“AI is biggest shift of our lifetimes,” says Google boss Sundar Pichai. And he’s not wrong. However, leveraging all the potential advantages that machine learning can offer, and is yet to offer, requires data and power, and lots of it. For data centres this brings its own challenges and as a UPS manufacturer we are working hard to be part of the solution explains John Kreyling, Managing Director, Centiel
Following a report issued in April 2025 by the International Energy Agency (IEA), Fatih Birol IEA Executive Director, said: “Global electricity demand from data centres is set to more than double over the next five years, consuming as much electricity by 2030 as the whole of Japan does today.”
The fact is that AI is so power demanding that one of the main considerations when it comes to locating data centres is energy availability. The sheer cost of this power can be a deal breaker and the cost of energy needed to run AI data centres in the UK, is massive compared to overseas markets, so those responsible for AI data centre builds are weighing up alternative options.
The government released the ‘Quarterly Energy Prices’ on the 26 June 2025, that includes a comparison on the Non-Domestic Market Price (Industrial and Commercial). Chart 5.3 compares 14 European countries (EU14) that shows the Median electricity price to be just under 15pkWh with the UK having the highest electricity price at circa 26pkWh. The next most expensive was Ireland at circa 18pkWh so this highlights a significant challenge for the UK to attract Data Centres.
The other question is, is the grid resilient enough? We saw how reliant facilities are on the electricity supply following the recent Heathrow Airport incident, which resulted in the airport needing to shut for 16 hours following a fire at an electrical substation. Of course, data centres have in-built resilience and the UPS is part of that so this is far less likely to occur, but if they can’t get enough energy from the UK grid, what then?
All this means that other countries investing in their Grids to provide high and secure power become attractive and a real consideration for new AI data centres. For example, Middle East regions are becoming even more viable to solve potential supply issues.
As UPS manufacturers we have worked on addressing the challenge of efficiency to minimise power usage and also of course, UPS by their very nature are designed to mitigate grid instability. However, AI data centres will continue to be restricted on a choice of locations due to power cost and availability for the foreseeable future. Renewables will be able to contribute but are unlikely to satisfy the vast increase in energy demand.
The other challenge we face as an industry is how rapidly AI is moving and being adopted. There have been several stories in the mainstream media recently about investment in nuclear as a reliable source of power that makes the UK self-sufficient. However, a new power plant would take around 15 years to build and commission and AI is moving faster than that. It’s moving quicker than the infrastructure. Within a decade, the technology will have changed again and other solutions will be available.
It seems likely that planning applications for AI data centres in the future will need to show how facilities can contribute their assets to help manage grid stability by putting power back. The vast battery banks within data centres provide an opportunity here to show the facilities are not just a drain but can contribute back to grid services during times of lighter load.
There is also the potential for heat recovery, where excess heat from data centres can be captured and diverted to be used to power or heat small towns or residential communities.
At the moment there is no clear answer to the challenges set out. Our latest UPS StratusPower is AI data centre ready, as far as it can be. AI is not a constant load, its dynamic, and peaks quickly and the UPS supports that all without compromising resilience. It offers a market leading 97% online efficiency plus availability of 99.9999999% equating to milliseconds of downtime per year.
If UPS can support grid instability while efficient systems help reduce energy costs and offer less drain on grid this will all help. However, at the moment we are all in a situation where everyone is making lots of small gains – and by adding these together hoping to make a bigger gain.
With AI applications developing so rapidly, how data centres will manage AI’s need for power in the future is not certain. As UPS manufactures we have done all that we can, but perhaps ultimately, it will be a question of asking the AI itself to solve the challenge it has created by its very existence?
For further information about Centiel’s AI ready UPS click here.
