10.00 – 10.25
The explosive growth of AI workloads is completely turning the underlying data center infrastructure upside down. Rack densities are skyrocketing from a few tens of kW to 400+ kW per rack, meaning Direct Liquid Cooling (DLC) is no longer an option but a necessity. At the same time, the massive east-west network traffic (NW) between GPU clusters is forcing us to rethink the hall layout: more compact clusters, heavier floor loading, and a clear distinction between air and liquid cooling. Planning regarding SPACE, POWER & COOLING becomes crucial in this context, as a single AI hall draws as much power as an entire conventional data center.
In this session, we lift the veil slightly on all the changes on the horizon — and why planning today is essential if you want to run your own AI workloads on-premise in the near future.
Speaker: Stefan Boutsen – Hewlett Packard Enterprise