NVIDIA, Emerald AI build ‘AI factories’ for the grid
NVIDIA and Emerald AI announced collaborations with energy companies to develop AI 'factories' that can act as flexible grid assets — blending data center performance with grid‑responsive operation. That trend means future commercial and campus buildings will need integrated, AI‑driven energy strategies and closer collaboration with MEP and grid planners. (nvidianews.nvidia.com)
The collaboration named AES, Constellation, Invenergy, NextEra Energy, Nscale Energy & Power and Vistra as energy partners, and the announcement was made at CERAWeek 2026 in Houston on March 23, 2026. (nvidianews.nvidia.com) The technical blueprint for the effort is the NVIDIA Vera Rubin DSX AI Factory reference design, which embeds a DSX Flex software library intended to connect AI factory operations to power‑grid services. (nvidianews.nvidia.com) Emerald AI’s Conductor platform is described as orchestrating computational flexibility — scheduling workloads alongside onsite generation, battery systems and other behind‑the‑meter resources to deliver grid‑responsive power flexibility while protecting priority AI workloads. ( ) The partners say the reference design supports hybrid “bridge power” models using co‑located generation and storage to bring capacity online sooner, while the DSX architecture can also operate without co‑located energy to pursue larger, faster interconnections. (nvidianews.nvidia.com) Emerald AI previously identified a 96 MW Aurora AI Factory under construction in Manassas, Virginia as the first implementation of this standard and estimated that nationwide adoption of power‑flexible AI factories could unlock roughly 100 GW of existing grid capacity. (prnewswire.com) NVIDIA CEO Jensen Huang framed the approach as designing “energy, compute, networking and cooling as one architecture,” and Emerald AI CEO Varun Sivaram said DSX Flex plus Conductor can both monetize AI outputs as “AI tokens” and provide measurable relief back to the grid. (nvidianews.nvidia.com)
