Artificial Intelligence

Study: AI Reasoning Models Emit 50X More CO2 Than Others

Published JUN 20 2025

Story

Above: App logos of AI companies. Image credit: Nicolas Economou/NurPhoto via Getty Images

The Facts

Researchers from the Munich University of Applied Sciences tested 14 large language models and found that reasoning-enabled AI models produce up to 50 times more carbon dioxide emissions than concise response models when answering questions.
EL PAÍS English
The Washington Post
Frontiers Science News
Their study, published in Frontiers in Communication on Thursday, revealed that reasoning models generate an average of 543.5 tokens per question compared to just 37.7 tokens for concise models and each additional token increases carbon emissions regardless of answer accuracy.
ScienceDaily
Independent
Questions requiring complex reasoning in subjects like abstract algebra and philosophy produced up to six times higher emissions than straightforward topics such as high school history.
Live Science
The New York Times
Data centers powering AI models consumed 4.4% of U.S. electricity in 2023 and are projected to reach 6.7% to 12% by 2028, with the four largest AI companies seeing a 150% increase in emissions since 2020.
Time
EL PAÍS English
Model choice significantly impacts emissions, with DeepSeek R1 requiring the same carbon footprint as a round-trip flight from London to New York to answer 600,000 questions. In comparison, Qwen 2.5 can answer 1.9 million questions with similar accuracy.
ScienceDaily
Independent
Engineering and Technology Magazine
A single ChatGPT query uses approximately 10 times more energy than a Google search, though experts question the transparency of energy consumption figures provided by AI companies like OpenAI.
Time
Wired

Sources Split

LEFT

CENTER

RIGHT

ESTAB
CRIT

CENTER

PRO
ESTAB

The Spin

Techno-skeptic narrative

The environmental toll of AI is staggering and accelerating at an alarming pace. As AI systems proliferate and become more sophisticated, they're consuming electricity at unprecedented rates, with data centers potentially accounting for 4-6% of global electricity by 2030 — equivalent to powering entire nations. The world is witnessing an environmental catastrophe unfolding in real time, disguised as progress.

Ctech

Tech Brew

Techno-optimist narrative

Despite AI's voracious energy appetite, this technology emerges as a paradoxical ally in the fight against climate change. AI systems reduce emissions by orders of magnitude compared to humans performing the same tasks. From optimizing fuel-efficient transportation routes to accelerating climate-resilient crop development, AI's computational prowess could ultimately prevent more emissions than it generates.