The numbers are nothing short of staggering. Take Sam Altman, Open AI\u2019s CEO. He reportedly wants 250 gigawatts of new electricity\u2014equal to about half of Europe\u2019s all-time peak load\u2014to run gigantic new data centers in the U.S. and elsewhere worldwide by 2033. <\/p>\n
Building or expanding power plants to generate that much electricity on Altman\u2019s timetable indeed seems almost inconceivable. \u201cWhat OpenAI is trying to do is absolutely historic,\u201d says Virun Sivaram, Senior Fellow for Energy and Climate at the Council on Foreign Relations. The problem is, \u201cthere is no way today that our grids, with our power plants, can supply that energy to those projects, and it can\u2019t possibly happen on the timescale that AI is trying to accomplish.\u201d<\/p>\n
Yet Sivaram believes Altman may be able to reach his goal of running multiple new data centers in a different way. Sivaram, in addition to his position at the CFR, is the founder and CEO of Emerald AI, a startup that launched in July. \u201cI founded it directly to solve this problem,\u201d he says\u2014not just Altman\u2019s problem specifically, but the larger problem of powering the data centers that all AI companies need. Several smart minds in tech like the odds of Sivaram\u2019s company. It\u2019s backed by Radical Ventures, Nvidia\u2019s venture capital arm NVentures, other VCs, and heavy-hitter individuals including Google<\/a> chief scientist Jeff Dean and Kleiner Perkins chairman John Doerr.<\/p>\n Emerald AI\u2019s premise is that the electricity needed for AI data centers is largely there already. Even big new data centers would confront power shortages only occasionally. \u201cThe power grid is kind of like a superhighway that faces peak rush hour just a few hours per month,\u201d Sivaram says. Similarly, in most places today the existing grid could handle a data center easily except in a few times of extreme demand.<\/p>\n Sivaram\u2019s objective is to solve the problem of those rare high-demand moments the grid can\u2019t handle. It isn\u2019t all that difficult, at least in theory, he argues. Some jobs can be paused or slowed, he explains, like the training or fine-tuning of a large language model\u00a0for academic research. Other jobs, like queries for an AI service used by millions of people, can\u2019t be rescheduled but could be redirected to another data center where the local power grid is less stressed. Data centers would need to be flexible in this way less than 2% of the time, he says; Emerald AI is intended to help them do it by turning the theory to real-world action. The result, Sivaram says, would be profound: \u201cIf all AI data centers ran this way, we could achieve Sam Altman\u2019s global goal today.\u201d<\/p>\n A paper<\/a> by Duke University scholars, published in February, reported a test of the concept and found it worked. Separately, Emerald AI and Oracle<\/a> tried the concept on a hot day in Phoenix and found they could reduce power consumption in a way that didn\u2019t degrade AI computation\u2014\u201ckind of having your cake and eating it too,\u201d Sivaram says. That paper is under peer review.<\/p>\n No one knows if Altman\u2019s 250-gigawatt plan will prove to be brilliant or folly. In these early days, Emerald AI\u2019s future can\u2019t be divined, as promising as it seems. What we know for sure is that great challenges bring forth unimagined innovations\u2014and in the AI era, we should brace for plenty of them.\n<\/p>\n