A Utah data center battle has become a test of how far AI builders will go to frame local opposition as a national security threat.
Kevin O'Leary wants America to see his proposed Stratos data center in Utah as a necessary weapon in the AI race with China. Many residents in Box Elder County see something much simpler: a 40,000-acre project that could change their land, power supply and local politics for decades.
That gap is now the story. O'Leary, the Shark Tank investor backing the project through O'Leary Digital, has claimed that opposition to the development is being helped by Chinese propaganda and foreign-linked influence campaigns. The Trump administration and some industry voices have echoed the broader idea that resistance to data centers is not always organic. According to a report from The Washington Post, those claims are being made with scant public evidence, even as local groups say their concerns are rooted in water, electricity, noise, land use and whether their communities will actually benefit.
The Stratos proposal is not a small industrial park. Reports and local project materials describe it as a massive AI data center campus planned for roughly 40,000 acres in northwest Utah, with a projected power need of up to 9 gigawatts at full buildout. That scale makes the project a symbol of the new AI economy. It also makes it a target for people who believe the costs are being pushed onto towns that did not ask to become infrastructure hubs for the next phase of computing.
The timing matters. Utah Gov. Spencer Cox issued an executive order on May 29 directing state agencies to tighten their review of large data center projects, a clear sign that the Stratos fight has moved beyond one county meeting. When a governor steps in, the dispute is no longer just about zoning. It becomes a question of how fast a state should move when artificial intelligence collides with energy, water and public trust.
Data centers used to be boring real estate. They sat outside the public imagination, serving cloud computing, enterprise software and everyday internet traffic. AI changed that. Training and running large models requires enormous clusters of chips, electricity and cooling capacity, which means the industry now has to negotiate directly with power developers, county commissions and residents who can see the physical footprint arriving at their doorstep.
That is where the China argument becomes useful for developers. If a data center is framed as a frontline asset in a technology race, local objections can start to look less like democratic pushback and more like obstruction. It is a powerful message, especially in a political environment where AI supremacy is increasingly treated as a national project. But it can also backfire. People who are worried about their water supply or electricity rates do not usually respond well to being told they are part of a foreign plot.
O'Leary has argued that America needs to move quickly to build the compute capacity that AI companies and national defense will require. That point is not frivolous. The United States is trying to keep its lead in advanced chips, cloud infrastructure and frontier AI systems while China builds its own capacity. The companies that control compute will have real leverage in the economy, and states that cannot provide enough power and data center capacity may lose investment.
Still, speed is not the same thing as consent. The most durable infrastructure projects usually earn public trust before they ask for public sacrifice. When a project is this large, the questions become practical very quickly. Who pays for transmission upgrades if any are needed around the site? Who owns the power generation? How much water will cooling require? How many permanent jobs arrive after construction crews leave? What happens to nearby communities if demand forecasts change?
AI Infrastructure Needs A Better Pitch
The industry has a communication problem because its benefits are often national while its burdens are local. A hyperscale data center may support AI models used by millions of people, but the noise, land use and power demand sit in one county. That imbalance is why protests are spreading beyond Utah. Similar fights have appeared around projects tied to Google, Microsoft, Meta and other large technology companies as communities weigh tax revenue against grid strain and environmental pressure.
For entrepreneurs and investors, the lesson is uncomfortable but clear. You cannot build the physical internet of AI with a software startup playbook. Moving fast through local government may get a project approved, but it can create a legitimacy problem that follows the project for years. Residents do not care that AI is strategic if they believe the deal was shaped without them.
The foreign influence question should not be dismissed entirely. China, Russia and other governments do try to exploit American divisions online, and data center opposition would be an obvious narrative to amplify. But amplification is different from origin. A campaign can be useful to Beijing and still be rooted in real local grievances. Treating every skeptical town hall as a proxy battle risks weakening the very public support the AI industry needs.
That matters because data centers are becoming political infrastructure. They are tied to energy policy, industrial policy, defense policy and local tax bases. The winners in this market will not only be the companies with the best chips or the biggest financing partners. They will be the ones that can explain, in plain terms, why a community should accept the tradeoffs and what protections are in place if the promises fall short.
O'Leary's Utah fight is therefore bigger than one celebrity investor and one giant campus. It is an early warning for the AI buildout ahead. If developers lead with suspicion instead of transparency, they may find that the biggest bottleneck in artificial intelligence is not chips or power. It is trust.
Also read: Utah is raising the price of AI data center ambition • Blue Origin faces a long pause after New Glenn destroys its launch pad • GitHub is making Copilot costs harder for founders to ignore
