AI data centre projects are often discussed in terms of possibilities. Faster computing. Smarter systems. Entire industries transformed by machines capable of processing information at unprecedented speeds.
But every technological leap carries a physical footprint, and sometimes that footprint becomes impossible to ignore.
That is the question now surrounding Project Stratos, a massive AI-focused data centre development linked to Shark Tank investor Kevin O’Leary and his infrastructure company, O’Leary Digital.
Planned for Utah’s Hansel Valley, the project has become the center of a growing debate not about artificial intelligence itself, but about what it takes to power it.
At the heart of that debate is a striking claim from a Utah physicist: the completed facility could release heat equivalent to the energy of 23 atomic bombs into the local environment every day.
There are no nuclear weapons involved. No radiation. No catastrophic explosion.
Yet the comparison has captured attention because it forces a broader question into view: how much energy can a landscape absorb before it begins to change?
The Scale of Kevin O’Leary’s AI Data Centre
For years, data centers existed largely in the background of the digital economy. Most people never thought about where their cloud storage lived or where AI models processed their requests. That has changed.
The rise of generative AI has dramatically increased demand for computing infrastructure. Behind every AI-generated answer, image, or prediction sits an enormous network of servers requiring electricity, cooling systems, and constant operation.
Project Stratos represents one of the most ambitious examples of that trend.
The proposed development in Box Elder County is reportedly part of a larger Wonder Valley initiative, envisioning dozens of data centre buildings spread across a vast desert landscape. The completed AI data centre could consume as much as nine gigawatts of electricity.
Numbers like that are difficult to visualize. Nine gigawatts is not merely a large amount of power; it is the kind of energy consumption typically associated with major metropolitan regions. And according to critics, the electricity consumed is only part of the story.
The larger concern is what happens afterward.
How the AI Data Centre Generates Massive Heat
Physics has a simple answer. Energy does not disappear.
When data centres run, much of the electricity they consume eventually becomes waste heat. Servers process information, cooling systems work continuously, and the resulting thermal energy must be released somewhere.
That reality prompted Utah State University physicist Dr. Rob Davies to conduct an analysis of the project. His findings have become the focal point of the controversy.
Davies estimates that, beyond the nine gigawatts needed to power operations, an additional seven to eight gigawatts could ultimately be released as waste heat. Combined, that creates what he describes as a thermal load approaching 16 gigawatts.
The comparison that followed was impossible to ignore.
According to Davies, the total energy discharged into the surrounding environment each day would be roughly equivalent to “23 atomic bombs worth of energy” being deposited into the area.
The phrase is dramatic, but its purpose is less about fear and more about scale. It is an attempt to communicate the magnitude of continuous heat generation from a facility unlike anything the region has previously experienced.
Why the AI Data Centre Location Matters
The debate becomes more complex when geography enters the conversation. Hansel Valley is not an empty, featureless space.
Scientists describe the area as having a basin-like topography capable of trapping air and environmental effects within the region. It also sits near the Great Salt Lake, an ecosystem already facing significant environmental challenges due to declining water levels and changing climate conditions.
For critics, that context matters.
Davies has questioned what happens when a sustained flow of thermal energy is introduced into a high-desert valley that already experiences environmental stress. The concern is not simply that the facility generates heat.
It is whether years of continuous heat release could alter local weather patterns, affect air circulation, or place additional strain on surrounding ecosystems. Those fears are shared by other researchers.
Ecology professor Ben Abbott of Brigham Young University reviewed Davies’ analysis and offered one of the starkest assessments of the potential consequences.
According to Abbott, the scale of thermal output being discussed could represent the difference between Utah’s semi-arid environment and conditions more closely resembling the Sahara Desert.
Whether such outcomes would ultimately occur remains subject to scientific debate. But the fact that respected researchers are asking the question has elevated the conversation beyond a typical infrastructure dispute.
Developers Defend the AI Data Centre Project
The people behind Project Stratos argue that many concerns overlook the project’s actual design. Paul Palandjian, CEO of O’Leary Digital, has emphasized that the development would utilize closed-loop cooling systems intended to significantly reduce water consumption.
Water use has become another major point of concern in Utah, where drought and resource management remain ongoing issues.
Palandjian has argued that the facility’s potable water requirements would be substantially lower than some public estimates and, in certain cases, less than the agricultural activity already taking place on the same land.
Developers have also noted that the project would not appear overnight. Construction is expected to occur in phases over more than a decade, potentially allowing environmental assessments and adjustments as the development expands.
Their position reflects a broader argument increasingly heard across the technology industry: advanced computing infrastructure is necessary to support economic growth, innovation, and AI development.
The challenge is determining how that infrastructure can coexist with environmental priorities.
A Debate Bigger Than One Data Centre
What makes the Stratos controversy notable is that it extends beyond Kevin O’Leary or a single project. Around the world, governments, companies, and communities are grappling with similar questions.
AI systems require enormous computational resources. Those resources require data centres. Data centres require electricity, cooling, land, and infrastructure.
The chain is straightforward.
In response to growing concerns, Utah Governor Spencer Cox recently announced a higher standard for future data centre development in the state. The framework reportedly considers factors such as water resources, air quality, wildlife impacts, utility costs, and overall quality of life.
That move suggests policymakers increasingly recognize that AI infrastructure is no longer just a technology issue. It is an environmental issue, an economic issue, and a community issue.
For now, Project Stratos remains a vision rather than a finished reality. The warnings, projections, and counterarguments are all part of an ongoing conversation about what large-scale AI development should look like. Yet the debate leaves a lingering question.
As society races to build the infrastructure required for artificial intelligence, will the true challenge be creating smarter machines or learning how to manage the physical world that powers them?
