Google, Nvidia, Anthropic bet on smarter grids for AI’s power crunch

What if the fix for AI's power crunch isn't building more power plants but simply using the grid smarter? That's the bet behind a new coalition just formed by some of the biggest names in tech and energy.

Grid software startup Emerald AI has founded the AI Energy Management Alliance alongside Google and Nvidia, with Anthropic joining in too.

A group of utilities, including AES, Constellation, National Grid, and NRG Energy, rounds out the coalition. The goal is ambitious: making demand response a standard part of how new data centres get built, rather than an afterthought.

Utilities have been paying plants to reduce electricity consumption during periods of high demand for decades. What’s new here is the player that employs the concept.

Data centres can utilise such services by turning to their backup generators, but Emerald AI proposes a greener alternative: software that stops non-critical computing or moves the workload to a data centre with available grid power.

AEMA claims this approach could unlock an additional 100 gigawatts of data centre capacity on the grid. That's not a wild outlier either; a Goldman Sachs study last year found that simply capping max grid usage at 90% for a few hours at a time could free up 76 gigawatts nationally.

The possibility is there since grids have been made keeping peak loads in mind and hence are underused during other times of the year.

Google too has its own version of the technology being developed, while Enel X has its unique offering of using backup power supplies of data centres.

Emerald AI, however, takes things a step further by linking utilities to data centres in such a way that they are able to respond to demand in much the same way a battery would.

This unique positioning helped Emerald AI secure $150 million in Series A funding from Energise Capital and DCVC with a valuation of $1.05 billion.

Emerald AI itself doesn't claim to be a silver bullet either. As their chief scientist, Ayse Coskun, told TechCrunch, the technology helps mitigate the need for more power generation, not eliminate it.