AI Compass
Compass

Energy and sustainability

What a request actually consumes, how to compute it, and which figures hold up in sustainability reporting.

·2 min read·By Fachredaktion Technik
DETAIL
3 sections

The idea

Every request makes a graphics card work briefly. Consumption is power times time. A 400-watt card working for one second uses 400 watt seconds, about 0.11 watt hours. If it serves twenty requests concurrently in that second, each accounts for roughly 0.0056 watt hours.

Batching, which lowers cost, therefore lowers consumption per request in the same proportion.

Orders of magnitude

OperationOrder of magnitude
Web search~0.3 Wh
Short answer from a mid-sized model0.1 to 1 Wh
Long answer from a large model2 to 10 Wh
One generated image1 to 5 Wh
One kilometre by car~1,900 Wh

Measuring it yourself

  1. 01

    Record power draw

    Through the card management interface, averaged over the measurement window, not as a peak.

  2. 02

    Measure throughput at the same time

    Requests and tokens per second in the same window.

  3. 03

    Convert to per request

    Power divided by throughput gives energy per request.

  4. 04

    Add the overheads

    Processor, power supply, cooling. The data centre PUE covers this.

Energy per request

E [Wh] = (P / R) · PUE / 3600

Power divided by throughput gives energy per request, multiplied by the data centre factor.

P
mean card power draw in watts
R
requests per second
PUE
data centre efficiency factor, typically 1.1 to 1.6

Worked through: P = 400 W, R = 12 requests per second, PUE = 1.3: E = (400/12) × 1.3 / 3600 = 0.0120 Wh per request. At a million requests a day that is 12.0 kWh per day.

From energy to emissions

Emissions

CO₂e [g] = E · I

Consumption times the emission factor of the electricity actually drawn gives operational emissions.

E
energy consumption in kWh
I
emission factor of the electricity mix in g CO₂ equivalent per kWh

The factor I varies substantially: below 50 g per kWh in some European grids and above 400 g in others. The same service can therefore be eight times more emission-intensive depending on location. Reporting must state the location-based factor, not a European average.

What holds up in a report

  • Measured energy consumption of your own equipment, not extrapolated from vendor figures.
  • The PUE of the data centre actually used, with a reference period.
  • The location-based emission factor, with source and year.
  • For bought-in services: the provider's figures, explicitly labelled as such.
  • Training and operation reported separately.

Not defensible: estimated per-request consumption from third-party publications, vendor peak power figures used as operating values, and any figure without a utilisation statement.

The most effective levers

LeverOrder of magnitude
Increase batching5 to 20 times per request
Smaller model for easy requests5 to 20 times on the affected share
Quantisation to int8~2 times
Cap prompt and answer lengthLinear
Site choice by electricity mixUp to 8 times on emissions

The last row changes emissions, not consumption. It therefore belongs reported separately and must not be mixed with efficiency measures.

Was this page helpful?
Energy and sustainability