Capacity Allocation
Capacity allocation is the process a foundry uses to divide a fixed quantity of wafer output among competing customers in a given period, and it becomes the decisive factor whenever total demand exceeds what the fab can actually produce, which is most of the time at the leading edge.
Contracts, not fairness, set the order. A signed take-or-pay capacity agreement is honored first because breaking it carries legal and reputational cost; strategic accounts with a long production history and shared demand forecasts come next; spot-market orders, placed without any standing commitment, sit last and absorb nearly all of the actual shortfall when supply runs tight.
The 2020-2022 global chip shortage made this visible in public. Automakers who had historically ordered on the spot market found themselves at the back of TSMC's and Samsung's allocation queues behind smartphone and PC customers with standing volume agreements, and several automakers responded by signing their own direct capacity agreements afterward specifically to avoid being last in line again.
| Customer tier | Typical commitment | Shortage-era treatment |
|---|---|---|
| Take-or-pay | multi-year volume floor, often prepaid | protected, served first |
| Strategic / long-term | multi-quarter forecast sharing | protected, served second |
| Spot market | order-by-order, no standing commitment | absorbs almost all shortfall |
| New / unqualified customer | no production history at that fab | last in line, longest lead times |
Allocation decisions are also political, not purely mechanical. A foundry weighing two equally sized orders will favor the customer more likely to sign next year's capacity agreement too, which means a company's allocation priority is really a function of its expected lifetime value to the fab, not just its current order size.
Being allocated capacity is not the same as having wafers in hand. Allocation sets a customer's place in the queue for a given process step and tool set; actual delivery still depends on yield, cycle time, and whether upstream steps hit their schedule, so a well-allocated customer can still see slipped lead times if the fab itself is behind.
Read capacity allocation through a *queue-priority* lens rather than a *fairness* lens: the fab is not deciding who deserves wafers most, it is protecting the contracts that let it justify building the next module, which is why a take-or-pay agreement is worth more during a shortage than its stated price ever suggests, and why spot-market customers are the first ones to feel every capacity crunch in this industry.