Standardized cost-per-unit benchmarking for semiconductor manufacturing.
Raw wafer cost conflates manufacturing performance with factor-cost geography and equipment vintage. Reference Plant re-costs each process module at common reference factor prices — isolating how a fab is run from where it sits and how old its tools are — for like-for-like comparison across fabs, technology generations and regions.
Methodology originated at Infineon Technologies and generalized across manufacturing verticals.
Why raw cost-per-unit is not comparable
Raw cost ranks by geography, not by performance
Two fabs running the same 28 nm process at different sites can differ several-fold on raw wafer cost, driven predominantly by labour and electricity prices rather than manufacturing performance. Direct comparison of raw cost is therefore not meaningful.
At reference factor prices the ranking inverts: the higher-yield, newer-equipment fab is the lower-cost operation once labour and energy geography is removed.
Raw € / good wafer
Standardized € / good wafer
Source: 28 nm logic; reference factor prices €35/h, €0.12/kWh. Productivity indicators vary 3–5× across comparable fabs: Leachman & Hodges (1996). Synthetic peer set · calibrated to public benchmarks.
How standardization works
How standardization works
01
Re-cost the physical activity
Take the plant's reference route — labour-minutes, energy-kWh, equipment and material euros per wafer for its product — the part that is physics, not price.
02
Price it at one reference
Value that same activity at reference factor prices (€35/h labour, €0.12/kWh energy). Local labour and energy context is removed.
03
Read the normalization bridge
A waterfall from actual to standardized €/unit shows exactly how much of the gap is geography and how much is equipment vintage.
Operational KPIs, modelled with realistic correlations
Die yield
% good dies / wafer
Defect density D₀
defects / cm²
Equipment OEE
A × P × Q
Cycle time
days / mask layer
WIP turns
throughput / WIP
DPPM
ppm shipped
Calibrated to public benchmarks
The peer set is synthetic — generated, not measured — but it isn't invented. Distributions, correlations and factor-price ranges are calibrated to the canonical public literature, so the numbers behave the way a fab benchmarking expert expects.
Your real data would never leave your control
For paying customers we build client-side encryption: your KPIs are encrypted in your browser before storage, keyed from a secret we never see. This demo stores only synthetic data — the architecture is laid out in full.