Capacity

Fab capacity isn't a single number like "how big is the building" — it's set by whichever tool set is the slowest step in the entire process flow, because a fab can only move wafers through as fast as its tightest bottleneck allows, no matter how much idle equipment sits everywhere else.

Capacity Is Set by the Bottleneck, Not the Average wafer flow through five process areas — pipe width = throughput capacity Deposition 12,000 wafers/mo Litho (EUV) 6,000 wafers/mo ← bottleneck Etch 11,000 wafers/mo Implant 10,500 wafers/mo Metrology 10,000 wafers/mo ↑ narrowest pipe The fab's real output is 6,000 wafers/month — set by litho, not the average Adding deposition or etch tools does nothing for total output until the litho bottleneck itself is relieved.

This is why capacity is quoted per technology node, not as one fab-wide number. EUV lithography scanners cost well over $150M each and have a hard limit on wafers-per-hour; a fab can have abundant deposition, etch, and metrology throughput and still be capped by however many EUV-capable layers its scanner fleet can expose in a month. Capacity planning in a real fab is fundamentally about identifying and expanding that one bottleneck tool set, not about uniformly scaling everything.

Capacity is also node-specific, which is why "50,000 wafers a month" is never the whole story. A 200mm line running mature 180–90nm processes and a 300mm line running advanced 65–28nm processes aren't interchangeable or additive in any simple way — a customer needing a 7nm chip can't be served by spare 180nm capacity no matter how idle that line is. That's also why foundries lean on partner capacity (TSMC, Samsung) for leading-edge nodes they don't run in-house themselves: capacity access at the node a customer actually needs is the real constraint, not total fab floor space.

When demand exceeds the industry's bottleneck capacity — as it did industry-wide during the 2020–2022 shortage — allocation, not price, becomes the rationing mechanism. Customers with long-term capacity agreements and reserved allocation get prioritized ahead of spot-market buyers, which is exactly why capacity reservation exists as a commercial product in its own right, not just a scheduling courtesy.

ChipFoundryServices operates 50,000 wafer starts per month across 200mm (30,000/month, 180nm–90nm) and 300mm (20,000/month, 65nm–28nm) fabs, with access to leading-edge 16nm–7nm capacity through foundry partnerships with TSMC and Samsung. Downstream, our packaging and test lines support 10M units/month wire bond, 1M units/month flip chip, and 10M units/month final test. We support everything from 5-wafer prototype runs to 10,000+ wafer/month production, with capacity reservation and long-term agreement options for customers who need guaranteed allocation.

Take capacity further

Ask the copilot about this term, or have our engineers assess it against your process.

Ask CFSGPT about this → Request a technical assessment → Open the CFS app