Silicon Respin Cost

A silicon respin is a second (or third) pass through mask fabrication and wafer processing after first-silicon testing reveals a functional bug that software or firmware cannot work around, and its cost is dominated by two things happening at once: a brand-new mask set (or at minimum, the specific layers that changed) has to be fabricated from scratch, and the schedule slips by however long that fabrication and subsequent wafer processing takes, typically several months. This is the realized, worst-case outcome that extensive pre-tapeout verification exists specifically to avoid, and it is the reason a respin is treated as a serious program-level event rather than a routine engineering fix, unlike a firmware patch that can ship the next day.

A Bug Found in Silicon Restarts the Clock everything after the bad decision point has to happen again Design finalized Mask set fabricated First wafers processed First-silicon test bug found here bug requires a mask change — everything downstream restarts New mask set fabricated again, from scratch Wafers processed again Re-test, hopefully pass Cost = a full new mask set + wafer re-processing + several months of lost schedule Everything before the design-finalized step does not have to repeat — everything after it does. This is exactly the "mask re-spin" cost flagged in the previous row's verification warning

Not every bug forces a full respin, which is why engineering teams triage first-silicon failures before committing to the expensive path. A bug that can be worked around in firmware, disabled via a feature fuse, or avoided by operating the part in a narrower voltage/frequency range can often ship as-is or with a minor workaround, at a fraction of a respin's cost; only a bug that is fundamentally unfixable without changing the silicon itself, a broken state machine, an incorrect logic gate, a timing path that cannot close, forces the full mask-and-wafer cycle to repeat.

First-silicon testing finds a functional failure
the critical triage moment for the whole program
Triage: can this be fixed without changing silicon?this decision determines whether a respin happens at all
Workaroundable in firmware/fuses
ship with a software fix — no respin needed
Requires a silicon-level fix
no workaround exists — a respin is unavoidable
New mask(s) fabricated, wafers re-processed, re-tested
full mask-set cost and multi-month schedule hit repeats

This is why pre-tapeout verification spending, however large it looks in isolation, is consistently cheaper than the respin it is trying to prevent, since a respin combines the entire mask-set cost covered in the previous row with a multi-month schedule delay that can cost a product its market window entirely, not just its development budget, making the verification-versus-respin trade a case where spending more earlier is almost always the cheaper path in expectation.

Failure typeFix pathRelative cost
Firmware-correctable bugSoftware patch, no hardware changeLow — ships on original schedule
Fuse/configuration workaroundDisable or reconfigure the affected featureLow to moderate — some feature loss
Silicon-level logic/timing bugFull mask and wafer respin requiredHigh — full mask cost plus months of schedule slip

Silicon respin is the realized downside that mask-set-cost economics and pre-tapeout verification spending both exist to avoid — it is the single most expensive way a chip program can discover that a bug was not caught early enough.

Take silicon respin cost further

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