OpenAI is calling on the federal government to close the “electron gap” between the United States and China, by patching up “vulnerabilities in the U.S. industrial base” — and installing 100 gigawatts of new capacity on the grid each year, starting in 2026.
In an 11-page letter, sent yesterday to the White House Office of Science and Technology Policy, the company outlined its vision for keeping pace with China’s massive infrastructure buildout. This includes everything from embracing load flexibility and accelerating new transmission lines, to establishing a strategic reserve of “the raw materials essential to AI infrastructure.”
Adding that amount of capacity will require doubling down on an approach China itself has been using for two decades: investment in the electric tech stack, also known as the electro-industrial stack. This is the vertically integrated network of mining, refining, manufacturing, and grid infrastructure that underpins the emerging electricity-based economy.
“Every stage of the AI buildout — from critical minerals and semiconductor fabrication to grid components and data center construction — relies on globally concentrated inputs or constrained manufacturing capacity,” the letter explained.
China currently dominates nearly every one of these sectors. The U.S., meanwhile, has invested heavily at the top of that stack, and its focus on algorithms, chips, and cloud platforms has made it the global leader in developing artificial intelligence.
But maintaining that leadership, OpenAI wrote, will require taking a leaf out of China’s book and “treating capacity as the foundation of industrial competitiveness.”
This is an echo of the language of the Trump administration, which has long emphasized the importance of U.S. leadership in both artificial intelligence and manufacturing. But in this case, the rhetoric is being deployed to encourage a rapid capacity buildout that — to read between the lines — will inevitably be dominated by renewables, as well as some gas. Distributed energy and batteries are some of the fastest-to-deploy ways of getting new capacity on the grid, followed by utility-scale renewables and gas, though the latter is becoming more challenging as the backlog for gas turbines gets longer and the equipment gets more expensive.
Getting to 100 GW
OpenAI’s letter outlines three key recommendations for enabling the addition of 100 GW: strengthening America’s industrial base by investing in manufacturing, unlocking more energy by removing or modernizing regulations, and funding workforce development. At the same time, the letter adds, the federal government should “ensure that frontier AI systems protect American national security interests,” including by “removing unnecessary regulatory barriers to safety testing and government adoption.”
In order to strengthen the industrial base, OpenAI argued, the Trump administration should expand the Advanced Manufacturing Investment Credit to include the semiconductor supply chain, grid components like transformers and the specialized steel for their production — as well as to AI servers and data centers themselves.
“Broadening coverage of the [credit] will lower the effective cost of capital, de-risk early investment, and unlock private capital to help alleviate bottlenecks,” the letter argued.
But countering China will require more federal financial support than tax credits alone. OpenAI also called on the government to deploy grants, cost-sharing agreements, loans, or loan guarantees in industries where China’s own investments are distorting markets, like copper, aluminum, electrical steel, rare earth elements, and semiconductor inputs. The Loan Programs Office in particular, OpenAI said, should be leveraged to this end.
The letter points to several opportunities for the federal government to accelerate transmission build out and permitting, including by shortening timelines for Clean Water Act and Clean Air Act certifications, and by leveraging AI itself.
It also pays particular attention to the potential for flexible loads, and DOE’s recent instructions to FERC to speed up the interconnection for large loads that can be curtailed. Those new rules could be built on the “Connect and Manage” model used in Texas, the letter explained. “Our data centers are designed to be curtailable — reducing their draw or even returning power during peak demand, helping to protect reliability and avoid higher costs for consumers,” it added.
But there’s another key area where China has a massive headstart: labor.
“To close the electron gap specifically, the U.S. needs not only steel and concrete, but also a skilled workforce at a scale not seen in generations,” the letter said, pointing to mechanical, electrical, and plumbing trades. Electrical engineers in particular are already in short supply in the U.S. thanks in large part to the AI boom.
According to an analysis that OpenAI commissioned of its own AI infrastructure buildout plans — which already include six Stargate sites under construction in Texas, New Mexico, Ohio, and Wisconsin — the company’s projects alone will require “an estimated 20% of the existing workforces in skilled trades such as specialized electricians and mechanics.”
The company’s proposed approach includes connecting AI companies with community colleges and trade schools via local “AI Hubs.” These would be federally funded, but managed at the state and local level. “Ideally communities would participate in the co-design of these sessions, to provide the expertise on how best to tailor training for their residents,” the letter added.
