The Critical Role of UPS for AI Applications
The global explosion of Artificial Intelligence (AI) is revolutionising the data centre industry.
This blog explores the critical role that Uninterruptible Power Supplies (UPS) for AI applications play in powering modern AI driven services. It also examines how UPS for AI can help to support the wider infrastructure in relation to this emerging technology.
The eruption of generative AI platforms has been a significant catalyst for change in the data centre industry. Investment into AI is accelerating at a previously unmatched pace, with billions being ploughed into the technology.
While existing data centre capacity can support a number of AI applications, traditional facilities were never built to handle the unique demands of AI workloads, which are denser and more power-hungry than conventional applications. From the smallest of edge data centres through to the largest of hyperscale facilities, the demands of cloud processing are now driving the need for a fresh perspective when it comes to data centre design.
With the increasing use of Graphics Processing Units (GPU’s) and Central Processing Units (CPU’s) for AI applications, it is inevitable the blueprint of our data centres is having to change. Greater emphasis is being placed on design, construction and operation to accommodate the increases in power consumption, heat generation and density of hardware powering AI applications.
Despite there being no single, authoritative viewpoint on what format this should follow, it is clear data centre infrastructure is to play a critical factor in the successful integration of this emerging technology. So much so, that demand is now driving the need for new data facilities too, resulting in a dramatic growth across the data centre space.
According to an analysis by leading independent publication, UK Tech News, nearly 200 planning applications for data centres were submitted across more than 250 local authorities in England and Wales between 2020 and 2024. At least 38 separate applications for construction were made in 2024 alone, representing a 40.7% rise compared to the previous year.
This highlights the anticipated level of uptake in AI technologies, as businesses strive to stay competitive in a rapidly evolving marketplace, where less efficient, less capable solutions are quickly left behind.
Maintaining the functionality of all of this data centre equipment, together with the clean and consistent power supplies that enable optimal performance, will be no less of a concern. While the maximum power drawn from an everyday CPU-based server is typically around 200 watts (W), a GPU-based server consumes more than double that, at around 700 W.
Not only that, but high-performing AI computing hardware is also exceptionally susceptible to the damage that an unstable electrical source can create, making it particularly vulnerable to electrical disruptions.
Failure in any section of the electrical system can cascade, affecting multiple racks and servers. This interrupts real-time AI processes that rely on continuous data processing, corrupting active datasets and disrupting critical workflows.
The loss of accurate data undermines future AI performance and can render specific workload results unusable. Just a single electrical anomaly or transient failure can interrupt this process, requiring the application to start retraining from scratch. This can be a particularly costly exercise due to the significant volume of data involved.
Unexpected downtime can also halt e-commerce platforms, streaming services and other revenue-generating activities, resulting in millions in lost revenue due to the sheer scale of affected customers. Maximising equipment uptime is therefore paramount in the bid to avoid wasting both time and significant resource.
Uninterruptible Power Supplies (UPS) for AI applications provide the stable and consistent power source that protects sensitive equipment from power fluctuations and outages. The UPS guarantees that AI powered platforms and technology has the ability to remain operational and that complex algorithms and training processes are not disrupted.
As many data centres also often serve industries with strict regulations, such as finance, healthcare and defence, UPS for AI applications help to ensure these remain online. Many facilities are often required to operate under service-level agreements (SLA’s) which have strict uptime requirements due to the critical nature of application they protect.
For example, AI-driven applications in smart grids, autonomous vehicles or medical diagnostics that rely on constant uptime to power essential systems. Any electrical failure within these can cause real-world harm or liability, resulting in SLA violations that incur hefty financial penalties.
The use of UPS for AI also extends to protecting the infrastructure that powers the technology. AI workloads rely on highly specialised cooling systems due to the significant amount of heat this type of hardware generates.
Electrical issues can take cooling systems offline resulting in overheating, or potentially triggering automatic shutdowns that prevent damage and permanent hardware failure. Identifying the source of the problem can sometimes take hours to diagnose and then repair.
The combination of massive power requirements, expensive hardware and critical operations makes electrical failures in data centre AI applications particularly devastating in terms of costs and impact. Operators must therefore be able to employ end-to-end solutions that encompass infrastructure, cabling, cabinet design and stabilised power if they are to provide the dedicated support that is required for emerging AI technologies.
Power Control offer the Legrand Keor FLEX UPS for AI applications to mitigate service disruptions and prevent downtime. This UPS provides the stable and consistent power supply that prevents data loss and potential damage to sensitive AI hardware. By bridging the gap during any power anomaly, the UPS ensures consistency in electrical supply, avoiding any disruption to AI training models and operations.
While there is currently no consensus on the design and construction requirements of next-generation data centres, it is important that we anticipate shifts in the industry to ensure we have the UPS infrastructure in place to support current and anticipated future needs.
If you would like further information on how Power Control can support your data centre with UPS for AI applications, please call a member of the team on 01246 431431 or email info@powercontrol.co.uk .