Behind every query lies an infrastructure of steel,
electricity, and water — growing faster than the grid can go green.
A single ChatGPT query consumes up to 10 times the electricity of a Google search. Multiply that by billions of daily queries.
Training GPT-4 consumed an estimated 50 million kWh — enough to power 4,600 average US homes for a year. Each new generation of AI requires exponentially more compute. The energy cost of training AI has doubled every 3.4 months since 2012.
Cooling servers requires enormous amounts of water — most of it evaporated into the atmosphere. This is happening in drought-stricken regions competing with farms and households.
Daily water consumption of a town of
50,000 people
Data centers don't appear in an abstract cloud. They land in real neighborhoods — and not always wealthy ones. Across America, communities are fighting back.
Data center cooling systems run 24/7, generating noise exceeding 80 dBA — comparable to a leaf blower running indefinitely outside your bedroom window. Noise mitigation remains largely voluntary across the industry.
A class-action lawsuit filed in February 2026 alleges Clean Water Act violations at an Amazon data center in Canton. Expansion was halted pending independent audits — a rare instance of legal accountability.
Local residents and the NAACP filed notice to sue under the Clean Air Act in Memphis, Tennessee — arguing a proposed data center would worsen already-dangerous air quality in a historically overburdened community.
Nearly one-third of California data centers sit in census tracts already in the top 10% most polluted in the state. This is not coincidence — it is the established pattern of an industry that follows the path of least resistance.
In 2024, former members of Trump's inner circle announced the most ambitious data center project ever proposed — in the territory the President is trying to acquire.
"We're gonna get Greenland."
Whether this is coordination or opportunism, the effect is the same: a US economic and infrastructural foothold in a Danish autonomous territory, at a moment of deliberate geopolitical pressure. For Danish readers, this is not abstract geopolitics. This is about Greenland.
The industry says yes. The data says: not yet — and the gap between promise and reality is widening.
"We are not optimizing AI for sustainability.
We are optimizing sustainability rhetoric for AI adoption."
Is the AI we are building worth the world it is costing?