The artificial intelligence revolution is reshaping global energy consumption at a pace that has caught utilities, governments and environmental advocates off guard. Tech sector carbon emissions continued their upward trajectory in recent years, driven by rapid advances in AI and data infrastructure, according to a report released today by the International Telecommunication Union and the World Benchmarking Alliance.
The surge reflects a fundamental shift in how the world's largest technology companies operate, with AI workloads demanding exponentially more power than traditional computing tasks and forcing difficult choices between climate commitments and competitive advantage.
Data centers now consume electricity at a rate 12 percent higher each year since 2017, four times faster than global electricity growth1. Leading tech firms announced $300 billion in AI infrastructure spending this year alone, with BlackRock forecasting annual spending could approach $1 trillion by 20302.
"Every $100 billion spent on new data centers will result in something like $100 billion spent on power over one decade of operation," analysts noted2. AI applications may consume 23 gigawatts of power by early 2025, potentially surpassing Bitcoin's notorious energy footprint3.
The International Monetary Fund projects AI-driven demand could generate cumulative emissions of 1.7 gigatons of CO2 from 2025-2030, equivalent to Italy's energy-related emissions over five years45. Under constrained renewable energy scenarios, U.S. electricity prices could increase 8.6 percent while global carbon emissions rise 1.2 percent4.
Tech companies face mounting pressure to reconcile climate pledges with AI ambitions. Microsoft's water consumption jumped 34 percent, largely driven by AI research cooling requirements1. Google cited AI as a factor in facing a "power capacity crisis" while expanding data center capacity2.
Some companies are exploring nuclear power and geothermal solutions, though these remain limited3. Geographic optimization and workload shifting to regions with renewable energy offer partial solutions, but "renewable energy simply isn't scaling fast enough to match AI's growth," said Eric Masanet, a sustainability researcher3.
The International Energy Agency suggests concerns about AI accelerating climate change may be "overstated," noting data centers will account for less than 1.5 percent of total energy-related emissions1. AI applications could potentially reduce global greenhouse gas emissions by up to 4 percent through energy optimization and efficiency gains2.
Yet this optimism faces a fundamental challenge. "AI is a tool, potentially an incredibly powerful one, but it is up to us – our societies, governments and companies – how we use it," said IEA Executive Director Fatih Birol1.
