The U.S. Department of Energy is said to be backing a $100 billion private investment to turn a former uranium enrichment site in western Kentucky into the Paducah American Energy Hub, a plan that folds the AI buildout frenzy into one of the country’s most fraught nuclear-era locations. If it proceeds as described, Brookfield Asset Management would build a 1.8-gigawatt AI and high-performance computing campus while NextEra Energy develops 2 gigawatts of natural gas-fired generation and 2.6 gigawatts of battery storage. The pitch is simple: power the machines, cool them efficiently and keep the grid from footing the bill.
What makes the project stand out is not just its size, but its timing. The market is starting to question whether the capital needs of artificial intelligence can keep outrunning the costs, delays and community pushback that come with them. A Kentucky project on a Cold War nuclear footprint is the latest example of how far the industry is willing to go to secure electricity, land and political cover. It also lands as investors grow more cautious about heavy AI infrastructure spending, even as the biggest names in the business keep pressing ahead.
According to the report, the phased buildout is targeted for 2032 and is expected to create roughly 8,000 construction jobs and 600 permanent operations jobs. The project would pair dedicated power supplies with a closed-loop cooling system meant to recirculate water and reduce outside consumption. That is a deliberate contrast with the site’s earlier life as the Paducah Gaseous Diffusion Plant, which was originally designed to pull up to 30 million gallons of water per day from the Ohio River.
That resource-efficiency message matters because data centers have become a political flashpoint. AI campuses need not just land and chips, but vast amounts of power, cooling and local tolerance. Kentucky’s pitch, at least as framed in the report, is that the facility could even feed excess energy back to the regional grid, easing rather than worsening costs for local families and businesses. That is the kind of promise communities are hearing more often as developers try to blunt resistance to big industrial projects.
The Paducah plan arrives at a moment when the market is souring on the scale of AI infrastructure spending. That is a shift from the mood that greeted Stargate, the private-sector venture unveiled last year by Donald Trump with Oracle’s Larry Ellison, OpenAI chief executive Sam Altman and SoftBank chief executive Masayoshi Son. Stargate was framed as a $500 billion effort over four years to build hyper-scale AI data centers and energy infrastructure across the United States. Trump called it the largest AI infrastructure project in history, and the early development phase began in Abilene, Texas.
But the financing has not looked nearly as easy as the headlines suggested. In January, Business Insider reported that JPMorgan Chase had trouble syndicating $38 billion in construction debt tied to the Stargate initiative as institutional investors became more cautious about concentrated tech exposure. Banks spent more than six months trying to place the debt packages linked to campuses in Texas and Wisconsin, but buyers were reluctant to absorb huge single-tenant AI risks. Oracle leases the facilities, and S&P cut Oracle in July to BBB-, one notch above junk, over its data center spending. The result: investors wanted more return, and the banks had to sell at a discount.
That is the kind of financing drag that can slow even the biggest ambitions. AI developers are not just building software; they are building industrial systems that need long-dated debt, utility access and faith that usage will justify the spend. When the credit market gets more nervous, the whole timeline can wobble.
The pressure has not stopped at financing. In April, OpenAI officially paused its Stargate UK data center project, citing high industrial electricity prices and regulatory uncertainty. The plan had been set with hardware firms Nvidia and Nscale and was supposed to deploy thousands of advanced AI chips, starting with 8,000 GPUs and scaling up to 31,000, across sites in North-East England, including Cobalt Park near Newcastle and Blyth. OpenAI did not cancel the project, but said it would only resume when the right conditions allow for long-term infrastructure investment.
That pause matters because it hits more than one company. The report says it dealt a major blow to Nscale, which had been scaling up operations and considering a stock market listing, and to Britain’s effort to become a sovereign AI power. It also reinforces a bigger point: even when demand for AI compute remains intense, the supporting infrastructure can still get bogged down by electricity costs, planning rules and financing risk. That is where the fantasy of unlimited scale starts to look like a utility bill.
Last month, Bloomberg reported another setback for the broader AI buildout push: Oracle’s $18 billion Project Jupiter in New Mexico had stalled. Oracle issued a force majeure notice to lender Blue Owl Capital after it failed to secure key state and federal land-use permits for a planned natural gas plant. Oracle then moved toward solid oxide fuel cells, an experimental backup that has not been deployed on that scale. The pattern is hard to miss. Projects keep getting announced at eye-watering size, then run into the less glamorous realities of permits, infrastructure and local opposition.
That backlash is no longer anecdotal. A recent Gallup poll cited in the report found that roughly 71% of Americans oppose the construction of AI data centers in their localities. The same report says pushback in the United States has already resulted in more than $130 billion of projects being blocked or delayed, with more than 500 active data center bans or moratoriums nationwide. Those numbers help explain why developers now emphasize jobs, water efficiency and grid support almost as much as they emphasize computing power.
For Brookfield and NextEra, the Kentucky project is a chance to show that AI infrastructure can be framed as a regional development story rather than an energy drain. For the market, though, it is another reminder that the AI trade is entering a harder phase. The winners now may not be the firms that announce the biggest numbers first, but the ones that can actually secure power, financing and local approval without watching the timeline slip.
Paducah could become a model if the project gets built on schedule. It could also become another example of how difficult it is to turn AI ambition into steel, turbines and concrete. Either way, the story is no longer just about demand for computing. It is about whether the capital market, the power grid and the public can keep up with it.
