Nvidia, Google and the startup Emerald AI have formed an industry consortium to push data centers toward behaving more like grid assets than fixed loads.
The AI Energy Management Alliance launched Wednesday with three founding members and 18 other backers, including Anthropic, chipmaker Analog Devices and several utilities. Its first task is drafting best practices for flexible facilities.
A flexible data center trims its power draw when the grid is strained. Operators can shift training runs away from forecast peaks, or draw from on-site batteries instead of the network. Storage cuts both ways: excess charge can be pushed back onto the grid when a utility is short.
Each founder brings a piece of the stack. Google says its demand response capacity passed 1 gigawatt in March. Nvidia offers DSX Flex, software that tunes a facility’s grid usage, on-site generation and batteries using signals from utility infrastructure, for operators running its Rubin cards. Emerald AI, which raised $150 million in August with Nvidia as an investor, sells Emerald Conductor, which lowers cluster consumption under stress while limiting the hit to AI workloads.
Nvidia and Emerald AI are building a 100-megawatt facility in Manassas, Virginia that they call the world’s first power-flexible AI factory, with Conductor inside.
The alliance lands as data center electricity demand turns into a political problem. Communities from Virginia to South Africa are pushing back on large campuses, and utilities are being asked to serve loads that arrive faster than transmission lines can be built.
