Big Tech is charging ahead with mind-boggling new AI data centre plans and demanding unprecedented electricity supply, even as concerns grow about who will bear the grid costs and whether the massive buildout will ever pay off.

OpenAI recently wrote to the United States’ Donald Trump administration, urging [pdf] it to target 100 gigawatts of new electricity generation per year and warning that power supply limits could threaten a “once-in-a-century opportunity.”

An “unlocking of electrons” will prevent the People’s Republic of China from overtaking the U.S., OpenAI Chief Global Affairs Officer Christopher Lehane wrote in the letter to the White House Office of Science and Technology Policy. While China added 429 gigawatts of new capacity  in 2024, the U.S. added 51gigawatts.

The race for dominance is driving the entire sector. “AI is a sport of kings,” Gil Luria, an analyst at D.A. Davidson, told Reuters. OpenAI CEO Sam Altman “understands that to compete in AI he will need to achieve a much bigger scale than OpenAI currently operates at.” The ChatGPT developer is in the early stages of planning an initial public offering (IPO) by late 2026 or early 2027 that could value the company at up to US$1 trillion, one of the biggest IPOs of all time, the news agency reported.

AI investment now makes up about 40% of U.S. GDP growth. Morgan Stanley said capital expenditure on AI infrastructure could surpass US$3 trillion in the next three years, but fear is mounting that AI companies have overplayed their hand.

“It’s becoming clear that their plausible revenues are an order of magnitude short of what they would need to make money and reward their investors,” renowned energy analyst and RMI co-founder Amory Lovins told The Energy Mix.

Other big AI companies are echoing the OpenAI strategy, Politico reported. In a quarterly earnings call last week, Meta CEO Mark Zuckerberg dismissed investor concerns about overbuilding the technology. He called for “aggressively front-loading building capacity,” arguing that if there was an overshoot, the company would “grow into that and use it over time.”

The U.S. Department of Energy is moving to meet the expected power demand surge. Citing AI- and manufacturing-driven load growth, Energy Secretary Chris Wright recently directed the Federal Energy Regulatory Commission (FERC) to speed up power grid connections for data centres. Former FERC Chair Mark Christie responded to the unusual move in a LinkedIn post, suggesting it raises questions of whether the interconnection of large load customers would threaten reliable power service, and whether residential and other customers would be forced to bear the costs of the large-load interconnections. Another question is whether Wright’s move could be seen as an “intrusion” on state regulatory authority, he added.

Lovins suggests state utility regulators can avoid the problems of unreliable grids, stranded assets, and unfair cost burdens for other electricity users by requiring data centre developers to “guarantee power payments, backed by bonds or insurance that can market-price the risk of default.”

“Smarter grid strategies can support data centre growth without overbuilding,” he adds. Improving demand flexibility, enabling efficiency gains, and co-locating new data centres with clean energy at underused gas plants are some of the ways utilities “can deliver fast, cheap, clean, reliable power to support AI progress on a solid foundation.”

But the letter from OpenAI’s Lehane specifically calls for the U.S. federal government to take more of a leadership role over states and local municipalities. He also provides details of OpenAI’s expansive plan to build 30 gigawatts of computing resources worth $1.4 trillion, adding one gigawatt per week.

Lehane makes a laundry list of changes he’d like to see implemented as quickly as possible to “close the electron gap” between the U.S. and China. It includes more investment in manufacturing, the removal or “modernization” of regulations, and increased adoption of AI for national security purposes.

OpenAI also pitched a “Classified Stargate” initiative to meet national security needs, “mobilizing private capital alongside government partners to establish accredited, classified data centres purpose-built for government AI.”

In Canada, OpenAI’s head of U.S. and Canada public policy and partnerships, Chan Park, recently met with AI Minister Evan Solomon to discuss a potential partnership, CBC News reported. OpenAI promoted its new “OpenAI for countries” to help spread “democratic AI.”

Data sovereignty analysts warn that OpenAI, and any American AI company building in Canada, would be subject to U.S. laws.

Guillaume Beaumier, an assistant professor of political science and international studies at l’École nationale d’administration publique in Quebec City, told CBC News that he was “struck by OpenAI using digital sovereignty as an international selling point.”

“I think we can be very critical of the extent to which it’s going to be able to achieve and promote greater Canadian sovereignty,” he said.

U.S. legislation limits foreign countries’ control over their data stored by American companies. Microsoft executives admitted as much to a Senate inquiry in France, saying the company “cannot guarantee” data sovereignty to other countries, The Register reported.