Hidden Cost of Smart AI: 50× More CO₂ for a Single Question

Every time we ask an AI a question, it doesn’t just return an answer—it also burns energy and emits carbon dioxide.

German researchers found that some “thinking” AI models, which generate long, step-by-step reasoning before answering, can emit up to 50 times more CO₂ than models that give short, direct responses. These emissions don’t always lead to better answers, either.

AI Answers Come at a Hidden Environmental Cost

No matter what you ask an AI, it will always generate an answer. In order to do this, whether the response is accurate or not, the system relies on tokens. These tokens are made up of words or fragments of words that are transformed into numerical data so the AI model can process them.

That process, along with the broader computing involved, results in carbon dioxide (CO₂) emissions. Yet most people are unaware that using AI tools comes with a significant carbon footprint. To better understand the impact, researchers in Germany analyzed and compared the emissions of several pre-trained large language models (LLMs) using a consistent set of questions.

“The environmental impact of questioning trained LLMs is strongly determined by their reasoning approach, with explicit reasoning processes significantly driving up energy consumption and carbon emissions,” said first author Maximilian Dauner, a researcher at Hochschule München University of Applied Sciences and first author of the Frontiers in Communication study. “We found that reasoning-enabled models produced up to 50 times more CO₂ emissions than concise response models.”

Reasoning Models Burn More Carbon, Not Always for Better Answers

The team tested 14 different LLMs, each ranging from seven to 72 billion parameters, using 1,000 standardized questions from a variety of subjects. Parameters determine how a model learns and makes decisions.

On average, models built for reasoning produced 543.5 additional “thinking” tokens per question, compared to just 37.7 tokens from models that give brief answers. These thinking tokens are the extra internal content generated by the model before it settles on a final answer. More tokens always mean higher CO₂ emissions, but that doesn’t always translate into better results. Extra detail may not improve the accuracy of the answer, even though it increases the environmental cost.

Accuracy vs. Sustainability: A New AI Trade-Off

The most accurate model was the reasoning-enabled Cogito model with 70 billion parameters, reaching 84.9% accuracy. The model produced three times more CO₂ emissions than similar sized models that generated concise answers. “Currently, we see a clear accuracy-sustainability trade-off inherent in LLM technologies,” said Dauner. “None of the models that kept emissions below 500 grams of CO₂ equivalent achieved higher than 80% accuracy on answering the 1,000 questions correctly.” CO₂ equivalent is the unit used to measure the climate impact of various greenhouse gases.

Subject matter also resulted in significantly different levels of CO₂ emissions. Questions that required lengthy reasoning processes, for example abstract algebra or philosophy, led to up to six times higher emissions than more straightforward subjects, like high school history.

How to Prompt Smarter (and Greener)

The researchers said they hope their work will cause people to make more informed decisions about their own AI use. “Users can significantly reduce emissions by prompting AI to generate concise answers or limiting the use of high-capacity models to tasks that genuinely require that power,” Dauner pointed out.

Choice of model, for instance, can make a significant difference in CO₂ emissions. For example, having DeepSeek R1 (70 billion parameters) answer 600,000 questions would create CO₂ emissions equal to a round-trip flight from London to New York. Meanwhile, Qwen 2.5 (72 billion parameters) can answer more than three times as many questions (about 1.9 million) with similar accuracy rates while generating the same emissions.

The researchers said that their results may be impacted by the choice of hardware used in the study, an emission factor that may vary regionally depending on local energy grid mixes, and the examined models. These factors may limit the generalizability of the results.

“If users know the exact CO₂ cost of their AI-generated outputs, such as casually turning themselves into an action figure, they might be more selective and thoughtful about when and how they use these technologies,” Dauner concluded.

Reference: “Energy costs of communicating with AI” by Maximilian Dauner and Gudrun Socher, 30 April 2025, Frontiers in Communication.
DOI: 10.3389/fcomm.2025.1572947