Grid‑responsive AI data centers
Nvidia and Emerald AI are partnering with leading U.S. utilities to build grid‑responsive, AI‑powered data centers — “AI factories” designed to optimize power use and support renewable integration, not just raw compute. The move aims to tame AI’s rising energy footprint by letting data centers flex with grid needs, signaling Big Tech is treating compute as an energy system problem as much as a hardware one. (digitimes.com)
NVIDIA and Emerald AI announced the collaboration at CERAWeek on March 23, 2026, naming AES, Constellation, Invenergy, NextEra Energy, Nscale Energy & Power and Vistra as participating energy partners. (nvidianews.nvidia.com) NVIDIA said the effort will use its Vera Rubin DSX AI Factory reference design and a DSX Flex software library to connect AI factories to grid services and interconnection workflows. (investor.nvidia.com) NVIDIA and Emerald say DSX Flex is slated for commercial deployment in 2026 at an NVIDIA AI Factory research center in Virginia as part of the company’s Vera Rubin/DSX rollout. (engtechnica.com) The coalition’s earlier “Aurora” reference project is a 96‑MW Aurora AI Factory in Manassas, Virginia, described as the world’s first power‑flexible AI facility and targeted to open in the first half of 2026. (prnewswire.com) A Duke University Nicholas Institute analysis found that modest, short‑duration curtailments by flexible large loads could allow roughly 100 gigawatts of additional new load to be absorbed by the existing U.S. grid. (nicholasinstitute.duke.edu) Emerald says it has run DSX/Conductor tests across five commercial data centers and took part in a UK demonstration with National Grid, EPRI and partners proving high‑performance GPUs can adjust power draw in real time without disrupting key workloads. (datacenterdynamics.com) Emerald AI disclosed a $24.5 million funding round to accelerate its Emerald Conductor orchestration platform, with backers including Radical Ventures, Amplo, CRV and Neotribe. (datacenterdynamics.com) NVIDIA’s and Emerald’s press materials state the approach aims both to speed interconnections and to let AI facilities operate as grid assets that can provide measurable relief while producing AI “tokens” and intelligence. (nvidianews.nvidia.com)
