Details
- NVIDIA and Emerald AI unveiled a collaboration at CERAWeek 2026 to transform AI factories into flexible grid assets using NVIDIA Vera Rubin DSX reference design and Emerald AI Conductor platform for real-time energy orchestration.
- Partners include energy firms AES, Constellation, Invenergy, NextEra Energy, Nscale Energy & Power, Vistra; infrastructure providers GE Vernova, Schneider Electric, Vertiv; and innovators Maximo, TerraPower, Adaptive Construction Solutions.
- AI factories dynamically adjust power usage based on grid conditions, integrating compute, onsite generation, storage, and controls to accelerate grid connections, boost tokens per second per watt, and support reliability without overbuilding infrastructure.
- Builds on NVIDIA's five-layer AI infrastructure with energy as foundation; follows trials at five data centers worldwide and precedes 2026 commercial deployment at NVIDIA's 96MW Aurora facility in Virginia.
- Emerald AI raised $25M for expansion; approach could unlock 100GW US grid capacity, with demos proving 100% alignment to power targets using NVIDIA Blackwell GPUs without compromising AI workloads.
Impact
Power-flexible AI factories address surging data center demand by turning loads into grid stabilizers, potentially unlocking 100GW capacity and accelerating connections amid constraints. This shifts R&D toward integrated energy-compute designs, boosting efficiency metrics like tokens per second per watt by orders of magnitude. Over 12-24 months, it could drive hyperscaler adoption, ease GPU-power bottlenecks, and influence regulation favoring flexible infrastructure.
