Not a single study
Story checked
AI Agents Are Thirsty for Power | WIRED (opens in a new tab)
wired.com · 2026-09-13
Short answer
Not a single studyNo clear answer.
The article covers a research area rather than reporting a specific new paper, so there is no single study to check it against.
Checked against the study summary. The full text wasn't available, so some details couldn't be settled either way.
Share this check
The story
AI Agents Are Thirsty for Power | WIRED
wired.com · 2026-09-13
The story’s checkable claims.
Read the original story (opens in a new tab)NewsLink checks it
This story isn't reporting one study.
The article covers a research area rather than reporting a specific new paper, so there is no single study to check it against.
Open claim evidenceSearch result
No single source study.
The article covers a research area rather than reporting a specific new paper, so there is no single study to check it against.
Evidence layer
Claim by claim
What the story asserts. With no source study, there is nothing to check these against.
Reading mode
Scan verdicts. Open evidence only when needed.
Claim 1 of 8Not checkedAI agents, rather than simple chatbot queries, are increasingly central to Silicon Valley’s AI strategy and are part of what is driving the power buildout for new data centers.View evidenceHide evidence
Not checked
NewsLink could not identify one source study, so this claim was not compared against a paper.
Claim 2 of 8Not checkedAgentic systems can make hundreds of small prompts or run for hours on a task, making them much more energy-intensive than single chatbot queries.View evidenceHide evidence
As statedhundreds of small prompts; run for hours
Not checked
NewsLink could not identify one source study, so this claim was not compared against a paper.
Claim 3 of 8Not checkedOpenAI said a swarm of more than 10,000 agents sending 2.7 million messages solved a longstanding math problem, though mathematicians pushed back on the company’s claims.View evidenceHide evidence
As statedmore than 10,000 agents; 2.7 million messages
Not checked
NewsLink could not identify one source study, so this claim was not compared against a paper.
Claim 4 of 8Not checkedThe article says there is a major dearth of reliable information about the energy use of AI agents, and that the power demands could vary enormously depending on the task.View evidenceHide evidence
As stateda massive gulf in power use
Not checked
NewsLink could not identify one source study, so this claim was not compared against a paper.
Claim 5 of 8Not checkedClimate scientist Zeke Hausfather calculated that his average daily Claude session may consume more energy than two refrigerators, but he said this still is not a world-ending amount and is a net new source of emissions.View evidenceHide evidence
As statedmore than the energy needed to power two refrigerators
Not checked
NewsLink could not identify one source study, so this claim was not compared against a paper.
Claim 6 of 8Not checkedMeta recently rolled out a personal AI agent called Muse that the company says is built to work for billions of people worldwide and will keep a dedicated cloud computer for each user.View evidenceHide evidence
As statedbuilt to work for billions of people worldwide
Not checked
NewsLink could not identify one source study, so this claim was not compared against a paper.
Claim 7 of 8Not checkedThe article says a future in which many people use agents could help explain massive data-center projects such as Hyperion in Louisiana, which it says will be powered by 10 natural gas plants.View evidenceHide evidence
As stated10 natural gas plants
Not checked
NewsLink could not identify one source study, so this claim was not compared against a paper.
Claim 8 of 8Not checkedIn response to a reader question, the article says small modular nuclear reactors could be a good carbon-free option for data centers, but none are operating commercially in the US and only one model has been licensed for sale.View evidenceHide evidence
As statednone operating commercially in the US; one model licensed for sale
Not checked
NewsLink could not identify one source study, so this claim was not compared against a paper.
Method layer
No single source study was found.
NewsLink still surfaces nearby research so you can inspect the field instead of a single paper.
Nearby research
No exact source, but these papers are close.
PubMed, Crossref, Europe PMC · 15 candidate papers
ELISA (Embedding-Linked Interactive Single-cell Agent): an interpretable hybrid generative Artificial Intelligence agent for expression-grounded discovery in single-cell genomics.
Briefings in Bioinformatics · 2026 · PubMed
Advancements in Multi-Agent Large Language Model Systems for Next-Generation AI
Advances in Computational Intelligence and Robotics · 2025 · Crossref
Small Language Models for Developing Agentic AI in Healthcare: A Comprehensive Systematic Review and Critical Analysis.
2026 · Europe PMC
Transforming large language models into medical specialists via knowledge injection.
Cell Reports. Medicine · 2026 · PubMed
An open-weight Large Language Model agent for condition monitoring and predictive maintenance in wind energy systems
Energy and AI · 2026 · Crossref
And 10 more candidates considered.