How to calculate recording storage for an IP camera system — bitrate by resolution, what each recording mode actually saves, and the two places a correct calculation still leaves you short of the retention you promised.
Storage is bitrate multiplied by time. Everything else is a modifier on the bitrate.
Multiply by camera count and retention days for the total. Then read the two sections at the bottom of this page before you order drives, because that total is not the capacity you buy.
| Resolution | Bitrate | Per camera / day | 30 days |
|---|---|---|---|
| 720p | 2 Mbps | 21 GB | 0.63 TB |
| 1080p / 2 MP | 4 Mbps | 42 GB | 1.27 TB |
| 4 MP | 8 Mbps | 84 GB | 2.53 TB |
| 5 MP | 10 Mbps | 105 GB | 3.16 TB |
| 4K / 8 MP | 16 Mbps | 169 GB | 5.06 TB |
| 8 MP (high complexity) | 24 Mbps | 253 GB | 7.59 TB |
| 12 MP | 32 Mbps | 338 GB | 10.1 TB |
Scene complexity moves these more than any spec sheet admits. A camera watching a still corridor encodes far below its rated bitrate; the same camera pointed at a car park with moving foliage and headlights can sit at the ceiling all night. Size on the busy scene.
Most storage quotes assume continuous recording because it is the safe default. It is also the most expensive assumption in the design, and often the wrong one.
| Mode | Multiplier | 24× 4 MP for 30 days | vs continuous |
|---|---|---|---|
| Continuous | 1.0× | 60.7 TB | — |
| Scheduled (12 h/day) | 0.5× | 30.4 TB | −50% |
| Motion-triggered | 0.3× | 18.2 TB | −70% |
| AI analytics-filtered | 0.25× | 15.2 TB | −75% |
| Event / alarm only | 0.15× | 9.1 TB | −85% |
The AI row is the one worth explaining to a customer. Analytics-filtered recording keeps only footage where the system classified something meaningful — a person, a vehicle, a line crossing — rather than any pixel change. It cuts storage roughly like motion detection while removing the false triggers that make motion recordings unreviewable: rain, headlights, branches, shadows moving across a wall.
The 0.3× multiplier assumes roughly 30% activity. A loading dock during shift change is not 30% active — it can approach continuous, and the storage plan built on motion savings runs out weeks early. Where retention is contractual or evidentiary, size on continuous and treat any mode savings as headroom, not budget.
| Group | Qty | Resolution | GB/day ea. | 30-day total |
|---|---|---|---|---|
| Perimeter | 10 | 4 MP | 84 GB | 25.3 TB |
| Entrances | 6 | 4K | 169 GB | 30.4 TB |
| Interior | 8 | 1080p | 42 GB | 10.1 TB |
| Raw recording total | 24 | 65.8 TB |
65.8 TB is the footage. It is not the array. Read on.
Two reductions sit between raw storage and usable capacity, and skipping either is how a "30-day system" turns into a 22-day system in month two.
Applied to the example: 65.8 TB of footage in RAID 6 with 15% headroom needs roughly 103 TB raw — about 13 × 8 TB drives, not the nine a naive division suggests. Specify surveillance-rated drives; desktop drives are not built for continuous sequential write and fail early in this duty cycle.
H.265 typically halves the bitrate for equivalent quality, which is why the table above assumes it. The catch is decode load: some VMS platforms and older client workstations struggle to display many H.265 streams at once, so the saving on storage can reappear as a cost on the viewing side. Verify both ends support it before you size on it.
Proportionally, yes — the bitrate scales linearly with frame rate. Dropping 30 fps to 15 fps halves storage. For most surveillance the evidentiary value of 30 fps over 15 is minimal; for cash handling, gaming floors, and licence-plate capture it is not.
Retention, always — then buy to it. Retention is usually the contractual or regulatory number, and it is the one the customer will hold you to after an incident. Working backwards from an existing array tells you what you will get, not what you owe.
The bandwidth calculation replaces the capacity one: sustained upload has to carry the aggregate bitrate, which for the example above is roughly 200 Mbps continuous. Most sites cannot, which is why hybrid designs keep full-rate local recording and push only events or a lower-rate substream to the cloud.
Design Right AI computes retention, bitrate and array capacity from the cameras already in your design — resolution and recording mode per group, not one blanket assumption — and carries the result into the proposal alongside the bill of materials.