Amazon is breaking all kinds of records on the back of the AI boom, and not the good kind.
According to new reporting by Axios, Amazon’s AI buildout has fueled an unprecedented leap in the company’s emissions, a black mark on a corporation that pledged to reach net-zero carbon emissions by 2040 not even a year ago.
In its newly released environmental report, Amazon disclosed that its total electricity use jumped by a whopping 34 percent across 2025, fueling a 16 percent rise in greenhouse gas emissions. That’s a new record for the corporation, which noted that its data centers also hoovered up nearly 2.5 billion gallons of water last year.
Meanwhile, carbon emissions from “indirect sources” — corporate speak for “manufacturing supply chains” — were up 10 percent compared to last year, and over 21 percent compared to 2019.
Notably, the data table disclosing all of this information is tucked into the 46th page of the 51 page document. Amazon does a good job of trying to pre-empt these findings, with buzzy initiatives aimed at fluffy targets, like a pledge to combat the rise of “super pollutants.”
“Some pollutants are far worse for the climate than carbon emissions,” the report boldly declares — a statement clearly meant to downplay the company’s overall record-breaking carbon footprint.
Though Amazon brags that the pledge aims to commit $100 million to eliminate super pollutants, it fails to disclose how much of that funding will actually come from Amazon. First signed back in March, the super pollutant pledge is a collective greenwashing effort by corporations like Google, Figma, Salesforce, and JPMorgan, and there’s no telling how much or how little Amazon will contribute.
Reading the report, it’s clear that initiatives like these represent an effort by Amazon to appear sanctimonious while silently ratcheting up pollution for its own gain. That kind of corporate greenwashing isn’t new for Amazon, but the immense scale of AI-driven pollution certainly is.
More on Amazon: Amazon Investigating Its Own Employees for Daring to Question AI Data Centers