Is the future arriving on a power grid already straining at the seams? We’re told artificial intelligence will revolutionize everything from healthcare to transportation, but the breathless coverage rarely mentions the sheer electricity required to run these revolutions. The real story here isn't the dazzling potential of AI – it's the rapidly escalating, and largely hidden, energy cost of making it happen, and who ultimately foots the bill. For months, the industry treated energy consumption as a proprietary secret, a detail too messy to share with investors or the public. Now, thanks to dogged reporting and a little public pressure, the numbers are starting to leak, and they’re far from reassuring.
The Data That Forced Transparency
For six months, James O’Donnell and Casey Crownhart at WIRED meticulously pieced together the energy puzzle surrounding AI, sifting through reports and interviewing experts to quantify the previously opaque energy demands of these systems. Their investigation didn’t just point fingers; it quantified the problem. It wasn’t about vague warnings of increased energy use, but about the concrete energy cost of a single prompt, scaled up to illustrate the potential impacts of AI’s current and future demand. This wasn’t a theoretical exercise. The team’s work directly spurred action. In the wake of their reporting, companies like OpenAI, Mistral, and Google – previously tight-lipped about their energy usage – began releasing details about their models’ consumption of both energy and water. This shift from secrecy to disclosure is significant, but it’s also a reactive measure, a damage control response to scrutiny rather than proactive environmental stewardship.
Based on the original technologyreview.com report.
Beyond the Server Farm: The Hidden Costs
The initial shock comes from the sheer scale. We’re not talking about a marginal increase in energy demand. The energy required to train a single AI model can now rival the lifetime emissions of five cars. But the impact extends far beyond the server farms themselves. The source of that energy matters immensely. A model trained on renewable energy has a drastically different climate footprint than one powered by coal. And here’s where the picture gets murkier. While some companies are investing in renewable energy credits, the actual percentage of AI workloads running on truly clean energy remains stubbornly low. The industry’s reliance on grid electricity, often a mix of fossil fuels and renewables, means that every query, every image generated, every line of code written contributes to overall carbon emissions. This isn’t a problem for “the future” – it’s happening now.
The Water Problem Nobody’s Talking About
The focus on energy consumption often overshadows another critical resource: water. AI data centers require massive amounts of water for cooling, particularly in warmer climates. Google, for example, revealed significant water usage figures alongside their energy data, highlighting the strain AI is placing on local water resources. This is especially concerning in regions already facing water scarcity, where AI development could exacerbate existing challenges. The water issue isn’t just an environmental concern; it’s a social justice issue. Data centers are often located in or near marginalized communities, meaning these communities bear a disproportionate burden of the environmental costs of AI. The industry’s response has been largely silence, with few concrete plans to mitigate water usage or invest in water conservation technologies.
What Happens When AI Gets… Bigger?
The current energy and water demands of AI are substantial, but they represent just the beginning. As models become more complex, as AI permeates more aspects of our lives, the energy footprint will inevitably grow. The 2026 awards, to be presented in New York City on May 19, will likely celebrate further advancements in AI, but will they also acknowledge the environmental cost of those advancements? The industry is currently focused on a race to build bigger, more powerful models, with little consideration for the sustainability implications. We’re hurtling towards a future where AI is ubiquitous, but unless we address the energy and water challenges head-on, that future could be powered by unsustainable practices and exacerbate existing inequalities.
Here’s what to watch for: in the next 18 months, expect a surge in “greenwashing” from AI companies, with increasingly elaborate claims about their sustainability efforts. The real test won’t be the press releases, but the independent audits of their energy sources and water usage. Specifically, look for data on the actual percentage of renewable energy powering AI workloads, and whether companies are investing in technologies to reduce water consumption beyond simply purchasing water credits. If those numbers don’t improve dramatically, the AI revolution will come at a price far higher than most of us are willing to pay.
