Upstream: Extraction, processing, and manufacturing

Upstream, silicon's ore-to-metal ratio improves substantially across the historical record. The model assumes this ratio remains broadly stable throughout the projection period. Quartz quality nevertheless varies considerably. While some producing regions have deposits exceeding 98% purity, many rely on lower-grade quartz that requires additional beneficiation before refining.¹⁴⁸ Chart 37 shows that ore input continues growing throughout the projection period, meaning the absolute scale of extraction—and the energy demand associated with subsequent processing—continues to expand through to 2050. This reflects patterns associated with metallurgical-grade silicon prior to further processing. Metallurgical-grade silicon (98–99% purity) must undergo extensive purification before reaching the 6N purity required for solar photovoltaics or the 9N purity needed for semiconductor wafers.

The more distinctive upstream dynamic concerns where silicon loses material after refining. Chart 38 shows absolute upstream losses resulting from the assumptions on upstream efficiency built into the model at each stage. Losses at semi-fabrication are large and grow steadily across the projection period, while manufacturing-stage losses remain almost negligible throughout. This distribution maps directly onto silicon's purity cascade: semi-fabrication is where metallurgical-grade silicon is purified into polysilicon and where polysilicon is converted into wafers, with both stages generating significant material losses.^{149 150 151} Kerf loss during wafer slicing alone produces substantial quantities of fine silicon powder, while purification steps discard material that does not meet increasingly stringent specifications.¹⁵² Manufacturing, by contrast, largely assembles already-purified material into finished devices and therefore generates comparatively little further loss.¹⁵³

The chart shows losses are concentrated at the semi-fabrication stage, a pattern that is particularly important for silicon because it identifies where material efficiency improvements and circularity interventions are likely to have the greatest impact. Vertical integration across ingot, wafer, cell and module production can allow scrap and off-specification material to be recovered directly as feedstock,¹⁵⁴ while process improvements, such as machine-learning-based monitoring and diamond-wire sawing, can reduce kerf loss during wafer production.¹⁵⁵ Together, these approaches address both recovery and prevention of material losses.

This concentration of losses also has important energy implications. Material lost during semi-fabrication has already absorbed the energy required for furnace reduction and, in many cases, partial purification. Each tonne of kerf powder therefore represents not only lost material but also energy already embedded in the production process.