Susana Schwartz

8.2K Like transformers and switchgear, prefabricated busway systems (and their tap-off units) are in high demand. They represent yet another part of essential electrical infrastructure that cannot be built fast enough to keep up with data center demand. Designed to efficiently transmit large amounts of electrical power, they carry up to 6,300 amperes in a single run, meeting the extreme power demands of high-density GPU clusters.

Today’s top story discusses some of the challenges top-tier operators encounter when managing frontier AI models and large-scale cloud services, in which faults are guaranteed and idle time can be catastrophic. When trying to monitor 100,000 to well over 1,000,000 GPUs in LLM training workloads, training faults not only waste compute time and progress, but trigger inference faults that can destroy user experience, violate SLAs, and destroy unit economics.

8.2K When Nvidia this week announced that the “Rubin generation of NVIDIA AI infrastructure is the world’s first to achieve 100% liquid cooling,” some media outlets inaccurately touted it as a “100% reduction in water.” There were Nvidia comments that touted “zero water consumption” and “the elimination of massive amounts of power usage and pretty much all water usage.” I have since requested an interview with Nvidia to clarify what the design can and cannot do, but in the meantime, remain...