// Decision Engine 03

Size your power budget
for edge AI deployments

Input device count, PoE class, UPS runtime requirement, and cooling approach. The engine calculates total power draw, recommends PSU sizing, UPS capacity, and PoE switch spec.

// Define requirements

01
Edge AI Platform
// Primary inference device (determines per-unit power draw)
Coral TPU — 2W
RK3588 — 10W
Hailo-8 — 8W
Jetson Orin Nano — 10W
Jetson AGX Orin — 40W
Custom / Other — 25W
02
Device Count
// Number of inference nodes in the deployment
1
2
4
8
16
32
// Total system power scales linearly with device count
03
Camera / Peripheral Power
// PoE camera or peripheral power per node
No PoE cameras
PoE (802.3af) — 15W/cam
PoE+ (802.3at) — 30W/cam
PoE++ (802.3bt) — 60W/cam
USB camera — 5W
04
UPS Runtime Requirement
// How long must the system run on battery backup?
No UPS required
5 minutes (graceful shutdown)
30 minutes (short outage)
2 hours (extended outage)
8 hours (full shift)
05
Thermal / Cooling Environment
// Installation cooling context
Passive / fanless enclosure
Forced air (fan-cooled)
Ventilated cabinet / rack
Outdoor / high ambient (>40°C)
// Select all parameters to continue
// Calculating power envelope…

// Power Budget Recommendation

// Power Budget Summary
SIZING CONFIDENCE
← Hardware Selector

// Power bill of materials

// Estimated power draw by component
ComponentPer Unit (W)CountTotal (W)

// Infrastructure recommendations

// machine-readable output — application/json

    

What this Power Budget Planner decides

This tool estimates the electrical envelope of an edge AI deployment based on five decision inputs: platform class, device count, camera or peripheral load, UPS runtime target, and cooling environment. It is designed for engineers sizing PSU headroom, PoE budgets, backup runtime, and thermal overhead before finalizing deployment infrastructure.

// Inputs considered
01
Platform + Count

The selected edge compute platform and number of deployed nodes define the base system load.

02
Peripheral / PoE Load

Camera and peripheral power adds directly to the total deployment draw and often determines switch budget requirements.

03
Runtime + Cooling

UPS runtime targets and thermal environment change the practical PSU size, backup requirement, and thermal margin needed in production.

// How recommendations are scored

The engine calculates platform load, peripheral power, and cooling overhead to estimate total system draw. It then applies headroom for peak load and startup behavior, recommends a PSU class, estimates required UPS capacity for the requested runtime, and suggests PoE switch capacity when camera power is present.

  • Base compute-device power draw by platform class
  • Total deployment power as device count scales
  • Camera and PoE contribution to overall system load
  • Cooling overhead for passive, forced-air, cabinet, and outdoor conditions
  • PSU sizing with deployment headroom and UPS runtime planning
// What the output includes
  • Total system draw: combined power of compute, cameras, and cooling overhead
  • PSU size: recommended power-supply class with headroom
  • UPS capacity: required watt-hours for the selected backup runtime
  • PoE switch recommendation: switch class sized for camera power and port count
  • Infrastructure notes: operational guidance for outdoor, passive, or backup-heavy deployments
  • Machine-readable JSON: a structured result for copying, sharing, or downstream reuse
// Worked examples
// Example 01
Small fanless deployment
1 low-power device, no PoE cameras, no UPS, passive cooling → a compact PSU is often enough, but thermal margin still matters in sealed enclosures.
// Example 02
PoE-heavy camera edge node
8 devices with PoE+ cameras and 30-minute runtime → camera load often dominates total system draw, which pushes both PSU and UPS sizing higher than compute load alone suggests.
// Example 03
Outdoor high-power deployment
Multiple AGX-class devices outdoors with extended backup runtime → total power, heat, and UPS capacity increase quickly, requiring larger supplies and environmental planning.
// Example machine-readable output
{
  "schema": "edgeaistack/power-budget/v1",
  "inputs": {
    "platform": "jetson_nano",
    "device_count": 4,
    "poe_class": "poe_15",
    "ups_runtime": "30min",
    "cooling": "forced"
  },
  "computed": {
    "platform_load_w": 40,
    "camera_poe_w": 60,
    "cooling_overhead_w": 5,
    "total_system_w": 105,
    "psu_spec_w": 150,
    "ups_capacity_wh": 55
  },
  "recommendations": {
    "psu": "150W DIN Rail PSU (24VDC)",
    "ups": "APC BR1500G Back-UPS Pro 1500VA",
    "poe_switch": "Netgear GS308PP (8-port PoE+)"
  }
}
// FAQ

Why is PSU size larger than calculated system draw?

Production PSU sizing should include headroom for startup surges, peak load, power-supply aging, and future expansion. A PSU sized exactly to average draw is usually too tight for real deployments.

Does passive cooling change the recommendation?

Yes. Passive cooling affects sustained operating feasibility because enclosure temperature and thermal dissipation become practical deployment constraints. The planner includes cooling overhead for this reason.

When is a UPS required?

A UPS is required when the system must survive short or extended outages, or when controlled shutdown is necessary to protect local storage, system state, and uptime expectations.

Does PoE load matter more than compute load in some systems?

Absolutely. In camera-heavy deployments, peripheral and PoE draw can exceed the power required by the inference device itself, which is why total system planning must include both.