According to foreign media reports, Nvidia (NVDA-US) and start-up Emerald AI announced on Monday (23rd) that they will jointly build a new generation of "AI factories" with a number of power companies ". The project aims to combine energy and computing architecture to speed up the connection of data centers to the grid and alleviate the pressure on the power system caused by the rapid growth of artificial intelligence demand.
Participating power companies include AES, Constellation, Invenergy, NextEra Energy(NEE-US), Nscale Energy & Power, and Vistra. Nvidia said the AI plants will be designed as "elastic assets" that can run high-intensity computing tasks while supporting grid stability by adjusting power demand and combining on-site power generation and energy storage systems.
According to the plan, the relevant facilities will adopt NVIDIA Vera Rubin DSX reference architecture, with DSX Flex software, so that the calculation load can be dynamically adjusted according to the power supply situation. Through its Conductor platform, Emerald AI coordinates computing needs with the operation of on-site power generation, batteries and other energy resources to improve power efficiency and system flexibility.
the two companies said that by deploying local power generation and energy storage facilities before the grid is officially connected to the grid, they can be used as a transitional power source to accelerate the AI infrastructure on-line. Subsequently, these resources can also be fed back to the grid to support the operation of the overall power system. This model helps reduce data center construction time and reduces the pressure on the grid to over-expand in response to peak demand.
Nvidia pointed out that traditional power grids are designed to meet peak demand, but most of the time there is idle capacity. AI plants with flexible scheduling capabilities are expected to unlock the potential supply capacity of the U.S. grid and improve overall energy efficiency through more efficient power consumption patterns.
With the rapid expansion of AI computing demand, data center power consumption has risen sharply, and power supply and infrastructure have become key bottlenecks in the development of the industry. Nvidia CEO Huang Renxun said that the AI factory will become the "engine of the era of wisdom". Only by integrating energy, computing, network and heat dissipation into a single architecture can the next wave of AI applications be supported.