NVIDIA Introduces DSX MaxLPS to Optimize AI Factory Power Efficiency

AI factories are increasingly operating as power-constrained industrial systems. [1] The critical question for infrastructure teams is no longer how many GPUs can fit within a single data center footprint, but exactly how much AI output each available watt can produce. [1] To address these hard physical limits, the NVIDIA DSX MaxLPS architecture is designed specifically to maximize performance per watt across large-scale AI deployments. [1]

NVIDIA's framing of AI factories as industrial systems highlights a shift in data center economics, emphasizing that raw compute must be balanced against hard power limits. By focusing on performance per watt with the DSX MaxLPS, NVIDIA acknowledges that energy efficiency is now a primary bottleneck for scaling AI operations. What the sources don't address: the specific performance per watt metrics or hardware specifications of the DSX MaxLPS compared to previous generations.