New deal, same old fuel
The AI boom has a shiny, futuristic vibe — until you follow the power cord. According to Global Energy Monitor, the U.S. has overtaken China in new gas-fired power plant construction, with gas capacity under development jumping 50% since January, from 252GW to 378GW.
That’s not a typo. The country is building twice as much gas-fired capacity as China and more than anywhere else on Earth. If all those projects get finished, the U.S. gas fleet would expand by roughly two-thirds. The bill? More than $647 billion.
The hidden price tag on your chatbot
The pitch for all this infrastructure is simple: AI needs electricity, and lots of it. But gas isn’t exactly a free lunch. GEM says completing these projects could lift U.S. power emissions by as much as 20%, while also locking utilities and ratepayers into decades of fuel-price whiplash.
In other words, the AI race may be creating a very unsexy side quest: building a mountain of fossil-fuel infrastructure to keep the servers humming.
Big tech, meet public backlash
Alphabet, Microsoft, Meta, Oracle, Amazon, OpenAI, and xAI all signed the White House’s voluntary Ratepayer Protection Pledge in July, a pretty on-the-nose reminder that people are getting nervous about data centers driving up residential electricity bills. And now even Republicans are whispering about political blowback if local communities keep pushing back.
That’s the awkward twist here: companies want the compute, communities want the jobs, but nobody wants the higher bill or the emissions. So the AI buildout is starting to look less like a pure tech story and more like a full-on utility, policy, and infrastructure battle.
Big picture: AI still looks like the growth engine of the decade, but the energy math is getting messy fast. The winners may be the companies that can secure power without turning every data center into a climate and regulatory headache.