Skip to content

Analytics

The numbers behind the index

Reported figures are solid bars; projections and estimates are lighter. Hover any bar for the basis and scope.

How to read the labels:Company figureCompany figure. Published by the company itself. We take it as given and record the scope next to it.Measured independentlyMeasured independently. Measured or estimated by researchers or analysts with no stake in the result.ecogpt.co estimateecogpt.co estimate. Our own estimate from published factors. The working is shown next to the number; treat it as an order of magnitude, not a measurement.DerivedDerived. Arithmetic on a disclosed or independent input using a published conversion factor.Not disclosedNot disclosed. The company publishes nothing on this and no defensible estimate exists.— hover any label.

Global data-centre electricity

IEA base case — 2024 reported, 2030 and 2035 projected

International Energy Agency

US data centres, share of national electricity

Berkeley Lab — 2028 shown as the low and high end of its range

Lawrence Berkeley National LaboratoryCongressional Research Service

Where 2024 demand sat

Share of global data-centre electricity

International Energy Agency
  • United States45%
  • China25%
  • Europe15%
  • Rest of world15%

Index score distribution

All companies, current edition

Google (arXiv 2508.15734)Mistral AIInside Climate News

Energy per prompt (log scale)

Green = company-disclosed · teal = independent estimate. Long inputs dominate: a 100k-token prompt costs ~160× a median one.

Google (arXiv 2508.15734)Sam Altman (blog)Epoch AILuccioni, Jernite, Strubell (arXiv 2311.16863)Hugging Face, Salesforce, Cohere, Carnegie Mellonheise online

Water per prompt (log scale)

Scope explains the 170× gap: Mistral counts water used to generate the electricity, Google counts cooling only

Google (arXiv 2508.15734)Sam Altman (blog)Mistral AI

Fleet water-use effectiveness

Litres of water per kWh of IT energy — lower is better

AmazonMetaMicrosoftGoogle (arXiv 2508.15734)Koomey AnalyticsLawrence Berkeley National Laboratory

AI chips in service, by company

Company statements and filings are solid; Epoch AI H100-equivalent estimates and EcoGPT order-of-magnitude estimates are lighter. Hover for the unit.

Network WorldTechCrunchData Center DynamicsSemiAnalysis

Carbon baked into the hardware

Thousand tonnes of CO₂e emitted making each fleet's chips and servers, at 514 kg per accelerator slot — before the first prompt

arXiv 2607.05408NVIDIA developer blogNetwork World

Clean electricity share

Green = company-disclosed or independently measured · teal = estimate. Units differ (annual matching vs hourly) — hover for the scope.

GoogleTencentiTiger (Alibaba ESG report)IBMData Center Dynamics

Reported company emissions

Latest reported year, millions of tonnes CO₂e — boundaries differ, so compare trends rather than bars

Data Center DynamicsESG DiveMetaGoogleAxiosData Center DynamicsTencent