Electricity for manufacturing AI chips
Electricity consumed to manufacture AI chips from Nvidia (2023) then from Nvidia and AMD (2024), in GWh
Electricity for manufacturing AI chips
The analysis
- The electricity used to manufacture AI chips was multiplied by 4.5 in one year (+351%), partly because AMD only enters the calculation in 2024.
- It is consumed mostly in Taiwan and South Korea, on grids that are still very carbon-intensive.
- This is a bottom-up study by an NGO: the only public estimate of this item.
The figures
- 2023 : 218 GWh.
- 2024 : 984 GWh.
Assumptions
- Manufacture of silicon wafers (compute chips and memory) for Nvidia's AI chips in 2023, for Nvidia and AMD in 2024 (5 families).
- 2023 counts only Nvidia, 2024 adds AMD: the increase mixes real growth and a change of scope.
- Excluding the giants' in-house chips (Google's TPUs) and generic server components.
- Emissions calculated using the local grid only.
Continue with hardware
The whole Hardware topic →Charts and maps
Chart
HardwareLife cycle assessment of Mistral Large 261%for materials, the manufacture of servers and chipsChartHardwareManufacture of an AI server, by component42%of manufacturing: memory, the largest itemChartHardwareTraining emissions, 4 published assessments20timesthe emissions of GPT-3, for Llama 3.1ChartHardwareThe footprint of digital devices99%of a smartphone's footprint comes from its manufacture
HardwareLife cycle assessment of Mistral Large 261%for materials, the manufacture of servers and chipsChartHardwareManufacture of an AI server, by component42%of manufacturing: memory, the largest itemChartHardwareTraining emissions, 4 published assessments20timesthe emissions of GPT-3, for Llama 3.1ChartHardwareThe footprint of digital devices99%of a smartphone's footprint comes from its manufactureOn the Projet Celsius blog ↗
WaterAcross the hub
AI Impact DatabaseAll the figures, their methodology and their sources.
AI Ranking10 assistants rated out of 100: performance, ethics and environment.
AnalysesThe signed pieces: analyses, viewpoints, "Where does this figure come from?".
Charts and mapsYou are hereThe reference visuals, with their data and sources.
ToolsOpen calculations, your assumptions on the left, the result on the right.
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® ↗
