Rebound parameterisation for consequential LCA decision support in imec.netzero

Context

New GPU generations and other efficiency improvements can reduce the energy required per AI inference. However, in practice, lower cost and higher performance can also increase demand (more utilisation, faster adoption, shorter replacement cycles, or new use cases). This demand response - often referred to as rebound - can partly or fully offset the expected environmental savings. Within the SSTS program, imec.netzero is developing a decision-support layer based on consequential life cycle assessment (CLCA) to make these trade-offs explicit.

This work supports real-world decisions in imec.netzero by making it possible to test “what-if” scenarios (e.g., GPU upgrades) with a consistent, documented representation of demand response, so expected savings are not overstated.

Objective

The objective of this internship is to build a practical, transparent rebound representation that can be used in imec.netzero CLCA studies. This will be achieved by:

• Reviewing and synthesising literature on rebound effects relevant to digital services, ICT, and efficiency-driven technology transitions.
• Defining a small set of reusable rebound channels (e.g., increased service demand, higher intensity/quality, faster adoption, shorter replacement cycles, and constraint effects).
• Translating the evidence into a compact parameter library (clear definitions, units, plausible ranges, and uncertainty notes).
• Packaging these parameters into a set of standard demand-response scenarios (e.g., fixed / low / medium / high) that can be directly used in CLCA calculations.

Responsibilities

You will actively engage in the evidence gathering and modelling of rebound scenarios, and in preparing implementation-ready inputs for the imec.netzero CLCA decision layer. This will involve working closely with the sustainability research team to ensure correct interpretation of CLCA concepts, as well as with the developers to ensure the outputs can be integrated smoothly.

In practice, this will involve:

You will turn literature findings into a small set of parameters and scenarios that can be used directly in CLCA runs.

• Extracting quantitative values and definitions from key sources and logging them in a structured template.
• Harmonising assumptions (scope, boundaries, units) so parameters are comparable and reusable.
• Drafting short, clear documentation explaining what each parameter means and when it applies.
• Testing the scenario set on a simple illustrative case (e.g., a GPU upgrade example) to confirm the workflow is usable.

Skills and Learning Objectives:

Applicants are expected to have a general background in engineering, environmental assessment, data science, or a related field. Experience with structured data handling (Excel/CSV) is required; Python is a strong advantage. During the internship, you will gain proficiency and enhance your skills in the following key areas:

• Consequential LCA and decision-support modelling
• Rebound effects and scenario design under uncertainty
• Transparent parametrisation and documentation practices
• Working with researchers and developers to translate concepts into tool-ready inputs