According to the IMF and other global observers, AI is becoming an increasingly large contributor to energy demand, driven by massive data centres, high-intensity model training, and real-time inference at scale. The question is no longer whether AI will impact the climate, it's how we ensure it helps rather than harms.

Training a single large language model can consume as much energy as a small town uses in a week. As deployment scales across healthcare, finance, manufacturing and government, global data centre energy demand is expected to more than double by 2030.

This growth is unsustainable without significant changes to infrastructure, regulation, and design philosophy. Ironically, some of the most promising use cases for AI are in making other systems more energy-efficient:

We're also seeing innovations in neuromorphic computing and event-driven AI models (e.g. the Spiking Neural Network) which consume drastically less energy than traditional deep learning models, in some cases, approaching the efficiency of biological systems.

In these scenarios, AI doesn't just "offset" its own footprint. It could eliminate meaningful power usage altogether, particularly when deployed on-device or in ultra-low-power environments like edge IoT.

So, should AI have an energy label? Yes, and urgently. Consumers now expect energy ratings on fridges, washing machines and televisions. Shouldn't we expect the same transparency for AI? Imagine an AI system with an Energy Impact Rating, clearly indicating:

This would:

For AI to continue driving economic growth without derailing net-zero targets, we need a shared response:

AI won't magically reach a zero-carbon footprint. But if we pair it with smart legislation, innovative engineering, and greater public transparency, it can become one of the greatest climate tools we have. The question is: will we demand it?