Casey Crownhart argues in The Download that Google’s AI energy figure is not the whole picture

Casey Crownhart argues in The Download that Google’s AI energy figure is not the whole picture

On August 28, 2025, Casey Crownhart argues in MIT Technology Review’s The Download that Google’s estimate for a typical Gemini app query should not be used to dismiss concern about AI’s energy demand. Google announced that a typical query uses about 0.24 watt-hours of electricity, which Crownhart compares to running a microwave for one second. Crownhart welcomes more openness from major AI players about estimated energy use per query, but says some people are drawing the wrong conclusion from this figure. The excerpt introduces Crownhart’s explanation of why, but does not include it. It therefore gives no basis for assessing what the estimate covers or how it relates to total energy demand. Crownhart’s point here is that a per-query estimate should not, by itself, lead readers to conclude that AI’s energy demand is nothing to worry about.

Publication: 56f15a0e-1a76-9ad9-2be0-3d095d95498e

attribution — pass: The reading attributes the argument to Casey Crownhart, whose byline is on the lead item. It places it in MIT Technology Review's The Download. The issue comments on items in its own words, so it is one piece, not a bare listing. The source notes the item comes from The Spark; naming The Download as where it appears is accurate.

Google’s still not giving us the full picture on AI energy use —Casey Crownhart

summary — pass: The reading names the actors (Crownhart, Google) and states her thesis. It explains the 0.24 watt-hour figure through her microwave comparison and keeps her pushback ('not the right takeaway'). It focuses on the lead item rather than the unrelated must-reads, which fits its title. It says plainly that the excerpt omits her full explanation. The question follows from that gap without leading the reader.

That’s not the right takeaway here. Let’s dig into why.

date — pass: The reading's date, August 28, 2025, matches the UTC day of originalPublished (2025-08-28T12:10Z). The modification timestamp is earlier than publication, so there is no later edit. Later-dated sidebar teasers, such as the 377 math findings, are neither used nor relied on. The reading uses no hindsight framing, and the publisher page needs no archive disclosure.

support — pass: All three claims are quoted verbatim from the source. The summary's statement that the excerpt gives no basis for judging what the estimate covers is an accurate limitation, not an invented claim. The reading adds no benefits, costs or forecasts beyond the source.

I welcome more openness from major AI players about their estimated energy use per query.

media — pass: No media was selected and no verified media candidates were supplied, so nothing is misidentified.