Artificial intelligence is often described as “cloud-based,” immaterial, or effortless. This description is misleading; an epic marketing deception. AI is deeply material: every response, image, or prediction depends on large-scale infrastructure. Servers consume megawatts of electricity, data centres require massive amounts of water for cooling, and semiconductors rely on rare and contested minerals. AI is not abstract. It is material, extractive, and energy-intensive.
Energy Demand and Infrastructure Lock-In
AI’s energy requirements are growing rapidly. Training a single large model can consume as much electricity as thousands of households in a year¹. Once deployed, these systems run continuously to serve millions of users¹. Even when companies claim to run on renewable energy, the scale of infrastructure locks in decades of high-energy use¹. Efficiency improvements reduce energy per calculation but do not prevent total energy consumption from increasing². Cheaper computation drives more demand: bigger models, more queries, more energy².
According to the International Energy Agency, global electricity use by data centres — driven largely by AI workloads — is projected to more than double from 415 terawatt-hours in 2024 to around 945 terawatt-hours by 2030, roughly equivalent to the annual electricity consumption of a country the size of Japan¹. In the United States, hyperscale data centres already consume around 2% of national electricity, with some metropolitan regions seeing over 10% of local grid capacity dedicated to cloud and AI infrastructure, illustrating how rapid AI growth can stress energy systems even before global totals double¹.
Deliberate Invisibility
AI interfaces are intentionally designed to hide these material realities. The “cloud” metaphor, and dashboards; sustainability reports do not appear in daily user experience. This invisibility is not accidental; it is functional. By hiding energy, water, land, and labour costs, corporations avoid scrutiny and ensure the expansion of AI infrastructure remains politically uncontested.
Market Control and Digital Power
This invisibility supports a broader concentration of power. Large AI companies do not simply provide services; they increasingly control digital markets. By owning models, cloud infrastructure, proprietary data, and access to compute, a small number of firms are able to set the terms under which AI can be developed and used. High energy and infrastructure costs act as barriers to entry, preventing meaningful competition and locking governments, universities, and smaller firms into dependency on corporate platforms. Decisions about how much computation is used, for what purposes, and at what environmental cost are therefore concentrated in private hands, where growth and market dominance are prioritised over ecological or social limits.
Geopolitical and Environmental Implications
AI’s material footprint extends beyond energy. Rare earth minerals, water, and land are extracted and distributed through global supply chains governed by purchasing power and legal authority³. For example, Neodymium, a rare earth element critical for high-performance permanent magnets used in data centres, is mined primarily in China, the United States, and Australia³⁴. Benefits are concentrated while environmental and social costs are externalised onto vulnerable communities and ecosystems³⁵. Structural inequities are not addressed by efficiency improvements, green certification, or auditing³.
Opportunity Costs and Climate Consequences
Even when AI contributes to climate-related work, such as optimising energy systems, its energy consumption competes with other urgent decarbonisation efforts¹. Every watt powering large AI models is a watt unavailable for public transport, electrified housing, or other critical climate solutions¹. Rapid expansion risks reinforcing energy scarcity and undermining planetary stability¹.
Intelligence and Ecological Limits
Intelligence in ecological and social systems is relational and adaptive, balancing use with sustainability. Current AI systems prioritise scale, speed, and control over sustainability. Unlimited growth and extraction are not intelligent; they are destructive. AI expansion erodes the ecological and social conditions necessary for human and planetary survival.
Policy and Ethical Implications
AI energy consumption must be recognised as a planetary boundary issue. Large-scale AI should be treated as heavy infrastructure, subject to limits on energy, water, and supply-chain impact. Smaller, slower, and community-focused AI systems could serve practical needs within ecological limits. Governance requires transparency, accountability, and attention to environmental and social justice.
Artificial intelligence is not artificial. It is material, extractive, and energy-intensive. Its growth is neither neutral nor invisible. Unless these realities are acknowledged and addressed, AI will continue to threaten the planetary systems that sustain life and knowledge.
---------------------------------------------------------------------------------------------------------------------------
Annotated Footnotes
1. International Energy Agency (IEA), “Energy Demand from AI,” 2025
2. Chen, X., et al., “Electricity Demand and Grid Impacts of AI Data Centers,” arXiv, 2025
Link
3. Wikipedia, “Neodymium” (accessed 2026)
Link
4. Wikipedia, “Neodymium magnet” (accessed 2026)
Link
5. China Briefing, “China’s Rare Earth Elements Dominance,” 2024
Link
No comments:
Post a Comment