Schneider Electric Races to Meet AI Data Centre Demands
AI is forcing data centre operators to build faster than the industry can support. Racks that once drew 10kW now pull 120kW, while construction timelines that previously spanned months must now compress to weeks. Marc Garner, Cloud and Service Provider Segment President at Schneider Electric, says the market is caught between accelerating demand and deepening constraints.
“There’s really two main trends that I see almost everywhere I go. It’s all around speed and scale,” Marc says. “Operators today are trying to build out the capacity as fast as possible to keep pace with the technology demand changes, but also to keep pace with the sheer demand that's coming on the back of AI as a principal growth driver.”
The workforce shortage hits first. The data centre industry is turning to prefabrication to address this gap, moving construction off-site where supply chains are more predictable and installation times shorter. Schneider Electric is also working towards standardisation, shifting from engineered-to-order to configured-to-order processes and removing lead time bottlenecks, allowing operators to deploy the same technology across multiple sites without redesigning each time.
Power, however, is the deeper problem, with grid congestion in Amsterdam, Ireland and Singapore making access a competitive differentiator in today’s data centre market. “For years, we’ve always gone to where the connectivity is,” Marc notes. “I’d argue now that there's an opportunity to build the connectivity faster than you can build the power.”
Schneider Electric and NVIDIA: Developing reference designs for high-density racks
The power densities that AI requires create thermal challenges that existing cooling infrastructure cannot handle. Schneider Electric has partnered with NVIDIA to develop reference designs for the chipmaker’s GB200 and GB300 offerings.
“The idea really is around shortening design cycles and increasing standardisation,” Marc says. The companies’ reference designs provide electrical and thermal architecture for racks that run from 50kW to 120kW today, scaling to 300kW to 600kW with potential to reach as much as one megawatt per rack.
The thermal demands of these systems pushed Schneider Electric to acquire Motivair, completing its cooling portfolio from chiller to chip level. The acquisition gives Schneider the full range of thermal management technologies needed for high-density AI infrastructure, from the cooling distribution units that deliver liquid to the rack through to the air handling systems that manage ambient temperatures.
Marc explains that liquid cooling will extract most heat from processors, but air cooling remains necessary for the broader data centre environment.
“While liquid cooling is going to extract the vast majority of the heat from the chip, there’s still a lot of heat that’s being generated in the white space within the data centre,” he says.
Beyond the technical challenges of power and cooling, the workforce shortage remains, with Marc saying the industry needs around 300,000 new professionals currently.
To help combat this, Schneider is deploying condition-based monitoring systems that help to reduce the risk of downtime while optimising maintenance activity and minimising wasted site visits from data centre personnel. The technology uses software to identify potential failures before they occur, allowing fewer engineers to manage more infrastructure.
“AI is a solution for data centre growth as well,” Marc says. “This all starts to help bridge the skills gap that we see in the market today.”
