As the need for power increases due to AI and HPC workloads, data center managers are concentrating on the challenge of providing sufficient power, cooling, and scalability for high-density racks as the infrastructure changes. Although flexibility in the unused space is essential, is there a comparable focus on the adaptability of the backup power system that supports it?
Distributed architectures enable operators to incrementally add smaller generator sets to meet growing demand, rather than committing to significant backup generation capacity upfront, according to Christer Hedström from Volvo Penta. The architecture for backup power can be considered as less important compared to the more prominent infrastructure found in the data hall.
Ultimately, these systems are intended to stay in a state of readiness, prepared to assume control if the main power source becomes unavailable. However, as AI complicates forecasting future capacity needs, the adaptability and expandability of backup power infrastructure should be equally considered alongside the systems it safeguards.
AI is transforming the backup power dynamic. According to the International Energy Agency (IEA), global electricity consumption by data centers is predicted to nearly double, increasing from 485TWh in 2013 to around 950TWh by 2030. Additionally, electricity consumption in facilities focused on AI is anticipated to triple during the same timeframe.
This rising power demand is consequently placing more pressure on electricity networks. Operators face extra difficulties due to grid connection limitations and doubts about upcoming power supply, which add to the existing complications.
