Solve Water Before You Break Ground, or Else

Water is emerging as the hard constraint that power never was for datacenter developers, and the evidence landed from multiple directions today. A new academic study from UC Riverside, Caltech, and RIT documents that local water supply is a non-negotiable bottleneck for AI infrastructure siting, with OpenAI's scrapped Stargate expansion in Abilene, Texas cited as Exhibit A. Simultaneously, a Tower Tech executive argues the industry's pivot to air-based cooling simply shifts water consumption upstream to power generators, doubling or tripling electricity demand in the process. For developers treating water as a secondary checkbox, today's research says the market is about to correct that assumption violently.

Google and TotalEnergies signed two 15-year power purchase agreements for 1 GW of solar capacity from two Texas projects: the Wichita solar farm (805 MWp) and Mustang Creek facility (195 MWp), expected to deliver roughly 28 TWh over the contract period. This is TotalEnergies' largest US renewable deal to date.

Google is reportedly investing $40 billion in Texas datacenter expansion through 2027, and this 1 GW covers only a fraction of that demand. Additional gas, nuclear, or storage contracts are inevitable. Early reports suggest these projects still face ERCOT's severe interconnection backlog; any queue delays push Google toward supplemental gas procurement, undermining the renewable supply strategy.

Bottom line: watch permitting and interconnection timelines on Wichita and Mustang Creek as leading indicators of whether Texas solar can actually keep pace with hyperscale demand.

Tower Tech President Mathu Solo makes the case bluntly: air-cooled systems demand two to three times the kilowatt-hours of water-based alternatives. That extra electricity still gets cooled with water at the power plant. "It's robbing Peter to pay Paul." The UC Riverside study reinforces the point: unlike electricity, water can't be transferred across long distances, making local supply a site-selection gate that no amount of capital can easily overcome.

Ohio's Cedar Bog fight illustrates what happens when developers don't clear that gate early. A $1 billion, ~460,000-square-foot hyperscale facility near Urbana faces near-universal opposition because it sits directly atop the aquifer feeding one of Ohio's most ecologically sensitive areas. An estimated 300-400 residents packed a public meeting; zero spoke in favor.

Bottom line: developers who fail to resolve water sourcing before site acquisition are handing communities a rallying point that's harder to defuse than any power constraint.

Datacenter backlash has moved beyond local hearings into campaign ads. AdImpact monitoring has detected ad spending targeting datacenter supporters in a Texas congressional primary, a North Carolina county commissioner race, and a Georgia gubernatorial primary. Narrative Strategies polling shows 36% of Americans say data centers do more harm than good, with a plurality believing their families stand to lose.

Mayors from Phoenix to Chattanooga are publicly challenging tech companies over grid strain, water depletion, and NDA-shielded secrecy. Phoenix Mayor Kate Gallego revealed that if every requesting facility gained approval in APS's service area, demand would hit 19,000 MW, more than double the grid's record peak.

For Texas developers: the political cost of poor community engagement is rising fast. Developers who articulate clear benefit-sharing models and participate openly in local processes will separate themselves from the pack. Those who don't are giving candidates ammunition.

Bottom line: the question isn't whether datacenter politics will intensify before November; it's which developers will have already locked in community support before the wave hits.

Forney, Texas is voting March 17 on an ordinance requiring Conditional Use Permits for all datacenters, with noise enclosure mandates and 100-foot residential setbacks. City staff explicitly framed the move as defensive: without standards, developers could "attempt to sidestep restrictions."

This mirrors a statewide pattern of North Texas municipalities racing to adopt controls before proposals arrive. The CUP requirement gives Forney leverage to impose conditions project-by-project. What it doesn't address is whether city staff have the technical capacity to evaluate power and water impacts during review. Smart developers should view Forney's framework as an engagement opportunity, not an obstacle. Show up early, bring data, and shape the CUP conditions before they're set.

Bottom line: municipalities that write their own rules before developers arrive will be harder to negotiate with later; early engagement is the play.

Across Colleton, Marion, and Spartanburg counties in South Carolina, residents are discovering the same pattern: code-named LLCs, NDAs silencing elected officials, and critical power and water details withheld until after zoning votes. Eagle Rock Partners' "Project Boulder" in Colleton County proposes a 1,000-MW facility on 859 acres, and state officials can't even account for how many data centers currently operate in the state.

Virginia's Dominion Energy is building a 944-MW gas plant in Chesterfield to serve 600+ data centers, exploiting a reliability carve-out in state law that environmental groups are challenging in court. The pattern is the same everywhere: demand arrives faster than regulatory frameworks can track it.

Developers who rely on opacity are borrowing against future goodwill. The South Carolina lawsuits and Virginia litigation show the bill comes due. Could it happen in Texas? If developers don't engage transparently with communities and regulators, the answer is yes.

Bottom line: secrecy-dependent siting strategies are a depreciating asset; every NDA that leaks makes the next project harder to permit.