AI
Rack density – the electrical load drawn by a single server cabinet housing the hardware used to train frontier models – is roughly doubling every year. As these individual racks demand more power, the total energy requirements for new facilities have grown, triggering a flood of grid connection requests that utility networks simply cannot process fast enough.
This dramatic escalation was the core focus for Steven Carlini, Chief Advocate of Data Centres and AI at Schneider Electric, in his opening remarks for a panel session on Grid and Energy Considerations for Supporting the AI Factory Revolution during a Schneider Electric press event in Buffalo, New York in May.
“ From a geographic perspective, there’ s a lot of backup for grid prediction requests,” he noted.“ Grid connection queues in the US and the UK have a 500 % backlog.” This backlog is forcing data centre operators, utilities and energy technology providers to rethink how these facilities are powered. Moving forward, the industry is balancing a delicate act of meeting immediate power demands while sticking to longterm zero-carbon sustainability goals.
“Grid connection queues in the US and the UK have a 500 % backlog”
Steven Carlini Chief Advocate of Data Centres and AIr Schneider Electric
Navigating the interconnection bottleneck Securing a connection to the local power grid has become the biggest obstacle for new data centre developments. With backlog queues swelling, many operators are looking to alternative energy sources to get facilities online.
88 August 2026