For more than a decade, the security industry has been talking about “the cloud”, writes Jonathan Rickard, pictured, Senior Sales Engineer at the video surveillance product manufacturer i-PRO.
It promised limitless scalability, centralised access, and freedom from on-prem maintenance. For many applications, it’s delivered exactly that. But when it comes to mission-critical security, where bandwidth is tight, uptime is non-negotiable, and privacy regulations are getting increasingly strict, the cloud alone isn’t always the best answer. The latest innovations in video security aren’t necessarily happening in distant data centres. They’re happening at the edge, inside the camera itself.
When the Cloud Isn’t Enough
Streaming 4K video 24/7 to the cloud sounds appealing until you run the numbers. A single camera can generate terabytes of data per month. Multiply that by a few hundred or a thousand cameras, and bandwidth and storage costs quickly spiral out of reach for many organizations. Latency is another limitation. If you’re trying to detect a vehicle entering a restricted area or identify a person in distress, waiting for a cloud server to respond can take seconds you simply don’t have.
And then there’s the issue of uptime. What happens when your connection drops during a critical event? For hospitals, transport hubs, or utilities operating under strict continuity requirements, the answer has to be “nothing changes.” Security can’t depend on an external link.
Edge computing turns that model on its head. Instead of sending raw video to the cloud for analysis, AI-enabled cameras process data on-site. They interpret what they see in real time and send only metadata, a lightweight description, such as “vehicle detected at 14:02” or “person loitering near entrance.” That means lower bandwidth, faster detection, and less reliance on remote infrastructure.
Smarter Devices, Not Bigger Servers
Modern edge-AI cameras are more than sensors; they’re computers in their own right. Equipped with advanced chipsets such as Ambarella SoCs (System on Chip), they run multiple analytics streams simultaneously. Object detection and descriptions of unique attributes, licence-plate recognition, PPE compliance, crowd counting, it all gets processed within the camera.
This edge intelligence reduces latency to near zero and allows cameras to operate even when external networks are offline. For industries that require segmentation or air-gapped networks such as defence, healthcare, and critical infrastructure, this is critical. It also gives organisations more control over where their data lives, helping them comply with data sovereignty requirements and regional privacy laws like the GDPR.
The edge isn’t replacing the cloud; it’s making it more accessible for all the right reasons. The cloud still excels at centralised management, large-scale trend analysis, and remote access. But real-time intelligence happens best where the action is.
Containerisation: The App Store for Edge AI
The real revolution isn’t just that cameras can run AI on the edge. It’s also that they can learn new skills without being upgraded or replaced.
Using container technologies like Docker, developers can now package AI applications with everything they need to run such as code, libraries, and settings, into lightweight, self-contained units called containers. These “apps” can be deployed directly onto compatible cameras, much like installing an app on your smartphone.
Want to add fall detection in a care facility? Deploy the app. Need to count pallets at a warehouse or detect smoke outside? Spin up another. Each container runs independently, so updates or new functions don’t disrupt existing operations. For integrators, this modularity means they can tailor installations to each client’s exact requirements without reinventing the wheel every time.
Because Docker is an open, well-established standard in the IT sector, the same container can run across any supported device or cloud platform. That openness invites innovation and lowers the barrier for developers entering the physical-security space. It also enhances cyber resilience: each container is isolated from the camera’s core functions, reducing the risk that a vulnerability in one application could compromise the device. Containerisation effectively transforms the camera into an open-platform edge computer. And in an industry that’s long been divided between closed and open ecosystems, it represents a powerful way forward.
Generative AI Meets the Edge
The next leap is already underway. Generative AI, the same technology driving natural-language models, is now finding practical use in on-premises systems. An example is i-PRO’s Active Guard 3.0, which enables security teams to search across video using free text. Typing phrases like “person who fell” or “red fire truck” to locate relevant footage instantly. The system combines metadata generated at the edge with a generative-AI engine running entirely on-premises. This ensures that no sensitive footage leaves the network, maintaining compliance for sites where internet connectivity is restricted or prohibited.
Generative AI has the potential to make complex forensic searches as intuitive as typing into a search bar. The fact that it’s running locally, without the need for an external cloud connection, shows how far edge processing has come and how it’s redefining what’s possible inside a single camera or server.
Hybrid Is the Sweet Spot
Despite all this innovation at the edge, the cloud still has a vital role. The most resilient architectures will remain hybrid; edge-first and cloud-supported. Real-time analytics and decision-making stay local, while the cloud provides oversight, coordination, and long-term storage.
A transport authority, for example, might store 30 days of footage on-site for immediate access, then archive older video to the cloud. A retailer could run people-counting analytics on-camera but aggregate data centrally for strategic planning. Hybrid systems give each organisation the flexibility to balance performance, compliance, and cost.
Most importantly, hybrid doesn’t mean half-measures. Instead, it’s about placing workloads where they make the most sense. As networks evolve and 5G or private LTE become more common, we’ll see even tighter integration between edge and cloud systems. But for the foreseeable future, the edge will remain where real-time intelligence lives.
Cybersecurity and Data Sovereignty Still Rule
No discussion of edge computing is complete without cybersecurity. The “install it and forget it” mentality has no place in a world of networked cameras. Devices must be patched, monitored, and protected like any other endpoint. i-PRO’s approach includes secure boot, encryption, and compliance with evolving international standards such as ISO/IEC 42001 for ethical AI governance.
Edge processing can actually strengthen cybersecurity posture by minimising the data that leaves the network. Instead of streaming video to third-party servers, only metadata or event triggers are transmitted. This reduces the attack surface and helps maintain compliance with regional data-sovereignty laws, especially important for government, healthcare, and education clients operating under strict privacy mandates. For critical infrastructure or air-gapped environments, edge AI provides full functionality with zero internet dependency. That’s not just a performance advantage; it’s a risk-reduction strategy.
What’s Next: Collaboration at the Edge
As processing power grows, so too will collaboration between devices. Technologies such as Docker Swarm already enable distributed workloads, allowing multiple cameras to share computing tasks dynamically. Imagine a network of cameras that balance analytics between them, automatically compensating if one goes offline. This type of decentralised, self-healing architecture is the logical evolution of the hybrid model; fast, resilient, and inherently scalable.
At the same time, we’ll see edge devices integrating generative AI not only for search but for summarisation and prediction. Cameras could eventually produce natural-language reports or automatically flag anomalies based on learned behaviour. The hardware is ready and the software is catching up fast.
Conclusion: Intelligent by Design
The future of video security isn’t about choosing between edge or cloud. It’s about designing systems that use both intelligently, where each layer plays to its strengths. Edge computing puts intelligence exactly where it’s needed: next to the lens, at the moment an event unfolds. The cloud, meanwhile, provides the big-picture perspective and coordination that make complex operations manageable. Together, they form a flexible, secure, and future-proof architecture.
For security professionals, the takeaway is simple: Don’t buy into one-size-fits-all thinking. Assess where real-time decision-making matters most, where compliance rules apply, and where the cloud truly adds value. The smartest systems will blend all three; edge, cloud, and human oversight, into a seamless, adaptive whole. The next wave of innovation in physical security won’t come from bigger data centres. It will come from smarter devices on the edge, teaching AI new tricks and redefining what intelligent security really means.
