Raspberry Pi has been publishing life cycle assessments (LCAs) of its boards for years. The classical approach is straightforward: weigh each component, map it to datasets like ecoinvent, and sum up the embodied carbon. That works well for metals, plastics and passives.
For silicon chips, it turned out to be the wrong lens.
Why weight fails for semiconductors
A microcontroller or SoC isn’t just a lump of material. Its embodied carbon depends strongly on:
- process node (e.g. 40 nm vs 16 nm),
- die area,
- number of process steps,
- energy intensity of the fab,
- manufacturing yield.
Treating Raspberry Pi’s SoCs as generic semiconductor mass, mapped to a broad industry average, produced numbers the team felt were conservative but not correct – systematically overstating the footprint of the actual chips they ship.
MicroGreen and ACT: a better model for silicon
To fix this, Raspberry Pi teamed up with researchers at Cornell and Columbia, using MicroGreen, a tool built on top of ACT (Architectural Carbon Modeling Tool) from Harvard and Meta.
Instead of inferring carbon from weight, MicroGreen/ACT models semiconductor emissions from the parameters that really drive them:
- die size and process technology,
- yield and scrap,
- fab‑level emissions and energy mix.
Working directly with Udit Gupta, Ariel Goldner, and Xuesi Chen, the team tuned the model to the exact Raspberry Pi chips, not a generic dataset.
Researchers at Cornell Tech and Columbia note that applying MicroGreen to Raspberry Pi hardware “grounds our research in reality and dramatically increases the impact our work can have.”
The impact is substantial: by modelling silicon properly, Raspberry Pi’s overall product LCAs dropped by around 30%, with no hardware changes at all – just more accurate accounting.
For engineers, this matters because it turns carbon into a first‑class design metric, alongside performance and power, and ensures sustainability decisions are based on physics, not rough averages.
Source: Raspberry Pi
