Until now, experts have mostly guessed about the energy consumption from AI use. But this topic is of interest in the context of the fight against the climate crisis. This is reported by UNN with reference to MIT Technology Review and Vlaamse Radio TV.
Details
Google has finally broken its silence, providing data on the energy consumption of its Gemini models (Google's AI-powered assistant).
The average AI query in Google consumes 0.24 watt-hours. This figure may seem insignificant, but experts point out that in total, it's much more.
0.24 watt-hours doesn't seem like much. Indeed, it's comparable to a second of microwave oven operation, or an electric car driving a few meters, or a few minutes of Netflix viewing.
"But it's very relative, and it all depends on how you look at it," the expert noted, emphasizing that "queries are entered constantly, by everyone at the same time, throughout the day."
Trees may become more flammable due to climate change - study12.08.25, 17:24 • [views_3598]
The work can also be compared to such everyday examples:
- 8 seconds of TV viewing;
- 30 seconds of LED lighting (if 8 W);
- 15 seconds of microwave oven use at minimum power;
- one hour of Netflix viewing = 150-200 Wh = 625 to 830 AI queries
- 45 seconds of electric toothbrush use.
(lamp)
2 minutes of smartphone charging in standby mode
1 minute of digital alarm clock operation
Previously, precise data on energy consumption was something of a "black box." But Google's new report is clearly not accidental. The internet giant is sending a message to consumers:
Queries require less and less energy
Therefore - "act, because it's not that bad."
It should be noted that the study is quite broad, covering different periods. According to Google, the average query in May 2025 required 33 times less energy than in May 2024.
Google also calculated the climate impact:
The average query produces 0.03 grams of CO2.
Recall
UNN reported that the operational carbon emissions of leading technology companies increased by an average of 150% from 2020 to 2023. This is due to increased investment in artificial intelligence and data centers.