Much of the discussion around AI cooling focuses on what happens at the chip and rack, but deploying high-density infrastructure at scale creates challenges across the entire thermal system. Conventional cooling, CDUs and two-phase technologies are all evolving, while thousands of GPUs are placing new demands on cooling plants, power infrastructure and heat rejection. The next phase of AI cooling design will require a broader view of how these systems work together at facility and campus scale. This session will explore:
- How novel and evolving cooling technologies are reshaping facility-wide design
- What thousands of GPUs change upstream at the cooling plant
- How absorption cooling and on-site generation could bring power and cooling closer together
- Whether the next efficiency gains will come from the chip or the wider thermal system
