2025’s AI Boom Came With a Hidden Environmental Cost

Artificial intelligence had a spectacular year in 2025. Faster models. Smarter assistants. AI everywhere — from education and healthcare to marketing and music.

But while the tech headlines celebrated innovation, a quieter story emerged in the background: AI’s rapidly growing carbon and water footprint.

Recent research has shown that the explosive growth of large AI models in 2025 drove significant increases in CO₂ emissions and water use, largely due to the energy demands of data centres that train and run these systems.

Why AI Uses So Much Energy

Training modern AI models requires:

• Vast data centres running continuously

• Thousands of high-performance GPUs

• Massive cooling systems to prevent overheating

All of this consumes electricity — and in many regions, that electricity still comes from fossil fuels.

Even once trained, AI models continue to draw power every time they are queried. As AI tools become embedded in daily life, those “small” requests add up very quickly.

The Water Problem Few People Talk About

Cooling data centres doesn’t just require electricity — it often requires huge quantities of fresh water.

Some estimates suggest that training a single large AI model can consume millions of litres of water, particularly in systems that rely on evaporative cooling. In water-stressed regions, this creates a direct competition between tech infrastructure and local communities.

Efficiency Gains… But Not Fast Enough

To be fair, AI companies are working on:

• More efficient chips

• Better model optimisation

• Locating data centres near renewable energy sources

However, efficiency gains are currently being outpaced by demand. As AI use doubles, triples, and embeds itself into every workflow, total environmental impact continues to rise — even if each individual task becomes slightly more efficient.

This Is Not an Anti-AI Argument

AI has genuine potential to:

• Improve climate modelling

• Optimise energy grids

• Reduce waste and inefficiency

• Support scientific research

But pretending AI is “immaterial” or “cloud-based” in the environmental sense is no longer credible.

The Question for 2026 and Beyond

The real issue isn’t whether AI should exist — it’s how transparently and responsibly it is powered.

If AI is going to reshape society, then:

• Its energy sources must be scrutinised

• Its water use must be measured and reported

• Its environmental costs must be part of public debate

Innovation without accountability is not progress — it’s just outsourcing the damage.