PoE switches are sold with a total power budget number: 370W, 520W, 740W. Most deployments treat this number as the limit they’re working within. The problem is that the nameplate budget and the power actually available to connected devices are not the same number — and the gap is large enough to cause deployment failures.
Where the power goes
A PoE switch draws power from its power supply and uses it for two things: running the switch itself, and delivering power to connected PDs (powered devices). The switch’s own consumption — processors, memory, fans, uplink ports — comes out of the power supply budget before any PoE delivery happens.
A 370W switch might consume 30–50W for its own operation, leaving 320–340W for PoE delivery. This is not hidden information, but it is frequently not factored into deployment calculations that simply divide the nameplate budget by the number of ports.
Per-port maximums and how they interact
IEEE 802.3af delivers up to 15.4W per port. 802.3at delivers up to 30W. 802.3bt delivers up to 90W. These are the standards, but the actual power delivered to a device — the power at the PD, not at the PSE — is lower after cable loss. A device rated for 25.5W consumption may require 30W from the switch port to actually receive 25.5W after the cable.
This means sizing based on device consumption ratings rather than port delivery maximums will produce an undersized budget.
The calculation that fails
The common approach: take the number of cameras (or APs, or VoIP phones), multiply by their rated power consumption, check that it’s under the switch budget. This fails because: it uses device consumption rather than port delivery requirements. It doesn’t account for switch self-consumption. It assumes all ports will never simultaneously draw maximum power — which is fine for diverse mixed deployments but wrong for uniform high-power deployments. And it doesn’t include headroom for power spikes during device startup, which can be 20–30% above steady-state draw.
The calculation that works
Start with the switch’s actual PoE delivery capacity — typically 85–90% of nameplate budget after self-consumption. For each port, use the IEEE maximum for that port’s standard, not the device’s rated consumption. Sum the per-port maximums for all ports you intend to actively use. Add 20% headroom for startup peaks and future capacity. If the result exceeds the switch’s PoE delivery capacity, the switch is undersized for this deployment.
For a 48-port deployment with 802.3at cameras averaging 15W consumption: the correct sizing assumes 30W per port delivery requirement, 48 ports × 30W = 1440W delivery requirement. A 740W switch delivers approximately 650–680W after self-consumption. This deployment needs either two switches or a higher-budget single switch — not a 740W unit.
The budget number to ask for
These are different numbers, and suppliers who can’t distinguish between them are not the right partners for a deployment where PoE reliability matters.
The calculation is not complicated. The failure mode is treating it as simpler than it is.