bloomberg.com web signal

Amazon Emissions Rise 16%, Google's Climb 18% on AI Buildout

TL;DR

  • Amazon emitted about 81 million metric tons of CO2e in 2025, a 16% increase from 2024 and 58% above its 2019 baseline.
  • Google reported an 18% jump in greenhouse gas emissions, the largest annual increase the company has published to date.
  • Amazon's emissions from purchased electricity rose 34%, tied to data center growth and delivery-network electrification.

The two most quoted numbers out of this week's disclosures land in the same direction and in the same order of magnitude, which is what makes them worth sitting with. Amazon's carbon footprint rose about 16% year over year in 2025, to roughly 81 million metric tons of CO2 equivalent, according to Bloomberg's reporting. Google, on the same rough beat, reported an 18% jump, the largest annual increase it has published.

The interesting part is where the pressure is coming from. On the Amazon side, emissions from purchased electricity rose 34% in 2025, which the company ties to data center growth alongside fuel used by its delivery operation. Google's own environmental report points the finger at manufacturing AI hardware, chips and servers in particular, on top of the electricity that data centers consume once those chips are racked. Same underlying story told from two different vantage points.

For a practitioner leaning on either cloud, the practical version of this is that the sustainability numbers your procurement team quotes from a hyperscaler's site are increasingly a moving target. Amazon's total is now 58% above its 2019 baseline, the same year it pledged to hit net zero by 2040. Google, which has framed its own climate targets as 'ambition-based' and used the word 'moonshots' in the latest report, has softened the language in the same direction the numbers are moving, while still holding a 2030 net-zero date.

The honest caveat is that these are single-year prints, filtered through the companies' own reporting. What the coverage does not give you is a clean split between AI-specific load and everything else inside these data centers, or how the two firms' large carbon-credit portfolios would look under stricter location-based accounting rules. Take the specifics as reported, not settled.

The direction is the part worth watching. If Amazon and Google are both moving 16 to 18 points the wrong way in a single year while the AI buildout is still early, the interesting bets shift toward the vendors that make either the electrons cheaper or the inference cheaper. Clean-power developers, cooling and efficiency startups, and independent carbon-accounting auditors all get a bigger role, whether or not the hyperscalers' deadlines survive contact with reality.