Nvidia is moving beyond selling processors to addressing the electricity needed to run them, launching an energy-management coalition Thursday with Google and Emerald AI and planning up to two gigawatts of AI data-center capacity in Australia by 2027. The AI Energy Management Alliance aims to steer AI data centers' power appetite more precisely and plug them into existing grids as flexible loads rather than rigid, round-the-clock drains. CEO Jensen Huang, speaking at a Goldman Sachs conference on September 10, said the global AI infrastructure buildout is still in its early innings, naming available construction land, supply-chain throughput and existing power constraints as key obstacles. The Australian project, developed with partners including Firmus, CDC, NEXTDC and AirTrunk, rests on Nvidia's in-house DSX platform and, according to Reuters, would more than double the entire continent's current computing load. Nvidia also unveiled the Vera Rubin NVL72 platform in the MLPerf Inference v6.1 benchmark and expanded its open-source CUDA-Q platform with an orchestration layer called CUDA-Q Logical. Two days earlier, in Amsterdam, it extended its AI for Media platform. Huang said Thursday the company expects to sell twice as many chips next year as in the current one.