The short answer is 100 meters, about 328 feet, for the entire channel from the switch port to the device. That figure includes the patch leads at both ends, which is the part people forget.
Why the limit exists
It is not arbitrary and it is not a safety margin somebody invented. Ethernet over copper depends on the receiving end hearing the signal clearly enough to reconstruct it, and the signal degrades with distance. Past roughly that length, error rates climb.
Which is why exceeding it does not produce a clean failure. It produces an intermittent one. The link comes up, works fine on a mild day, and then starts dropping under load or when the temperature changes. Diagnosing that from the symptom is miserable, which is why it is worth avoiding at design time.
The four honest ways around it
Put a switch in the middle. An intermediate cabinet resets the 100 meter clock, and this is the usual answer in a large building. It needs power and somewhere lockable to live.
Run fiber for the long leg. Fiber does not have the same distance problem, and a media converter or a fiber-capable switch at the far end takes it back to copper for the last stretch. This is the usual answer for a separate building, a gate, or a yard.
Move the equipment. Sometimes the rack is simply in the wrong place, and relocating it turns three marginal runs into three comfortable ones.
Use a wireless link. Appropriate for a detached structure with line of sight and no practical route for cable. It introduces a dependency worth being deliberate about rather than casual.
What is not an honest way around it
Running it long and hoping. Extenders stacked on extenders. Joining two cables in the middle of a run with a coupler in a ceiling void where nobody will find it in three years.
Where this bites on camera jobs
Gates, yards, detached garages, the far corner of a warehouse, and anything across a parking lot. These are exactly the positions that matter most on a security system, so they are worth resolving on the drawing rather than on the day.
Measure the route, not the map
The distance that matters is the path the cable actually takes, and it is always longer than the straight line on a drawing. Cable goes up into a ceiling, along a joist line rather than diagonally, around a duct, down a wall, and then leaves slack at each end for termination.
A run that measures sixty meters on a floor plan can easily be ninety in practice. That is the difference between comfortable and marginal, and it is why estimating from a drawing is where the error usually enters.
Then subtract the patch leads. A three meter lead at the rack and a two meter lead at the device both count toward the same hundred meters, which is the part that catches people who correctly measured everything else.
The practical habit is to design to about ninety meters and treat anything beyond that as a run needing a decision, rather than aiming at the limit and hoping the route cooperates.
Long runs get resolved during design at Elite Security, because the alternative is finding out in year two. More like this in the blog, and low-voltage cabling is where to start.