Home / Indexes / Technical Performance / AI Energy & Water per Query
Thank you for reading this post, don't forget to subscribe!
Technical Performance · Index
AI Energy & Water Use per Query 2026
A single AI text prompt uses roughly 0.2–0.3 watt-hours of electricity — far less than the alarming numbers that circulated in 2023, but the fleet-level total is what’s exploding. Every figure is linked to its primary source and dated.
Key takeaways
- ~0.24 Wh per prompt. Google’s Aug 2025 report puts a median Gemini text prompt at 0.24 Wh; Sam Altman cited ~0.34 Wh per ChatGPT query; Epoch AI estimates ~0.3 Wh for GPT-4o. (Google / Altman / Epoch)
- Water is tiny per query, big at scale. Google reports ~0.26 mL per prompt; the totals add up across billions of queries. (Google)
- The macro number that matters. The IEA puts data-center electricity at ~415 TWh in 2024, rising to ~945 TWh by 2030 — roughly a doubling, with AI the main driver. (IEA)
- “Methodology wars.” Per-query estimates vary 10×+ depending on model size, PUE, grid mix, and whether one-time training is counted — treat any single figure with caution.
0.24 Wh
~945 TWh
0.26 mL
Per-query energy estimates compared
The headline figures cluster around 0.2–0.3 Wh — close to a Google web search — but an older, widely-circulated estimate of 2.9 Wh per query (from 2023–24) is roughly 10× higher. The gap is the “methodology war” in a single chart: Google’s Gemini at 0.24 Wh, Epoch’s GPT-4o estimate at ~0.30 Wh, and Sam Altman’s ~0.34 Wh figure for ChatGPT all sit near a reference Google search (~0.30 Wh), while the older estimate towers over them.
| Source | Watt-hours per query |
|---|---|
| Google Gemini (2025) | 0.24 |
| Epoch GPT-4o est. | 0.30 |
| ChatGPT — Altman | 0.34 |
| Google search (ref.) | 0.30 |
| Older 2023–24 est. | 2.90 |
“Each query is cheap; the fleet is not. The real story is data-center demand more than doubling by 2030.”
Why the estimates disagree so much
Model size & routing. A small distilled model costs a fraction of a frontier model; “per query” hides which model answered. Boundary choices — some figures count only the GPU; others add cooling, networking, idle capacity, and data-center overhead (PUE). Training vs inference — training is a large one-time cost; amortizing it across queries changes the per-query number dramatically. And grid & location: a query on a coal-heavy grid emits far more CO₂ than the same query on hydro — even at identical energy use.
The numbers in full
| Metric | Value | Source & date |
|---|---|---|
| Energy, median Gemini text prompt | 0.24 Wh | Google, Aug 2025 |
| Water, median Gemini text prompt | 0.26 mL | Google, Aug 2025 |
| Energy, ChatGPT query | ~0.34 Wh | Sam Altman, Jun 2025 |
| Energy, GPT-4o query (est.) | ~0.3 Wh | Epoch AI |
| Data-center electricity, 2024 → 2030 | ~415 → ~945 TWh | IEA, 2025 |
| GPT-3 training emissions (one-time) | ~552 tonnes CO₂ | Patterson et al., 2021 |
Frequently asked
How much energy does one AI prompt use?
Current vendor and research estimates cluster around 0.2–0.3 watt-hours for a typical text prompt — Google reports 0.24 Wh, Sam Altman cited ~0.34 Wh, and Epoch AI estimates ~0.3 Wh. That’s comparable to a Google search and far below older 2.9 Wh estimates. (Google / Altman / Epoch)
If each query is small, why the concern?
Scale. The IEA projects data-center electricity roughly doubling to ~945 TWh by 2030, with AI the main driver. Billions of queries plus training and idle capacity add up even when each query is cheap. (IEA)
How much water does an AI prompt use?
Google reports about 0.26 mL per median Gemini text prompt — a fraction of a millilitre per query, though the totals add up across billions of queries and during training. (Google, Aug 2025)
Cite this page
The AI Index (2026). AI Energy & Water Use per Query 2026. Retrieved Jun 20, 2026, from report-ai.org/indexes/technical-benchmarks/ai-energy-water-per-query/
Related: Technical Performance · Compare year over year · Large Language Model · Tokens
On this page
Primary sources
- Google — technical report, Aug 2025 · 0.24 Wh · 0.26 mL per prompt
- IEA — Energy & AI, 2025 · 415 → 945 TWh by 2030
- Altman / Epoch / Patterson — ~0.34 Wh · ~0.3 Wh · GPT-3 training CO₂
10×
Gap between current ~0.24 Wh estimates and the older 2.9 Wh per-query figure — the “methodology war” in one number.
Share