Two years ago, I asked the president of one of America’s largest public power utilities what worried him about the data centers lining up to connect to the grid. Each wanted the same thing, he replied: electricity delivered immediately, on an expensive grid overbuilt to guarantee power around the clock. He remarked that utilities and communities were asked to serve as “foot soldiers” of the AI buildout. The data centers, I thought, could serve as foot soldiers in their own way.
I founded Emerald AI after that conversation to enable data centers to support the grid and their local communities, easing power use in the rare hours when the grid is stressed while protecting the performance of critical AI work. By flexibly consuming energy, AI data centers could become good citizens of the power grid, protecting energy affordability and reducing the risk of blackouts for communities. What’s more, America can connect flexible AI data centers much more quickly to the power grid, advancing U.S. competitiveness with China in AI frontier innovation.
That’s why today, my company, along with Google and NVIDIA, is founding the AI Energy Management Alliance (AEMA), which launches with 18 member companies who lead the AI and energy industries. These include the AI frontier lab Anthropic, the utility National Grid, and power producers AES and NRG. The premise is that if data centers change how they operate to support the communities that host them, they should be rewarded with faster access to power.
The benefits of flexible data centers start with affordability. Electricity bills have risen nationwide, and residents are anxious about whether surging AI demand could raise rates further by triggering expensive grid upgrades. But flexible data centers that are good grid citizens spare the peak demand moments that force the costliest upgrades, better using the grid we already have. The Brattle Group estimates that each 10 percent gain in utilization lowers rates by about 3.4 percent.
Today, a new U.S. data center can wait a decade or more for a grid connection—endangering U.S. competitiveness in AI and national security. This is because utilities must have enough capacity to serve everyone when demand is highest, including on sweltering afternoons when air conditioners are running full tilt. Much of that capacity goes unused at other times–the grid is only 50% utilized on average. If AI data centers were just moderately flexible during the grid’s worst hours of the year, America could unlock 100GW on our existing grid for flexible data centers.
There are many technologies to make AI data centers flexible power users, more responsive to the needs of the grid. Batteries can carry facilities through peaks, on-site generation can support demand, and software can slow, pause, or shift less urgent computing. Our new Alliance is technology-neutral and champions policies that reward any data center that can measurably provide relief to the grid.
There is already a strong track record of proof that AI data centers can be good grid citizens. Google runs a nationwide demand-response portfolio of roughly a gigawatt. My company, Emerald AI, and NVIDIA have completed six global demonstrations of flexible data centers. Later this year in Virginia, NVIDIA, Digital Realty and Emerald AI will turn on the world’s first power-flexible AI factory at nearly 100 megawatts, designed to prove that a data center can be a precise, controllable load, instead of drawing constant, unyielding power 24/7 and putting unprecedented strain on a rigid grid.
Data center developers who cannot get power are already building campuses that generate their own electricity. But the cost and complexity of powering an off-grid data center makes AI more expensive, while inevitably raising costs for everyone else because private systems and the public grid compete for the same turbines and transformers. This behind-the-meter approach also deprives utilities of the anchor customers that flexible data centers represent, high paying customers whose revenue could have funded upgrades and held rates down for everyone else.
State and federal regulators are already looking at flexible data centers as a potential solution. In June, the Federal Energy Regulatory Commission directed the six regional grid operators it oversees to make room for large customers willing to limit their draw in return for faster connections. Texas’s grid operator is finalizing rules letting controllable data centers connect sooner, and Silicon Valley Power, a California municipal utility, has launched the nation’s first flexible-load interconnection program.
AEMA will press this case in state capitals and in Washington. Our ask of governors, regulators and federal policymakers is simple: offer data centers that commit to flexibility a faster and larger grid connection — and hold them to it. Rarely does a country get to have more of everything at once: more AI innovation and a stronger American hand in the technology race, reined in energy bills for the communities that host the industry, and a more reliable grid with the means to improve itself. AI data centers are ready to provide the flexibility; we can all benefit if those who run and regulate the grid make room for it.
The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune. Varun Sivaram‘s opinions are his alone and not on behalf of AEMA or other founding members.