Connect | Virginia // DCD Talks building cooling around live workloads with Joe Capes | LiquidStack

The plan will change. Rack densities will rise, deployment dates will move and capacity may be added in a different order than expected. But once an AI facility is live, the cooling cannot be paused and rebuilt every time requirements change.  

In this session, Joe Capes will look at what happens when liquid cooling grows from individual units into one multi-megawatt system. He’ll explore how operators can add capacity in stages, maintain individual components and keep the rest of the system running as the site expands. The goal is simple: a cooling system that can grow around a live AI workload without making every new phase another redesign.