AI BriefingAI Factories with Power Flexibility Offer a Solution for Stabilizing Global Energy Grids
Key point
AI factories equipped with power flexibility can adjust their usage during power demand surges to stabilize energy grids and reduce infrastructure build-out costs.
Details
In situations where power demand spikes suddenly, such as during halftime of UK football matches, technology that allows AI factories to autonomously adjust their own power usage and ease the burden on energy grids is drawing attention.
Emerald AI collaborated with NVIDIA, EPRI, National Grid, and Nebius to demonstrate how power-flexible AI factories can autonomously adjust their power usage during peak hours. Through this technology, AI factories can connect to the grid quickly without large-scale infrastructure upgrades, and consumers can benefit from the effect of curbing electricity rate increases.
In a proof-of-concept experiment conducted at the Nebius AI Factory in London, the following technical results were achieved:
- High-performance workload operation using 96 NVIDIA Blackwell Ultra GPUs and the NVIDIA Quantum-X800 InfiniBand platform
- Real-time GPU power telemetry collection and control via the Emerald AI Conductor Platform
- In a power demand surge scenario, high-priority AI workloads were maintained while flexible tasks were throttled, reducing power usage by 30% within 40 seconds
As a result of this test, Emerald AI proved its technical value by achieving 100% compliance with more than 200 power targets set by EPRI and National Grid.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Stanford and NVIDIA Release 'Contrastive Language Model' CLM with 9x Faster Inference