NVIDIA and Emerald AI have joined forces with AES, Constellation, Invenergy, NextEra Energy, Nscale Energy & Power and Vistra to develop a new generation of AI factories designed to connect more rapidly to power networks while operating as flexible grid assets. The collaboration brings together expertise across technology, energy and infrastructure, illustrating how cross-industry coordination can accelerate AI deployment in the United States while enhancing power system reliability. These next-generation facilities are intended not only to generate high-value AI outputs but also to function as responsive grid assets that can support electricity networks during periods of demand fluctuation.
At the core of the initiative is the NVIDIA Vera Rubin DSX AI Factory reference design, which incorporates the DSX Flex software library to enable seamless interaction between AI infrastructure and grid services. The approach allows for faster deployment by initially relying on co-located generation and storage as bridge power for hybrid AI factories. Over time, these same resources can be leveraged to supply electricity back to the grid, improving interconnection timelines and reinforcing the role of AI facilities as dynamic grid assets. The DSX architecture also enables configurations without co-located energy, offering flexibility for larger and faster grid connections.
Emerald AI’s Conductor platform will play a central role in managing this ecosystem by coordinating computational workloads with onsite generation, battery systems and other behind-the-meter resources. This orchestration ensures that AI compute operations maintain service quality while delivering grid-responsive flexibility. By aligning power usage with grid needs, operators can meet energy targets, safeguard critical workloads, reduce reliance on temporary bridge power and accelerate interconnection processes. The system is also designed to minimize the need for infrastructure sized around peak demand, helping to manage long-term system costs while strengthening the contribution of AI infrastructure as grid assets.
The broader concept reflects a shift in how energy systems are utilized. While existing grids are typically designed to meet peak demand and remain underused for much of the time, power-flexible AI factories offer a way to unlock up to 100 gigawatts of capacity across the U.S. system. These facilities convert electricity into AI tokens, models and intelligence, creating high-value outputs while contributing to grid stability. However, the companies note that relying solely on isolated generation and storage can lead to inefficiencies, including underutilized resources and higher costs per AI token.
To address this, AES, Constellation, Invenergy, NextEra Energy, Nscale Energy & Power and Vistra are working to expand generation capacity aligned with rising demand. Their collaboration includes evaluating optimized energy solutions, including hybrid projects that integrate co-located power with grid connectivity. By combining large-scale AI demand with flexible operations and intelligent energy management, the initiative aims to strengthen reliability and expand infrastructure efficiently. The deployment of DSX Flex is expected to scale commercially later this year at the NVIDIA AI Factory Research Center in Virginia, marking a key step in advancing AI infrastructure that operates as integrated grid assets.
