AI Data Centres
Data centres don’t exactly have a reputation for excitement. For most people, the words probably conjure up images of grey warehouses filled with servers and cables. But the reality is very different, and in some cases, straight out of a spy movie. In some facilities, visitors are weighed on the way in and out to prevent tailgating and ensure hardware has not mysteriously disappeared. Then you’ll find yourself walking across sticky flooring designed to remove contaminants from shoes before entering the sensitive environment. In colocation facilities, where multiple tenants are housed, customers are escorted directly to their cabinets so they do not accidentally… or intentionally… wander. For buildings that store computers, data centres can have surprisingly intense security procedures.
And there’s a good reason for that. Data centres are among the most secretive buildings in modern society because they hold some of the world’s most valuable infrastructure. On the surface, they may just look like rows of cabinets filled with blinking lights. But knowing what sits inside those cabinets can reveal far more than people realise. Depending on who the customer is, infrastructure could support healthcare systems, financial networks, government platforms, or critical national services. Even understanding where systems are located, who occupies a facility, or how networks are arranged can reveal sensitive operational information. In the wrong hands, visibility itself becomes a security problem.
But physical security is only one side of the challenge.
Data centres do not simply need to protect information. They also need to ensure systems remain online continuously. Unlike many industries, downtime is not viewed as an inconvenience. It is viewed as a failure.
The cost of downtime
According to Uptime Institute’s 2024 outage survey, more than half of respondents said their most recent significant outage cost over $100,000, while one in five reported losses exceeding $1 million. The financial impact alone explains why operators approach resilience with near obsession. But avoiding downtime requires far more than secure buildings and restricted access cards.
Like layers of an onion, the systems keeping data centres online are built in layers too.
Redundancy everywhere
One of the most important concepts in data centre design is redundancy. At its simplest, redundancy means avoiding single points of failure by ensuring backup systems exist throughout the facility. Rather than relying on a single component, operators build additional capacity so that if one system fails, another immediately takes over. Take substations as an example. Instead of relying on a single grid connection, facilities may establish multiple independent power feeds. If one connection fails, another continues supplying electricity. The same philosophy exists almost everywhere throughout the building.
Different levels of redundancy exist depending on the amount of resilience operators want to achieve. At the most basic level sits N, where facilities deploy exactly enough equipment to support operations. If a component fails, systems fail with it. N+1 adds an additional backup unit beyond what is required, allowing a spare to immediately take over if equipment goes offline. At the highest level sits 2N, where entirely separate systems are built, each capable of supporting the full load independently.
Building redundancy at these levels becomes one of the primary factors determining classifications such as Tier III and Tier IV facilities. Redundancy itself also appears across almost every system inside a data centre. Cooling systems are duplicated. Network pathways are duplicated. Hardware infrastructure can be duplicated. Backup systems themselves often contain additional layers of redundancy.
But nowhere is this approach more important than power.
Backup for the backup
Most people assume power simply arrives from the grid and flows into a data centre. In reality, power systems inside these facilities resemble a sequence of overlapping safety nets. The first line of defence is usually the uninterruptible power supply, or UPS. As we explored previously in our article: UPS: The hidden systems keeping data centres online, UPS infrastructure instantly responds to power disturbances, ensuring servers continue operating while larger backup systems prepare to take over. Increasingly, modular UPS architectures are introducing redundancy into the UPS layer itself, allowing individual modules to fail without affecting wider operations.
But UPS systems are designed to buy time, not provide indefinite power. That responsibility falls to backup generation. Traditionally this role has been dominated by diesel generators, which start automatically during outages and provide power for extended periods. Increasingly however, operators are adding battery energy storage systems, or BESS, into their infrastructure. Batteries, UPS systems, and generators can now work together as overlapping resilience layers rather than isolated systems.
Once again, the same principle appears: backup for the backup.
The redundancy beyond the data centre
The interesting thing is that redundancy does not stop at the walls of individual facilities. Data centres themselves can also be redundant. Information can be replicated across multiple sites, ensuring that if one facility experiences an outage, fire, or disaster, systems can continue operating elsewhere. Redundancy, in other words, is not merely a backup plan. It is a design philosophy built around keeping critical systems running under almost any circumstances.
