Kevin O'Leary Utah Data Center Collides With Local Democracy

Kevin O’Leary’s Utah data center was pitched at a staggering scale, then local opposition stopped its rapid march toward construction. The proposed campus covered roughly 40,000 acres and envisioned up to nine gigawatts of on-site power. That capacity would exceed twice Utah’s average electricity use.

The Stratos Project, which O’Leary also promoted as Wonder Valley, was supposed to represent American ambition in the global AI race. Instead, it became a warning about what happens when infrastructure planning moves faster than public consent. County officials entered the process late, major engineering questions remained open, and residents had little time to understand the proposal.

A months-long investigation by The Verge’s Josh Dzieza reconstructs how the project advanced through Utah’s political system. His reporting shows that Stratos did not stumble because demand for AI computing disappeared. It stalled because speed, secrecy, and speculative development created a political backlash that money could not quickly resolve.

The Kevin O'Leary Utah Data Center Was Bigger Than Its Tenant List

Stratos combined an enormous physical promise with unresolved questions about who would use the campus and how its systems would work.

The project took shape in Hansel Valley, a rural part of Box Elder County near the northern arm of the Great Salt Lake. Local real estate developer Austin Pritchett began approaching landowners in late 2025. Former congressman and television commentator Jason Chaffetz then introduced him to O’Leary.

The location offered one decisive asset. A 42-inch interstate natural gas pipeline crossed the area, creating a potential route to large-scale electricity generation. That access made the site attractive to developers pursuing “powered land,” meaning property packaged with potential energy access and development approvals.

Powered land has become valuable during the AI infrastructure boom. Developers can assemble land, utility plans, permits, and tax arrangements before signing a major technology customer. The resulting package can then be marketed to cloud companies or AI laboratories that need capacity quickly.

That model also shifts considerable risk into the planning stage. A project can look advanced because it controls land and has government support, even when it lacks a disclosed anchor tenant. Stratos illustrates that gap.

O’Leary’s team met Utah Gov. Spencer Cox and state Senate President J. Stuart Adams in January 2026. Adams also chaired the Military Installation Development Authority, known as MIDA. The state-created authority can manage land development connected to military objectives.

On February 17, the developers announced Wonder Valley as a proposed 7.5-gigawatt campus. Later descriptions placed its potential generation between 7.5 and nine gigawatts. O’Leary repeatedly called it the world’s largest data center project.

The scale deserves context. A gigawatt equals one billion watts of electrical capacity. Nine gigawatts would represent power on the scale of several large generating stations operating together. The proposed campus would produce that energy primarily on-site, rather than depending entirely on Utah’s existing electrical grid.

Yet the data center plan remained preliminary. No major technology tenant had been publicly identified. Officials had not disclosed a specific military customer, contract, or mission. The final generation technology, construction schedule, and total water requirements also remained unsettled.

The distinction between a project area and a completed facility became politically important. Utah’s own Stratos materials said much of the 40,000-acre area would remain open space. The actual buildings and power systems would occupy only a portion of the land.

That qualification did not erase the project’s scale. The approved area established where extensive data centers, generation facilities, manufacturing, and commercial development might eventually occur. It also transferred significant land-use authority to MIDA.

The project’s supporters presented this flexibility as responsible phased development. Critics saw a vast entitlement package with too few technical commitments. Both interpretations arose from the same fact: Stratos was an ambitious framework, not a construction-ready data center.

The Wonder Valley data center therefore entered public debate with its strongest claim and its greatest weakness intertwined. Its size attracted national attention, but the absence of settled operational details made that size difficult to defend locally.

A Forty-Day Approval Rush Created the Backlash

The decisive problem was not simply how large Stratos was, but how little time county officials and residents received to evaluate it.

State-level conversations were already underway before Box Elder County commissioners received a formal briefing. According to the county’s official project timeline, MIDA first presented the proposal to commissioners on March 25, 2026.

That date left only 40 days before the county’s final May 4 vote. The timetable covered a project area comparable to a small city, alongside a multi-gigawatt power complex.

County consent mattered because MIDA needed approval to include private land in its project area. Once that consent became effective, MIDA would exercise broad authority over development inside the area.

The project’s development agreement describes MIDA as holding municipal-style police powers within the project area. It also says county land-use ordinances would not govern that land in the usual way.

That arrangement placed extraordinary weight on the county’s initial decision. Commissioners were not merely approving another zoning application. They were consenting to a governance structure that would limit their control over later development.

Public notice did little to communicate the stakes. An April 22 meeting agenda referred to an interlocal agreement involving the “Stratos Project Area.” It did not clearly identify a giant AI data center campus.

Residents began learning the details as the approval process accelerated. More than 100 people attended an April 27 meeting, where commissioners delayed their decision. Nearly 3,000 responses later arrived through the county website.

The county moved the final meeting to the Box Elder County Fairgrounds. Approximately 1,500 people attended on May 4, according to the county. Commissioners ultimately approved resolutions supporting the project area and its interlocal agreement.

Project supporters argue that this vote did not approve immediate construction. Utah’s official position is that Stratos still requires detailed planning, regulatory permits, and environmental review. That distinction is legally meaningful.

Politically, however, residents were reacting to more than a building permit. They understood that the vote would establish the project area and place MIDA in control of land-use decisions. Reversing that structure would be harder than rejecting an ordinary application.

Tax terms added another point of friction. The agreement offered extensive incentives intended to attract a project competing nationally for capital and tenants. The Verge reported that server equipment would effectively avoid property taxes, while much of the remaining tax revenue would support project infrastructure.

Supporters said the arrangement would make private development responsible for roads, water systems, and other facilities. They also predicted construction employment, permanent technical jobs, and new public revenue.

Opponents questioned whether those benefits justified the loss of local control. They also asked whether a speculative campus without a named anchor tenant deserved incentives designed around its maximum potential.

The conflict was not a simple fight between environmentalists and technology advocates. Conservative voters, farmers, ranchers, local officials, and environmental groups all challenged aspects of the process.

Their concerns differed, but the procedural complaint united them. People believed a decision with decades of consequences had arrived almost fully formed. The short review period made every unresolved technical question look more serious.

The Stratos Data Center Project Became a Governance Fight

Stratos pressured Utah leaders because the project converted an infrastructure debate into a test of political accountability.

MIDA was created to support military installations and related economic development. For Stratos, officials argued that advanced computing and energy infrastructure could serve national-security goals.

The authority’s involvement offered speed and statewide coordination. It also bypassed parts of the normal local land-use system. That tradeoff became the central political conflict.

Project documents described hyperscale data centers, energy generation, storage, and advanced manufacturing. Hyperscale facilities are very large computing sites built to support cloud platforms or major AI workloads.

The documents also referred broadly to military support. However, officials did not publicly identify a military tenant or contract. At an April meeting, representatives could not specify how much of the development would serve military operations.

That ambiguity mattered because MIDA’s authority gave the project advantages unavailable through ordinary county zoning. Residents wanted to know what military purpose justified those advantages.

The Kevin O’Leary Utah data center also arrived during a statewide push for more electricity. Cox had announced Operation Gigawatt in 2024, seeking to double Utah’s power generation. State lawmakers later enabled large energy users to build on-site generation.

Stratos appeared to fit that strategy. It promised private capital, dedicated generation, and infrastructure intended for AI workloads. Officials could present it as a way to capture economic activity without forcing the existing grid to supply every watt.

However, on-site generation does not eliminate local impact. A gas power complex still requires fuel infrastructure, emissions controls, water planning, and regulatory oversight. It can also affect regional air quality even when it operates outside the public grid.

The project’s projected load made those issues impossible to treat as ordinary permitting details. At full scale, Stratos would become one of Utah’s most consequential industrial developments.

Public opinion moved sharply against the original proposal. A May survey of 802 Utah voters found that 53 percent opposed the 40,000-acre project. Nearly seven in ten respondents said data-center economic benefits did not outweigh natural-resource costs.

The poll also showed bipartisan concern. Opposition reached almost 85 percent among Democrats, 62 percent among independents, and 36 percent among Republicans. Only 45 percent of surveyed Republicans supported it.

That reaction forced political leaders to adjust. Adams called for a smaller project area and new environmental commitments. O’Leary then agreed to remove about 20,050 acres from the 41,200-acre zoning area.

His commitments included preserving open space, reducing water consumption, directing excess water shares toward the Great Salt Lake, and publishing project updates. The proposed first phase was also limited to one gigawatt.

Those concessions did not restore the original political momentum. Two county commissioners who supported Stratos later lost Republican primaries. Adams also lost his primary contest.

The results demonstrated the project’s unusual coalition of critics. Utah’s data center backlash could not be dismissed as resistance from one party or one advocacy group.

The primary opponent in this story is therefore not O’Leary against a competing developer. It is an accelerated, state-led approval model against local democratic consent.

Stratos needed speed to compete for AI tenants. Residents needed time to assess a multi-decade transformation of their county. The project could not satisfy both demands once the process became public.

Water and Power Claims Never Reached Stable Ground

The environmental debate intensified because Stratos changed its plans while the public was still trying to identify the project’s real footprint.

Water became the most visible concern. Hansel Valley lies in a dry region near the Great Salt Lake, where declining water levels carry ecological and public-health consequences.

Data centers use water differently depending on their cooling systems. Power plants can also require water for cooling, emissions controls, and other operations. Without a final technical design, no single consumption estimate could settle the debate.

Applications associated with the site initially sought a combined 1,911 acre-feet of water. State officials later said developers expected to seek around 13,000 acre-feet. One acre-foot equals roughly the volume needed to cover one acre with one foot of water.

The initial water-rights request attracted nearly 4,000 formal protests. Developers withdrew it on May 6 and said they intended to refile after adjusting the request to match the project’s phases.

That withdrawal did not end the water dispute. It nullified the existing protests, each of which required a filing fee. Any revised application would start a new review process under amended Utah law.

House Bill 60 changed what the state engineer may consider when reviewing water applications. The law narrowed the role of broad public-welfare concerns, focusing the process more directly on water availability and impacts on existing users.

The timing attracted scrutiny because the law took effect just before the withdrawal. There is no conclusive public evidence that Stratos caused the legislation. Still, a new application would face a different standard from the one residents expected when filing their objections.

Air quality presented a separate regulatory challenge. A large natural gas generation complex would need permits governing pollutants and emissions-control systems. Diesel backup generators might require additional review.

The Utah Department of Environmental Quality stressed that the project had not cleared those processes. Its commissioner said public review had not yet begun and that a development of this size would not simply rush forward.

The state’s permitting requirements include air-quality, water-quality, and drinking-water oversight. Those reviews would examine actual engineering plans rather than the project’s promotional maximums.

Supporters said the campus could use less water than existing agriculture on the same land. They also proposed heat recovery, water conservation, and open-space protections.

Those ideas remain commitments, not independently verified operating results. Construction has not advanced far enough to test them. The final balance among gas generation, solar capacity, storage, and other technologies also remains unresolved.

The same uncertainty applies to electricity. On-site generation could shield household customers from some direct grid costs. It would not automatically prevent pipeline constraints, emissions, transmission interactions, or regional price effects.

A one-gigawatt first phase would be far smaller than the original nine-gigawatt vision. It would still represent an unusually large industrial power project.

This uncertainty explains why environmental questions became inseparable from governance. Residents were asked to accept the project area before developers could provide final designs. Developers argued that those details naturally belong to later permitting.

Both positions describe a real feature of large development. Early approvals are necessary before companies spend heavily on engineering. However, early decisions can also transfer leverage before communities understand the final impact.

For Stratos, the trust needed to manage that sequence had already eroded. Every revised estimate looked like evidence that the original announcement had outrun the underlying plan.

O'Leary's Attack on Critics Deepened the Debacle

The project’s public campaign shifted from defending Stratos to questioning its opponents, and that strategy created a second crisis.

O’Leary initially framed Wonder Valley as a national-security response to China’s AI ambitions. He argued that the United States needed enormous computing and energy capacity to remain competitive.

That framing connected a Utah land project to a genuine strategic concern. AI laboratories and cloud providers are seeking vast supplies of electricity, chips, land, and cooling capacity.

The message changed as local opposition grew. O’Leary began claiming that foreign-linked interests had helped organize resistance to the project. He repeated allegations involving China during television appearances and online posts.

Some industry-aligned organizations had produced reports about foreign funding connected to American advocacy groups. Those reports did not establish that China directed Utah residents or local organizations opposing Stratos.

A funding investigation found that the cited materials generally showed indirect philanthropic connections. They did not provide direct evidence of a foreign operation controlling local sentiment.

The distinction was crucial. Foreign influence campaigns exist, and divisive infrastructure debates can attract automated or coordinated content. That possibility does not prove that residents attending county meetings are acting for a foreign government.

O’Leary later stated that he lacked evidence connecting several named Utah critics to China. Fox hosts also aired corrections and apologies concerning the claims.

Alliance for a Better Utah, Elevate Strategies, and associated individuals filed a defamation lawsuit. O’Leary and Fox News sought dismissal, arguing that viewers would understand the statements as opinion or advocacy.

The legal dispute remained active in late September 2026. A judge had not resolved the core claims when The Verge published its investigation.

The lawsuit does not determine whether Stratos itself is technically viable. It does show how far the public battle moved away from permits, engineering, and economic development.

Calling critics foreign agents also weakened the project’s political case. The opposition included local landowners, Republican activists, environmental advocates, and residents concerned about county authority.

Dismissing that coalition made it harder to address its shared complaint. Many opponents believed leaders had hidden the project until approval became urgent.

The strategy also obscured legitimate arguments for data center construction. The United States does need more computing capacity if companies continue expanding advanced AI systems. New campuses will require faster planning and significant energy development.

Those needs do not excuse unsupported claims about residents. In fact, the national importance of AI infrastructure raises the standard for public credibility.

The Stratos data center project demonstrates that public communication is part of infrastructure execution. A developer can secure land options, political support, and favorable tax structures, yet still lose the consent needed for long-term construction.

O’Leary’s celebrity initially gave Wonder Valley national visibility. His confrontational response later amplified opposition, produced legal exposure, and kept attention on the project’s weakest procedural choices.

What Happens Next Will Test the Entire AI Building Boom

Stratos remains a proposal with legal and regulatory paths ahead, but its original vision has lost the political conditions that made rapid development possible.

The first signal to watch is the referendum litigation. Residents attempted to challenge Box Elder County’s resolutions through a public vote. County officials rejected those applications after classifying the resolutions as administrative actions rather than referendum-eligible legislation.

Opponents sued over that decision. A ruling favoring them would create a direct route for voters to revisit county consent. A ruling against them would preserve the existing approvals but would not restore political support.

The second signal is a revised water application. Its requested volume, phasing, and technical assumptions will reveal whether the smaller project differs materially from the original plan.

A narrowly defined first phase would strengthen the developer’s claim that the campus can proceed responsibly. Another large or poorly documented request would reinforce concerns that public concessions changed the presentation more than the infrastructure.

The third signal is an anchor tenant or binding development commitment. Stratos was assembled through the powered-land model, with approvals and energy access preceding a disclosed customer.

A credible tenant would show that the project answers specific computing demand. Continued silence would support the view that Wonder Valley remains a speculative entitlement package seeking a buyer.

Permits will matter just as much. Air-quality applications should identify generation technologies and emissions. Water filings should clarify consumption. Site plans should show which acreage will be developed and which will remain open.

Those records will let residents compare the revised project with the original nine-gigawatt pitch. Until then, claims from both supporters and critics remain partly dependent on assumptions.

The Kevin O’Leary Utah data center also offers a wider warning for the AI sector. The industry treats speed as a competitive requirement because model development consumes rapidly increasing amounts of computing capacity.

Land, grid connections, pipelines, turbines, transformers, and permits operate on much slower schedules. Local politics can move quickly, but not always in the direction developers expect.

Other data center proposals face similar resistance over electricity bills, water consumption, tax incentives, pollution, and land use. Stratos stands out because all those tensions appeared inside one unusually large project.

The lesson is not that communities will reject every AI campus. It is that scale changes the political contract. A facility drawing city-sized resources cannot be presented as a routine commercial development.

Developers need to identify who bears the costs, who receives the benefits, and which commitments are enforceable. Governments must explain incentives and authority before asking for irreversible consent.

Communities also need evidence that later regulatory reviews can meaningfully alter a project. Public hearings lose credibility when residents believe the important decisions have already been made.

Stratos once promised to show how quickly the United States could assemble enormous AI infrastructure. It now shows how speed can destroy the trust required to finish that infrastructure.

The project is not a clean story of technology losing to obstruction. Nor is it proof that every giant AI campus is destined to fail. It is a case study in sequencing.

O’Leary and his partners secured attention, state support, and a development framework before building local legitimacy. When residents finally entered the process, they encountered a project whose scale seemed settled while its impacts remained uncertain.

That reversal should concern cloud providers, AI companies, public officials, and communities evaluating the next campus. Computing capacity cannot operate without physical and political foundations.

The next question is straightforward: will Stratos return with a named customer, verifiable engineering, and a process Utah residents trust? Until those pieces appear, Wonder Valley remains less a blueprint for the AI boom than a warning about how that boom can stall.