The growth of GPU infrastructure is creating a difficult challenge for operators of existing data centers. Facilities originally designed around air cooling and much lower rack densities are increasingly being asked to support hardware with very different power and thermal requirements, often without additional floor space or the option of building from scratch. Containment, rear-door heat exchangers, direct liquid cooling and other retrofit approaches can extend the capabilities of these facilities, but every site eventually reaches a limit. This session looks at the practical decisions involved in introducing high-density infrastructure into brownfield environments and where the constraints begin to outweigh the benefits of further retrofit. Join this discussion to explore:
- Where operators should start when an existing facility suddenly needs significant GPU capacity
- How rack containment can extend the life of existing facilities
- Where the biggest constraints emerge across cooling, space, power, pipework and building design
- How much density operators can add before a retrofit effectively becomes a rebuild
