Two years ago, I asked the president of one of America’s largest public utilities what worried him about the data centers lining up to connect to the grid. Everyone wanted the same thing, he replied: electricity delivered immediately, via an expensive network that was developed to ensure power supply around the clock. He noted that utilities and municipalities have been asked to act as “foot soldiers” in expanding AI. I thought the data centers could serve as foot soldiers in their own way.
After that conversation, I founded Emerald AI to enable data centers to support the grid and their local communities, reducing power consumption during the rare hours when the grid is overloaded while protecting the performance of critical AI work. By flexing energy consumption, AI data centers could become good citizens of the power grid, protecting energy affordability and reducing the risk of blackouts for communities. Additionally, America can connect flexible AI data centers to the grid much more quickly, increasing U.S. competitiveness against China in breakthrough AI innovations.
That’s why today my company, together with Google and NVIDIA, is launching the AI Energy Management Alliance (AEMA), launching with 18 member companies that are leaders in the AI and energy industries. These include the AI Frontier Lab Anthropic, the utility National networkand power generators AES and NRG. The premise is that if data centers change the way they operate to support the communities in which they are located, 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 concerned about whether rising AI demand could further increase rates by triggering costly grid upgrades. But flexible data centers that are good net citizens avoid the moments of peak demand that require the most costly upgrades, making better use of the network we already have. The Brattle Group estimates that every 10 percent increase in occupancy reduces rates by about 3.4 percent.
Today, a new U.S. data center can wait a decade or more for a connection to the network—threatening U.S. competitiveness in AI and national security. That’s because utilities must have enough capacity to serve everyone when demand is highest, including on sweltering afternoons when air conditioners are running at full blast. Much of this capacity remains unused at other times – the network is only 50% full on average. If AI data centers were only moderately flexible during the worst hours of the year, America could unlock 100 GW in our existing network for flexible data centers.
There are many technologies available to make AI data centers flexible power users that can better respond to the demands of the network. Batteries can power facilities during peak periods, on-site power generation can meet demand, and software can slow, pause, or relocate less urgent computing operations. Our new alliance is technology neutral and advocates for policies that reward any data center that can provide measurable relief to the network.
There is already a strong track record proving that AI data centers can be good netizens. Google operates a nationwide demand response portfolio of around one gigawatt. My company, Emerald AI and NVIDIA have conducted six global flexible data center demonstrations. Later this year, NVIDIA, Digital Realty and Emerald AI will launch the world’s first energy-flexible AI factory in Virginia at nearly 100 megawatts. It aims to prove that a data center can be a precise, controllable load, rather than drawing constant, unrelenting power around the clock and placing unprecedented strain on a rigid grid.
Data center developers that don’t have electricity 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 and inevitably increases costs for everyone else, as private systems and the public grid compete for the same turbines and transformers. This behind-the-meter approach also deprives utilities of anchor customers that represent flexible data centers, high-paying customers whose revenue could have funded upgrades and kept rates low for everyone else.
State and federal regulators are already exploring flexible data centers as a possible solution. In June, the Federal Energy Regulatory Commission ordered the six regional grid operators it oversees to make room for large customers willing to limit their electricity use in exchange for faster connections. The Texas grid operator is finalizing rules that will allow controllable data centers to be connected more quickly, and Silicon Valley Power, a California municipal utility, has launched the country’s first flexible load connection program.
AEMA will advance this case in state capitals and Washington. Our call to governors, regulators, and federal policymakers is simple: Provide faster, greater connectivity to data centers committed to flexibility—and keep them there. Rarely does a country have more of everything at once: more AI innovation and a stronger American hand in the technology race, curbed energy bills for the communities where the industry is based, and a more reliable grid with the means to improve itself. AI data centers are ready to provide the necessary flexibility; We can all benefit if those who operate and regulate the grid give it space.
The opinions expressed in Fortune.com comments are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Assets. Varun Sivaram‘S The opinions are his alone and not those on behalf of AEMA or other founding members.