Energy-Efficient AI Networks Lead to Dramatic Reduction in Environmental Impact

Download this paper to learn about network power consumption of large scale AI training implementations and the impact of various emerging approaches to increase power efficiency. The paper details how specific optical innovations can potentially reduce network power consumption by up to 80%, saving more than 17 megawatts of power, and reducing the carbon footprint by the equivalent of 10,000 metric tons of CO2 per AI training cycle.