How Data Centres and AI Models Are Reshaping Global Energy Demand — and What Sustainable AI Must Look Like
By Amb. Canon Otto
Convener, Global Sustainability Summit
Commentary in partnership with Cleancyclers
Artificial Intelligence is often presented as invisible. Clean. Weightless. A product of code, algorithms, and innovation floating somewhere in the cloud.
But there is nothing invisible about the infrastructure that powers AI.
Behind every search query, recommendation engine, facial recognition system, and large language model lies a vast physical system: data centres consuming enormous amounts of electricity, water, land, and materials. The digital future is being built on very real environmental foundations—and those foundations are carrying a growing carbon cost.
At Cleancyclers, and through ongoing reflections on SustainabilityUnscripted, we believe it is time to speak plainly: AI is not automatically sustainable simply because it is digital.
The Carbon Reality Behind the Cloud
Data centres are the industrial factories of the digital age. They operate 24/7, require constant cooling, and rely on uninterrupted power. As AI models grow larger and more complex, their training and deployment demand exponentially more energy.
Training a single advanced AI model can consume as much electricity as hundreds of households use in a year. Multiply this across thousands of models, global cloud providers, and constant retraining cycles, and the footprint becomes impossible to ignore.
The uncomfortable truth is this: we are rapidly decarbonising some sectors while quietly carbonising our digital infrastructure.
The question is not whether AI will continue to grow. It will.
The question is whether we are designing AI systems that respect planetary limits—or quietly outsource their environmental cost to future generations.
Digital Does Not Mean Dematerialised
One of the great myths of the modern economy is that digital systems replace material ones. In reality, they often shift and concentrate material demand elsewhere.
AI requires:
- Energy-intensive data centres
- Rare earth minerals and metals
- Water for cooling
- Hardware with limited lifespans
- Complex global supply chains
When servers are replaced every few years, the result is not just innovation—it is electronic waste, resource extraction, and embedded emissions.
This is where the work of Cleancyclers becomes especially relevant. Waste is not an afterthought of technology; it is a design outcome. If we do not plan for circularity in digital infrastructure, we simply delay the problem until disposal—when it becomes more expensive and more damaging.
AI Efficiency Is Not the Same as AI Sustainability
Many technology companies speak about “efficiency”—faster processing, smaller chips, optimised algorithms. Efficiency matters, but efficiency alone does not equal sustainability.
A more efficient system that is deployed at massive scale can still increase total emissions. This is the rebound effect in digital form.
True sustainability asks harder questions:
- Is the AI application necessary?
- Does it solve a real social or environmental problem?
- Is it powered by renewable energy?
- Is the hardware designed for reuse and recovery?
- Is the data centre integrated into a circular system?
Through conversations on SustainabilityUnscripted, we consistently return to this point: technology without governance simply accelerates existing problems.
The Role of Circular Thinking in Sustainable AI
At Cleancyclers, we approach sustainability at the system level. Whether we are dealing with plastics, metals, or electronic components, the principle is the same: design for the full lifecycle.
Sustainable AI must embrace circular thinking:
- Modular hardware that can be upgraded, not discarded
- Responsible e-waste recovery and recycling
- Heat recovery from data centres for surrounding communities
- Integration of renewable energy at scale
- Transparent reporting of digital carbon footprints
Creativity plays a central role here. When creativity is applied to systems, waste becomes opportunity. Data centres can become energy hubs. Hardware waste can become material input. Digital growth can coexist with environmental responsibility—but only by design, not by accident.
Why This Conversation Matters Now
AI is being embedded into governance, finance, healthcare, education, security, and climate modelling itself. It will shape how societies function for decades.
If we ignore its environmental footprint now, we lock in high-carbon digital infrastructure that will be difficult—and expensive—to reverse.
As Convener of the Global Sustainability Summit, I have seen how quickly emerging technologies outrun policy frameworks. Sustainability cannot afford to be reactive. We must design standards for sustainable AI before scale becomes dependency.
From Innovation to Responsibility
The future of AI should not be judged only by intelligence or profitability, but by responsibility.
Responsible AI is not just ethical in how it treats data and people—it is ethical in how it treats the planet.
Through Cleancyclers, SustainabilityUnscripted, and the ongoing dialogues convened by CanonOtto, one message is consistent: sustainability must extend to every system we build—including the digital ones.
The hidden carbon cost of artificial intelligence is no longer hidden.
What remains is a choice.
Will AI deepen our environmental debt—or help us design our way out of it?
That answer will depend not on technology alone, but on intent, creativity, and the courage to build differently.
