Skip to content
ChartHardwareOrganisations · % of manufacturing emissions
Published on 20 October 2026

Manufacture of an AI server, by component

Nvidia HGX H100 board with 8 GPUs, 1,312 kgCO2e, share of each component

Manufacture of an AI server, by component

02040Memory42 %Integrated circuits25 %Cooling18 %Assembly9 %Electromechanical4 %Other3 %
Source: Nvidia, HGX H100 Product Carbon Footprint. projetcelsius.com/en/ai · CC BY 4.0AI Impact, by Projet Celsius

The analysis

  • Memory accounts for 42% of manufacturing: it is the largest item, ahead of the compute chips.
  • The next generation, the HGX B200, weighs 2,274 kgCO2e, 1.7 times more, of which 49% for memory.
  • Over its lifetime, an AI server emits far more through its electricity than through its manufacture; the order is reversed for metals.

The figures

  • Memory: 42 %.
  • Integrated circuits: 25 %.
  • Cooling: 18 %.
  • Assembly: 9 %.
  • Electromechanical: 4 %.
  • Other: 3 %.

Assumptions

  • Manufacturing, from cradle to factory gate, excluding use.
  • A data sheet published by the manufacturer itself.
  • The complete server (processors, storage, chassis) and the building are not included.

Published on 20 October 2026MethodologyAll sourcesLicence CC BY 4.0

A question about a figure?

An error, a source to add, a methodology question, a press request: write to us. Guillaume Pakula, Clément Reynaud, Sébastien Pierfederici and Alexandre Veilly will reply.

Write to us →

Projet Celsius, which runs this hub, also helps companies measure and reduce the footprint of their AI use.

Accounting for AI in your Bilan Carbone® ↗
Projet Celsius, environmental strategy consultancy

We support companies that account for AI in their Bilan Carbone®

Visiter Projet Celsius ↗