Why adaptability beats capacity for data centers
Reported by TechRadar1 source ↓
A data center operator argues that facilities need to adapt to AI workloads’ rising power density, cooling demands and environmental constraints, rather than focus only on total capacity.
Key points
- The operator says AI rack power needs are rising beyond traditional data center designs
- A site reconfigured 40 racks at 5 kVA into 20 racks at 10 kVA in 30 days, TechRadar reports
- The operator says modular design lets it add power and cooling as customer workloads grow
- Its facilities use air cooling with in-row units and refrigerant-based cooling, the operator says
Details
TechRadar’s article is written by a data center operator, who says customers increasingly ask whether facilities can meet AI workloads’ power-density requirements on a short timeline. The operator says some decisions that previously took years now need to be made within months, and argues that a facility’s ability to accommodate future density matters more than its available capacity today.
The operator gives an example of a rack that supported 5 to 10 kVA two years ago being expected to support 15 kVA within 18 months, with 25 to 30 kVA planned in the next few years. The operator also says a POD was reconfigured in 30 days, changing 40 racks at 5 kVA into 20 racks at 10 kVA for a customer.
The operator says its modular design allows power and in-row cooling to be added as IT loads increase. It describes using air cooling and refrigerant-based systems, which it says keep water consumption low; the article provides no independent measurements of energy or water use.
Written from coverage as of
Images
- TechRadar