Newsbite analysis · AI, Cloud, Data, Energy Tech, Hardware · New Product
NVIDIA Newsroom
NVIDIA Vera Rubin and DSX Platform Advancements Showcase Energy Efficiencies of Optimizing Tokens Per Watt for AI Factories
At the AI Infra Summit, NVIDIA's Ian Buck highlighted the need for energy optimization in AI factories, introducing the Vera Rubin systems and technologies aimed at maximizing performance per watt. The company's DSX MaxLPS platform reportedly delivers up to 1.4 times more tokens per megawatt, benefiting AI infrastructure by allowing better resource management based on real-time grid signals. These advancements allow AI factories to dynamically adjust power usage, improving efficiency while meeting performance demands.
Act nowConfidence: HighShort term horizon
Published 15 Sept 2026, 18:40 UTC10 min read
Lead image — NVIDIA Newsroom
Brief
NewsBite reading:NVIDIA Unveils Energy-Efficient AI Factories at AI Infra Summit
TL;DR
- NVIDIA introduces DSX MaxLPS for AI factories, enhancing tokens per megawatt.
- Collaborations with partners like Emerald AI demonstrate grid-responsive power management.
- NVIDIA Vera Rubin NVL72 systems improving AI performance significantly.
- MaxLPS can increase GPU capacity by up to 40% without additional power lines.
- AI workloads benefit from dynamic power allocation and energy efficiency.
Interesting facts
- The event attracted over 8,000 attendees, reflecting growing interest in AI infrastructure.
- NVIDIA’s technologies aim to convert AI factories into flexible grid resources.
- MaxLPS achieved a 24% increase in token throughput with efficient power usage strategies.