AI cameras run in realtime. Does your connectivity?

Twenty 4K cameras on a single site is a fairly typical picture of an AI camera rollout for security or monitoring today. A single camera streaming continuously already requires 5 to 10 Mbps. Install twenty of those cameras at one location, and traffic climbs to a few hundred Mbps, running day and night without a break.

Keep the same SIM card and configuration that you’d use for a sensor reporting a meter reading once an hour, and the problem usually doesn’t surface for a few weeks: slow footage, missed alerts, and an unexpectedly high bill.

Why does video streaming place heavy demands on the connection?

The video stream is the largest share of the traffic: a steady few Mbps per camera, around the clock. An incident alert, by contrast, is small, sometimes no more than a few kilobytes plus a short clip, but it needs to reach the control room within a few hundred milliseconds.

A brief stutter in the video stream rarely gets noticed. An alert that arrives late can be the difference between acting in time and missing the moment entirely.

Why doesn’t every country need the same SIM configuration?

At a fixed location, it’s easy to see what a local SIM profile delivers. With a traditional SIM profile, traffic first runs through the home operator’s core network before reaching the control room or cloud platform. A local breakout shortens that route by sending traffic through the nearest network directly, skipping the detour back to the home country. That often saves tens of milliseconds. For a single alert, that’s barely noticeable, but for a system that has to respond continuously to what’s happening on camera, every millisecond counts.

Still, what ultimately determines the best profile is the country where the control room or cloud platform sits, not where the camera itself is mounted. If the control room for a site in the Netherlands is nearby, a local SIM profile is the most logical choice. If the control room for a location in France happens to sit in Germany, a regional profile is more practical. And if a site in Poland has its own regional control room, a local profile again makes the most sense there. Settling on one standard configuration for the entire deployment isn’t realistic.

How do you choose the right operator for an AI camera location?

The operator that performs best at a given location can only be determined through a practical test, carried out at the exact spot where the camera will be mounted. A coverage map alone isn’t enough. The largest operator in a country doesn’t automatically offer the best network on a specific industrial site, along a rail line, or at a remote location.

A coverage test measures actual throughput, latency, and stability across different operators. The immediate surroundings matter too: a warehouse, building, piece of equipment, or mast can weaken the mobile signal on part of a site, while a connection just a few dozen meters away holds up fine.

Based on these measurements, Thingsdata, as a reseller, selects the operator with the best combination of coverage, speed, latency, and stability for each location. When extra certainty is needed, multiple operators can be used at the same time. That means not every AI camera location has to run on the same operator, even when the entire deployment is managed centrally under a single agreement with Thingsdata.

Why is remote SIM management important for an international rollout?

Once a deployment spans multiple locations, physically swapping a SIM card quickly becomes impractical. If something needs to change in the configuration at one of the sites, eSIM technology based on the SGP.32 standard handles that remotely: the profile is downloaded, activated, or replaced without anyone needing to visit the site. Thingsdata Control also allows that same connection to be monitored continuously, so deviations stand out before they turn into an outage.

That’s useful at the start of a rollout. It’s just as valuable months later, when an operator at a particular location underperforms and the profile needs adjusting.

When do you need more than one network at a single location?

An outage hits a critical location harder than a regular one. If the network goes down at a perimeter or rail camera, the alert disappears right when it’s needed most. Thanks to Thingsdata’s multi-operator portfolio as a reseller, that kind of backup is usually already in place: a second operator, a router with multiple connectivity options, or a backup profile. Which solution fits best depends on how much downtime is acceptable. A few minutes is no issue at one site; at a level crossing or the fence around a power plant, that’s a different story.

What does GDPR mean for footage from an AI camera?

Footage from an AI camera can quickly end up containing personal data: faces, license plates, behavior. How the cloud environment handles that is one part of the picture. Where that traffic is routed (through which countries, for example) and where the breakout takes place fall under the same GDPR responsibility as storing the footage itself. That needs to be documented per country and per site.

What should a connectivity partner deliver for AI cameras?

For an AI camera that needs to run 24/7, downtime is a direct risk. A ten-minute outage on a perimeter camera means ten minutes without detection. That’s why Thingsdata works with multiple network operators at every location, so the network offering demonstrably the best coverage and stability is the one used there.

With eSIM technology based on the SGP.32 standard, that choice doesn’t have to be locked in at the moment of installation. Thingsdata Control allows the connection to be monitored continuously, and a profile can be switched to a different operator remotely, without a technician having to visit the site. If a network does go down, the platform can switch to an alternative operator within minutes.

Want to know how to best connect your AI cameras for the rollout, with the lowest possible risk of downtime? Contact us at +31-85-0443500 or email info@thingsdata.com.