32-camera Frigate NVR build
Frigate Hardware for 32 Cameras: CPU, AI and Storage
Thirty-two cameras are no longer a casual mini-PC project. Frigate can support large installations, but the design needs deliberate decode capacity, detector throughput, memory, storage, network switching, power and cooling. The most important variable is still simultaneous activity. A warehouse with constant motion is very different from thirty-two mostly quiet perimeter views, so this guide sizes from workload rather than declaring one fixed workstation mandatory.
Quick answer
What this Frigate workload needs
Use a modern Core-class or workstation-capable host with at least 16GB RAM and usually 32GB or more when enrichments and other services are active. Hardware decoding is mandatory in practice. For detector load, Frigate's guidance treats 12+ active-camera scenarios as high activity and 40+ commercial installations as candidates for modern discrete GPUs, so a 32-camera design should evaluate multiple accelerators or a capable GPU.
Live Amazon hardware
Current products that fit this decision
These listings come from this sprint's dedicated Amazon catalogue. Barebones systems, accessories, camera kits, enclosures, replacement batteries, unrelated product classes and obvious multipacks are excluded by the normalizer.
Buying decision
Match the hardware to the bottleneck
At thirty-two cameras, optimize for reliability and serviceability. A SFF tower or compact workstation with proper PCIe expansion, cooling and storage connectivity can be better than an ultra-small mini PC. Use a managed PoE switch with generous power budget, model multi-drive storage, and size the UPS around cameras and networking as well as the server.
Interactive decision tool
32-Camera Frigate Build Planner
Use the workload inputs to size a practical Frigate hardware tier before comparing live products. Results are planning guidance; verify current Frigate support, camera bitrates and the exact hardware configuration before deployment.
Compatibility checklist
Four checks before you purchase
Use hardware video decoding
A detector does not decode camera streams. Verify Intel, AMD or NVIDIA hardware acceleration separately.
Size by simultaneous activity
Busy overlapping cameras can require more detector throughput than a larger but quiet camera count.
Use supported detector hardware
Compare Hailo, OpenVINO, supported GPUs and other current Frigate detector paths before buying.
Calculate storage from bitrate
Retention depends on recording bitrate, duty cycle and days kept—not a fixed TB-per-camera rule.
Thirty-two cameras require activity-based detector sizing
A 32-camera site may have only a few detections at once, or it may have constant overlapping motion across entrances, aisles and parking areas. Frigate's planning guide explicitly warns that camera count alone is insufficient and classifies 12+ highly active deployments as needing robust detection capacity. Build around peak simultaneous regions. For a busy site, multiple Hailo devices or a modern GPU may be more appropriate than one small accelerator.
Hardware decoding is non-negotiable at this scale
Decoding dozens of camera streams in software can overwhelm CPUs that would otherwise be adequate. A supported iGPU or discrete GPU should carry the FFmpeg decode workload. Frigate also warns that many consumer GPUs have practical concurrent-stream limits, so verify the chosen GPU's decoding capabilities rather than assuming any modern card can process unlimited feeds. The recording and detection stream design should be optimized before hardware is finalized.
A workstation-like host can be easier to engineer
Ultra-small mini PCs are attractive for low power, but a 32-camera system may need PCIe slots, multiple storage interfaces, a discrete GPU, 10GbE or multiple accelerators. A compact workstation or SFF desktop can provide those resources with better cooling and serviceability. The extra chassis volume is not waste when it prevents a tangle of external adapters and gives the NVR a clean upgrade path.
Memory should leave room for enrichments
Frigate recommends 16GB for many-camera systems, but 32 cameras with semantic search, face recognition, LPR or other containers often justify 32GB or more. Memory should be monitored in production because model choice and feature use can change consumption. Buy a platform with replaceable DIMMs if possible. A high-camera NVR is infrastructure that may gain features over several years, so soldered minimum memory is a poor long-term trade.
Detector design may combine specialized hardware
There is no rule that all AI work must run on one device. Frigate can use supported detector hardware separately from enrichment acceleration. A dedicated Hailo device may handle object detection while an NVIDIA or Intel GPU handles video decoding and enrichments. That separation can be efficient, but it adds configuration. Choose a clear architecture and document which device owns decoding, detection and enrichment tasks before production rollout.
Storage planning should model multiple drive failures and replacement
Thirty-two continuous streams can require very large recording volumes. Use the calculator to estimate raw footage, then choose a multi-drive architecture that balances capacity, rebuild time and serviceability. Surveillance or enterprise HDDs make economic sense for long retention. RAID may improve availability, but it does not replace backup or an export policy for critical incidents. Large arrays should also have airflow appropriate for sustained writes.
NAS recording needs a robust network fabric
Network storage is possible, but a 32-camera system turns the NAS and uplink into core NVR dependencies. If recordings cross the network, design for sustained throughput, switch resilience and failure behavior. 10GbE may become useful when the same path carries recordings, multiple live clients and backups, although the exact requirement depends on aggregate bitrate. Local storage remains the simpler baseline when the chassis can accommodate it.
PoE power is a major electrical load
Thirty-two cameras can consume hundreds of watts depending on model and features. A 48-port PoE switch or multiple switches may be needed, and the total PoE budget must exceed worst-case camera draw with reserve. PTZ motors, heaters and IR illumination can raise consumption above idle values. This power also becomes heat in the network rack, so ventilation and UPS runtime should be planned together.
UPS architecture may need separate units
A single desktop UPS may not provide useful runtime for a 32-camera system once the PoE switch, server, storage and router are included. It may be cleaner to split compute/storage and PoE/networking across separate UPS units or use a higher-capacity rack solution. Decide the required behavior during outages: keep recording, preserve only critical cameras, or shut down safely. Then size from measured wattage.
Managed switching improves visibility
At thirty-two cameras, a managed switch provides practical operational benefits: VLANs, per-port power cycling, link monitoring and traffic visibility. Remote PoE cycling can recover a failed camera without visiting the rack. Segment camera traffic and limit unnecessary internet access where appropriate. These features do not increase Frigate inference speed, but they make a large surveillance deployment far easier to maintain.
Plan for stream limits and codec compatibility
Large sites often mix camera models and codecs. Frigate documents H.264 as the most broadly compatible path, while H.265 provides better compression but can have playback limitations depending on clients. Decoder stream limits can also matter. Standardize camera settings where possible and test representative streams before buying all hardware. A 32-camera deployment is the wrong place to discover that the selected GPU or browser path behaves differently than expected.
Operational monitoring is part of the build
A high-camera-count Frigate system should expose metrics for CPU, GPU decode, detector latency, memory, storage health and recording errors. Establish normal baselines after installation and alert on sustained changes. The best hardware design is one that makes bottlenecks observable. That lets future upgrades target the actual constraint instead of replacing expensive components based on guesswork.
Questions people ask
Frigate Hardware for 32 Cameras questions
What hardware do I need for 32 Frigate cameras?
Use a modern host with strong hardware decoding, robust detector capacity, at least 16GB RAM and usually 32GB+ for enrichment-heavy deployments, plus large recording storage.
Do 32 cameras require a discrete GPU?
Not universally, but a GPU becomes increasingly attractive for high simultaneous activity, enrichments and decoding. Multiple dedicated accelerators are another option.
Can I use a mini PC for 32 cameras?
A powerful mini PC can work in some architectures, but expansion, cooling, storage and GPU connectivity may make an SFF workstation easier to engineer.
How much storage do 32 cameras need?
Potentially dozens of terabytes or more. Calculate from average bitrate, recording duty cycle and retention days.
How much RAM for 32 cameras?
16GB is the Frigate recommended baseline for many-camera systems; 32GB or more is sensible when enrichments or other services share the host.
Can Hailo handle 32 cameras?
It depends on simultaneous activity and model choice. Busy systems may need multiple accelerators or a more powerful detector path.
Do I need 10GbE?
Not always. It becomes more useful when recordings live on NAS, backups are large, or many clients share the same uplink.
What PoE switch size for 32 cameras?
A 48-port PoE switch or multiple switches is common. Verify total PoE wattage, uplinks and spare ports.
What UPS size for 32 cameras?
Calculate the combined wattage of cameras, PoE switches, server, storage and networking. Large systems may need multiple UPS units or a rack solution.
Should I use RAID for Frigate recordings?
RAID can improve availability, but it does not replace backup or incident export. Choose the level based on capacity, failure tolerance and rebuild risk.
Official references and methodology
Verify current Frigate support before deployment
Cloudzat treats 32 cameras as a workload tier, not a guarantee of performance. Recommendations combine Frigate's official guidance on simultaneous activity, hardware decoding, detector support, RAM and storage with a dedicated Amazon catalogue. Verify exact camera bitrates, supported accelerators and seller configurations before deployment.
- Frigate - Planning a New Installation
- Frigate - Recommended Hardware
- Frigate - Video Decoding
- Frigate - Object Detectors
- Frigate - Camera Setup
- Frigate - Enrichment Hardware Acceleration
- Frigate - Installation
As an Amazon Associate, Cloudzat may earn from qualifying purchases. Prices, seller terms, exact configurations, camera firmware and Frigate support can change.