What Happened?
The massive expansion of AI and cloud infrastructure is fundamentally changing the requirements for perimeter security at data centers. According to the Futurum Group, the five largest US cloud and AI infrastructure providers have committed to capital expenditures (CapEx) of 660 to 690 billion US dollars for 2026 — a significant increase from around 380 billion US dollars in 2025. A substantial portion of these funds is flowing into new and expanded sites, which today frequently span 20 hectares or more, whereas earlier campus areas were often only a few hundred square meters in size.
This expansion of footprint presents security managers with new challenges: fence lines run alongside access roads, substations, generator fields, and open terrain. Large sections are poorly lit and can hardly be adequately secured through remote monitoring alone.
The Details
As soon as activity at the fence line shifts toward power supply, cooling, or other critical subsystems, uninterrupted operation is at stake. This distinction is highly relevant financially: according to the Uptime Institute's 2024 outage analysis, more than half of all severe data center outages now incur costs exceeding 100,000 US dollars; roughly one in five outages even exceeds the 1 million US dollar mark.
On such large sites, the number of false alarms triggered by wildlife, vegetation, or adjacent traffic also increases dramatically. While more cameras and detectors raise the volume of data reaching the control center, they do not automatically improve the situational awareness of security personnel.
As a solution, modern concepts rely on a multi-layered sensor approach:
- Ground radar (e.g., Flir R-190/R-290) monitors open terrain sections around the clock, regardless of weather conditions. It provides an object's location, direction of movement, and speed — crucial for distinguishing, for example, a target heading toward a substation from a passerby moving parallel to the fence.
- Integration platforms such as Flir Nexus automatically steer PTZ cameras onto the detected target, sparing the operator from manually searching through video streams.
- Multispectral cameras (e.g., the Flir FH series) combine thermal and visual sensors. Thermal imaging technology keeps targets in focus even in complete darkness, fog, or extreme weather without requiring additional perimeter lighting.
- Multispectral PTZ cameras (Flir PT series AI SR) automatically follow a radar track, confirm the target thermally, and, if needed, deliver high-resolution images for identification.
Radar thus handles detection, thermal imaging confirms the target under difficult conditions, and optical sensors provide the final detailed assessment.
Context
Why is this approach relevant? If unfiltered nuisance alerts continuously reach the control center, personnel spend a large portion of their shift verifying irrelevant events. The consequences are alarm fatigue, rising operating costs, and declining trust in the alarm system as a whole.
High-quality analytics classify people and vehicles before an alarm is even generated. By the time an alert reaches the operator, a validated dataset is already available — including target type, position, direction of movement, and the associated video stream. Another important factor is where data processing takes place: decentralized pre-processing directly at the sensor (edge computing) reduces the load on network bandwidth as well as computing and licensing capacity, which would otherwise increase significantly with centralized processing as the site grows larger.
It's also important to note that modern perimeter security does not require a complete replacement of existing security technology. The sensor components are designed for cross-vendor operation and can be integrated into existing VMS and access control systems. Management software such as Flir Latitude VMS or Cameleon Command and Control consolidates the data into a unified operator interface, ensuring that humans retain decision-making authority without having to piece together the situational picture from isolated individual alerts.
Practical Tips
Concrete recommendations for planning large-scale perimeter security can be derived from the requirements described above:
- Match sensor technology to topology: Open sections call for radar surveillance, while gates, blind spots, or utility yards require combinations of fixed thermal/visual cameras and nearby PTZ units.
- Prioritize by sector: In open terrain, the focus is on early detection and tracking; in dark or hard-to-monitor sections, on thermal verification; at gates and critical infrastructure, on high-resolution optical detail for final assessment.
- Plan before installation: Simulation software such as Flir Raven allows surveillance fields to be modeled during the planning phase in order to identify coverage overlaps, blind spots, and the optimal sensor mix. Integrating multi-layered verification is significantly more cost-effective during the planning stage than retrofitting an already-installed fence line.
- Incorporate existing systems: Since new sensor components can be integrated across vendors, existing VMS and access control systems can continue to be used rather than completely replaced.
Outlook
Regulatory and business conditions vary by region and will help shape the direction of future perimeter security projects. In the EU, the Energy Efficiency Directive (EED) obligates data center operators with an IT connected load of 500 kW or more to report on energy consumption, with requirements becoming increasingly stringent in countries such as Germany and Ireland. Perimeter sensor technology that does not rely on energy-intensive outdoor lighting directly supports these efficiency and sustainability requirements. Similar grid availability requirements are emerging in projects in the United Kingdom.
In the US, meanwhile, the primary focus is on grid resilience, as the logical perimeter of a data center today also encompasses the upstream energy infrastructure on which the campus depends. Given the scale of investment in AI infrastructure described above, it is reasonable to expect that multi-layered sensor concepts combining radar, thermal imaging, and optical verification will increasingly become the standard for hyperscale sites — not least because they give security personnel the critical head start needed to detect and prevent disruptions to operations at an early stage.
