← Back to the proposal
Sources & Verification
Are these numbers real?
Every claim in the proposal, checked against primary sources — independently, in July 2026 — with the link and an honest verdict on each. The core numbers hold up. Four needed a correction; those have now been fixed in the paper and the PDF, and are documented here in full.
The short answer: yes — and the four corrections are now applied.
The load-bearing quantities are real and sourced: La Osa's 3,385 acres, 3 GW (as-proposed) and $33B; the aquifer's 8-million acre-foot shortfall; the grid's 280 MW district peak; Griffin's 2,685 acres / 550 / 500 MW; the $2.68 water-bill figure; and the Gift Clause case law. Each links to a primary filing, a government source, or established reporting below.
Verification turned up four things to fix — a wrong utility, a mislabeled number, a case described backwards, and a cooling term. All four are now corrected in both the on-page proposal and the downloadable PDF; the record of what changed is kept in the box below for transparency.
⚠ What changed after the paper was written
The proposal is dated May 28, 2026 and flagged the May 27 La Osa outcome as unconfirmed. It has since resolved: at the May 27 hearing the Board continued the vote 3–1 to August 26, 2026, and the developer (Vermaland) cut La Osa ~80% — from 59 data-center buildings to about 11, and from ~3 GW toward ~1 GW. So the paper's La Osa scale figures were accurate as proposed; the live project is smaller. The framework applies at any size.
✎ Corrections found in verification — now fixed in the paper
All four below have been corrected in the live proposal and the downloadable PDF. Kept here as the record of what changed.
- Project Midway — utility and number. The paper lists Midway as “≈225 MW baseline grid power · SRP.” In the record it is in Electrical District No. 3 (ED3) territory, not SRP, and the applicant had no power contract; the “225” is 225 permanent jobs, not megawatts — there is no published MW figure. Pinal Post
- Marigold — which body votes. The ~600 / 400 / 675 MW figures are right, but the “Board vote targeted Summer 2026” is SRP’s own board approval, not a Pinal County Board of Supervisors vote. Pinal Post
- Rodgers v. Huckelberry — described backwards. The paper cites it as having “voided a county building a facility for one tenant and leasing it back below market.” The trial court did void it, but the Court of Appeals reversed and ruled for Pima County — the below-market World View lease was upheld, and the court expressly did not reach the Gift Clause. It supports economic-development leases; it doesn’t condemn them. FindLaw — Rodgers v. Huckelberry
- Microsoft comparison — cooling term. The “hundreds of thousands of gallons per day” water draw is real (WI records show ~234k–702k gal/day), but Microsoft’s Mount Pleasant/Racine campus is marketed as closed-loop “zero-water-evaporation” liquid cooling, not “air-cooled.” The water figure holds; the label should say closed-loop. WPR
What's cited, and what's ours. The framework itself — the two pillars, the three-layer exit, the load-not-labels rule, and every policy recommendation — is the author's original design and stands on its own reasoning. Everything factual underneath it — the water science, the cooling engineering, the case law, the local project numbers — is sourced here. Conclusions are ours; facts get credit.
✓ Verified confirmed against a primary / authoritative source
⚠ Changed real when written, changed since
✎ Corrected was off — now fixed (see the box above)
◐ Caveat real mechanism, modest magnitude
The projects
La Osa is 3,385 acres, carries 33 PAD stipulations; P&Z recommended the rezone 7–2 on April 16, 2026; CPA approved Nov 19, 2025.
✓ VerifiedPinal Post and AZBEX report the 3,385-acre figure, the 33-stipulation PAD overlay, the April 16 2026 P&Z recommendation, and the Nov 19 2025 Comprehensive Plan approval.
La Osa: up to 3 GW of compute and a reported $33 billion — as proposed.
⚠ ChangedData Center Dynamics and AZ Big Media confirm $33B / 3 GW at proposal; both the GW and the building count were cut ~80% in late May 2026 (see the orange box). 3 GW was the original ceiling.
La Osa grid + water interface: WAPA 115kV + APS 230kV, Electrical District No. 4, Global Water.
✓ VerifiedPer the application: WAPA owns the 115kV lines crossing the property, APS the 230kV line on the southern parcels; ED4 is the delivering district; Global Water is the named service provider (early phases on wells until demand justifies a plant).
La Osa: up to 2,000 MW of on-site natural gas.
⚠ ChangedThe 2,000 MW figure appears as the assumption in the project’s water-memo addendum; the developer’s release says “up to 3 GW,” and at the April 2026 hearing the applicant’s attorney called the number “not set.” Real cited figure, not a fixed capacity — and now subject to the 80% downsizing.
Griffin Energy (PZ-PA-014-25): 2,685 acres · 550 MW solar / 550 MW battery / 500 MW gas · CPA approved Nov 19, 2025 · feeds the W Holdings campus.
✓ VerifiedAZBEX and the Casa Grande Dispatch confirm Griffin’s (Copia Power) 2,685-acre re-designation, the 550/550/500 MW figures, the Nov 19 2025 Board approval, and that it is intended to serve the W Holdings data-center campus.
W Holdings “Energy Generation & Technology Campus” (PZ-PA-015-25): ~495 of 2,495 acres · replaces ≈10,600 planned homes · ED3 territory.
✓ VerifiedAZBEX confirms the 2,495-acre campus, the “+/-10,600 planned single- and multi-family units” it replaces, and the SR-347 / W Louis Johnson Dr location (inside ED3’s service territory). Sources round the redesignated parcel to “~500 acres”; the exact 495 figure isn’t separately published.
Project Midway (PZ-PA-013-25): 215 acres · ~225 MW · SRP.
✎ CorrectionAcreage and the Nov 19 2025 5–0 CPA approval are correct. But the utility is ED3, not SRP, and “225” is 225 jobs, not megawatts — no MW baseline was published, and the applicant had no power contract. See the corrections box.
SRP Marigold Energy Center: 600 MW solar / 400 MW BESS / up to 675 MW gas; Board vote targeted Summer 2026.
✎ CorrectionThe MW figures are confirmed (≈1,275 MW combined, near Stanfield). But the Summer-2026 vote is SRP’s own board, not the Pinal County Board of Supervisors — the paper attributes it to the wrong body.
Project Bella: 352 acres · 400–440 MW BESS / 480–520 MW gas · comprehensive plan approved & rezoned.
✓ VerifiedThe Board voted 5–0 on Aug 6, 2025 to rezone 351.8 acres for Project Bella (Seguro Energy Partners) — 480–520 MW gas (10 GE LM6000 turbines) plus 400–440 MW BESS, ~1.5 mi south of Casa Grande.
Frontier East: 160 acres · 50 MW solar / 300 MW BESS · comprehensive plan approved, near the City of Maricopa.
✓ VerifiedFrontier East (Sawtooth DevCo, case PZ-PA-011-24): ±50 MW solar / up to ±300 MW BESS on ±159.68 acres near Maricopa, redesignated to Green Energy Production and reaching land-use approval.
Water — the closed basin
ADWR projected unmet groundwater demand of more than 8 million acre-feet, and stopped approving new groundwater-based Assured Water Supply applications. Pinal AMA, 1980 Groundwater Code.
✓ VerifiedStraight from ADWR: on June 28, 2021 its Pinal AMA model showed unmet demand exceeding 8 million acre-feet over 2016–2115, and the Department stopped approving new groundwater-based AWS applications in the Pinal model domain. The Pinal AMA is an Active Management Area under the 1980 Groundwater Management Act.
A same-design Microsoft air-cooled campus draws into the hundreds of thousands of gallons per day.
✎ CorrectionThe water figure is real — Wisconsin records show ~234,000 gal/day (Phase 1) up to ~702,000 gal/day (full campus) at Microsoft’s Mount Pleasant/Racine site, Microsoft citing ~350,000 gal/day on hot days. But the campus is marketed as closed-loop “zero-water-evaporation” liquid cooling, not “air-cooled.” Keep the number; fix the cooling term.
The grid
ED3 peaked near 280 MW in August 2025 — so a 3 GW campus is more than 10× that district's peak.
✓ VerifiedInMaricopa reports ED3’s 280 MW system peak as of Aug 7, 2025, across ~36,000 meters. 3 GW = 3,000 MW ≈ 10.7× that peak — the arithmetic holds.
ED4 generates no power and must buy wholesale and resell; and there are 2-to-7-year waits for transformers and turbines.
✓ VerifiedElectrical District No. 4 (Eloy) doesn’t generate — it buys wholesale (Parker-Davis/WAPA plus supplemental) and resells to ~2,850 users. Gas-turbine lead times run ~1–7 years (OEMs quoting 5–7; GE Vernova backlog into 2029) and large transformers ~2.5–3+ years — consistent with “2-to-7-year.”
The utility bill
Global Water's rate settlement: ~$2.68/month for the median user, new rates requested effective November 1, 2026.
✓ VerifiedGlobal Water Resources’ own April 29, 2026 filing with the Arizona Corporation Commission: a ~$2.68 estimated monthly increase for the median GW-Santa Cruz user (5,500 gal/mo), new rates requested effective Nov 1, 2026.
The science & engineering
Chronic groundwater overdraft causes land subsidence and earth fissures — documented specifically in south-central Arizona / the Pinal AMA.
✓ VerifiedThe USGS documents that after ~109 million acre-feet of groundwater was withdrawn (1915–1975) across Pinal and Maricopa Counties, “the land surface has subsided … and earth fissures have developed.” It’s a measured Arizona phenomenon, not a projection.
Evaporative (“wet”) cooling consumes water by evaporating it; closed-loop / air-cooled designs consume little-to-no water on-site.
✓ VerifiedThe LBNL / U.S. DOE 2024 Data Center Energy Usage Report: “the best facilities today use air-cooled or closed-loop liquid cooling systems that need almost no water at all,” while evaporative towers lose water to evaporation. DOE FEMP confirms the mechanism directly. (Honest caveat both raise: closed-loop can use ~10% more electricity, shifting some water use off-site to power plants.)
The data-center industry is actively shifting away from evaporative cooling toward closed-loop / air cooling.
✓ VerifiedMicrosoft states all new datacenter designs from Aug 2024 use closed-loop cooling that “consume zero water for cooling,” avoiding >125M liters/yr per datacenter; the independent LBNL report corroborates the broader shift. (The specific liter figure is Microsoft’s own modeled estimate; the trend is independently supported.)
Reclaimed / treated-effluent water can serve data-center cooling, reducing potable and groundwater draw.
✓ VerifiedEPA’s Quincy, Washington case study documents RO-treated industrial wastewater “returned to the data center for use as cooling water,” offsetting ~138 million gallons/year of potable groundwater demand.
Water offsets — fallowing farmland, retiring grandfathered groundwater rights, and managed aquifer recharge — are recognized mechanisms to put water back / free it up.
✓ VerifiedUnder Arizona’s Groundwater Management Code, permanently retiring a grandfathered irrigation right generates extinguishment credits; fallowing irrigated acreage frees water for other use; and managed aquifer recharge is an established replenishment tool. Retiring the right is codified at A.R.S. § 45-463.
Solar arrays double as a stormwater-catchment surface — runoff shed off the panels channels to on-site infiltration and reuse.
✓ VerifiedDOE/NREL’s PV-SMaRT work and a Penn State field study (Journal of Hydrology) confirm rain sheds off panels and concentrates at the driplines, and that runoff is directed to infiltration basins/trenches — dripline soil moisture ran ~19% higher than adjacent land. As the paper itself says, the harvested volume is modest (stewardship, not the water solution).
Panel shade cuts soil evaporation and retains moisture (an agrivoltaics finding).
✓ VerifiedA peer-reviewed agrivoltaics field study (Barron-Gafford et al., PLOS ONE) found soil under PV panels stayed significantly moister through the season because shade cuts evaporative loss — water-use efficiency up to ~328% higher in shaded rainfed plots.
Solar arrays cut the heat-island effect.
◐ CaveatThe shade mechanism is real — EPA notes shaded surfaces run 20–45°F cooler and shading is core to heat-island mitigation. But for solar PV specifically, the measured air-temperature effect is small (~0.2–0.3 K in one peer-reviewed urban simulation) and context-dependent — panels can even warm surroundings in some conditions. Real, but a modest co-benefit, not a strong cooling claim. Worth softening in the paper.
The legal ground
The Gift Clause (Ariz. Const. art. 9 §7) and its two-part test — Wistuber (1984), Turken (2010), and Schires v. Carlat (2021), which held jobs / growth / future tax revenue don't count as consideration.
✓ VerifiedArt. 9 §7 is the Gift Clause (azleg.gov). Wistuber v. Paradise Valley USD, 141 Ariz. 346 (1984) set the two-part test; Turken v. Gordon, 223 Ariz. 342 (2010) (CityNorth) refined the consideration prong to fair market value of what was promised; Schires v. Carlat, 250 Ariz. 371 (2021) held anticipated jobs, economic development, and future tax revenue are not consideration.
Gilmore v. Gallego (2024) as the current Gift Clause precedent; A.R.S. § 11-1101 (development agreements) and § 33-261 (perpetuities).
✓ VerifiedGilmore v. Gallego (2024) is on the Arizona Supreme Court’s own site. A.R.S. § 11-1101 is titled “Development agreements” and authorizes county development agreements; A.R.S. § 33-261 is the “Rule against perpetuities.”
Rodgers v. Huckelberry voided a county building a facility for one tenant and leasing it back below market.
✎ CorrectionReal case (Pima County / World View, 2 CA-CV 2017-0091). But the holding is the opposite: the Court of Appeals reversed and ruled for the county — the below-market lease was upheld and the Gift Clause was not reached. As cited, it argues against the paper’s point rather than for it. See the corrections box.