The Impact of AI Video Surveillance on Storage

Many companies are adopting AI video surveillance to streamline video analysis and classification. These modern cameras continuously generate metadata that is stored alongside raw video data. This additional data can provide invaluable context and reduce human toil, but it also increases the storage footprint of an organization’s surveillance data.  

With AI surveillance systems generating ever-increasing amounts of data, delaying critical decisions on surveillance storage can come at a significant cost to the company. IT teams must understand the main drivers of AI surveillance storage growth to plan and budget effectively. 

What makes AI video surveillance different? 

Traditional Closed-Circuit Television (CCTV) and AI video storage are both designed to monitor a scene, offering insight into the objects, people, and events within. However, AI video surveillance systems differ from their predecessors in several significant ways, including: 

• Active vs. passive monitoring: Traditional CCTV passively generates a record of the monitored environment, producing video files that can be analyzed later. AI, on the other hand, actively processes video to identify objects, behaviors, and events, storing its findings as metadata alongside the raw video files. 

• Higher resolution: AI object detection algorithms require higher-resolution video than traditional CCTV, which increases the storage footprint even before metadata is added. 

• Continuous recording: AI surveillance systems perform continuous analysis, which requires continuous recording and storage of full video clips. In contrast, traditional CCTV can record and store motion-only clips to reduce the amount of storage required. 

• Generated metadata: AI active analysis of recorded video produces information about the objects and individuals within the scene. These insights are stored as metadata alongside the video streams. 

How much capacity does AI video surveillance storage require? 

AI video surveillance systems do not produce a fixed amount of data per camera. Total storage needs depend on several factors, including: 

• Video resolution: AI object detection requires clear, high-resolution images, making 4K the baseline for object detection. While higher resolution improves detection accuracy, it can multiply the required file sizes, especially for large camera deployments. 

• Frame rates: Higher frame rates help capture fast-moving objects that could be missed between frames, but they also increase file sizes and storage requirements. 

• Activity levels: AI surveillance cameras continuously analyze video for objects, behaviors, and events, storing the results as metadata alongside the video stream. More activity in the scene produces more metadata. 

• Retention windows: Typical surveillance footage is stored for 30-90 days, but multiple factors can extend this window, including video retention for AI model training, incident analysis, legal holds, and regulatory compliance. 

For example, a single 4K camera running 24/7 might generate 100-200 GB of data per day, depending on the level of compression and frame rate. Scaling this up to a 50-camera deployment with a 30-day retention window, an organization needs to budget hundreds of terabytes of storage before considering potential redundancy or AI metadata. 

However, this 30-day retention window is the absolute minimum that an organization may be able to consider. Regulatory requirements may extend the retention window across the board, and organizations may need to hold certain footage to support incident reviews, legal holds, model retraining, and AI analysis for much longer—if not indefinitely. 

Storage challenges for IT teams 

IT teams building storage for AI surveillance footage face several key challenges, including: 

• Storage scalability: On-prem surveillance storage scales poorly with AI storage. Adding capacity requires buying hardware, provisioning space, and managing infrastructure that depreciates. Additionally, ensuring that adequate storage will be available when needed requires overprovisioning, which increases the associated expenses. 

• No monetization opportunities: Surveillance storage costs continue to rise whether the organization uses the data or not. Unlike many other data types, surveillance footage typically cannot be monetized to offset those costs. 

• Egress fees: Major cloud providers often charge egress fees when data leaves their environment. When access to stored data costs extra, teams are less likely to use the data they are already paying to store. 

• VMS compatibility: Many organizations already have existing video management systems (VMS) in place. Surveillance storage infrastructure needs to integrate with these existing platforms without the need for custom engineering work. 

• Expanding deployments: AI camera deployments often begin as pilot programs and scale quickly once they show a return on investment (ROI). This rapid growth can make capacity planning difficult as deployments expand from a few cameras to hundreds in a short period. 

Planning storage for AI surveillance 

When planning AI surveillance storage, hidden costs can add up quickly. To avoid surprise charges and incompatible solutions, follow these best practices: 

• Pre-calculate storage requirements: Calculating storage requirements before deployment helps to determine the volume of storage required and whether the provider can scale to meet an organization’s needs. Key inputs include the video resolution and frame rate, number of cameras, and target retention period. 

• Define usage tiers: Separate active footage from archived footage. Recent clips that are accessed frequently need different cloud infrastructure than footage kept only for regulatory compliance or AI reanalysis. 

• Calculate TCO: The cost of surveillance storage is often more than the base per-GB rate. A true total cost of ownership (TCO) calculation considers the base per-GB rate, egress and application programming interface (API) fees, redundancy costs, and the labor cost to manage on-prem hardware. 

• Consider purpose-built solutions: Cloud storage designed for AI video surveillance workloads often offers benefits such as elastic capacity, data immutability, access management, and predictable monthly costs. This flexibility eliminates the need to implement additional infrastructure on top of generic cloud buckets or to manage hardware refresh cycles and up-front costs. 

• Consider VMS compatibility: S3-compatible cloud storage integrates directly with most major VMS platforms. This compatibility eliminates the need to implement or purchase custom connectors or middleware to connect existing VMS platforms with storage infrastructure.  

Conclusion 

AI video surveillance can consume a large amount of storage, much more than traditional CCTV systems. While this additional data brings benefits, organizations that fail to plan for these demands can face costly infrastructure scrambles in the future. 

A strong AI video surveillance system is built for both current and future needs. The right storage solution scales easily, enables fast retrieval, and avoids extra charges whenever footage is accessed. 

Wasabi Hot Cloud Storage provides surveillance teams scalable, S3-compatible storage with no egress fees; that means organizations don’t have to pay a premium every time they access their own surveillance data.