⚡ Power Clock — Counting Since Jan 1, 2026
The Real-Time Cost of AI Energy
Every AI query, every training run, every inference request consumes electricity. Watch it accumulate — live.
AI Electricity Consumed
in 2026 So Far
in 2026 So Far
Terawatt-Hours (TWh)
Global AI-specific data center energy consumption in 2026.
Growing 26% year-over-year.
Equivalent to powering all of Poland for a year.
Estimated Energy Cost
in 2026 So Far
in 2026 So Far
USD Billions ($B)
Electricity bill for global AI infrastructure.
~$1,015 per second.
Borne by hyperscalers, cloud vendors, and data center operators.
CO2 Equivalent
in 2026 So Far
in 2026 So Far
Million Metric Tons CO2e
Carbon footprint of AI power consumption.
Equivalent to 14 million gasoline cars
driven for a full year.
Water Consumed
in 2026 So Far
in 2026 So Far
Billion U.S. Gallons
Freshwater used to cool AI server farms.
~1.8 liters per ChatGPT conversation.
Microsoft alone used 700M gallons in 2023.
AI Queries Processed
in 2026 So Far
in 2026 So Far
Billions of AI Queries
Estimated queries across ChatGPT, Gemini, Claude and peers.
Each uses ~10x more energy
than a standard Google search.
U.S. Homes
Equivalent
Equivalent
Million U.S. Homes Powered
If redirected, annual AI electricity could power
every home in Illinois + Ohio combined.
And growing every year.
~5.07 MWh
AI electricity / second
+26% YoY
Annual AI energy growth rate
31% of all
Data center power that is AI
~158/sec
New AI queries, worldwide
Live Grid Data
Connecting to live data...
⚡ Wholesale Avg (Live ISOs)
--
$/MWh — spot market price
📍 Retail Avg (EIA Monthly)
--
¢/kWh — residential rate
🌫 CO₂ Intensity (National)
--
gCO₂/kWh — grid carbon intensity
💲 AI Query Cost (cloud avg)
--
per query at current grid price
⚡ Regional ISO Wholesale Prices ($/MWh)
ISO-NE
CT, MA, ME, NH, RI, VT
--
NYISO
New York
--
ERCOT
Texas
--
PJM
PA, OH, IL, NJ, MD, VA...
EIA est.
MISO
MN, IA, IL, IN, MI...
EIA est.
CAISO
California
EIA est.
💬 Energy Cost Per AI Query (at current grid prices)
🤖 ChatGPT (GPT-4o)--
🤖 Claude 3.5 Sonnet--
🤖 Gemini 1.5 Pro--
🏠 AI Standards On-Prem--
Energy cost only — does not include hardware amortization, labor, or API margin. On-prem runs at ~30% of cloud energy cost per token.