After a 40,000-acre data center proposal in Box Elder County drew the biggest civic backlash Utah has seen in years, Governor Spencer Cox took the path that wasn't on offer in most state capitols this spring: he didn't impose a moratorium, and he didn't wave the project through. On May 29, 2026, Cox signed Executive Order 2026-03, directing five state divisions — Air Quality, Drinking Water, Water Quality, Water Rights, and Wildlife Resources — to apply 'higher standards' to large data center projects, with explicit protections for the Great Salt Lake, utility ratepayers, and air quality, plus a transparent public comment requirement. The order is effective immediately.
The case for stricter review
Start with the strongest version of the regulators' argument, because the numbers make it. The Stratos project, the catalyst for the order, would span 40,000 acres in Hansel Valley — larger than Bryce Canyon National Park — and could draw 7.5 to 9 gigawatts at full build, more than double Utah's current statewide electricity consumption of roughly 4 GW. It would consume around 13,000 acre-feet of water a year, equivalent to the basic needs of about 20,000 households, in a watershed that drains toward a Great Salt Lake already at historic lows. At a May 4 county commission hearing, roughly 1,100 residents — a remarkable turnout for an unincorporated jurisdiction — voiced opposition before commissioners approved the project unanimously. A regulator who looked at those numbers and shrugged would not be doing the job ratepayers and downstream water users pay them to do.
The legitimate concern is not data centers as such. It is that hyperscale build-outs impose costs — on aquifers, on the grid, on neighbors — that the developer may not face directly unless the state structures the deal so they do.
What the order actually does (and doesn't)
The Cox order does not introduce a permitting veto, a tonnage cap, or a moratorium. It instructs agencies that already hold relevant authorities to apply them with more rigor: water rights review, drinking-water separation from cooling water, discharge permits, air quality modeling, and wildlife mitigation. As Cox put it, 'Utahns have expressed legitimate concerns regarding the potential impacts of large data centers on water resources, air quality, utility rates, local communities and quality of life. And those concerns must be carefully considered.' Industry, the state motto and a signal of intent, stays — paired now with explicit environmental stewardship.
That is what proportionate regulation looks like. The state is not telling Kevin O'Leary's consortium it cannot build; it is telling the developers they must build under rules that price in the externalities. Crucially, the order also tells utility regulators that ratepayers should not subsidize the dedicated transmission, generation, and water infrastructure a multi-gigawatt campus requires. This is not new ground.
The federal–state ratchet
On March 4, 2026, the White House announced a voluntary Ratepayer Protection Pledge under which Amazon, Google, Meta, Microsoft, OpenAI, Oracle, and xAI committed to 'build, bring, or buy' the generation they need and absorb the cost of new transmission upgrades, with the explicit aim that data centers' energy needs 'will not increase household electricity costs for American citizens.' The pledge is sensible, but it is unenforceable: no signature, no consequence.
States have begun filling the gap with binding rules. Ohio's Public Utilities Commission approved a tariff for AEP Ohio that places data centers in a distinct rate class with a take-or-pay minimum equivalent to 85% of their declared load, long-term contract terms, collateral requirements, and exit fees equal to three years of minimum bills. California's AB 2383 authorizes special customer classifications for very large loads, but without the take-or-pay teeth. Utah's executive order is structurally similar to Ohio's logic — make the developer pay for what the developer needs — but with broader scope. It addresses water and air, not just kilowatt-hours.
The pro-innovation case for this kind of order
A counterintuitive truth: clear ex ante standards are friendlier to legitimate capital than the absence of standards. The investor case against Stratos was never 'Utah is hostile.' It was 'Utah's rules for a project of this scale were undefined, so every parcel-level fight became a referendum.' That is bad for the company, bad for the county commissioners absorbing the political risk, and bad for the next, smaller AI workload that would happily site in Utah if the cost menu were knowable. Cox's order moves toward a menu. A hyperscaler that can absorb its own water and power costs — which the seven federal-pledge signatories said they can — has nothing structural to fear.
The risk worth flagging
The honest concern with EO 2026-03 is vagueness. 'Higher standards' is a direction, not a number, and agencies could over-rotate into open-ended review that effectively becomes a soft moratorium. The cure is implementation discipline: published thresholds (a megawatt floor that triggers heightened review, water-use intensity benchmarks), defined timelines, and a single coordinating office so applicants are not chasing five divisions independently. Utah should also publish a model term sheet for ratepayer protection, as Ohio effectively did, so projects can underwrite to it.
Done with that discipline, this is a model other states should copy. Cox has shown it is possible to take an 1,100-person public meeting seriously without conceding that AI infrastructure has no place in the American interior. Innovation does not require subsidized water or socialized grid costs. It mostly requires rules that hold.