What Internet Speed Do You Need for AI-Powered CCTV?

The right internet speed for AI-powered CCTV depends on where analysis happens, how many streams leave the site and each stream's actual bitrate. There is no honest single speed that suits every installation. A four-camera shop using selected substreams has different needs from a multi-building site transmitting high-resolution video continuously.

Start with upload capacity, not just the large download number advertised on a broadband plan. If video travels from your site to a remote service, it uses upstream bandwidth. Internal network capacity also matters even when processing occurs locally, because cameras still send video to a recorder, gateway or analysis device.

Bitrate is the useful starting point

A camera's bitrate describes how much data its encoded video uses over time. Resolution, frame rate, codec, image complexity and quality settings all influence it. Variable bitrate can rise when a scene becomes busy, while constant bitrate aims for a configured level. Read the current camera or recorder settings and observe actual traffic rather than estimating from resolution alone.

To create a planning figure, add the bitrates of all streams that will travel over the connection at the same time. Then allow headroom for bitrate variation, network overhead and the site's normal business traffic. Do not plan to consume the full measured upload capacity continuously; congestion can affect alerts, calls, cloud applications and other essential services.

A simple example

If four selected streams each average 2 Mbps, their combined video load is roughly 8 Mbps before overhead and headroom. That is a planning illustration, not a universal recommendation. The real streams might be lower or much higher, and the design may send video only under certain conditions. Validate figures with measurements from the proposed configuration.

Cloud and on-site processing use links differently

Cloud analysis typically requires relevant video to leave the site, so sustained upload performance and internet resilience receive close attention. Edge or on-site analysis can keep more traffic on the local network, although it may still use the internet for management, notifications, updates or selected clips. Hybrid designs combine these approaches.

Ask where full streams, event clips and metadata travel. Also ask what continues during an internet interruption. The answer should form part of the system design rather than being assumed. The Surnexo how-it-works overview gives context for assessing cameras, connectivity and alerts together.

Measure the site under real conditions

Run tests at several times, including when staff, guests and cloud backups are active. Record upload throughput, latency, packet loss and stability. A headline speed can conceal congestion or an unreliable wireless hop. For remote properties, consider the behaviour of fixed wireless, cellular or satellite connections during poor weather and peak usage.

Inspect the local path too. A fast fibre service cannot correct an overloaded switch, weak Wi-Fi camera signal, damaged cable or misconfigured recorder. Wherever practical, use stable wired connections for fixed cameras and ensure network equipment can handle the aggregate traffic.

Information to collect

• Number of concurrent camera streams for analysis.
• Actual bitrate, codec, resolution and frame rate for each stream.
• Average and peak upload use from other site services.
• Local switch capacity and any wireless links in the video path.
• Required behaviour during an outage or degraded connection.
• Available backup connection and power arrangements.

Ways to manage bandwidth responsibly

Use a fit-for-purpose stream rather than automatically sending the camera's maximum setting. Many devices provide a substream that may suit a particular analytics task, while the recorder retains a higher-quality main stream. Adjustments must be tested against the actual detection goal; reducing quality too far can remove the visual detail the model needs.

Prioritise important cameras, schedule lower-priority analysis where appropriate and apply quality-of-service controls with qualified network support. Avoid exposing cameras to the public internet. Use secure connectivity, restricted accounts and network segmentation. Confirm stream support on the camera compatibility page before building calculations around an unavailable profile.

Design for failure, not only average speed

Decide which functions must continue if the connection drops. Existing local recording may remain available even when remote analysis or notifications pause, depending on the design. Document the expected behaviour, monitoring and escalation. Critical sites may justify a second connection, but resilience choices should reflect risk and budget rather than a generic rule.

A short pilot provides the best validation. Connect representative streams, monitor network use across normal operations and confirm alert delivery. If you provide camera settings and a recent view of site traffic, Surnexo can scope this more accurately. Request a demo and ask for a quote based on your connection, cameras and priority outcomes.