What is kerf? And why it is not your part spacing.

Every cutting process destroys some material. The slot it leaves behind is the kerf. It decides how accurately parts come out and how close together they can sit, but those are two different jobs, handled in two different places: one in CAM, one in the nesting layout.

What kerf actually is

When a beam, an arc or a bit follows a programmed line, it does not cut a line. It cuts a slot with a width, and that width is taken from your material, roughly half on each side of the path. How wide the slot is depends on the tool (nozzle, lens, bit), the material, its thickness, power and feed rate. CutLaserCut, a UK laser cutting service, puts CO2 laser kerf at roughly 0.08 mm and up, growing with material thickness, and notes a detail worth knowing: curved paths cut slower than straight ones, so the kerf on curves comes out slightly wider. There is no universal number. There is a range per process, and a measurement you can do yourself.

Typical kerf widths, with sources

Two rows deserve a flag: the fiber and CO2-on-metal figures come from Accurl, a machine manufacturer, and are marketing material rather than independent measurements. The plasma range is better grounded, because manufacturers publish cut charts per material thickness and amperage. Hypertherm's charts for a Powermax45, as quoted in a PlasmaSpider thread, span roughly 0.6 mm (0.024") on 22 gauge up to about 1.4 mm (0.056") on 4.8 mm (3/16") plate. If you own the torch, the chart in your manual beats any number on the internet.

Measure your own kerf in five minutes

Charts get you close. Calipers make the number yours, because kerf shifts with your machine, your consumables and the exact material batch. The test:

1. Cut a small test square, nominal 20 × 20 mm.
2. Measure the cut-out piece. Say it comes to 19.82 mm.
3. Kerf = nominal minus measured = 0.18 mm.

Repeat the test when you change material, nozzle, lens or speed. All of them move the number, and the five minutes are cheaper than a batch of parts that measure wrong.

Kerf compensation is a CAM job, done per part

If the machine cuts exactly on the drawn outline, every part comes out a full kerf too small across each dimension, because the slot eats half a kerf on each side. The fix is an offset: the toolpath moves half a kerf to the waste side, outward on outer contours, inward on holes, so the finished part measures nominal. That offset is applied by your cutting software, per part: LightBurn calls it kerf offset, SheetCAM applies it through the tool's kerf setting, router CAM calls it cutter compensation. It belongs there because it depends on the toolpath, which only the CAM knows.

Nesting spacing is the other number

Spacing is a layout decision, not a per-part correction. It answers a different question: how far apart do parts sit so that two cuts never merge, the web between them survives, and heat has somewhere to go. That distance must contain the kerf (half from each neighbouring cut) and then some. The rules of thumb per process, with sources, are in the part spacing guide; the short version is that plasma wants webs of at least one full kerf, lasers want about one material thickness, and routers need room for the bit plus tabs.

The mistake to avoid is typing one of these numbers into the other field. Kerf compensation with the spacing value makes parts the wrong size. Spacing set to the kerf value puts cuts so close that the webs burn away. Keep them separate: kerf lives in CAM, spacing lives in the nest.