS model and the broader Texas advantage are not risk-free.
Groundwater conservation districts (GCDs) have broad statutory authority to deny or condition production permits based on management plans and desired future conditions.[42] The Texas Supreme Court has confirmed that landowners have a constitutionally protected ownership interest in groundwater but has equally affirmed the state’s authority to regulate production through conservation districts.[43] No reported GCD proceeding has yet addressed groundwater withdrawals by a data center, but the tools exist, and a denial or restrictive condition in a key West Texas district could, depending on the circumstances, materially affect project feasibility.
TCEQ’s pending rulemaking on land-application permits for treated produced water is the most consequential near-term regulatory milestone. If the rules are more restrictive than expected, projects already in development may face delays or redesign. The statute does not appear to establish an explicit grandfathering mechanism, and those evaluating projects under the new framework should monitor the rulemaking closely to assess transition provisions.[44]
Long-term water supply agreements face duration mismatch risk. Data centers operate for 20 to 30 years. Produced water supply depends on continued oil and gas production and is subject to volume decline curves and drilling activity levels. Groundwater permits are subject to GCD management plans that are revised periodically. Surface water rights are subject to pro rata curtailment during drought[45] and to temporary suspension or adjustment by TCEQ.[46] In Arizona, the reasonable-use doctrine governing groundwater may limit enforceability of long-term contracts during shortage.[47] Contracts should be structured with curtailment allocation provisions (including pro rata sharing with municipal and agricultural users during declared drought, with makeup volumes from storage or alternative sources within a defined period), alternative-source contingencies, and force majeure protections that account for GCD permit denial, TCEQ rule changes, and production decline.
Community opposition has not yet materialized in Texas at Arizona-scale intensity. But Texas municipalities have the zoning tools[48] and water regulation authority[49] to restrict data center development if public sentiment shifts. Planned Development overlay districts, already in use in Texas cities like Roanoke and Athens, could be deployed to impose site-specific water or infrastructure conditions.[50] The relative absence of organized opposition to date should not be taken as an indication that similar dynamics could not emerge in Texas.
One additional risk warrants emphasis: even where water rights are legally clear and contractually enforceable, the physical reliability of the water supply itself is a central consideration for data center investment. Prior appropriation rights and long-term supply contracts define legal entitlement but do not guarantee actual availability in years of severe drought, snowpack deficit, or reservoir drawdown. As the hydrology of the Colorado River Basin and other Western systems continues to trend drier, the gap between paper entitlements and deliverable supply will likely widen. This is not a novel legal risk; it is an operational one. A data center that cannot receive its contracted water allocation during a drought cannot cool its servers, regardless of the priority date of its water rights. The physical reliability of supply must therefore be modeled alongside legal certainty, with contingency sources, storage arrangements, and curtailment protocols treated as design requirements rather than afterthoughts. In this respect, the water reliability challenge is analytically consistent with the water quality risks discussed above: both reflect the principle that Mother Nature operates independently of contractual frameworks, and that durable data center investment requires engineering around natural variability, not just legal clarity.
At the federal level, the EPA’s announced revision of effluent limitation guidelines (ELGs) for oil and gas extraction, consistent with the Unleashing American Energy Executive Order, could expand federal pathways for beneficial reuse and surface discharge of treated produced water.[51] But no proposed rule text or timeline has been published, meaning the 1970s-era ELGs still remain in effect. Developers should accordingly plan on existing federal requirements and treat any revision as upside.
VI. Conclusion
The states, developers, and teams that assemble integrated water-and-power solutions first are likely to hold the most durable competitive position in the AI infrastructure buildout. Texas appears well positioned, with regulatory pragmatism, alternative water sources, institutional capital, power-supply flexibility, and developing legal infrastructure, to play a leading role. Other Western states are building pieces of the framework: Arizona through WIFA, New Mexico through the Strategic Water Supply Act, Colorado through recycling mandates, and Nevada and Oregon through geothermal and renewable energy incentives. None has yet assembled the full integrated framework.
Policymakers across the West face the same core tension: attract AI investment without compromising municipal supply reliability, agricultural water security, or environmental flows. Texas has chosen supply-side innovation; other states are testing demand-side tools. Both approaches warrant data-driven evaluation, and each state’s circumstances, including hydrology, existing appropriation, and community priorities, will shape the right balance.
The water exists, the treatment technology is proven at pilot scale, and institutional capital is beginning to organize around the opportunity. What remains is the commercial and legal architecture to connect them, and the discipline to build responsibly. Data center development that comes at the expense of agricultural communities or environmental flows is unlikely to be sustainable over the long term. A durable approach will likely require transparent valuation of competing water uses, independent monitoring of treatment adequacy, and meaningful engagement with host communities, in addition to supply-side innovation.
Texas built the modern energy economy. It is now building the AI infrastructure economy. The developers, investors, operators, and advisors who move first to integrate water and power solutions for this new class of demand will not only capture a significant commercial opportunity but will also help establish the model that the rest of the Western United States, and the country, will eventually follow. The window is open. With TCEQ’s implementing rules for produced water beneficial reuse imminent and Colorado River post-2026 operating guidelines due by October, the regulatory landscape that will govern the first generation of these projects is taking shape now. The teams that act first will define what the next generation of American infrastructure looks like.
RJ Colwell is a Senior Associate at Davis Graham & Stubbs LLP in the Energy & Mining Group. He advises energy companies, data center developers, and investors on transactions, regulatory compliance, and project structuring across the water-energy nexus, with particular depth in energy M&A, oil and gas transactions, FERC and state energy regulatory matters, produced water, and behind-the-meter and co-located power supply for data centers. He can be reached at rj.colwell@davisgraham.com.
Brian Annes is a Counsel at Davis Graham & Stubbs LLP in the Energy & Mining Group. His practice focuses on energy transactions, project development, and regulatory matters, with particular experience in produced water infrastructure, geothermal and energy transition project structuring, and the intersection of energy and water resources in Western U.S. basins. Brian can be reached at brian.annes@davisgraham.com.
Jeff Simonetti is CEO and Founder of The Aqua Exchange and Senior Vice President at Capitol Core Group. He has more than two decades of experience in water rights transactions, government affairs, and regulatory strategy across the Western United States, with particular expertise in water acquisition, transfer execution, and valuation for agricultural, municipal, and industrial users. He can be reached at jeff@theaquaexchange.com.
Patrick Giardina is Principal Managing Partner at Simon-Gardner. He brings extensive experience in infrastructure development, public-private partnerships, capital formation, and asset-backed structuring for energy, water, and land assets, including large-scale renewable energy projects and agricultural land platforms. He can be reached at patrick@atlanticgardner.com.
The views expressed are those of the authors and do not necessarily represent the views of their respective organizations.
This article is intended to provide a general overview of selected legal, regulatory, and commercial considerations relevant to water and energy infrastructure for data center development in the Western United States. It does not constitute legal, financial, or investment advice, and the appropriate approach will depend on the specific facts, jurisdictions, and circumstances applicable to each project. Readers should consult qualified legal and technical advisors before acting on any information discussed in this article.
[1] Texas Grid Faces High-Stakes Bet on How Big Demand Will Get, Hart Energy (Mar. 25, 2026), https://www.hartenergy.com/technology-and-innovation/he-texas-grid-ercot-demand/; see also Lexis Practical Guidance, How New Law Transforms Large-Load Power Projects In Texas (2026).
[2] Arcelia Martin, Data Center Explosion Creates Planning Problems for Texas, Tex. Tribune (Oct. 30, 2025), https://www.texastribune.org/2025/10/30/texas-ercot-power-grid-data-centers-puc/.
[3] Houston Advanced Research Center (HARC), Thirsty Data and the Lone Star State (Jan. 2026); see also Arman Shehabi et al., Lawrence Berkeley Nat’l Lab., 2024 United States Data Center Energy Usage Report (2024) (commissioned by U.S. Dep’t of Energy; estimating U.S. data centers directly consumed approximately 17 billion gallons of cooling water in 2023, with direct consumption projected to rise substantially through 2028).
[4] Colorado River Compact, Nov. 24, 1922, codified as approved by Boulder Canyon Project Act § 13, 45 Stat. 1057, 1064–65 (1928); see also Arizona v. California, 373 U.S. 546 (1963) (adjudicating apportionment of Colorado River water among Lower Basin states under the Compact and Boulder Canyon Project Act framework).
[5] Colorado River Storage Project Act, ch. 203, 70 Stat. 105 (1956) (codified at 43 U.S.C. §§ 620–620o).
[6] Treaty for the Utilization of Waters of the Colorado and Tijuana Rivers and of the Rio Grande, U.S.-Mex., Nov. 14, 1944, 59 Stat. 1219.
[7] Upper Colorado River Basin Compact, Nov. 24, 1948, 63 Stat. 31.
[8] Bureau of Reclamation, Draft Environmental Impact Statement: Post-2026 Operations of the Colorado River System (Jan. 2026); see also Press Statement of Interior Secretary Doug Burgum (Feb. 2026).
[9] 43 U.S.C. § 1552.
[10] Colorado River Compact art. IX; see also 43 U.S.C. § 620m.
[11] In re General Adjudication of All Rights to Use Water in the Gila River Sys. & Source, 201 Ariz. 307 (2001).
[12] JLL, North America Data Center Report: Year-End 2024 (2025) (reporting Phoenix inventory of 804 MW and 1,004 MW under construction, with an additional 3,684 MW planned); see also CBRE Research & Data Center Solutions, North America Data Center Trends H2 2025: Phoenix Market (2025), https://www.cbre.com/insights/books/north-america-data-center-trends-h2-2025/phoenix-data-center-market (confirming continued strong hyperscaler demand and power-delivery constraints limiting new supply).
[13] Tucson City Council, Vote to Reject Annexation of Project Blue Site (Aug. 2025); see Tucson Sentinel, “Project Blue Developers Close on Purchase of Pima County Land for Tucson Data Center” (Dec. 24, 2025), https://www.tucsonsentinel.com/local/report/122425_project_blue_close/; Data Center Dynamics, “Amazon-Linked Data Center Project in Tucson, Arizona, Requests Power Despite Water Denial from Local Officials” (Mar. 2026), https://www.datacenterdynamics.com/en/news/amazon-linked-data-center-project-in-tuscon-arizona-requests-power-despite-water-denial-from-local-officials/.
[14] Chandler, Ariz., Ordinance No. 5033 (Dec. 5, 2022) (codified at Chandler, Ariz., Code § 35-2214); City of Chandler, “Chandler’s Data Center Ordinance Now in Effect” (Jan. 11, 2023), https://www.chandleraz.gov/news-center/chandlers-data-center-ordinance-now-effect; see also AZBEX, “More Cities Considering Data Center Restrictions” (June 18, 2025), https://azbex.com/legislation-issues/more-cities-considering-data-center-restrictions/ (describing industry characterization of Chandler restrictions as effectively halting new development).
[15] A.R.S. §§ 9-463.06, 48-2033.
[16] Tex. Water Code § 11.027 (Texas); A.R.S. § 45-151 (Arizona).
[17] Tex. Water Code § 11.039; 30 TAC § 288.22.
[18] Tex. Water Code § 11.053; Tex. Comm’n on Envtl. Quality v. Tex. Farm Bureau, 460 S.W.3d 264 (Tex. App.-Corpus Christi 2015).
[19] Silver v. Pueblo Del Sol Water Co., 244 Ariz. 553 (2018).
[20] Tex. Nat. Res. Code § 122.003.
[21] Tex. Water Code §§ 15.502, 15.504, 15.505.
[22] Tex. Water Code § 15.153(b)(1)(B).
[23] White House, Fact Sheet: President Donald J. Trump Advances Energy Affordability with the Ratepayer Protection Pledge (Mar. 4, 2026), https://www.whitehouse.gov/fact-sheets/2026/03/fact-sheet-president-donald-j-trump-advances-energy-affordability-with-the-ratepayer-protection-pledge/.
[24] Press Release, Constellation Energy, Constellation to Launch Crane Clean Energy Center, Restoring Jobs and Carbon-Free Power to the Grid (Sept. 20, 2024), https://www.constellationenergy.com/news/2024/Constellation-to-Launch-Crane-Clean-Energy-Center-Restoring-Jobs-and-Carbon-Free-Power-to-The-Grid.html.
[25] Tex. Utilities Code § 31.002(6)(J)(ii); § 37.001.
[26] Tex. Utilities Code § 39.151.
[27] See generally ASC 815 (governing accounting treatment of prepaid commodity contracts as derivatives).
[28] Cf. In re Enron Corp. Secs., 235 F. Supp. 2d 549 (S.D. Tex. 2002).
[29] Tex. Utilities Code §§ 37.0561, 35.004.
[30] Tex. Energy Poverty Research Inst., ERCOT Electricity Affordability Outlook: Forecasting Residential Electricity Prices and Burdens (2025-2030) (Jan. 2026), https://tepri.org/2025/12/affordabilitypaper/.
[31] A.R.S. §§ 49-1203.01, 49-1205, 49-1271.
[32] A.R.S. § 49-1304.
[33] N.M. Stat. Ann. §§ 72-12C-1 through 72-12C-4.
[34] N.M. Stat. Ann. § 72-12C-2(B).
[35] N.M. Admin. Code § 20.6.8.400 (Water Quality Control Commission); § 19.15.34.8 (Oil Conservation Division).
[36] C.R.S. § 34-60-134(5)(c)(I).
[37] C.R.S. § 37-90-137(7); 2 CCR 402-17. Such framework governs groundwater in formations that do not contribute to surface streams and are therefore regulated separately from the state’s surface water appropriation system.
[38] 2 CCR 402-17.
[39] In the groundwater context, the Rule of Capture is a common-law doctrine under which a landowner may pump groundwater beneath their property without liability to neighboring landowners, subject to regulation by local groundwater conservation districts.
[40] Tex. Water Code § 13.242; City of Carrollton v. Tex. Comm’n on Envtl. Quality, 170 S.W.3d 204 (Tex. App.-Austin 2005); City of San Antonio v. BSR Water Co., 190 S.W.3d 747 (Tex. App.-San Antonio 2005).
[41] See Tex. Real Estate Research Ctr., Tex. A&M Univ., The Future of Produced Water Recycling in Texas (Oct. 2025), https://trerc.tamu.edu/blog/the-future-of-produced-water-recycling-in-texas/ (reporting Permian Basin disposal costs of approximately $0.60 to $0.70 per barrel and treatment-for-agriculture costs of $2 to $4 per barrel, drawing on Texas Produced Water Consortium and U.S. Department of Energy assessments); see also Thunder Said Energy, Shale Water Costs: Transport, Treatment and Disposal (Mar. 2025) (estimating roughly $1 per barrel for disposal and $3 per barrel for full treatment to agricultural or cooling-quality water).
[42] Tex. Water Code §§ 36.113, 36.1132, 36.101.
[43] Edwards Aquifer Auth. v. Day, 369 S.W.3d 814 (Tex. 2012); Barshop v. Medina County Underground Water Conservation Dist., 925 S.W.2d 618 (Tex. 1996).
[44] See SB 1145, 89th Leg., R.S. (Tex. 2025) (enacted without express grandfathering provision for pending projects).
[45] Tex. Water Code § 11.039.
[46] Tex. Water Code § 11.053.
[47] Silver v. Pueblo Del Sol Water Co., 244 Ariz. 553 (2018).
[48] Tex. Local Gov’t Code § 211.003.
[49] Tex. Water Code § 11.121.
[50] Tex. Local Gov’t Code § 211.003 (authorizing municipal zoning authority). Texas cities including Roanoke and Athens have deployed planned development overlay districts in connection with large-load industrial development.
[51] 33 U.S.C. §§ 1311(d), 1314(b), (m); Waterkeeper All. v. United States EPA, 140 F.4th 1193 (9th Cir. 2025).