The pitch for cloud video surveillance is that it modernizes your security. The reality is narrower, and more useful, than the pitch: moving to the cloud changes where your video lives, who can reach it, and who patches the software. It does not change whether anyone is watching.
That distinction matters because the buying decision usually gets framed as old versus new. On-premise recorders are the past, hosted video is the future, and the upgrade is assumed to carry a security improvement along with it. Some of that is true. Storage, remote access, and maintenance genuinely do get better. But a camera that streams to a data center is still a camera that nobody is looking at, and the three things cloud fixes are not the three things that let an incident run for six minutes before anyone notices.
Here is what the architecture actually delivers, the bandwidth arithmetic that rarely appears in a vendor brochure, and the questions worth asking before you sign a multi-year subscription.
What cloud video surveillance actually means
Cloud video surveillance is any system that stores, manages, or delivers camera video through a hosted service instead of a server sitting in your building. The label covers at least three different architectures, and the differences between them drive most of the cost and performance surprises buyers run into later.
Three architectures, one label
- Direct-to-cloud. Cameras stream continuously to the provider's data center, which handles all recording and retention. Simplest to describe, heaviest on bandwidth.
- Edge-first hybrid. Cameras or a local appliance record on site, and only metadata, thumbnails, or requested clips travel upstream. Most commercial deployments land here.
- Cloud-managed on-premise. Video stays local on your recorder, but configuration, health monitoring, user access, and updates are administered through a hosted console.
Vendors in all three categories market themselves as cloud. When a quote says cloud video surveillance without specifying which of these it means, that ambiguity is worth resolving in writing before procurement, because it determines your network requirements, your hardware options, and what happens to your footage if the internet connection drops.
The three problems cloud video surveillance genuinely solves
Cloud video surveillance solves storage management, multi-site access, and software maintenance, and it solves all three convincingly. These are real operational burdens, and organizations moving away from on-premise recorders are usually right to.
Storage stops being a capacity planning exercise. On-premise retention is bounded by the drives you bought, which means someone has to forecast how many days of footage a compliance requirement or an insurance policy will demand, then buy hardware against that forecast and replace it when drives fail. Hosted retention converts that into a line item you can change without a purchase order.
Multi-site access collapses into one login. An organization running twelve locations on twelve separate recorders has twelve sets of credentials, twelve VPN paths, and twelve places to look when something happens. A hosted console makes that one interface, which is a meaningful improvement for anyone responsible for a distributed footprint.
Patching stops requiring a site visit. Firmware and software updates delivered centrally close security holes faster than a maintenance window that gets deferred three quarters in a row.
The market reflects that value. MarketsandMarkets sizes the video surveillance as a service segment at USD 5.88 billion in 2026, growing to USD 12.01 billion by 2032 at a 12.6 percent compound annual rate. Independent forecasts for the same segment run as high as 11.5 billion for 2026, a spread that says more about differing scope definitions than about disagreement on direction. The direction is not in dispute.
The bandwidth wall nobody puts in the brochure
The single most common cloud video surveillance deployment failure is running out of upload bandwidth, and the arithmetic is easy enough to do before you buy. Video surveillance is not like most cloud workloads. It writes a relentless stream from every camera, all day, whether anything is happening or not, and that stream has to fit inside the upload capacity of your internet connection.
Verkada, one of the larger cloud camera manufacturers, publishes the per-camera figures directly: a traditional cloud camera streaming continuously consumes 1 to 2 Mbps, while their own edge-storage design holds steady state at roughly 20 kbps per camera by keeping recordings on the device and sending only encrypted thumbnails and metadata upstream. The company is candid about why the distinction exists, noting that organizations have found continuous streaming from every camera consumes more bandwidth than their networks can accommodate.
Multiply those published figures by a realistic camera count and the constraint becomes obvious.
| Cameras on site | Continuous cloud streaming | Edge-first architecture | Difference |
|---|---|---|---|
| 10 | 10 to 20 Mbps sustained | 0.2 Mbps sustained | 50x to 100x |
| 25 | 25 to 50 Mbps sustained | 0.5 Mbps sustained | 50x to 100x |
| 50 | 50 to 100 Mbps sustained | 1 Mbps sustained | 50x to 100x |
| 100 | 100 to 200 Mbps sustained | 2 Mbps sustained | 50x to 100x |
The number that matters here is upload, not download, and most business connections are asymmetric. A plan advertised on its download speed can carry a small fraction of that figure upstream, which is why a site that streams video fine during a demo starts dropping frames once every camera is live and the staff are also running voice, point of sale, and video calls across the same circuit.
The practical test: before signing, ask the vendor for the sustained upload requirement per camera at the resolution and frame rate you actually intend to run, then compare that total against the upload figure on your current internet bill, not the download figure. If the answer requires a circuit upgrade, that recurring cost belongs in the comparison alongside the subscription.
What cloud video surveillance does not do
Cloud video surveillance does not watch your cameras, and no amount of hosted infrastructure changes that. Storage location is an infrastructure decision. Whether a weapon, a fall, a perimeter breach, or an after-hours intrusion produces an alert within seconds is a detection decision, and the two are entirely separable.
This is the gap that catches organizations after the migration. The footage is now retrievable from a phone, retention is generous, the console is clean, and the incident still ran its full course before anyone knew about it. A hosted recorder produces exactly the same after-the-fact clip as the on-premise recorder it replaced, delivered faster. That is the difference between recording an incident and preventing one, and it survives every storage architecture.
The staffing answer does not close the gap either. Human attention degrades on a predictable curve when the task is watching video for events that almost never happen, which is why operator effectiveness falls off sharply after about twenty minutes of continuous monitoring. Adding a hosted console to that problem gives a fatigued operator more feeds to miss, not fewer. Outsourcing the watching to a monitoring center changes who is fatigued rather than removing the constraint, a tradeoff worth understanding on its own terms when you compare remote video monitoring against AI video analytics.
Detection is a separate layer, and the useful question is not whether your video is hosted but whether anything is analyzing it in real time.
The hardware question worth asking before you sign
Ask whether the platform requires its own cameras, because that answer determines whether a cloud migration is a software change or a building-wide hardware replacement. Some cloud platforms are built around proprietary cameras with onboard storage, which is a coherent engineering choice and the reason their bandwidth numbers are as good as they are. It also means the migration path involves pulling working cameras off the wall.
For a facility with fifty or a hundred serviceable cameras, that is the dominant cost in the project, and it is frequently discovered late. Other platforms accept standards-based streams from cameras already installed, which keeps the existing infrastructure in play and confines the change to software and licensing.
Why this shapes the detection decision too: IntelliSee layers AI detection onto the cameras an organization already owns, with no hardware replacement required, which is what makes proactive detection an incremental addition rather than a capital project. The same principle applies to any platform you evaluate: the question of what runs on your existing cameras is usually a bigger number than the subscription line.
Cloud video surveillance, cybersecurity, and privacy
Moving video off-site changes your risk profile rather than removing risk, and both halves of that sentence deserve attention. Centralized patching genuinely reduces one of the most common exposures in physical security, because internet-facing recorders and cameras running years-old firmware are among the softer targets on a corporate network. Anyone who has audited a camera fleet knows that security cameras are themselves an attack surface, and a vendor pushing updates centrally is usually an improvement over a maintenance schedule nobody owns.
What it adds is a third party holding your video and a set of questions your security team should ask in writing: where the data is stored geographically, whether it is encrypted in transit and at rest, who at the provider can access recordings, what the breach notification commitment is, and what happens to your footage if you leave.
Privacy is the other half. Some platforms bundle biometric features by default, and in a growing number of jurisdictions that carries specific legal obligations. IntelliSee does not use facial recognition and does not store or retain video, which removes an entire category of compliance exposure rather than managing it. If you want the detail on how that works and why it is a design decision rather than a limitation, we covered whether AI security cameras use facial recognition separately.
How to evaluate a cloud video surveillance platform
Evaluate a cloud video surveillance platform on five questions, and take the answers in writing. First, which of the three architectures is it, and what happens to recording if the internet connection fails. Second, what is the sustained upload requirement per camera at your intended resolution and frame rate. Third, does it work with the cameras already installed, or does it require its own. Fourth, what does the platform detect in real time as opposed to record for later retrieval. Fifth, what are the data residency, encryption, access, and exit terms.
The fourth question is the one most likely to go unasked, because it sits outside the frame the category sets up. Storage, access, and maintenance are the problems cloud video surveillance was built to solve, and a platform can answer the first three questions perfectly while leaving your cameras exactly as passive as they were.
The takeaway
Cloud video surveillance is a strong answer to a storage and access problem and a non-answer to a detection problem. Treat them as two separate line items on the same project. Get the bandwidth arithmetic done before the contract, confirm whether your existing cameras stay on the wall, and decide separately whether anything will be watching the feed in real time.
Frequently asked questions about cloud video surveillance
How much bandwidth does cloud video surveillance need?
It depends entirely on the architecture. Verkada's published figures put continuous cloud streaming at 1 to 2 Mbps per camera and edge-first storage at roughly 20 kbps per camera, so a 50-camera site needs somewhere between 1 Mbps and 100 Mbps of sustained upload depending on which design you deploy. Ask for the per-camera figure at your intended resolution and frame rate, then check it against the upload speed on your internet bill.
Is cloud video surveillance more secure than an on-premise recorder?
It is better on patching and worse on custody. Centralized updates close firmware vulnerabilities faster than most on-premise maintenance schedules, which is a real gain given how often cameras and recorders run outdated software. In exchange, a third party holds your footage, so encryption, access control, data residency, and breach notification terms become contractual questions rather than internal ones.
Do I have to replace my cameras to move to the cloud?
That depends on the platform. Some cloud systems are built around proprietary cameras with onboard storage and require replacing existing hardware. Others accept standards-based streams from cameras already installed. For a site with a large existing camera count, this single question is often the largest cost variable in the project, so resolve it before comparing subscription pricing.
Does cloud video surveillance detect threats automatically?
Not by default. Cloud video surveillance describes where video is stored and managed, not whether it is analyzed. Real-time detection of weapons, falls, unauthorized access, loitering, or perimeter breaches is a separate analytics layer that can run on hosted or on-premise video alike. A platform can offer excellent cloud storage and no detection capability at all.
What happens to recording if the internet goes down?
In a direct-to-cloud design, an outage can mean a gap in your recording, because the cameras have nowhere to write. Edge-first and cloud-managed on-premise designs keep recording locally and synchronize when the connection returns. This is one of the most important practical differences between the architectures and should be confirmed in writing.
The layer that decides whether anything happens
Cloud video surveillance is worth adopting for the reasons it is actually good at: retention you do not have to plan, access that works from anywhere, and software that stays current. Those are genuine improvements over a closet full of aging recorders, and the migration is usually the right call.
Just do not let the migration stand in for the harder decision. Cameras that record to a data center and cameras that record to a hard drive down the hall are equally passive, and passive cameras are only useful after something has already happened. The layer that turns a camera into something proactive is detection, running in real time on the feeds you already have, surfacing a weapon or a fall or an intrusion within seconds so that a person can act while acting still matters.
That is the work IntelliSee does: turning passive cameras into proactive protectors, on the camera infrastructure an organization already owns, without facial recognition and without storing video. If you are already planning a cloud migration, it is the right moment to decide what will be watching. Talk with our team about what proactive detection would look like on your existing cameras.