The Rising Environmental Toll of AI Datacentres
Recent studies indicate that generative artificial intelligence systems consume orders of magnitude more energy compared to conventional computing methods. This surge in power usage is fueling a rapid expansion of datacentres worldwide, raising critical questions about sustainability and the viability of continued AI adoption.
Explosive Growth in Datacentre Energy Demand
According to the International Energy Agency, global datacentre power requirements are increasing at a rate four times faster than all other sectors combined. Projections suggest that by 2030, this demand could surpass the total electricity consumption of Japan. In Australia, energy market operators anticipate datacentre energy needs will triple within the next five years, potentially exceeding the power used by the nation's entire electric vehicle fleet by the end of the decade.
Authorities also warn of significant strains on drinking water supplies due to the cooling requirements of these facilities. This escalating environmental footprint has sparked concerns that datacentres might hinder national net-zero ambitions, particularly in regions like Australia where energy transitions are already underway.
Quantifying AI's Environmental Impact
Estimates vary, but research consistently shows that generative AI models—which produce text, images, and video—are substantially more energy-intensive than traditional computing. Some analyses suggest they use up to five times more power, with potential for even higher figures depending on the specific model or query type.
Professor Jeannie Paterson, co-director of the Centre for AI and Digital Ethics at the University of Melbourne, highlights a key issue: limited transparency from tech companies regarding the energy, water, and emissions associated with AI operations. "Training models and running datacentres is an energy-intensive task," she notes, underscoring the hidden costs behind consumer-facing AI tools.
Ketan Joshi, a climate analyst based in Oslo, points out that AI chatbots and similar software are "uniquely energy inefficient" due to the vast datasets and complex pattern-matching processes involved. He compares using AI for simple queries to "driving to the shops in an SUV instead of riding your bike"—adding unnecessary energy demand with minimal societal benefit.
A study published in the journal Patterns projects that AI's global carbon footprint could reach between 32.6 and 79.7 million tonnes of CO2 emissions by 2025, with water consumption estimated at 312.5 to 764.6 billion litres—akin to the worldwide bottled water market.
The 'QuitGPT' Movement and Consumer Resistance
As AI tools become embedded in everyday life—from workplace software and educational platforms to supermarket self-checkouts and facial recognition systems—the 'QuitGPT' movement is gaining traction. Initially focused on boycotting AI over surveillance and weapons use, this campaign now also addresses environmental concerns.
Joshi describes the integration of generative AI into major platforms like Meta, Google, and Microsoft as a tactic to "embed these systems into society and instil dependency," similar to the proliferation of single-use plastics in the 1970s. He advocates for opting out as a "meaningful act of resistance," though he expresses disappointment that QuitGPT often redirects users to alternative AI platforms rather than promoting complete abandonment.
Practical steps to reduce AI usage include:
- Unsubscribing from AI platforms and services
- Excluding AI results from search queries by adding "-AI"
- Avoiding energy-intensive tasks like text-to-video generation or AI-created images for non-essential purposes
Professor Paterson emphasizes that while avoiding AI entirely is challenging, individuals can still influence its development and application by voicing concerns and making conscious choices.
Local Impacts and Community Advocacy
Datacentres, often massive warehouse-like facilities with continuous lighting and air conditioning, are the physical backbone of the AI boom. Their proliferation has led to growing calls for industry accountability. A coalition of energy and environmental groups, including the Clean Energy Council and Australian Conservation Foundation, has proposed public interest principles requiring datacentres to invest in renewable energy and responsible water use.
Adam Bandt, chief executive of the Australian Conservation Foundation, argues that "big tech corporations should be forced to do their fair share" by building accompanying renewables and water recycling infrastructure. Communities near proposed datacentre sites have begun campaigning against these projects, citing potential disruptions to local wildlife, noise pollution, and resource strains.
Dr. Bronwyn Cumbo, a social researcher at the University of Technology Sydney, notes that datacentres are frequently clustered in industrial hubs, amplifying their local effects. She stresses the importance of community awareness and critical engagement to shape how AI integrates into society. "There is an inevitability to AI being part of our lives," she says, "but how it's part of our lives is something we can definitely control."
As debates over AI's environmental costs intensify, the QuitGPT movement highlights a broader societal reckoning with technology's footprint. Whether through consumer boycotts, regulatory pressure, or community action, the path forward requires balancing innovation with ecological responsibility.
