Data Center · Datacenter Dynamics
Why liquid cooling is a system, not a SKU: Rethinking AI data center design
Compiled by KHAO Editorial — aggregated from 3 outlets. See llms.txt for citation guidance.
◌ Single Source
As AI workloads reach 130 kW per rack, data center operators need end-to-end liquid cooling solutions to efficiently handle rising thermal loads.
Key facts
- McKinsey predicts that data centers designed for AI workloads will demand $5.2 trillion in capital investment over the coming years
- As AI workloads reach 130 kW per rack, data center operators need end-to-end liquid cooling solutions to efficiently handle rising thermal loads
- Schneider Electric, along with Motivair by Schneider Electric, is the leading liquid cooling provider, offering an end-to-end cooling portfolio that delivers efficiency, reliability, and scalability
- As AI continues to place greater demands on data center infrastructure, thermal management cannot continue to be treated as a SKU
Summary
AI chips and heat densification are causing a shift in data center design, with gray space replacing IT racks in white space as the dominant factor. As high-density AI racks become more common in data centers, the demand for liquid or hybrid liquid-air cooling solutions is accelerating. Liquid-cooling infrastructure for AI workloads is considerably more complex than air-cooling solutions since liquid cooling is a direct extension of the server itself. Data centers have traditionally relied on multiple suppliers for various infrastructure components, such as racks, power solutions, and cooling.