A research study by Resilio & partners

Environmental transparency in the AI supply chain

On the 21st and 22nd of June, The Geneva AI Summit 2027 will bring the global conversation on artificial intelligence to the heart of multilateralism.

To have sustainability at the center stage, we are setting up the largest benchmark of real-world AI usage, built with the ICT value chain to bring transparency on the true environmental impact of Artificial Intelligence

2027
PUBLIC RELEASE (GENEVA) - JULY
10
PARTNERS HAVE
JOINED THE PROJECT
SINCE MARCH 2026
ITU
USING THE L.1801 METHODOLOGY
GLO
GLOBAL REACH - A PLATFORM TO FEDERATE THE ICT SECTOR
THE OBJECTIVE

Why this study exists

01

Corporates are flying blind.

Companies adopting AI cannot quantify its environmental impact (as information is disconnected and fragmented), nor anticipate how it will affect their global carbon emissions and decarbonisation targets (SBTi, CSRD, etc.).

02

Existing data is disconnected from reality.

Public references's data (Ecologits, AI Energy Score, academic LCAs) come from lab benchmarks, not real deployments . They miss the actual energy, hardware, and infrastructure footprint of AI at scale.

Our approach

Three steps, from raw signal to usable model

01

Collect

Real-world usage data from companies across the AI value chain (model providers, infrastructure, cloud, and end-users).

02

Analyse

Resilio's R&D team processes the data using established LCA methodology (ITU's L.1801 framework, GHG Protocol, ...).

03

Model

A parametric model any organisation can use to estimate and plan the environmental footprint of their AI usage.

+
Our contribution

In deployment

We measure marginal impact and real consumption from actual production workloads, across the full stack.

Hardware
Training
Infrastructure
Scope of the study

Assessing the full ICT value chain

01

AI Hardware

Hardware supporting computation (GPU/TPU/ASIC), network and storage.

Who we are looking for
Hardware ManufacturersOEMsSemiconductor industry
02

Training AI models

Training across foundation and task-specific models, fine-tuning for specific purposes.

Who we are looking for
AI model providersResearch institutions
03

AI inference

Individual traces from real deployment workloads. Marginal impact metrics.

Who we are looking for
Cloud & data centre operatorsPrivate Cloud
04

End-user AI applications

How AI integrates into products and workflows in real production. Agentic loops, real applications, ...

Who we are looking for
AI-deploying enterprisesCTOs / CSOs
Project journey

The story so far

First Wave · CH and EU companies
  1. March 2026
    Q1 2026
    • Project kick-off
    • First partner onboarded
  2. July 2026

    First results

    Initial data and validation results

  3. Road to AI Summit LondonJUNE 2026· Event at the Swiss EmbassySwiss embassy coat of arms
Second Wave · Worldwide assessment
  1. Road to AI Summit ParisNOV 2026
  2. Jan 2027
    Q1 2027

    Pre-release to partners

  3. June 2027
    AI SUMMIT GENEVA 2027

    Public release

Why join

What partners get from the study

01

Shape the methodology

Contribute to the reference benchmark for AI environmental impact.

02

Get early access

Partners receive the parametric model and findings before public release.

03

Position your organisation

Recognised contributor at the Geneva AI Summit 2027.

04

Get free support

AI Tech companies can get free support to assess the impact of their AI services.

Partners & contributors

The consortium

Founding contributors across the AI value chain, from Frontier Models, Data Center providers to Large entreprises and Governments

Institutional partners

3 organisations

Academic partners

2 organisations

Private industry

6 organisations

About Resilio

Measuring the environmental footprint of the ICT sector

Resilio is a Swiss company specialising in assessing the environmental footprint of the digital sector, working with stakeholders across the value chain. From CIOs, Telecom and Cloud providers, to OEMs and Semiconductor companies.