OpenAI’s official “blueprint for U.S. AI infrastructure” involves artificial intelligence economic zones, tapping the U.S. Navy’s nuclear power experience and government projects funded by private investors, according to a document viewed by CNBC, which the company plans to present on Wednesday in Washington, D.C.
The blueprint also outlines a North American AI alliance to compete with China’s initiatives and a National Transmission Highway Act “as ambitious as the 1956 National Interstate and Defense Highways Act.”
In the document, OpenAI outlines a rosy future for AI, calling it “as foundational a technology as electricity, and promising similarly distributed access and benefits.” The company wrote that investment in U.S. AI will lead to tens of thousands of jobs, GDP growth, a modernized grid that includes nuclear power, a new group of chip manufacturing facilities and billions of dollars in investment from global funds.
Now that Donald Trump is President-elect, OpenAI has made clear its plans to work with the new administration on AI policy, and the company’s Wednesday presentation outlines its plans.
Trump plans to repeal President Biden’s executive order on AI, according to his campaign platform, stating that it “hinders AI Innovation, and imposes Radical Leftwing ideas on the development of this technology” and that “in its place, Republicans support AI Development rooted in Free Speech and Human Flourishing.”
OpenAI’s presentation outlines AI economic zones co-created by state and federal governments “to give states incentives to speed up permitting and approvals for AI infrastructure.” The company envisions building new solar arrays and wind farms and getting unused nuclear reactors cleared for use.
“States that provide subsidies or other support for companies launching infrastructure projects could require that a share of the new compute be made available to their public universities to create AI research labs and developer hubs aligned with their key commercial sectors,” OpenAI wrote.
OpenAI also wrote that it foresees a “National Transmission Highway Act” that could expand power, fiber connectivity and natural gas pipeline construction. The company wrote it needs “new authority and funding to unblock the planning, permitting, and payment for transmission,” and that existing procedures aren’t keeping pace with AI-driven demand.
The blueprints say, “The government can encourage private investors to fund high-cost energy infrastructure projects by committing to purchase energy and other means that lessen credit risk.”
A North American AI Alliance and investment in more U.S. data centers
OpenAI also foresees a North American AI alliance of Western countries that could eventually expand to a global network, such as a “Gulf Cooperation Council with the UAE and others in that region.”
The company also outlined its vision for nuclear power, writing that although China “has built as much nuclear power capacity in 10 years as the US built in 40,” the U.S. Navy operates about 100 small modular reactors (SMRs) to power naval submarines, and leveraging the Navy’s expertise could lead to building more civilian SMRs.
OpenAI’s infrastructure blueprint aligns with what Chris Lehane, OpenAI’s head of global policy, told CNBC in a recent interview. He sees the Midwest and Southwest as potential core areas for AI investment.
“Parts of the country that have been ‘left behind,’ as we enter the digital age, where so much of the economics and particularly economic benefits flow to the two coasts… Areas like the midwest and the southwest are going to be the types of places where you have the land and ability to do wind farms and to do solar facilities, and potentially to do some part of the energy transition — potentially do nuclear facilities,” Lehane said.
The infrastructure, Lehane explained, is contingent on the U.S. maintaining a lead over China in AI.
“[In] Kansas and Iowa, which sits on top of an enormous amount of agricultural data, think about standing up a data center,” Lehane said. “One gigawatt, which is a lot, taking, you know, 200-250 megawatts, a quarter of that, and doing something with their public university systems to create an agricultural-based LLM or inference model that would really serve their community but also make them a center of agricultural AI.”
Lehane cited an estimate that the US will need 50 gigawatts of energy by 2030 to support the AI industry’s needs and to compete against China, especially when the country approved 20 nuclear reactors over the past two years and 11 more for next year.
“We don’t have a choice,” Lehane said. “We do have to compete with that.”