AI data centres now consume roughly 1% to 2% of global electricity. A single ChatGPT query uses around 10 times more energy than a Google search. Microsoft’s water consumption for cooling AI infrastructure rose 34% in a single year. These are not small numbers.
And yet the same technology is being used to model climate patterns with unprecedented accuracy, optimise renewable energy grids, accelerate the discovery of new battery materials, and reduce emissions across supply chains. The relationship between AI and the environment is genuinely complicated — and the honest answer is that it could go either way.
The Energy Cost of AI
Training a large language model like GPT-4 is estimated to have consumed between 50 and 100 gigawatt-hours of electricity. That is roughly equivalent to the annual electricity use of 5,000 UK homes for training a single model.
Inference — running the model to answer queries — is lower energy per query but adds up enormously at scale. OpenAI serves hundreds of millions of queries daily. Google processes over 8 billion searches per day and is integrating AI into more of them.
The UK is not exempt from this. The government’s AI growth zone proposals in 2025 included plans for large data centre campuses in the East Midlands and the North East. Each facility will require hundreds of megawatts of power — equivalent to small towns — and significant water for cooling.
Whether this energy consumption is environmentally harmful depends entirely on where the electricity comes from. Microsoft, Google, and Amazon have all committed to powering their data centres with 100% renewable energy, though the definitions and timelines vary. The UK’s increasingly renewable electricity grid — which reached 50% renewable in 2025 — helps, but AI demand is growing faster than renewables can expand.
Where AI Is Genuinely Helping
Climate modelling is the clearest win. Traditional climate models run on supercomputers and take weeks to produce predictions. AI-enhanced models from Google DeepMind and the European Centre for Medium-Range Weather Forecasts are now generating 10-day weather forecasts in seconds with accuracy that matches or exceeds traditional methods.
More accurate forecasting means better grid management. When you know three days in advance that wind generation will drop on Wednesday afternoon, you can schedule backup generation or demand reduction accordingly. The National Grid in the UK has been piloting AI forecasting tools since 2023 with measurable reductions in backup fossil fuel use.
Battery materials discovery is another area where AI is accelerating progress. Microsoft Research announced in 2024 that AI screening of 32 million candidate materials led to the identification of a new solid-state battery material in months — a process that would have taken decades using traditional laboratory methods. Faster battery development means faster electrification of transport and heating.
Agricultural efficiency matters for emissions too. UK farming accounts for around 10% of national greenhouse gas emissions. AI systems that optimise fertiliser application — using satellite imagery and soil sensors to apply exactly the right amount in the right places — can reduce nitrous oxide emissions, a greenhouse gas 300 times more potent than CO2, by 20% to 30% on participating farms.
Smart Grid Management
The transition to renewable energy creates a fundamental grid management problem. Wind and solar are intermittent — they produce electricity when the wind blows and the sun shines, not necessarily when demand peaks. Managing a grid with high renewable penetration is vastly more complex than managing one dominated by controllable fossil fuel plants.
AI is central to solving this. Demand forecasting, generation forecasting, battery storage optimisation, and real-time balancing all benefit from machine learning approaches that can handle the enormous complexity of modern electricity systems.
Google reduced the electricity used for cooling its data centres by 40% using DeepMind’s reinforcement learning system, as I mentioned in our piece on reinforcement learning. The same techniques are now being licensed to energy companies and grid operators.
The Rebound Effect Problem
There is a well-documented phenomenon in economics called the rebound effect. When technology makes something more efficient, consumption of that thing often increases rather than decreasing overall, because efficiency makes it cheaper and easier.
AI efficiency improvements in energy systems could follow the same pattern. If AI makes renewable energy cheaper and more manageable, it could accelerate deployment — which would be genuinely positive. But if it makes energy consumption in general cheaper and more accessible, total consumption could rise even as efficiency per unit improves.
This is not a hypothetical concern. AI-assisted design tools are already making it cheaper to design and manufacture more complex products. If those products are electric vehicles or heat pumps, that is good for emissions. If they are more consumer electronics with short replacement cycles, the net effect is less clear.
What UK Businesses and Consumers Can Do
For businesses evaluating AI adoption, asking about the carbon footprint of the tools you use is increasingly reasonable. Major cloud providers publish carbon intensity data for different regions. Running AI workloads in data centres powered by renewable energy makes a measurable difference.
Choosing when to run AI workloads matters. Electricity grids vary in carbon intensity throughout the day — lower when renewables are abundant, higher when demand peaks and fossil fuels fill the gap. Some large companies are already shifting AI training runs to overnight periods when UK grid carbon intensity is typically lower.
For individuals, the environmental cost of AI is real but context-dependent. Using AI to replace a car journey — getting a detailed answer that would otherwise require a trip to a professional — is almost certainly net positive. Using AI to generate 500 variations of a marketing image that will mostly be discarded is harder to justify environmentally.
The honest position in 2026 is this: AI could be a powerful tool in addressing climate change, and it is also a significant and growing source of energy consumption. Whether it ends up net positive for the climate depends on decisions being made right now about how data centres are powered and how AI capabilities are deployed.
This article is for educational purposes only and does not constitute financial advice. Cryptocurrency investments involve significant risk. Always do your own research.
