Intelligence

ai

The ultimate carbon cost of a ChatGPT query

Source
arXiv — Computers and Society
Published
Last verified
19 Aug 2026
Confidence
High
Evidence
Original document retained
Reading time
1 min
Country
International
Relevant to
Research & Evidence, Finance & Investment, Risk & Compliance, Technology & Data

Executive summary

What happened, and why should leadership care?

Recent research from arXiv analyzes the carbon cost associated with each query to a Large Language Model (LLM), such as ChatGPT. It estimates an environmental cost of approximately $0.4 USD per query for future generations due to environmental disruptions, corresponding to emissions around 10 gCO2eq/query. The primary variable impacting this cost is the number of tokens processed per query, which can range significantly.

Why this matters

Why is this strategically important?

This research highlights the environmental footprint and future financial liabilities associated with the increasing adoption and usage of AI, particularly large language models. Understanding these costs is critical for developing sustainable technology strategies and integrating environmental considerations into AI development and deployment frameworks.

Key insights

What should be noted from the evidence?

  • A single LLM query (e.g., ChatGPT) incurs a carbon cost estimated at $0.4 USD for the future human population, primarily due to environmental disruptions.
  • This financial cost corresponds to approximately 10 gCO2eq of emissions per query.
  • The total number of tokens computed per query is the most significant variable, leading to a wide range of potential costs from $0.012 to $1.20 per query.

Evidence and confidence

How far can this assessment be trusted?

High confidence. Named institution, original document retained and analysis corroborated.

Analysis is prepared editorially by Aziz Shuaib Ausi. The original publication remains the authoritative record, and executive judgement remains entirely human.

Source

Where does this originate?

Reported by arXiv — Computers and Society · International. This briefing summarises the publication for executive use; the document itself is not reproduced here.

Read the original publication