Kerf Loss

Kerf loss is the silicon area destroyed, not just divided, when a wafer gets cut into individual die — the width of material the saw blade, laser, or plasma stream actually removes from every scribe street on the wafer. It is pure waste: area that holds no transistors before cutting and holds nothing at all afterward, so every micron of kerf width is area a designer could otherwise have used to make dies, or the streets between them, physically smaller:

Reticle floorplan
die size + scribe street width set at design time
Finished wafer
dies separated only by scribe streets
Dicing
blade saw, laser, or plasma etch
Kerf loss
cut width removed from every street
Singulated die
ready for pick-and-place

The street has to be wider than the cut, not equal to it. A scribe street's design width is kerf width plus an alignment margin on each side, because the saw or laser never tracks a perfectly straight line to zero tolerance — a mechanical diamond blade typically removes 20-40 micrometers per pass and needs several more micrometers of margin for blade wander, while a laser or plasma process can narrow that to single-digit micrometers, which is exactly why advanced packaging lines have shifted toward laser and plasma dicing as die sizes shrink and every micron of street width becomes a larger fraction of total wafer area.

Kerf Loss: the Street vs. the Cut a cross-section through one scribe street die area die area scribe street kerf margin margin street width = kerf + 2 margins 10mm x 10mm die, area lost to the street alone 30µm street ≈ 0.6% of die-cell area lost 60µm street ≈ 1.2% of die-cell area lost doubling street width roughly doubles that loss, die after die, across the whole wafer narrower kerf, narrower street, more die per wafer — the whole reason packaging houses pay for finer dicing

Thinner wafers make kerf loss worse, not better. As package height targets push wafer thickness down toward 50 micrometers or less for stacked and fan-out packages, a dicing blade or laser that chips or cracks the die edge becomes far more likely, so packaging houses have been moving toward narrower-kerf laser grooving and full plasma dicing specifically to keep edge damage under control on thin wafers, not just to reclaim street area — the two benefits happen to point the same direction.

Dicing-equipment suppliers compete directly on kerf width and edge quality. Disco Corporation is the dominant supplier of precision wafer dicing saws and laser systems used across the industry, Tokyo Seimitsu and ASMPT build competing blade and laser dicing platforms, and plasma dicing tools from suppliers like Panasonic avoid mechanical cutting altogether by etching through the wafer along the scribe streets; foundries and OSATs set the minimum scribe-street width in their design rules specifically to match whichever dicing method a given package type requires.

Read kerf loss through a *street-width budget* lens rather than a *cutting detail* lens — every micrometer a dicing process shaves off the street width is a micrometer a designer gets back to shrink the street itself, and at advanced nodes with die counts in the thousands per wafer, that reclaimed margin adds up to real die-per-wafer economics, not a rounding error.

Take kerf loss 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