Executive Summary
- Don’t forget your boots on the ground when it comes to resiliency; two in three data centre operators polled by the Uptime Institute said they found difficulty retaining staff
- While many operators would use AI to analyse sensor data or perform preventative maintenance, they draw the line at configuration changes and managing staff
- Meanwhile, forward-thinking operators are building energy resilience through various methods, from rethinking site selection, to engaging early with grid operators, to modular design
Data Centre Insight explores three potentially overlooked aspects of data centre resilience – from staff shortages to energy strategy.
Don’t think impactful outages are becoming more frequent – despite the stories
You see the big headline and pour one out for the IT teams going through outage hell. The last quarter of 2025 saw three outages which took out an awful lot of the internet: AWS in October, Cloudflare in November, and Cloudflare again in December.
However, impactful data centre outages – in other words, those significant enough to record even if the cost was a rounding error – keep going down relative to overall IT growth. That is according to the Uptime Institute’s 2025 Global Data Centre Survey, published in July.
Yet don’t think they are going away either. According to the Uptime Institute study, while impactful outages are becoming less frequent, one in 10 still cause ‘serious or severe’ disruption. This not only underlines the need for continued investment, the Institute notes, but also the changing parameters for failures. Growing workload demands and complexity – driven primarily by AI – as well as increasing numbers of extreme weather events will lead to greater impacts than in previous years.
Don’t think resilience is a purely technical conversation
Another finding from the Uptime Institute: almost two thirds of data centre operators polled said they found difficulty retaining staff, finding qualified candidates, or both. The Institute notes that this trend has largely been unchanged for the past couple of years – so is in part a small victory – yet a talent gap in the electrical and mechanical aspect of data centres continues to be ‘significant.’
Again, the Institute finds an interesting link back to AI. A majority of operators would allow the use of AI in data centre operations for analysing sensor data and preventative maintenance – but not for configuration changes, controlling equipment, or managing staff.
Don’t think of resilience just in terms of uptime numbers
Data centre resilience can be defined in more than one way. A report from professional services firm Aon in October explored global strategies for data centre energy resilience. The increasing workload demands of AI and edge computing is a ‘mismatch’ compared with slow-moving infrastructure development cycles, making for a ‘systemic risk’, Aon notes.
Forward-thinking operators are building their energy resilience through a few methods. They are rethinking their site selection to optimise energy, connectivity and risk exposure; engaging early with grid operators to improve site planning and energy access; and using modular and edge data centres to reduce grid dependency and boost efficiency.
Deals are being done which make sense for both data centre operator and energy provider. Aon cited the agreement between AWS and Talen Energy in June as a good example. The latter will eventually provide 1.9 gigawatts of its 2.5-gigawatt output from a Pennsylvania nuclear plant to AWS, expecting to earn upwards of $18bn from the contract.
“Power availability is no longer a background consideration – it’s a strategic asset,” the report concludes. “Developers who proactively manage energy risk will be better positioned to scale securely and sustainably.”
