Multi Project Wafer Service
Multi-project wafer service, universally abbreviated MPW, is a cost-sharing fabrication model where multiple chip designs from different, entirely unrelated customers get tiled onto the same mask set and processed together on the same wafer — each customer paying only for the fraction of the reticle field their specific design actually occupies, rather than the full cost of a dedicated mask set that only their own design would use. This cost-sharing structure is exactly what makes advanced-node access economically possible for startups, universities, and small design teams who could never otherwise justify the cost of a dedicated tape-out on their own.
The fixed cost of an advanced-node mask set — commonly running somewhere between roughly one million and twenty million dollars depending on node and mask count — is essentially identical whether one customer's design occupies the entire reticle field or just a small fraction of it, which is precisely the economic fact that makes cost-sharing across many unrelated customers mathematically compelling. Because that mask cost doesn't scale down proportionally with how little reticle area any single customer actually needs, splitting it across ten, twenty, or even fifty-plus simultaneous projects converts an individually prohibitive fixed cost into something each participant can genuinely afford — exactly the arithmetic that makes MPW viable as a business model in the first place.
Die allocation on an MPW run is sold in units ranging from a single square millimeter up to full reticle-field allocations, which lets customers with very different actual silicon-area needs all participate in the same shared run while each paying roughly proportional to the area they actually consume. This granular allocation flexibility is exactly what makes MPW practical across such a wide range of customer sizes simultaneously — a university research group needing only a tiny test structure and a startup needing a much larger prototype chip can both participate in the identical MPW run, each purchasing exactly the die allocation their specific project requires.
Every design sharing a given MPW run gets processed through the exact same standard process design kit and identical process flow, with zero customer-specific process accommodation, because introducing any customer-specific variation would immediately undermine the entire cost-sharing premise that makes MPW economically viable. This uniform-process requirement is a genuine constraint customers have to design around — an MPW participant can't request even a minor process customization specific to their own design, since doing so for one customer on a shared run would either have to apply to every other customer sharing that same wafer, or would require breaking the shared-run model entirely.
Delivery from an MPW run gives each customer specifically their own die from the shared wafer — diced individually, tested, or delivered still on-wafer depending on what that customer purchased — which requires the foundry or MPW service provider to precisely track which physical die locations on the shared wafer belong to which customer throughout dicing and packaging. This per-customer die tracking and separation is a genuinely nontrivial logistics requirement layered on top of the shared fabrication itself — the cost-sharing benefit at the wafer level has to be followed by careful, accurate separation at the post-fabrication stage, ensuring each customer receives exactly their own die and nothing belonging to any other customer sharing that same run.
MPW access is specifically what enables startups, university research groups, and small design teams to reach advanced process nodes that would otherwise be entirely out of reach given the mask-cost economics of a dedicated tape-out, and the resulting ability to run multiple MPW tape-outs per year at manageable cost directly supports faster design iteration than a single, far more expensive dedicated run would allow. This iteration-speed benefit compounds the direct cost-sharing advantage: not only does MPW make any single advanced-node tape-out affordable for a small team, it makes running several tape-outs across a year affordable too, which is exactly the kind of rapid design-iterate-refabricate cycle that research and early-stage product development specifically depend on.