Brightcon 2025, hackathon & courses in Grenoble and online

Semiconductors are essential to the development of information and communication technologies (ICT). This industry is characterized by rapid innovation and industrial secrecy, with state-of-the-art manufacturing processes kept confidential and proprietary. From an environmental perspective, the semiconductor industry is also a resource- and energy-intensive sector, with a significant footprint due to the use of ultrapure materials and high energy consumption. As a result, the integrated circuits (ICs) that contain semiconductors emerge as environmental hotspots in ICT life cycle assessments (LCAs), making it increasingly urgent to understand and mitigate their impacts. Still, LCA practitioners face significant challenges in modelling ICs due to limited access to data.

One critical example is the estimation of semiconductor die area in ICs, which accounts for most of the embodied impacts of ICT products. However, this parameter is hidden from practitioners due to a lack of transparency in IC datasheets. To access it, practitioners must rely on destructive and time-intensive reverse engineering techniques (such as chemical decapping or X-ray imaging), which are rarely available in applied settings. As a result, they often fall back on heuristics or rules of thumb to quickly estimate this sensitive variable, introducing uncertainty in LCA results.

To address this issue, we have developed the Open Integrated Circuits Database (OICDB), an open-source, open-contribution dataset designed to support more accurate estimation of die area in ICs, especially for practitioners with limited means. Using this dataset, we apply advanced predictive models that estimate die area based on limited IC features, enabling practitioners to balance accuracy with time and resource constraints.

In this contribution, we will discuss:

How does industrial secrecy in the semiconductor industry limit the development of accurate, representative LCA models?

The practical advantages and limitations of using OICDB, with a focus on the trade-off between uncertainty and modelling effort in ICT LCAs.