Climate Change and AI – How They Interact and What Are the Potential Outcomes?

August 8, 2026

The arrival of generalized artificial intelligence (AI) couldn’t be coming at a worse time, say some environmentalists. Why? Because the AI being created by Silicon Valley billionaires is an energy hog, requiring massive data centres. That means expanding energy production to meet demand, which is turning into a carbon emissions problem as burning coal and natural gas in thermal power plants becomes the fastest short-term solution.

Others argue that the widespread adoption of AI globally will help humanity adapt to climate change and even mitigate its causes.

So, will AI exacerbate climate change, or will it be our saviour?

Data Centres and AI

The growing demand for data centres containing thousands of energy-hungry computers is at the crux of this debate. Data centres need energy to run and to stay cool. Where do they get the power? If it comes from coal and natural-gas-fired power plants, then AI in data centres represents a significant carbon enerative AI, Large Language Models, AI training, and AI use by big companies and people like me who ask AI questions and receive answers require lots of computing power.

GPUs, TPUs and CPUs

AI developers, when calculating energy requirements, count the number of GPUs, TPUs and CPUs needed to do the job. What are these?

GPUs are Graphics Processing Units. These are computer chips used when training LLMs and running machine learning (ML) models.
TPUs are Tensor Processing Units. These are chips used by neural network applications.
CPUs are Central Processing Units. Every computer, including the one I am using to write this posting, has a CPU. For AI, CPUs coordinate and run installed AI programs.

All of these processors need electricity. The amount needed reflects how much training the AI LLMs require, the number of training runs needed, including failures, and the fine-tuning required before AI models are publicly released.

Measuring Data Centre Kilowatt Hours

Companies like NVIDIA, whose chips dominate America’s AI developers, have tools to monitor and measure electricity per processing unit, per computer and per rack. That covers the computers. But data centres have external power requirements for cooling, lighting, storage and networking. In calculating total amounts of energy per data centre, a good measure is to look at the power requirements for every CPU, GPU and TPU. If a CPU needs a kilowatt-hour (kWh) of electricity, add an equivalent or greater amount of energy needed for external requirements.

Measuring Data Centre Carbon Footprints

If you know the amount of energy an individual data centre requires, you can begin to calculate its carbon footprint. That completes the first measure.

Adding emissions from the source where the electricity is produced is the second. Operational emissions are lower when a data centre operates in an area where electricity is produced from renewables like hydro, geothermal, wind, solar, or nuclear power. If the data centre draws its electricity from coal and natural gas-heavy thermal power plants, then operational emissions are greater.

Measuring Embodied Data Centre Emissions

There are additional carbon emission measurements to consider. Called embodied emissions, they include carbon produced in the manufacturing of the computing infrastructure, other hardware components, the data centre build itself, and the transportation and other external emission sources needed to support its development and operational existence.

It is hard to find consensus in actual calculations based on current data centre operations. One that I recently came across tries to total the energy and carbon footprint requirements and output coming from a small data centre operating 1,000 CPUs, GPUs and TPUs. Its energy requirement calculation came to 84,000 kWh. Its carbon footprint estimate for operations amounted to 25,200 kilograms (25+ metric tons) of carbon dioxide ( $CO₂). Sources not included in the calculation are embodied emissions or emissions by electricity source.$

AI’s Potential to Help Deal with Climate Change

The International Energy Agency (IEA), in quantifying the current environmental carbon footprint for data centres, estimates a present-day contribution of about 0.5% to total carbon emissions from the energy they use today. It sees that number rising to 1.4% by 2030.

With more AI, does that mean faster climate change, or can more AI help mitigate global warming? Can we use its artificial brain power to design solutions to climate change while allowing us to energize it?

Some AI enthusiasts in Silicon Valley think that the technology is our answer to climate change. They see AI data centre operations producing insights that humans cannot discern. For example:

AI can discern patterns in its own usage to forecast energy requirements from wind, solar and other renewable sources, and balance supply and demand.
AI can look at the energy grid and help us improve grid transmission performance.
AI can optimize data centres to reduce overall energy usage and address embodied emissions.

AI and Global Climate Change Externalities

Where AI can prove useful in mitigating or adapting to climate change is by focusing on important economic sectors such as agriculture, energy, healthcare, mining, transportation, manufacturing and service industries. AI can help to optimize food production, improve land use, address wildfire threats to global forests, identify new carbon sinks, improve weather forecasting and more.

AI as a technology accelerator can help us make how we produce the energy it and we need less negatively impactful on the environment. It can help us find new types of battery materials or invent new emission reduction technologies.

Widespread deployment of AI, through technological innovations, could reduce worldwide carbon emissions in 2035 by up to 1.4 gigatons states the IEA. The caveat for this prediction is that added data centres are powered only by renewable or net-zero energy sources.

Data centres generating AI could be a force for mitigating climate change. Yet currently, AI contributes to climate change through a proliferation of data centres and growing demand.

AI data centres can help save us from ourselves. As a technology accelerator that focuses on climate change challenges, AI operating from data centres can generate novel mitigation and adaptation solutions. Even though it may be obvious to AI, the one solution it cannot be programmed to consider is getting rid of us.