multi-project wafer (mpw)

Multi-project wafer (MPW) is a cost-sharing service where multiple chip designs from different customers are placed on the same reticle, dramatically reducing prototyping and low-volume production costs. Concept: instead of each customer paying for a full mask set ($1-15M+ depending on node), designs are tiled together on shared reticles—each customer gets a fraction of the wafer's die. Cost structure: (1) Full mask set (dedicated)—$100K (mature) to $15M+ (leading edge); (2) MPW slot—$5K-$500K depending on area, node, and number of wafers; (3) Cost savings—10-100× reduction in prototyping cost. How it works: (1) Customers submit GDSII within allocated area (typically 1×1mm to 5×5mm); (2) Foundry aggregates designs on shared reticle (shuttle run); (3) Wafers processed through full flow; (4) After fabrication, wafers diced—each customer receives their die. MPW providers: (1) Foundries directly—TSMC (CyberShuttle), Samsung (MPW), GlobalFoundries; (2) Brokers—Europractice, MUSE Semiconductor, CMC Microsystems; (3) Academic—MOSIS (educational and research). Use cases: (1) Prototyping—validate design before committing to full production; (2) Low-volume products—small markets don't justify full mask set; (3) Test chips—process characterization, IP validation; (4) Academic research—university projects at affordable cost; (5) Startups—first silicon at minimal investment. Limitations: (1) Limited die count—dozens to hundreds, not thousands; (2) Shared schedule—run dates fixed by foundry; (3) Limited customization—standard process options only; (4) Longer turnaround—aggregation adds to schedule. MPW democratized access to advanced semiconductor processes, enabling startups, researchers, and small companies to fabricate chips that would otherwise be financially prohibitive.