Across the United States, surging AI demand is straining electricity supplies and reshaping how digital infrastructure gets powered. A bold new concept proposes turning decommissioned US Navy nuclear reactors into clean, continuous power for data centers. The idea is as audacious as it is pragmatic, pairing retired military hardware with civilian innovation. It signals a broader shift toward firm, low‑carbon power for the world’s most energy‑hungry computing clusters.
From warships to workloads
Texas‑based HGP Intelligent Energy has submitted a proposal to the US Department of Energy to repurpose two out‑of‑service naval reactors for an AI data center at Oak Ridge National Laboratory in Tennessee. The plan targets between 450 and 520 megawatts of capacity—enough to support multiple large‑scale training campuses or a dense mix of inference nodes. These units would be drawn from reactors similar to Westinghouse A4W systems on Nimitz‑class carriers and GE S8G units on Los Angeles‑class submarines. Decades of maritime operation give these machines an unusual track record of high uptime and robust engineering.
According to the World Nuclear Association, the US Navy has run more than 100 reactors over 50 years without a major radiological incident. That operational history is central to the project’s claim of safety, reliability, and disciplined procedures. It is also a bridge between national security investments and civilian infrastructure needs.
Economics that challenge the status quo
HGP estimates a conversion cost of roughly $1–$4 million per megawatt, significantly below a greenfield nuclear plant or many first‑of‑a‑kind small modular reactors. Total private capital could reach $1.8–$2.1 billion, including new cooling, grid interconnection, and site hardening. A first phase could be online by 2029, aligning with the next wave of AI model scaling and regional grid constraints. Revenue‑sharing with the government and a dedicated decommissioning fund are built into the proposal’s financial model.
By reusing existing cores, the project would avoid immediate scrapping at the DOE’s Hanford site, giving this equipment a second life. Advocates argue that such economics could help anchor reliable, carbon‑free power near emerging compute hubs without years of conventional siting delays.
Safety, regulation, and public trust
While maritime safety records are encouraging, civilian use will demand fresh licensing, environmental reviews, and community engagement. Regulators will scrutinize fuel handling, emergency planning, and long‑term waste management. The company says it will establish a robust decommissioning fund and adhere to strict oversight. “We have the experience and partners needed to do this safely at scale,” said Gregory Forero, HGP’s chief executive.
Public acceptance remains a pivotal variable. Clear communication about passive safety features, containment design, and site security will shape local and national support. Lessons from naval operations—procedural rigor and redundancy—could translate into civilian confidence.
Technical hurdles and integration
Naval reactors are optimized for compact, maritime environments, not stationary campuses. Conversion will require tailored cooling systems, auxiliary power upgrades, and modern grid interfaces. Placement near Oak Ridge could leverage existing infrastructure, research partnerships, and workforce expertise. The aim is islandable, always‑on power with black‑start capability and minimal curtailment.
Matching thermal output to data center loads will also matter. Liquid cooling, heat reuse, and high power density designs could maximize efficiency and reduce grid stress. Integration with local transmission operators will be crucial to manage contingencies and load shaping.
Implications for the AI buildout
If successful, this approach could redefine the energy stack for hyperscale compute. Instead of chasing scarce grid capacity, operators could co‑locate firm, carbon‑free generation next to their clusters. That would insulate projects from volatile markets, speed deployment, and reduce lifecycle emissions. It might also temper pressure on regional grids already strained by electrification and industrial growth.
Major tech firms exploring SMRs may view repurposed naval reactors as a complementary path—faster than greenfield, cleaner than fossil backup, and more controllable than intermittent renewables alone. Even partial success could catalyze policies that streamline nuclear‑adjacent infrastructure for digital economies.
Key questions to watch
- Which regulatory pathway will govern the naval‑to‑civilian transition?
- How will long‑term waste and decommissioning liabilities be managed?
- What community benefits and local jobs will the project guarantee?
- Can costs stay within the proposed $1–$4M/MW range at scale?
- How will cybersecurity and physical security be integrated end‑to‑end?
Credit: BFMTV
The race to power AI is accelerating, and each year of delay compounds demand. Repurposed naval reactors offer an unusual blend of proven hardware and fresh ambition. If the economics and regulations align, they could become a template for resilient, low‑carbon compute power in the decade ahead. The stakes are high—and so is the opportunity to rethink how we fuel the world’s most advanced machines.
