Texas-based startup HGP Intelligent Energy has a radical proposal to solve AI’s crippling power crunch: repurpose the powerful nuclear reactors from U.S. Navy aircraft carriers to generate electricity for massive data centers. Their CoreHeld Project aims to deliver 520 MW of constant power by 2029, offering a potentially faster, cheaper alternative to building traditional nuclear plants.
The artificial intelligence boom is hitting a formidable wall: energy. Data centers required to train and run advanced AI models are so electricity-hungry they’re projected to drive up to 40% of U.S. power demand growth by 2035. Tech giants are desperately seeking reliable, around-the-clock power, often clashing with grid policies designed for intermittent renewable sources. “Because of this, tech companies are scrambling to find reliable, 24/7 sources of electricity,” reported New Atlas, which first detailed the proposal. The logical answer is nuclear power, but conventional plants are famously slow and expensive to build.
Enter HGP Intelligent Energy and its audacious plan, submitted as part of the White House’s Genesis Mission. Instead of constructing a new civilian reactor from scratch, CoreHeld would take two A1B-class naval pressurized water reactors—the very units built for the new Gerald R. Ford-class supercarriers—and install them in a land-based plant. The pilot project is slated for Oak Ridge National Laboratory (ORNL). The cost and speed advantages are compelling. The company estimates setup costs between $1.8 billion and $2.1 billion, significantly less than conventional reactors, with a much faster timeline because the naval design has over 70 years of proven operational history.
READ ALSO: https://modernmechanics24.com/post/china-bone-free-fish-easier-dining/
“We already know how to do this safely and at scale, and we’re fortunate to have a solid base of investors and partners who share that vision,” said HGP Chief Executive Officer Gregory Forero in a company statement. The plan also cleverly navigates a regulatory thicket. Traditional civilian reactors fall under the lengthy licensing purview of the Nuclear Regulatory Commission (NRC). HGP proposes a hybrid path where the Department of Energy (DOE) and the U.S. Navy—which already oversee these reactor designs—retain control, while the NRC expedites civilian licensing for the connected power plant.
However, the proposal isn’t without major technical and geopolitical hurdles. The core of the challenge is fuel. Naval reactors are designed for performance and longevity in warships, using highly enriched uranium (HEU) with 93% uranium purity. This is weapons-grade material, creating serious non-proliferation concerns and requiring extreme security measures. Civilian reactors use low-enriched uranium (LEU), which is not directly usable in these naval designs. According to New Atlas, a solution may be on the horizon, with plans to modify future naval reactors to use LEU fuel by 2030.
WATCH ALSO: https://modernmechanics24.com/post/chinas-humanoid-robot-embroidery-skills/
Security would be another extraordinary undertaking. The reactor designs are highly classified. Data center staff would have no access, while the maintenance crew would likely need DOE Q-clearance or be veterans of the Naval Nuclear Propulsion Program. Furthermore, these reactors are built to be sealed for their entire 25-year core life (with future designs aiming for 50 years), and are engineered to throttle power output dramatically to propel a ship—a feature not needed for a steady-state power plant.
If these challenges can be overcome, the payoff is immense. A 520 MW dedicated power plant would be a golden ticket for any tech company building an AI data center, providing a massive, carbon-free baseload. The CoreHeld Project is a testament to the extreme measures now being considered to feed AI’s insatiable energy appetite, looking not to untested technology, but to pulling proven, formidable power plants from the heart of American naval supremacy and planting them firmly onshore.
READ ALSO: https://modernmechanics24.com/post/aihelpscouplegetpregnent/
