Data centres
Supporting Data Centre Resilience with Long-Duration Energy Storage
Data centre backup is designed for the outage that lasts minutes. The pressure is increasingly on the network, for hours.
- Country
- An emerging data centre market
- Sector
- Data centre
The challenge
Data centres are among the most energy-intensive facilities in the modern economy, and their load keeps climbing — driven by cooling, and by high-performance computing that runs hot.
Their backup, though, is designed around a narrow case: hold the load for the minutes it takes a generator to start, and for the hours it takes the grid to return. The newer difficulty is not a failure at the site. It is a network that becomes constrained during seasonal peaks, when heat drives cooling demand and everything else on the network at the same time.
How LAVO answers it
Hydrogen storage separates the moment energy is generated from the moment it is used.
Electricity is converted to hydrogen when supply is comfortable, held in solid-state storage without pressure or loss, and converted back through a fuel cell when it is not. A lithium-ion buffer sits alongside it for immediate response, so the system answers a sudden change in load and a long, slow squeeze on the network with the same installation.
That gives an operator something a generator does not: a reserve that can be drawn on during a peak, rather than only during a failure.
Wider application in the sector
Resilience is the entry point, not the extent of it.
The same stored energy can be drawn on to shave a site’s peak, to firm on-site or contracted renewable generation so it counts toward a load that runs continuously, and to take part in grid services rather than sit idle between failures. In regions where connection queues are the constraint, storage of this duration is also what lets a site be built at the capacity it needs, rather than the capacity the network can currently offer — which is increasingly the question that decides where a data centre goes at all.
And as high-density AI and high-performance computing raise both the load and the cost of losing it, the reserve that matters is measured in hours of full operation, not minutes of orderly shutdown.
Project enquiry