Environmental Impact

Yet the same technology is also being deployed to protect the planet. Machine learning models analyze satellite imagery to detect deforestation and illegal fishing, process environmental DNA to track endangered species, and forecast sea-ice loss so Arctic communities can prepare. AI optimizes renewable energy grids, reduces agricultural waste, and helps climate scientists run simulations that once required supercomputers. The question is not whether AI affects the environment—it does, in contradictory ways—but whether we can scale the beneficial applications faster than the harmful ones.

The most immediate risk is energy. Hyperscalers spent more than $355 billion on AI infrastructure in 2025, with Amazon, Microsoft, Google, and Meta each committing tens of billions of dollars to new data centers. Much of that demand is still met by fossil fuels. The IEA estimates that roughly 40% of additional data center energy by 2030 will come from gas and coal in many regions, and a February 2026 report found that 74% of industry claims about AI's climate benefits were unproven. The carbon math is simple: more coal-powered AI means more warming.

Water is the hidden cost. AI data centers used an estimated 175 billion litres of water in 2023, a figure projected to triple by 2030. A large facility can consume millions of gallons per day, often in regions already facing drought. Unlike electricity, water use is hyper-local: one data center can stress aquifers, raise temperatures, and compete with agriculture and drinking supplies.

The net impact depends on choices. Conservation AI tends to use small, efficient models running on intermittent inference, while consumer generative AI runs massive models continuously. Location matters too: a query processed on a wind-powered grid is radically cleaner than one on a coal-heavy grid. Policy, transparency, and procurement criteria will determine whether AI becomes a climate solution or a climate accelerant.

For individuals and organizations, the most powerful lever is intentionality. Avoid generating content you do not need, choose AI providers that publish transparent sustainability reports, and favor efficient, purpose-built tools over massive general-purpose models. For policymakers, the agenda is clearer: require energy and water disclosure for large AI workloads, enforce hourly matching of renewable energy claims, and direct public AI investment toward climate adaptation and conservation rather than entertainment and surveillance. The European Union's AI Act and emerging U.S. state-level data center regulations are early signals, but the pace of infrastructure investment is outpacing governance.

This cluster examines that tension from two angles: a concrete case study of how AI both helps and harms polar bears, and a balanced analysis of whether AI is, on balance, ruining everything. Both articles share a common conclusion: the future of AI's environmental impact is not fixed by the technology itself, but by the decisions we make about how to build, power, and regulate it.