Small-Footprint, Net Zero Water and Power use, Locally-Owned Data Centers

A phased plan and policy case for distributed, locally-owned compute infrastructure over concentrated corporate mega-campuses, grounded in a Luna County case study

Prepared by Mary Meade for consideration by New Mexico state and local elected officials
August, 2026.

New Mexico should establish a formal state goal — backed by financing tools and permitting/regulatory support — to grow multiple small-footprint, locally or cooperatively-owned data-center facilities across the state's municipalities and counties, as a deliberate policy alternative to the current default of a small number of mega-scale campuses owned by out-of-state corporations. The state does not need to choose between data infrastructure and community control of water, land, and power; it needs to choose which ownership model it actively encourages. Right now, by default, it is encouraging the wrong one. This proposal lays out why that alternative is credible — not as an abstract ideal, but as a phased plan already partly underway, with the specific local evidence to show it is buildable in Luna County today.

The problem, and it is already here

This is not a hypothetical concern about the future. In Doña Ana County, Oracle's "Project Jupiter" data center is under development on roughly 1,400 acres — larger than New York's Central Park — with a planned capacity of 2.5 gigawatts and a reported $165 billion in investment. The project initially planned to draw on the order of 1 million gallons of water per day, acquired through water rights bought from a working sod farm, in a region that receives roughly 9 inches of rainfall a year. After public criticism, the developer switched from water-intensive natural-gas turbines to fuel cells and now claims the combined facility and power system will use approximately 11 million gallons of non-potable recycled water; critics and neighboring residents question whether that figure will hold. Whatever the final number, the structural pattern is the one this proposal is about: a single out-of-state corporation, negotiating from a position no single New Mexico county can match, making decisions about local water and land that the community has limited power to shape after the fact.

This pattern is not unique to New Mexico, and it is not incidental to how hyperscale data centers work. A mid-sized data center alone can use on the order of 300,000 gallons of water a day, and national data-center water consumption is projected to double or quadruple between 2023 and 2028 as this style of development accelerates; roughly a fifth of existing U.S. data centers already sit in watersheds under moderate-to-high stress. Because so few companies own so much of this infrastructure, they also carry outsized leverage over the communities that host them and over anyone downstream who depends on the compute they provide.

A plan that starts from where communities already are

This proposal does not require starting from zero. Many New Mexico communities are already in the middle of the first phase of this plan, without having named it that: getting rural broadband built. That is not a detour from the model below — it is Phase 1 of it, and it turns out to be the same historical playbook the state and country have used before. When private power utilities found it unprofitable to serve rural areas in the 1930s and simply didn't, the federal response was not to force them to serve — it was to fund an alternative ownership model. The Rural Electrification Act of 1936 financed new member-owned cooperatives, through loans rather than grants, to build and own their own distribution systems. Hundreds of these cooperatives still operate today, member-owned rather than shareholder-owned, and that same institutional lineage is now the primary vehicle bringing broadband to rural New Mexico. The entity a community builds to get broadband online now can plausibly be the same entity that runs shared compute infrastructure later — the way an electric cooperative's broadband subsidiary already reuses its parent's member-ownership and physical infrastructure. What follows is a three-phase plan built on that continuity.

Phase 1 — Broadband, using tools that already exist

The money for this already exists, and it flows through the Rural Electrification Act's direct descendant. The Rural Utilities Service (RUS) — the modern name for the agency the 1936 Act created — still makes broadband loans today. USDA's ReConnect program, run through RUS, funds rural broadband construction as grants, loans, or 50/50 combinations up to $50 million per award, and eligible applicants explicitly include state and local governments and their subdivisions and cooperatives or mutual organizations, not just private telecoms (USDA ReConnect overview). Layered on top of that, the federal BEAD program has allocated $42.45 billion across states specifically to close rural broadband gaps, prioritizing fiber-optic builds; most states are still in planning or early construction, with rural and mountainous areas expected to see service arrive as late as 2027–2029 (BEAD funding tracker).

Who is actually doing this kind of building nationally is instructive. More than 250 electric cooperatives are now deploying broadband, over 200 of them already serving customers. The consistent pattern is that the electric co-op stands up a broadband subsidiary — a separate entity with the same member-ownership DNA — and leverages assets it already owns: utility poles and right-of-way easements it does not have to negotiate for, because it has had them for decades (America's Electric Cooperatives). Where no existing electric cooperative serves an area, a county or multi-county public utility district, or a Joint Powers Authority formed specifically for the build, does the same job starting from scratch — following the model of UTOPIA Fiber in Utah, where eleven cities formed a bonded consortium in 2002, since grown to 19-plus municipalities, that owns the physical network while multiple private ISPs compete to serve customers on top of it under an open-access requirement written into Utah law (UTOPIA Fiber history). That open-access separation — public or cooperative ownership of the physical plant, competitive service on top — is what stops a broadband build from quietly becoming a new monopoly, which is the same failure mode this proposal is designed to avoid at the compute layer later. One obstacle worth flagging early: a number of states have passed laws restricting or preempting municipal or cooperative broadband, often at the urging of incumbent private providers. New Mexico's specific statutory landscape on this point should be confirmed before a governance vehicle is chosen.

Phase 2 — Regional interconnection

Once a community has broadband, the next constraint is backbone connectivity and peering — getting that local network connected to everything else affordably, without a single small municipality negotiating transit pricing alone. This is exactly the problem regional research-and-education networks (RENs) solve for universities. Internet2 and its regional affiliates work on a two-tier structure: individual institutions connect through a regional consortium rather than directly to the national backbone, and dues scale to what a member can actually pay (Internet2 membership levels). Governance is deliberately member-led: a small Board of Trustees sets strategic direction, while day-to-day program direction runs through member advisory groups (Internet2 governance). New York's BOCES system shows the same logic applied to a different kind of pooled demand — school districts too small to justify specialized infrastructure alone pool into regional "Co-Ser" agreements with state-formula cost-sharing, opting in per service rather than merging governance entirely (BOCES shared services). For a rural broadband cooperative, the equivalent move is joining or forming a regional interconnection consortium with neighboring counties or cooperatives, rather than each one separately buying transit.

Phase 3 — Shared local compute

This is the heart of the proposal, and it is genuinely a later phase — reliable local connectivity and regional backbone need to exist before shared compute infrastructure is useful to anyone. But the governance vehicle a community stands up for Phase 1 can plausibly carry through to Phase 3, the same way an electric cooperative's broadband subsidiary reuses its parent's member-ownership and infrastructure. A few design principles make this phase structurally different from the hyperscale model, all aimed directly at the water, land, and monopoly-pricing concerns above.

Anchor tenants aggregate real, modest demand instead of building for speculative growth. City or county government IT, the school district, the library system, and a local clinic or hospital don't individually need a data center, but together have enough steady load to justify a small shared facility — the same demand-pooling logic as BOCES and the RENs.

Ownership stays with a public or member-owned entity, not outside investors — a Joint Powers Authority, public utility district, or cooperative corporation, the same legal forms UTOPIA and the electric cooperatives already use. Customers and owners being the same people is the structural answer to monopoly pricing: there is no shareholder on the other side of the table with an incentive to extract.

Open-access separation continues at the compute layer: the consortium owns the physical facility, and member institutions or independent service providers run workloads on it, rather than the consortium itself becoming a new monopoly compute provider.

This governance form is not unsolved even in the compute space specifically. Community-owned data-center cooperatives already exist at smaller scale — Nubo Cooperative in Belgium runs cloud, email, and storage on locally-owned servers with customers as voting members; Infomaniak in Switzerland is a workers' cooperative; Germany's GAD ran cooperative data processing for member banks from 1963 to 2015 (RESET.org). None were municipal-anchor-tenant builds specifically, but the legal and governance mechanics translate directly.

Right-sizing and siting solve the resource-footprint problem directly, rather than mitigating it after the fact. A facility sized to real aggregated anchor-tenant demand, rather than speculative growth, with cooling matched to local water availability, avoids most of the water and land impact by construction. Siting a facility inside existing town or county utility infrastructure also opens up waste-heat reuse that a remote hyperscale campus never bothers with — Stockholm Exergi, a utility jointly owned by the City of Stockholm, buys waste heat from data centers as a tradable commodity and feeds it into the municipal district heating network (Eurelectric).

Technical feasibility: a Luna County case study

Phase 3 is no longer purely theoretical for this region — the pieces needed to power and site it already exist on the ground in Luna County.

Footprint and siting. A 2-megawatt-class facility — enough to serve a city government, school district, library system, and local hospital's shared compute needs — requires roughly 1 to 3 acres of total site footprint, including building, earthwork, access, and substation, versus the 1,400 acres committed to Project Jupiter. That footprint can be reduced further in visual and environmental impact by building partly or fully below grade: underground and bunker-sited data centers are an established category, not an experimental one — from Iron Mountain's former-limestone-mine facility in Pennsylvania to a hardened former nuclear command bunker in Newbury, England, now operating as a roughly 1.5–2 megawatt colocation facility. Burial pairs naturally with closed-loop ground-source cooling, a sealed underground pipe loop that rejects heat without evaporative water loss, unlike a cooling tower — essentially eliminating water consumption for cooling by design rather than mitigating a large one. Burial also frees the surface for renewables, though the noise-reduction benefit of burial applies mainly to fan, chiller, and generator noise; ventilation shafts still need acoustic treatment.

Solar. Luna County already hosts operating utility-scale solar — the Deming Solar Energy Center and Alta Luna — consistent with the Southwest's strong resource; utility-scale solar nationally averages roughly a 25% capacity factor, higher (around 31%) in the sunniest western regions (LBL, 2025). Matching a flat 2-megawatt load on an annual-energy basis takes roughly 30 to 80 acres of panels depending on mounting technology and exact local resource — far more land than the facility's own footprint, but modest by utility-solar standards, and land that can be sited to avoid competing with other local land uses.

Wind — verified locally. The Macho Springs Wind Farm — 50.4 megawatts, 28 Vestas V100 turbines, on roughly 1,980 acres of private ranch land still largely in grazing use — has operated in Luna County, roughly 20 miles northeast of Deming along Highway 26, since November 2011, fully contracted to Tucson Electric Power (Capital Power). Reported generation data puts its actual capacity factor at approximately 26%, and it is classified by at least one industry data source as "low wind" quality — notably more modest than the figure implied by the developer's own "14,000 homes served" marketing language, which would back-calculate closer to 33–35%. That gap is worth stating plainly: measured generation data, not marketing figures, should drive any real system sizing. Even at the more conservative, measured number, a real wind resource that has run as a bankable, utility-contracted asset for over thirteen years — on land that leaves the underlying ranching use largely intact — is strong evidence for pairing wind with solar in this region rather than relying on solar alone. What remains unverified is the hour-by-hour generation timing at this specific site — whether local wind is meaningfully stronger at night, as is well documented for wind farms further east in the Great Plains and West Texas corridor. Confirming that would take EIA-930 balancing-authority data, a request to the operator, or a NREL WIND Toolkit pull for the site's exact coordinates (32.5602, -107.5210).

Storage. Bridging an average night for a flat 2-megawatt load takes roughly 24 to 35 megawatt-hours of lithium battery storage — on the order of six to seven Tesla Megapack-class units, a normal scale for current utility storage. Multi-day resilience through the region's July–September monsoon season is the harder, less-solved case; long-duration technologies such as Form Energy's 100-hour iron-air batteries are moving from pilot to commercial deployment for exactly this duty cycle, including projects already sized to firm data-center loads elsewhere in the country, and are worth tracking as a pilot option for a New Mexico facility rather than assuming the problem is already solved.

What state action should look like

Set the goal explicitly. A legislative finding or executive directive establishing that New Mexico's default posture toward new large-scale compute infrastructure favors distributed, locally or cooperatively-owned facilities — comparable in spirit to a renewable portfolio standard, but aimed at ownership structure and facility scale rather than fuel mix.

Fund it through existing, proven mechanisms. A state loan or grant program modeled on the Rural Utilities Service's ReConnect structure — eligible applicants explicitly including counties, municipalities, and cooperatives — rather than a program that only large private developers can realistically access.

Grant clear bonding and joint-ownership authority. Explicit statutory authority for New Mexico municipalities and counties to jointly bond and co-own shared infrastructure, following the model Utah used to create the Utah Infrastructure Agency for UTOPIA Fiber, so no single small jurisdiction has to originate this structure alone or take on the legal risk of being first.

Require open-access terms as a condition of state support. Any facility receiving state financing or incentives should be required to operate on an open-access basis — the public or cooperative entity owns the physical infrastructure, and service is competitively provided on top of it — so state investment doesn't quietly recreate a new local monopoly instead of a corporate one.

Commission a real siting and engineering study. The figures in this proposal are planning-level estimates assembled from public data, not a completed engineering study. A state-funded feasibility study — ideally sited in Luna County, given the existing Deming solar and Macho Springs wind infrastructure already on the ground — would convert this from a policy direction into a buildable pilot.

Open questions and honest caveats

This proposal is a planning-level case, not a finished engineering design, and it is stronger for saying so plainly. Technical operations expertise is a real gap at every phase — cooperatives and BOCES-style consortia typically hire or contract this jointly rather than expecting each member institution to run infrastructure itself, and that should be planned for from Phase 1 rather than assumed to appear by Phase 3. Demand-aggregation risk — an anchor tenant overestimates its need, or drops out — is why UTOPIA's founding cities collectively backstop bond debt instead of each carrying it alone; the same collective-backstop logic should apply to any bonding for the broadband build. Whichever governance vehicle is chosen at Phase 1 — electric co-op subsidiary, county public utility district, or a new Joint Powers Authority — is worth choosing with Phase 3 in mind, since standing up a new entity later is real friction that both the electric cooperatives and UTOPIA avoided by extending an existing one. And as noted above, the hour-by-hour timing of the local wind resource, and New Mexico's specific statutory posture toward municipal and cooperative infrastructure, are both concrete, answerable questions that should be resolved early rather than assumed.

Closing

The choice in front of the state is not whether new data infrastructure gets built in New Mexico — Project Jupiter and projects like it show that decision has already been made by the market. The choice still open is who owns what gets built next, at what scale, and whether the communities hosting it have a real seat at the table rather than a permitting hearing after the site is already selected. New Mexico has the underlying resources — proven solar, a real if modest wind resource, and land — and it has the legal tools, largely borrowed from a century of rural electrification and two decades of municipal fiber, to build the alternative. It also, in Luna County, has a community already partway through the first phase of this exact plan. What's missing is a stated state goal to carry it the rest of the way.