The AI Dilemma Ahead of COP31: Data Centers to Decarbonization

As climate change accelerates, the explosive growth of AI is placing increasing pressure on energy systems, water resources and carbon budgets, making sustainable “green AI” policies and clean energy transitions a critical agenda for COP31 and beyond.

I'm still amazed by things most people take for granted such as having light by flicking a switch or taking a hot bath whenever I want. The electricity flowing into our homes and the water from our taps are things most of us now take for granted. But two forces are colliding to put this comfort at risk and understanding their collision may be one of the defining challenges of our time. 

The first force is climate change. The year 2024 was the warmest on record in 175 years of observation. The Paris Agreement asked the world to hold warming to 1.5°C. This is a threshold that scientists warn, if crossed permanently, triggers severe consequences to the world’s economy, biodiversity, social equity and dignity for all, as targeted in the UN Sustainable Development Goals. Many scientists now consider 1.5°C no longer achievable. So where are we actually headed? The UN's 2025 Emissions Gap Report finds that even if every country fully delivers on its current climate pledges, the world is still on track for 2.3–2.8°C of warming by 2100. The difference between 1.5°C and 2.5°C isn't just a degree on a thermometer. It's opening the gates towards a difficult future with more extreme weather conditions, more climate refugees, more failed harvests, more cities underwater and more destruction to biodiversity.

The second force is artificial intelligence. The massive computer systems that power AI need huge amounts of electricity and water. These systems are growing at a pace we've never seen before. Here is the hard question we, in the climate space, need to answer honestly: can we reach a 1.5°C world or a 2°C world while AI demand is exploding? Is this being addressed fast enough in global climate negotiations? 

As host to COP31, this conversation is particularly relevant to Turkey who is actively expanding its data center capacity. There are major developments planned for hyperscale data centers which would position the country as a regional hub for data storage. Strong government backing is accelerating this growth. As Turkey deals with increasing risks like drought, water scarcity and heatwaves, her ability to sustainably grow this ecosystem will depend on how well she balances this ambition with efficient resource management, renewable energy integration and transitioning to clean energy grids. 

The climate challenge is clear as electricity demand and carbon emissions remain tightly coupled. Today, around 60% of global electricity is still generated from fossil fuels. Decoupling this relationship through power sector decarbonization is the central challenge we must address.

In 2024, data centers worldwide consumed about 415 TWh of electricity, about 1.5% of all global electricity. To put that in perspective, that's more than the entire country of France uses in a year.  And this is only the beginning. The International Energy Agency projects energy consumption may triple in about a decade. A big reason? Every time we ask an AI chatbot a question, it uses far more power than a simple web search.  Whereas one Google search uses around 0.3 Watt-hour (Wh), one ChatGPT  question uses about 2.9 Wh, which is almost 10  times the electricity demand of the traditional searches. 

The carbon footprint hiding behind each search means more electricity and more emissions unless that electricity is clean. Globally, AI systems' carbon footprint estimates are between 32.6 and 79.7 million tons of CO₂ in 2025, a range roughly equivalent to New York City's annual emissions. By 2030, at the current rate of AI growth, US data centers alone could add 24 to 44 million metric tons of carbon dioxide annually which is the equivalent of putting 5 to 10 million additional cars on the road.

Did You Know? The world's largest tech companies pledged to reach net-zero emissions by 2030. Then AI arrived. Google's emissions climbed roughly 54% between 2019 and 2024; Microsoft's rose about 23% since 2020. Google has since quietly removed its 2030 net-zero pledge from its sustainability website — a sign of just how dramatically AI has rewritten the math on corporate climate commitments. 

Did You Know?

Google's annual data-center water use alone approaches one-third of Türkiye's total national water consumption.

Let's be fair and balanced here. In absolute terms, AI's share is still relatively small:

-Data centers use less than 3% of global electricity.

-They consume a tiny fraction of global freshwater.

Its the growth rate that is the real story. The problem is local concentration. Data centers and chip factories tend to cluster in specific regions, where they can gobble up 10–20% of the local power grid and water supply. 

The Key Insight AI doesn't threaten to drain the planet's resources on a global scale…. at least not yet. The real danger is that it's concentrating massive demand in specific places, straining local power grids, water supplies and carbon budgets that regular people depend on.

We in the Global Climate Academy urge that several key questions to be considered in shaping the agenda for COP31. 

• How can AI deployment be aligned with net-zero goals? 

• What policies can speed up transitioning to a clean grid (an electrical grid which uses sustainable sources instead of fossil fuels) which can sustainably power the increased needs of data center?

• Should there be global standards or transparency requirements for the carbon footprint of AI systems? 

• How can developing and emerging economies such as Turkey leverage AI without exacerbating energy demands and resource constraints? 

• What role should the private sector play in ensuring “green AI” practices? 

  
  

AI doesn't run on magic. It runs on enormous buildings full of powerful computer chips that work around the clock. These buildings are called data centers and they need two things in massive quantities: electricity and water. As upper middle-income economies join industrialized economies in growing more affluent, their energy consumption rises sharply. AI could support energy efficiency, renewable integration and climate resilience, but it also raises concerns around electricity demand growth and infrastructure readiness. Positioning AI as part of a “clean growth” strategy powered by renewable energy and efficient data infrastructure will likely be a key narrative going into COP31.

Lets note some important global efforts on AI 

• The UN Secretary-General’s AI  Advisory Body was recently launched to develop globally inclusive recommendations for governing artificial intelligence. We encourage this initiative to ensure that the rapid expansion of AI infrastructure and data centers will align with global climate goals and does not create unsustainable pressure on energy, water, and critical resources.

• WeDontHaveTime launched an Energy Efficiency Campaign where many of the top 10 industrial energy efficiency action is linked to AI.  The multiplier effect of this initiative can be critical if the solutions can spread rapidly and be made accessible to other industries.  

• Activation Group No:27 as part of the COP30 Action agenda focuses on artificial intelligence, digital public infrastructure and technologies. COP31 will provide an opportunity to follow-up on their progress. 

  
  

The Bottom Line

As climate change intensifies, the rapid expansion of AI could place additional pressure on already strained energy systems and increase carbon emissions if left unmanaged. The situation is not hopeless yet. Through strategic planning, stronger regulation, more efficient and less carbon-intensive technologies, a faster transition to clean energy and the exploration of innovative energy sources, governments and stakeholders can ensure that AI supports human progress without deepening the climate crisis.  

Climate Academy Global calls for governments and stakeholders to plan ahead by integrating AI and its climate and resource impacts into the COP processes before pressures become unmanageable.