NVIDIA and Emerald AI collaborate on grid-flexible AI factories

NVIDIA and Emerald AI have joined forces with major US energy companies to develop a groundbreaking AI infrastructure model aimed at enhancing the flexibility of the power grid. Unveiled at CERAWeek 2026, the initiative seeks to expedite the deployment of AI factories while assisting in stabilizing electricity networks grappling with escalating demand from data centers.

The collaboration involves key players such as AES, Constellation, Invenergy, NextEra Energy, Nscale Energy & Power, and Vistra, combining AI infrastructure design with advanced energy generation and grid management capabilities.

For readers of eeNews Europe tracking the swift growth of AI computing infrastructure and its implications for power systems, this project underscores how semiconductor and data center technologies are increasingly influencing energy infrastructure planning. It also sheds light on innovative architectural approaches that could alleviate grid bottlenecks as AI workloads expand.

At the core of the initiative lies NVIDIA’s Vera Rubin DSX AI Factory reference design, which integrates the DSX Flex software library to enable AI facilities to directly engage with grid services. These "flexible AI factories" are envisioned to dynamically adjust their power consumption based on grid conditions, aiming to streamline the connection process of large AI data centers to power infrastructure while offering services that bolster grid stability.

Emerald AI’s Conductor platform will synchronize compute workloads with onsite energy resources like batteries and generation assets. By coordinating both compute and power resources, the platform strives to uphold service quality for AI workloads while delivering grid-responsive flexibility.

AI factories designed as grid assets

“AI factories are the engines of the intelligence era, and like any great engine, every system must be designed together — energy, compute, networking, and cooling as one architecture,” remarked Jensen Huang, founder and CEO of NVIDIA. “NVIDIA and Emerald AI are collaborating to pave the way for a future where AI performance, efficiency, and grid responsiveness can be harnessed instantly.”

Faster power connections for AI infrastructure

One of the primary hurdles facing large AI data center projects is the prolonged timeline for grid interconnection. Many projects resort to temporary onsite power generation to kickstart operations while awaiting full grid access.

The new architecture enables AI factories to commence with co-located generation and storage, serving as "bridge power," and subsequently transitioning into flexible grid-connected assets. Once connected, these energy resources can also feed electricity back to the grid during peak demand.

According to Emerald AI, this approach could unlock up to 100 GW of capacity within the US power system by enhancing the utilization of existing infrastructure and curbing the necessity for extensive grid expansion.

Energy industry partners join the initiative

The participating energy companies will assess generation projects tailored to power these AI factories and expedite their integration into the grid. The strategy amalgamates new energy generation, flexible demand from AI workloads, and intelligent control systems.

Industry leaders believe this approach could help tackle the upsurge in electricity demand fueled by AI data centers. “Grid flexibility will be pivotal in meeting AI’s unprecedented demand while upholding system reliability,” noted Andrés Gluski, CEO of AES. “DSX Flex ingrains flexibility from the outset, enabling AI infrastructure to function as a grid asset that bolsters faster, more efficient expansion.”

The companies have already trialed AI power flexibility at five commercial data centers globally. A commercial-scale rollout is anticipated later this year at the NVIDIA AI Factory Research Center in Virginia, slated to be one of the initial power-flexible AI factories constructed using the Vera Rubin infrastructure.