This article was first published by Canada's National Observer.
The boom in artificial intelligence promises remarkable breakthroughs. Yet behind the optimism lies an uncomfortable truth: modern AI is expanding an energy footprint that was already too large. From training giant models to processing billions of prompts, the power draw is soaring — along with knock-on effects such as land for ever-bigger data centres and the water needed to cool them. Communities from Mexico to Ireland are already reporting electricity strain and water cuts.
We are, in effect, burning the candle at both ends of our climate commitments while sitting in a sauna that grows hotter by the day. What happens next depends on what we choose now.
Today’s progress playbook prizes quantity over quality: bigger models, more data, faster processing. That chase, measured in parameter counts and petaflops, celebrates scale even when the planet cannot sustain it. When “advancement” depletes natural systems and mortgages our children’s future, it is self-exhaustion, not growth. The task now is to reconcile technological ambition with planetary health. And three emerging trajectories hint at what comes next.
First, if it is business as usual, and we continue with the Old Normal, adoption keeps accelerating, and model sizes keep ballooning. Efficiency improvements arrive, but are swallowed by demand (classic Jevons Paradox). In this scenario, AI’s electricity consumption rivals that of nation-states; data centres drain local power grids and water supplies. Investment continues to reward brute-force scale, pushing decarbonization burdens onto the energy sector while avoiding deeper change. Climate crises intensify, with the harshest effects falling on the most vulnerable. And the verdict is an unsustainable future, with a tough path to recovery.
Second is the worst-case heat death scenario, in which a zero-sum race for AI supremacy, speed and scale override environmental rules and ethics. This trajectory makes scarcity the norm as clean water, air, electricity, land and food dwindle, inflaming geopolitical tensions and widening inequality. Economic gains concentrate among a few, while agency decay spreads as people cede ever more decisions to opaque systems. Society is left with costly, reactive measures that erode quality of life and accelerate planetary damage. The verdict in this scenario would be catastrophic and irreversible.
Finally, there is hope for a better future with humane hybrid ingenuity. A different path begins by shifting from “bigger” to “better.” Instead of raw scale, we optimize for purpose-built, efficient, and environmentally conscious systems. Regenerative intent is baked into blueprints and incentives, carried by a mindset oriented to shared prosperity.
However, the path to that future requires two foundational shifts — technological and ethical.
In the technological shift, researchers prioritize algorithmic efficiency and hardware co-design over parameter counts, including alternatives inspired by how natural intelligence works (e.g., neuromorphic approaches). “Less is more” becomes the engineering norm; AI is deployed as a decarbonization tool that optimizes logistics, energy grids and materials discovery — cutting emissions rather than spiking them.
When it comes to the ethical shift, governance ties AI development to transparent environmental metrics and prosocial outcomes. Investment flows to federated and edge AI to limit the need for colossal, centralized data centres. Cross-border, cross-sector, multi-generational alliances set standards for sustainable, prosocial AI. Natural and artificial intelligence are treated as complementary, amplifying creativity, compassion and kindness. By aligning goals and guardrails with human and planetary well-being, we lower the risk of runaway systems while creating a genuine win-win-win-win — for individuals, communities, nations and the planet. Verdict: sustainable, pro-people, pro-planet, pro-potential.
The ABCD of hybrid action
The energy challenge is an invitation to decide what kind of society we want to be in the 21st century — and to act accordingly. Four moves translate ambition into practice:
Assess damages. Demand full transparency on AI’s environmental footprint across the lifecycle: model design, training, inference, infrastructure construction, cooling and end-of-life. Publish standardized energy, emissions, water and land metrics so markets and regulators can compare systems like-for-like.
Build on opportunities. Champion research and investment in small, smart AI: compact architectures, sparse methods, transfer learning, retrieval-augmented approaches and domain-specific models that deliver accuracy with far less computing. Pair software innovations with efficient hardware and workload-aware scheduling.
Cultivate alliances. Establish international standards that link model deployment to environmental thresholds and prosocial outcomes. Incentivize federated and edge solutions where appropriate and share best practices across government, industry, academia and civil society.
Discover the Champion of Change in Yourself. Choose tools, vendors and investments that prioritize efficiency and human and planetary well-being. Shift procurement criteria and KPIs so “better” beats “bigger.”
We are navigating a dangerous hybrid tipping zone, delegating more choices to always-available digital assistants even as their aggregate footprint strains the physical systems that support life. But nothing forces us to keep moving in the same direction. By reframing the problem and acting on the ABCD, we can step out of the sauna of unsustainable growth and into a future where technology truly serves people and the planet.
