Face-free high-voltage transmission tower for AI Energy Management Alliance — electrical infrastructure, Sep 16 2026

AI Energy Management Alliance: flexible data centers for faster grid access

From Mexico — label the source first. Fortune published a Sep 16 op-ed by Varun Sivaram, founder and CEO of Emerald AI. It’s advocacy for a new industry group, not a neutral news brief. I’m reading it that way.

AEMA — flexible loads for faster interconnection

Per the Fortune op-ed: Emerald, together with Google and NVIDIA, is founding the AI Energy Management Alliance (AEMA), launching with 18 member companies. Fortune names Anthropic, National Grid, AES, and NRG among them. The thesis: if AI data centers flex power use when the grid is stressed, they should get faster interconnection — stay on grid flexibility, not chip silicon or lab-safety coordination.

Sivaram’s frame: software, batteries, and on-site generation can slow, pause, or shift less urgent compute so facilities aren’t rigid 24/7 loads. AEMA is technology-neutral in his telling — any measurable grid relief should count for policy rewards.

Cream-paper schematic: AEMA thesis — peak grid stress, flexible AI data-center load, faster interconnection as policy ask
Peak stress, then flex AI load, then faster interconnect. Original schematic.

~1 GW and ~100 MW — Fortune only

As Fortune/Sivaram states (advocacy, not my measurement): Google runs a nationwide demand-response portfolio of roughly ~1 GW. Later this year in Virginia, NVIDIA, Digital Realty, and Emerald plan to turn on a power-flexible AI factory at nearly ~100 MW. I’m not inventing extra MW/GW figures beyond that piece.

Brattle numbers — estimates or skip

The op-ed cites Brattle Group estimates that each 10% utilization gain lowers rates by about 3.4%, and that moderate flexibility could unlock ~100 GW on the existing U.S. grid. Those are cited estimates in an advocacy piece — labeled as such, not proven outcomes. Policy color in the same column: FERC directing regional operators to make room for large customers that limit draw for faster connections; Texas and Silicon Valley Power flexible-load programs mentioned as Fortune does.

My takeaway

From Mexico, the narrow read: Emerald + Google + NVIDIA are founding AEMA around flexible AI loads for faster grid access — 18 members including Anthropic, National Grid, AES, and NRG as Fortune lists — and the ~1 GW Google DR / ~100 MW Virginia flexible factory lines stay labeled as that Emerald CEO op-ed. Source: Fortune.