By Abbas Nazil
A new scientific study warns that the rapid expansion of artificial intelligence infrastructure could generate carbon emissions equivalent to putting up to ten million additional cars on the road by the end of this decade.
Researchers publishing in Nature Sustainability estimate that AI servers deployed across the United States could emit between 24 and 44 million metric tons of carbon dioxide annually by 2030.
The projected emissions stem from the explosive growth of generative AI tools now embedded in search engines, productivity software, creative platforms, and business operations.
Scientists from Cornell University and international partners caution that this surge could undermine years of progress toward emissions reduction goals.
The study also highlights a massive hidden cost beyond carbon emissions in the form of water consumption.
AI data centers are expected to use between 731 and 1,125 million cubic meters of water each year, an amount comparable to the annual household water use of up to ten million Americans.
These findings arrive as major technology companies publicly commit to achieving net-zero emissions targets by 2030.
According to the researchers, those pledges may be unrealistic without heavy reliance on carbon offsets, many of which remain uncertain or controversial.
The environmental burden of AI is driven largely by energy-intensive computing processes required to train and operate large language models.
Training a single advanced AI model can emit more than 626,000 pounds of carbon dioxide, roughly equal to the lifetime emissions of five cars.
Once deployed, these systems consume even more energy as they respond to millions of user queries every day.
Inference operations, which occur every time a user interacts with AI, account for roughly 60 percent of total AI energy use.
A single AI-powered query requires nearly ten times more electricity than a standard internet search.
If AI systems were to fully replace global search activity, electricity demand would rival that of millions of people combined.
Data centers already contribute up to 3.7 percent of global greenhouse gas emissions, surpassing even the aviation sector.
Nearly 40 percent of a data center’s electricity is used solely for cooling systems that prevent servers from overheating.
The environmental impact is intensified by construction demands, including massive quantities of steel, concrete, and specialized equipment.
The study suggests that relocating data centers to regions with abundant renewable energy and lower water stress could significantly reduce damage.
Midwestern states with strong wind resources emerge as potential solutions, though major infrastructure investment would still be required.
Even under optimistic renewable energy scenarios, millions of tons of emissions would remain by 2030.
Efficiency improvements alone, the researchers warn, will not offset the scale of AI growth underway.
