The ground is sized on the breaker, not on the wire

When to use this: pulling a new circuit, upsizing a feeder for voltage drop, running parallel sets, or checking whether an existing ground is adequate for a breaker that has been changed.

The equipment grounding conductor carries no current in normal operation. It carries fault current once, for a fraction of a second, and the breaker only opens if the impedance of that path is low enough. Nothing appears to be wrong with an undersized ground until the single moment it matters.

Its size comes from Table 250.122, indexed on the rating of the overcurrent device ahead of the circuit. Not the load. Not the conductor size. The common shortcut — “one size smaller than the hots” — is right often enough to feel like a rule and is not one.

Table 250.122

There is no 30 A row, and that catches people out

The table jumps from 20 A to 60 A. Where your device rating is not listed you take the next row up, so a 30 A or 40 A circuit lands on the 60 A row and needs 10 AWG copper — not something between 12 and 10.

Older editions of the code did have 30 A and 40 A rows. Anyone who learned this table years ago may reach for a smaller conductor than the current code permits, which is a good reason to look it up rather than remember it.

250.122(B): upsize the hots, upsize the ground

This is the rule that separates a correct answer from a plausible one, and the reason this page exists. If the ungrounded conductors are increased in size for any reason, the equipment grounding conductor must be increased proportionally.

It matters because it is not a corner case. Our wire size calculator will tell you to upsize almost any run past about 100 ft — that is what voltage drop does. Every one of those runs also needs a bigger ground, and it is routinely forgotten.

Worked through, for a 100 A circuit upsized from 3 AWG to 3/0 for a long run:

Note how close that lands. The requirement is 52,649 cmil and 3 AWG is 52,620 — short by 29 circular mils. So the answer is 2 AWG, not 3 AWG. Rounding the ratio, or estimating “about three sizes up”, gets this wrong.

Do the increase by circular mils, not by counting AWG steps. Each AWG step is roughly a 1.26× change in area, so “go up two sizes” means 1.6× — which is sometimes generous and sometimes short, depending where on the ladder you started.

The cap that stops it running away

250.122(A) also says the EGC is never required to be larger than the circuit conductors it grounds. This matters most on motor circuits, where 430.52 permits a breaker sized for locked-rotor inrush far above what the conductors would otherwise take: a 100 A breaker ahead of 12 AWG motor conductors would call for 8 AWG from the table, and the cap brings it back to 12 AWG.

There is a quieter property worth knowing: on an ordinary voltage-drop upsize the cap never fires. Scaling the ground by the same ratio as the phase conductors cannot overshoot them.

Parallel runs: one full-size ground per raceway

250.122(F) requires a full-size equipment grounding conductor in every raceway of a parallel set, each sized on the single overcurrent device protecting the whole set. Three parallel runs on a 400 A breaker need three 3 AWG grounds — not one, and not a third of one each.

The reason is simply that a fault can occur in any of the raceways, and the ground in that raceway has to clear it alone. Dividing the ground between runs would leave each one unable to do the job on its own.

EGC or GEC? They are not the same conductor

This is the most common misunderstanding in Article 250, and the two are sized from different tables on different inputs.

The GEC stops growing because earth is not a fault-current path. Soil resistance is measured in ohms, so a ground rod passes a few amps at best — nowhere near enough to open a breaker. Earth stabilises voltage to ground and gives lightning somewhere to go. The conductor that actually clears faults is the EGC, running back to the source.

People who believe otherwise sometimes drive a rod at an outbuilding and omit the equipment ground, which leaves a fault with no low-impedance path at all and a breaker that will not trip.

Table 250.66 — grounding electrode conductor

The aluminium column of the table is indexed on aluminium service conductors, whose thresholds are larger than the copper ones — an aluminium service reads a different row than a copper service of the same circular-mil area. The calculator handles this; reading the table casually does not.

One exception worth remembering: 250.66(A) caps the conductor to a ground rod at 6 AWG, and 250.66(B) caps the conductor to a concrete-encased electrode at 4 AWG, whatever the table says. A bigger conductor into soil achieves nothing.

Details that change the answer

• Green is reserved. 250.119 allows green, green with yellow stripes, or bare for equipment grounding conductors and nothing else. IEC practice differs — plain green is not used there — so imported equipment may carry colours that mean something else.
• Conduit can be the ground. 250.118 lists RMC, IMC and EMT among permitted types. Many designers pull a wire anyway, because raceway continuity depends on every coupling staying tight for decades and cannot be inspected behind a wall.
• Aluminium has placement rules. 250.120(B) prohibits it in direct contact with masonry or earth, in a wet location within a raceway, and within 450 mm (18 in) of earth.
• The neutral is not a ground. Bonding them anywhere downstream of the service disconnect puts normal load current on metal enclosures. Subpanels get separated bars.
• Flexible metal conduit has limits. FMC and LFMC qualify as an EGC only in short lengths with appropriately sized overcurrent protection — 250.118(5) and (6). Beyond those, pull a wire.

Once the ground is sized, check the whole bundle fits: our conduit fill calculator counts the EGC like any other conductor, and the wire size calculator is where the upsizing that triggers 250.122(B) usually comes from.

How this calculator is checked

Technical reviewer

Salamot Hok

Electrician · 10+ years of installation work in Bangladesh and the wider South Asian region

He reads the result the way an installer would: are the defaults values people actually meet, does the warning fire where you would stop and think, and is the answer something you could buy and fit? The code figures themselves come from the published standards cited below, not from him — that boundary is set out on his profile.

• The maths lives in a pure function with its own test suite, asserted against worked examples from published references and standards. A calculator does not ship until those tests pass.
• 3 sources cited by name and linked, so any figure on the page can be traced back to the document it came from.
• Last reviewed 17 August 2026. Review dates are advanced only when the page is actually re-read, never to look fresh.
• Unusable input returns no answer. Where the inputs do not describe a real design, the calculator says so and withholds the number rather than printing a plausible-looking wrong one.

The full process is written up in the methodology and editorial policy. Results are engineering guidance, not a code sign-off — see the disclaimer. If a result looks wrong, tell us; corrections are answered before anything else.