The environmental impact of AI is largely concentrated in its infrastructure. Data centres that power AI models require vast amounts of electricity, often sourced from carbon-intensive grids. Training a single large language model, for example, can emit as much carbon dioxide as the lifetime emissions of several cars. Water use is another issue – AI servers require extensive cooling, and some facilities use millions of gallons of water annually, sometimes in regions already facing water scarcity. Meanwhile, the frequent hardware upgrades needed for high-performance computing contribute to electronic waste and accelerate the extraction of rare earth materials.
Despite these challenges, organisations can still integrate AI with sustainability strategies, provided companies make concerted efforts to reduce harm while using AI to achieve broader environmental and social goals.
Best practices to aligning AI opportunities with sustainability goals
Below are six keys to aligning AI usage with sustainability goals – specifically the United Nations’ Sustainability Development Goals.
1. Prioritise green data centres
One of the most direct ways companies can reduce the environmental impact of AI is by selecting data centres that run on renewable energy. Cloud service providers are increasingly offering carbon-neutral or 100% renewable options for AI workloads. Companies should proactively choose these cleaner infrastructures and, where possible, shift compute-heavy tasks to regions with lower-emission energy grids. By doing so, businesses can drastically cut the carbon footprint of their AI operations without sacrificing performance.
In support of:
2. Track and report AI’s carbon footprint
Transparency is essential to aligning AI with sustainability goals. Companies should implement systems to measure and report the carbon emissions associated with training and deploying AI models. This includes assessing energy usage over the entire lifecycle of the model – from data processing and model training to storage and deployment. Emerging tools, such as machine learning carbon calculators, can help quantify these impacts. Public reporting of this data not only builds trust but also encourages continuous improvement and accountability.
In support of:
3. Use smaller, more efficient models where feasible
Not every use case requires a massive, general-purpose AI model. Companies should explore the use of smaller, purpose-built models that are more computationally efficient and less resource intensive. In many cases, models that have been compressed or optimised using techniques like pruning, quantisation, or knowledge distillation can deliver similar performance with a fraction of the environmental cost. This approach allows businesses to enjoy the benefits of AI while consuming significantly less energy and hardware resources.
In support of:
4. Apply AI to advance sustainability goals
Paradoxically, AI can also be a powerful driver of sustainability when applied to the right problems. Many companies are using AI to solve environmental and social challenges aligned with the SDGs. Examples include using AI to optimise water usage in agriculture, reduce waste in manufacturing, forecast energy demand more accurately, or track deforestation and wildlife populations. These applications directly contribute to goals such as clean water and sanitation (SDG 6), affordable and clean energy (SDG 7), sustainable cities (SDG 11), and climate action (SDG 13). When deployed responsibly, AI can become an enabler of sustainability rather than a threat to it.
In support of:
5. Extend hardware lifespans and reuse resources
AI infrastructure does not just consume energy – it also requires physical components like GPUs, servers, and cooling systems, all of which have environmental costs. Companies can mitigate this by managing the hardware lifecycle more responsibly. This includes extending the life of servers through maintenance and upgrades, sourcing refurbished equipment, and working with vendors who support circular economy practices, such as equipment take-back and recycling programs. Such measures reduce electronic waste and help conserve the rare materials needed to manufacture high-performance computing hardware.
In support of:
6. Use AI to optimise the business’s own sustainability
AI is not just a tool for external applications – it can be embedded into a company’s internal operations to make the business itself more sustainable. For instance, AI can help optimise logistics routes to reduce fuel consumption, manage building energy usage in real time, or detect inefficiencies in industrial processes. By integrating AI into day-to-day systems and decision-making, companies can cut waste, lower emissions, and operate more efficiently, contributing to SDGs related to industry, innovation, responsible consumption, and climate action.
In support of:
While AI does pose environmental risks – especially when used at scale – it is not inherently at odds with sustainable development. The challenge for companies is to be intentional: to recognise and mitigate AI’s environmental footprint while leveraging its capabilities to drive positive impact across society and the planet. By making thoughtful choices about infrastructure, transparency, model design, and application areas, businesses can continue to harness the power of AI without abandoning their commitment to the SDGs. In fact, when used wisely, AI may become one of the most effective tools available in the global pursuit of sustainability.
