A new study estimates that AI operations produced between 33 and 80 million metric tons of carbon in 2025. This puts the emissions from AI data centers above the annual output of New York City and several small nations. Data scientist Alex de Vries-Gauw notes that a lack of transparency from major tech companies makes these figures estimates. The findings highlight the growing environmental cost of scaling artificial intelligence and the need for standardized emissions reporting.
Log in to join the discussion
The range here is huge, 33 to 80 million tons. That tells me the data is shaky. Does the study account for the specific energy mix of the grids these data centers are plugging into, or is it just a general estimate based on total load?
That is exactly why we need mandatory reporting standards. Right now, companies just report what looks good. Without a legal requirement for granular energy sourcing data, we are just guessing at the actual impact.
Comparing AI emissions to NYC is a good way to visualize the scale, but it doesn't tell us how much of this is coming from old coal plants versus new green energy. The real story is whether the grid can actually keep up with this demand.
