One AI data centre will produce 23 atomic bombs worth of heat a day, physics professor claims
AN ENORMOUS new AI data centre that has been approved would dump the heat equivalent to 23 atomic bombs per day into the environment, a physics professor has claimed.
The Stratos Project would be three times the size of Manhattan and consume twice the electricity Utah currently uses, it has been claimed.
A Utah State University physics professor, Dr Rob Davies, has calculated that a single proposed data centre would dump heat equivalent to 23 atomic bombs per day into Hansel Valley.
Dr Davies said: “The thermal load from the proposed Stratos project is extreme. There is no way around the physics. This is the energy output of two-and-half New York City’s poured into a single confined desert basin, in a watershed that’s already in crisis.
"Of course it has effects. One of those effects is this: This facility imposes substantial drying on a watershed and ecosystem already in active collapse.”
Paul Morris, Executive Director of Military Installation Development Authority (MIDA), has denied that the project is negative environmentally, insisting that it is “sustainable”.
He said: “The combination of site location and design make Stratos one of the most sustainable and effective data centers in the country.
"This project reflects what’s possible when state and local leaders work together to advance strategic infrastructure that supports our military mission, strengthens energy resilience, and delivers long-term economic investment. We look forward to continuing that collaboration as this project moves forward.”
AI experts have warned that this issue could come to the UK, with Global Action Plan finding 84% of proposed UK data centres sit in areas the Environment Agency classifies as water stressed.
Thames Water estimates that under high growth, data centres would drive 30% of all new regional water demand, around 270 million litres a day. A single hyperscale facility requires the daily water equivalent of 10,000 households.
Colette Mason, Author & AI Consultant at London-based Clever Clogs AI, said the environmental costs are real with AI.
She added: "Ecological damage documented across continents. Governments and developers consistently failing to produce impact assessments, leaving university professors and charities to do the homework for them.
“72 of 94 UK builds not started, against demand the frontier companies insist is coming. This is the seedy underbelly of the AI bubble. The hyperscalers want us to subscribe to a bright future, but when the environmental costs are real, the assessments are missing, and the builds aren't happening, that future looks a lot less rosy.”
Mitali Deypurkaystha, Human-First AI Strategist & Author at Newcastle upon Tyne-based Impact Icon AI, said AI infrastructure is leading to less housing being built.
She added: "The London Assembly has warned that data centre growth is already delaying housing delivery in parts of the capital. I support investing in data centres because the UK needs to future-proof itself, but we cannot damage the present in pursuit of a promised future that may never fully arrive, especially when homes are being delayed by capacity constraints happening right now.
"AI firms requesting huge amounts of electricity without firm build commitments risks turning Britain’s energy grid into the digital equivalent of ticket touting, where infrastructure gets reserved, hoarded and locked away while ordinary communities are left waiting.
“I remember the frustration of watching concert tickets vanish in minutes, only to reappear on resale sites at double the price. There is a similar public unfairness when companies can sit on energy capacity indefinitely while housing developments, hospitals and local communities still need power today.”
Katrina Young, AI & Digital Transformation Strategist at KYC Digital, said
She added: "The Utah thermal analysis is striking not because a physicist ran the numbers, but because nobody else did. Not the developer. Not the regulator. Not the state. That points to a structural gap in how AI infrastructure is assessed.
"Through my work in sustainable technology, AI and digital transformation, the bigger concern is cumulative oversight. Planning frameworks were built to assess individual sites, not what happens when dozens of data centres compete simultaneously for grid capacity, water access and thermal load.
"The risk is not one facility alone. It is that approvals are being granted through systems that cannot fully model aggregate environmental pressure across regions. That is not simply an enforcement issue. It is an architectural gap in governance itself."
