Skip to main content

Vol. No. 56-1 Natural Resources

September 25, 2026

Eminent Domain, Public Use, and the Texas Data Center Boom

By: Katie Jeffress & Nicholas Paganini

Introduction

In the United States, eminent domain refers to the government’s power to take private property for public use, provided that the owner receives fair compensation.1 However, “public use” significantly limits the government’s authority.2

The public use requirement ensures that entities that are authorized to take land (e.g., cities and utilities) can only do so if there is a genuine public purpose, rather than a private benefit disguised as a public one.3 This article discusses the public use limitation on eminent domain as it relates to artificial intelligence (AI) data centers. Can eminent domain be used to benefit data centers, and, if so, in what situations?

Modern AI data centers present novel public use issues to the United States’ legal system. In both form and function, data centers are categorically different from any preceding commercial development. These data centers’ utility to the public is also in question, especially when examined alongside the scale and magnitude of their resource consumption.

It is unlikely that private land can be taken and given directly to an AI company to build a data center, as Texas and federal law forbid exclusively private takings.4 However, the sudden, massive demand for data centers in Texas and across the United States draws greater attention to using eminent domain in facilitating their construction. This article defines the limits of eminent domain’s use in aiding data center construction and explores the ways data centers could be considered to have a public use.

Data Centers Have Grown Rapidly

The emergence of generative AI seems poised to change the dynamics of land use in America. In 2025, the amount of money spent on data centers’ operations and computing capabilities surpassed the amount spent on all other commercial building development combined.5 Data centers and information processing technology accounted for approximately 92% of U.S. Gross Domestic Product (GDP) growth in the first half of 2025.6 This growth shows no signs of stopping, as major players, like Google, Amazon, Meta, and Microsoft, are expected to invest hundreds of billions more in 2026.7

Developing AI data centers is a resource-intensive process. A mid-sized data center consumes roughly as much water as a small town, and larger centers require as much water as a city of 50,000 people.8 The average data center uses as much electricity as 10,000 to 25,000 households, and the most extreme examples can draw more electricity than entire cities, and in some cases, states.9 These are neither small nor scarce projects. Axios reports that nearly 3,000 new data centers are currently under construction or planned across the U.S., adding to the more than 4,000 centers operating right now.10 Politicians, including Senator Bernie Sanders and Governor Ron DeSantis, have called for a temporary moratorium or the slowing of AI data center production to allow for legislators to catch up with demand.11

In Texas, developers are planning to cover hundreds of acres of land with energy and water-intensive warehouses. This expansive infrastructure growth would have devastating effects on the land itself, making it no longer suitable for agriculture, housing, or nature-related purposes. Thus, the use of eminent domain to develop AI data centers will contribute to significant ripple effects, affecting Texas’ environment and economy for years to come.

Eminent Domain Will Be an Essential Tool in Enabling Data Center Production

The inevitable growth of AI depends, largely in part, on data centers and the infrastructure required to support them. Eminent domain has been an essential tool for the proliferation of AI data centers in Texas and across the United States, as many of these centers could not have been built without pervasive eminent domain takings.12 Electricity transmission lines, substations, and water infrastructure—all essential for data center development—are candidates for the use of condemnation authority.

Consequentially, electricity providers are planning transmission projects that are explicitly justified by AI data center demands, and will likely involve eminent domain takings.13 The Electric Reliability Council of Texas (ERCOT), the state’s grid operator, has acknowledged that large data centers’ rapid expansion dramatically increases electricity demand and puts pressure on transmission planning.14 ERCOT has also noted that Texas needs 3,000 miles of new long-distance transmission lines, largely “to host more power-intensive cryptocurrency mines and data centers” as tech companies race to develop their AI capabilities.15

These long-distance lines will be built by transmission service providers, such as Oncor or CenterPoint.16 Under Texas law, those utilities have the authority to acquire easements for transmission lines and substations under eminent domain if voluntary deals fail.17

Based on these developments, it appears that eminent domain is being used to develop the infrastructure required to build out data centers. However, eminent domain’s ability to spur data center development is not unlimited. Any taking must satisfy the constitutional requirement of “public use”—a constraint that has been both expanded and policed through decades of judicial interpretation and legislative response—to be lawful.

Eminent Domain Power Is Limited by “Public Use”

Kelo v. City of New London is the most prominent, modern case addressing what public use in eminent domain means under the Takings Clause of the Fifth Amendment. In Kelo, the City of New London sought to redevelop a 32-acre area, which consisted of 115 privately-owned parcels, to stimulate economic growth after a new Pfizer pharmaceutical facility was built.18 Susette Kelo and other property owners sued New London after their properties were taken.19 These property owners argued that the city violated the Fifth Amendment’s takings clause—specifically, that the taking and selling of private property to private developers was not considered public use.20 However, SCOTUS held that these takings qualified as “public use,” despite the fact that the land was not going to be used by the public.21 The Court defined “public use” broadly to include not only direct use or ownership by the public, but also takings that serve a “public purpose.”22 The Court found that the government was not taking land solely to benefit a certain group of private individuals, but was rather following a comprehensive economic development plan with tax and job benefits, and that such justifications deserve deference to legislative judgments on what public needs justify takings.23

While the Kelo decision allows entities to use eminent domain for economic development under the federal Constitution, the Texas Constitution and state statute prohibit such an approach. According to the Texas Constitution, “public use” is the “ownership, use, and enjoyment of the property” by the government or another entity with eminent domain power (i.e., a common carrier).24 Explicitly rejecting the Kelo decision, the Texas Legislature passed Texas Government Code Section 2206.001, which prohibits takings that confer private benefits, are pretexts for private gain, or are primarily for economic development purposes.25 The language of the statute states, in pertinent part, that:

(b) A governmental or private entity may not take private property through the use of eminent domain if the taking:

(1) confers a private benefit on a particular private party through the use of the property;

(2) is for a public use that is merely a pretext to confer a private benefit on a particular private party;

(3) is for economic development purposes . . .26

The Supreme Court of Texas has further defined the public use test in Denbury Green Pipeline-Texas, LLC v. Texas Rice Land Partners, Ltd. Specifically, the Texas Supreme Court determined in this case that a common carrier’s use is considered public when there is a reasonable probability that, at some point after development, the use will assist one or more individuals unaffiliated with the condemner.27 In this case, Denbury Green Pipeline intended to condemn an easement across Texas Rice Land Partners’ land for a carbon dioxide pipeline.28 Texas Rice Land Partners denied Denbury access to survey its land on the grounds that the development would only serve the pipeline company’s private interests.29 Denbury asserted that it had eminent domain authority because the Texas Railroad Commission granted the entity common carrier status, and provided evidence that the gas would be utilized by two companies unaffiliated with Denbury. Ultimately, the Court sided with Denbury, holding that “there was a reasonable probability that, at some point after construction, the [CO2 pipeline] would serve the public by transporting CO2 for one or more customers who will either retain ownership of their gas or sell it to parties other than the carrier.”30

Three Different Roads to “Public Use”

The controlling law for “public use” and modern trends for AI data center construction raises three important considerations. First, how can the “broad” Denbury verdict on assistance to unaffiliated individuals be reconciled with the narrow constitutional and statutory limits on “public use?” Second, could an AI company qualify as a common carrier, expanding eminent domain power, to benefit data center construction? Third, regarding projects outside of Texas, could AI data centers serve an economic development purpose, thereby granting the projects a “public use?”

1. Contextualizing Denbury

The rule from Denbury does not necessarily greenlight any taking by a common carrier that benefits another entity. Although Denbury appears to allow such takings, the case and the Court’s holding must also be understood in the context of the Texas Constitution and Government Code.

The pipeline in Denbury served a common carrier purpose, as it was intended to move the CO2 associated with the two companies and the condemner.31 In contrast, while AI data centers may transmit data, they do not move the tangible products that are traditionally associated with common carriers, such as industrial gas. Instead, these centers consume resources, like water and electricity, to function, without supplying a physical product directly to customers.32 Whether a utility can condemn an easement to supply water or electricity directly to a data center is being debated, as the actions raises questions of public use since the easements would be primarily serving the data center and not local customers.33 According to the Texas Government Code, takings that benefit only one private user are an impermissible use of eminent domain authority.34 Any argument that the AI data center itself serves a public purpose would present a first impression case, requiring fact-specific inquiries on how the center operates.

2. Assessing if AI Companies can Be Common Carriers

Granting AI companies common carrier status could allow AI data centers to condemn land themselves, easing public use inquiries, and solving the apparent contradiction in the Denbury problem described above. If data centers were to gain common carrier status, they may be more likely to act on behalf of the public good and might be less liable to claims of particularized private benefit from a taking. However, it is unlikely that AI companies would be classified as common carriers because their activities fall outside the traditional scope of common carrier status.

The Supreme Court of Texas defines “common carriers” as “those in the business of carrying passengers and goods who hold themselves out for hire by the public,”35 and has instructed that “[w]hen determining whether someone who provides transportation is a common carrier, we look to their primary function.”36

Of the entities that exercise eminent-domain authority in Texas, AI companies share similar circumstances and impacts as pipeline operators.37 Similar to pipelines, AI data centers are privately owned infrastructure projects that serve as intermediaries in the production process, rather than direct public service providers. Pipelines transport oil or gas, rather than creating the resources themselves. Similarly, AI data centers process and transmit data, which then enables downstream economic activity.38 Texas law permits pipeline companies to condemn land only because they qualify as common carriers—entities that serve the public by making their infrastructure available to unaffiliated users.39

However, unlike pipelines, AI data centers do not transport public goods or services on equal terms, nor do they offer open access to their infrastructure.40 Rather, these centers operate as closed systems that serve proprietary ends, making them incompatible with common carriership. Data centers manipulate data to create AI models, which are marketed to the public. Further, the primary function of these centers is to create AI models for profit, not for data transportation.41 Eminent-domain authority likely cannot, therefore, be extended to AI data centers without eroding the common carrier definition.

A better argument may be made for “colocation” data centers, as these centers use shared infrastructure for multiple companies. However, this use still fundamentally differs from the traditional model, in which common carriers operate with open access to the public. Colocation centers do not hold themselves out to the public; instead, they are highly selective in who they partner with and have broad discretion to refuse service.42

For an AI company to plausibly qualify as a common carrier under Texas law, the company would need to operate as a neutral conduit for third-party computation.43 These operations, however, would likely undermine the current business goals that modern AI companies are built upon.

3. Assessing if AI Data Centers Can Provide Sufficient “Economic Development” to Justify a Public Use Determination

SCOTUS in Kelo acknowledged that its decision on public use established only a federal baseline for the requirement and did not preclude states from imposing stricter limitations on eminent domain usage for economic development purposes.44 This decision explains why, as stated above, economic development alone cannot sustain a public use determination for AI data centers in Texas.

Outside of Texas, this question is of greater consequence. There is a powerful national political movement to build AI infrastructure as far and fast as possible on the grounds of economic and strategic development,45 and political winds may shift within Texas to closer align with the federal approach. Specifically, the Trump administration is aggressively exploring ways to increase data center construction.46 In July 2025, President Trump signed an executive order titled Accelerating Federal Permitting of Data Center Infrastructure (EO 14318), which directed federal agencies to streamline permitting, expand incentives, and utilize federal land to support rapid construction of data centers and AI infrastructure.47 In December 2025, Trump also signed an executive order titled Ensuring a National Policy Framework for Artificial Intelligence, stating that he imminently intends to sue states over AI laws that limit the “United States’ global AI dominance.”48 The Trump administration’s position is forcefully anti-regulatory and was likely made with significant input from industry players.49 For these reasons, it is worthwhile to assess the arguments surrounding economic development justifications for data centers.

If economic development were a valid public use for AI data centers, governments could easily take private land for certain projects. In Kelo, SCOTUS held that economic development constitutes a valid “public use” under the Takings Clause of the Fifth Amendment.50 The Court reasoned that the term “public use” should be interpreted broadly to include “public purpose,” reflecting a longstanding policy of deference to legislative judgments in this area.51 Specifically, the Court emphasized that promoting economic development is a “traditional and long-accepted function of government” and that there is “no principled way of distinguishing economic development from the other [previously recognized] public purposes,” such as facilitating agriculture, mining, or eliminating blight.52

The Court in Kelo upheld the City of New London’s decision to take private property as part of a comprehensive economic development plan aimed at revitalizing an economically distressed area.53 The plan included constructing a hotel, restaurants, retail stores, residences, and office space, which the city believed would create jobs, increase tax revenue, and benefit the community as a whole.54 The Court deferred to the city’s determination that the development plan served a public purpose, although the plan involved transferring property to private developers.55 However, the Court cautioned that takings cannot be justified if they are a mere pretext to confer a private benefit.56

Considering the Court’s deferral to the city in Kelo regarding the project’s public purpose, state law and public opposition to AI data centers can provide information on whether a taking for a data center falls under an economic development claim.

It also must be noted that, although job creation is often cited alongside tax revenue as an economic justification for AI data center construction, the facilities themselves are weak long-term job creators.57 Modern data centers require a fraction of the personnel that previous centers needed, and nearly all of the jobs these centers create are temporary and rooted in their construction, not their operation.58

Despite SCOTUS rulings, AI data centers are unlikely to be justified as economic developments. Apart from public political opposition and state laws that currently outlaw economic development as a public use justification, data centers are likely to be substantially different from projects that have sustained public use determinations on economic development grounds and cannot be so justified.

Conclusion

The rapid expansion of AI data centers places unprecedented pressure on Texas land, water, and energy systems, but that pressure does not relax constitutional limits on eminent domain. While utilities may condemn properties to build infrastructure that serves the public, neither economic pressure nor industry necessity permits eminent domain usage to support private facilities built to exclusively benefit individual AI companies. If, however, data centers were common carriers, they would likely be able to condemn land, or, if these centers were justified by economic development-related purposes, governments could take land directly for data centers. Regardless, as Texas navigates rapid AI growth, the public use requirement will remain a chief consideration for all parties involved.

 

Katie Jeffress is an associate at Baker Botts LLP. Her practice focuses on a range of environmental issues at the state and federal level, including regulatory compliance, permitting, and litigation. Her experience includes air and water permitting, waste issues and enforcement matters. She earned her B.S. in Environmental Economics from Juniata College and her J.D. from the University of Texas School of law, with high honors.

Nicholas Paganini is a recent graduate from Texas Law originally from Lake Arrowhead, California. He attended the University of California, Irvine for his B.A. in Environmental Science before coming to UT to study law. He is interested in urban and environmental planning and intends to practice real estate law in Southern California.

  1. U.S. Const. amend. V.; Eminent Domain, Cornell L. Sch.: Legal Info. Inst. (Dec. 2022), https://www.law.cornell.edu/wex/eminent_domain [https://perma.cc/KA3X-44F2] (explaining that the Fifth Amendment of the Constitution states that only the government may exercise this power if they only provide compensation to the property owners).
  2. See Eminent Domain, supra note 1.
  3. Philip Nichols, Nichols on Eminent Domain § 4.7 (Julius L. Sackman et al., 3d ed. 2025).
  4. See Tex. Const. art. 1, § 17; U.S. Const. amend. V.
  5. Mark P. Mills, The Rise of AI: A Reality Check on Energy and Economic Impacts, Nat’l Ctr. for Energy Analytics (Nov. 13, 2025), https://energyanalytics.org/the-rise-of-ai-a-reality-check-on-energy-and-economic-impacts/ [https://perma.cc/FKW8-6XZ4].
  6. Nick Lichtenberg, Without Data Centers, GDP Growth Was 0.1% in the First Half of 2025, Harvard Economist Says, Fortune (Oct. 7, 2025), https://fortune.com/2025/10/07/data-centers-gdp-growth-zero-first-half-2025-jason-furman-harvard-economist/ [https://perma.cc/G28Y-SWCB].
  7. Why AI Companies May Invest More than $500 Billion in 2026, Goldman Sachs (Dec. 18, 2025), https://www.goldmansachs.com/insights/articles/why-ai-companies-may-invest-more-than-500-billion-in-2026 [https://perma.cc/8M29-58N5]; Noel Randewich & Shashwat Chauhan, Nvidia Briefly on Track to Become World’s Most Valuable Company Ever, Reuters (July 3, 2025), https://www.reuters.com/business/nvidia-set-become-worlds-most-valuable-company-history-2025-07-03/ [https://perma.cc/4KU7-KMLQ].
  8. Jon Gorey, Data Drain: The Land and Water Impacts of the AI Boom, Lincoln Inst. (Oct. 17, 2025), https://www.lincolninst.edu/publications/land-lines-magazine/articles/land-water-impacts-data-centers/ [https://perma.cc/5VMG-AXQM].
  9. Id. (“Meta’s Hyperion data center in Louisiana, for example, is expected to draw more than twice the power of the entire city of New Orleans once completed. Another Meta data center planned in Wyoming will use more electricity than every home in the state combined.”).
  10. Alex Fitzpatrick, America’s Data Center Growth Hot Spots, Mapped, Axios (Dec. 18, 2025), https://www.axios.com/2025/12/18/data-center-growth-map-states [https://perma.cc/79FP-D986].
  11. Jenna Ruddock, The Real Race for an AI Moratorium: Stopping Data Centers, Tech Policy Press (Dec. 17, 2025), https://www.techpolicy.press/the-real-race-for-an-ai-moratorium-stopping-data-centers/ [https://perma.cc/5CP2-T4AN].
  12. Arcelia Martin, Explosion of Data Centers Causes Planning Struggles for Texas Power Grid, Tex. Trib. (Oct. 30, 2025), https://www.texastribune.org/2025/10/30/texas-ercot-power-grid-data-centers-puc/ [https://perma.cc/SWW8-KMLP].
  13. Id.
  14. Id.
  15. Claire Hao, Why Texas Needs 3,000 Miles of New Power Lines, More than the Distance Driving Cross Country, Hous. Chron. (Jan. 27, 2025), https://www.houstonchronicle.com/business/energy/article/texas-transmission-crypto-data-centers-20023584.php [https://perma.cc/L5LF-NZB4].
  16. Id.
  17. Id.
  18. Kelo v. City of New London, 545 U.S. 469, 473 (2005).
  19. Id. at 475.
  20. Id.
  21. Id. at 476.
  22. Id. at 480.
  23. Id. at 479–81.
  24. Tex. Const. art. 1, § 17.
  25. Tex. Gov’t Code § 2206.001 (“Limitation on Eminent Domain for Private Parties or Economic Development Purposes”).
  26. Tex. Gov’t Code § 2206.001 (2023).
  27. Denbury Green Pipeline-Tex., LLC v. Tex. Rice Land Partners, Ltd., 510 S.W.3d 909, 914 (Tex. 2017).
  28. Id. at 911.
  29. Id.
  30. Id.; Common Carrier, Cornell L. Sch.: Legal Info. Inst. (June 2021), https://www.law.cornell.edu/wex/common_carrier [https://perma.cc/6TZQ-9Q3V] (defining a common carrier as “a person or a commercial enterprise that transports passengers or goods for a fee and establishes that their service is open to the general public”).
  31. Denbury Green Pipeline, 510 S.W.3d at 914.
  32. See Adam Bridgman & Max Adelman, #5. Data Centers Unplugged: Physical Assets for the Digital Economy, Ironspring Ventures (Oct. 15, 2024), https://ironspring.com/5-data-centers-unplugged-physical-assets-for-the-digital-economy/ [https://perma.cc/ES2P-ENXD].
  33. Nikshep Myle, Your Power Company Can Seize Your Land for AI Data Centers – Here’s How to Fight Back, Yahoo News (July 20, 2026), https://www.yahoo.com/news/us/articles/power-company-seize-land-ai-153147592.html [https://perma.cc/J9Z8-AQ24].
  34. Tex. Gov’t Code § 2206.001.
  35. Mt. Pleasant Indep. Sch. Dist. v. Estate of Lindburg, 766 S.W.2d 208, 213 (Tex. 1989).
  36. Id.
  37. Michael Hren, The Case for Heightened Scrutiny: AI Development, Pipelines, and Eminent Domain, Tex. Undergrad. L.J. (Apr. 20, 2026), https://www.texasulj.org/blog/the-case-for-heightened-scrutiny-ai-development-pipelines-and-eminent-domain [https://perma.cc/8AEJ-ZN52].
  38. Consumer Energy All., Data Center Toolkit (2025), https://consumerenergyalliance.org/cms/wp-content/uploads/2025/10/CEA-FL_Data-Center-Toolkit.pdf [https://perma.cc/N2YS-QCZN] (“Beyond jobs, data centers transform economies. They generate billions in investment, boost local tax revenue, and strengthen communities without straining services.”)
  39. Tex. Nat. Res. Code § 111.019(a).
  40. Private vs Public Data Center: What’s Best for Your Data?, Impact (Jan. 4, 2024), https://www.impactmybiz.com/blog/blog-private-vs-public-data-center/ [https://perma.cc/E8VC-SN6H].
  41. See Robert J. Szczerba, What Is a Data Center and Why Are They Suddenly Everywhere?, Forbes (July 9, 2026), https://www.forbes.com/sites/technology/article/what-is-a-data-center/ [https://perma.cc/ZUG2-BV5G].
  42. See David Zorn, A. Fletcher Mangum & Martina Arel, The Impact of Colocation Data Centers on the State and Local Economies of Texas (2003), https://static1.squarespace.com/static/63a4849eab1c756a1d3e97b1/t/66d760bf9e495b6e3fe844ea/1725391088430/The+Impact+of+Colocation+Data+Centers+on+Texas+2023 [https://perma.cc/WT37-S9CH].
  43. See Texas Rice Land Partners, Ltd. v. Denbury Green Pipeline-Texas, LLC, 363 S.W.3d 192, 202 (Tex. 2012) (requiring a “reasonable probability” that the pipeline will transport product for customers who retain ownership or sell to others); Tex. Nat. Res. Code § 111.002 (defining common carriers as entities transporting products “to or for the public for hire”).
  44. Kelo, 545 U.S. at 489.
  45. Exec. Order No. 14179, 90 Fed. Reg. 8741 (Jan. 23, 2025); Exec. Order No. 14365, 90 Fed. Reg. 58499 (Dec. 11, 2025); Exec. Order No. 14318, 90 Fed. Reg. 35385 (July 23, 2025).
  46. Exec. Order No. 14318, 90 Fed. Reg. 35385 (July 23, 2025).
  47. Id.
  48. Id.
  49. See id. (directing financial support, streamlining environmental reviews, and expediting permitting for large data centers).
  50. Kelo, 545 U.S. at 484.
  51. Id. at 480.
  52. Id. at 489.
  53. Id.
  54. Id.
  55. Id.
  56. Id.
  57. Jon Gorey, Data Drain: The Land and Water Impacts of the AI Boom, Lincoln Inst. Land Pol’y (Oct. 17, 2025), https://www.lincolninst.edu/publications/land-lines-magazine/articles/land-water-impacts-data-centers/ [https://perma.cc/VR29-AWJR].
  58. Va. Joint Legis. Audit & Rev. Comm’n, Data Centers in Virginia, https://jlarc.virginia.gov/landing-2024-data-centers-in-virginia.asp [https://perma.cc/K22K-Q2KJ] (last visited Aug. 28, 2026) (noting that a typical 250,000-square-foot data center may have approximately 50 full-time workers, compared with approximately 1,500 workers on site at the height of construction).