The pitch is unlike anything the nuclear industry has marketed before: take reactors built for the Navy, retire them from military service, and put them to work powering the artificial intelligence boom. Bloomberg reported that HGP Intelligent Energy is in talks to list through a special purpose acquisition company at a valuation of about $1.2 billion, joining the parade of energy ventures seeking public capital on the strength of AI’s appetite for electricity.
HGP’s model rests on a specific resource. The company proposes to use decommissioned naval reactors, the power plants that once drove submarines and warships, as the heart of facilities supplying electricity to AI data centers. The reactors would be paired with systems that can adjust output to follow the needs of both the power grid and the computing loads they serve, flexibility that nuclear plants have historically not been known for.
The business case is a product of the AI industry’s peculiar energy problem. Data centers need power around the clock, in quantities that strain grids designed for different consumption patterns, and they need it without carbon emissions to satisfy the climate commitments of their owners. The largest cloud companies have concluded that nuclear power is the only technology that can provide zero-carbon electricity on the scale and schedule they require, and they have signed agreements with utilities and startups to bring reactors online.
That conclusion has transformed the economics of nuclear energy. For two decades, the industry’s problem was finding customers willing to pay for new plants; today, the problem is finding enough reactors to meet the demand of technology companies willing to commit to decades of purchases. The shortage has pushed buyers toward every available source of nuclear capacity, from restarting retired plants to investing in startups building small reactors, and now to the idea of repurposing naval propulsion reactors.
HGP is part of a wave of companies built on the intersection of nuclear power and artificial intelligence. The most prominent examples have involved restarts and new construction: a utility bringing an idled Pennsylvania plant back to life to serve Microsoft’s data centers, cloud companies investing billions in reactor developers, and technology executives describing nuclear as the only path to the electricity their models require. The industry’s investors have rewarded the trend, with shares of nuclear-related companies rising as AI spending has grown.
The naval reactor concept adds a twist that makes financiers take notice. The reactors HGP proposes to use were built at government expense, and much of their engineering has been validated through decades of military operations. In theory, that means the company can acquire power plants whose design and safety case are already proven, at a fraction of the cost of new construction, and deploy them where AI data centers need them most.
The path is not simple in practice. Naval reactors were designed for ships, not for commercial power generation, and converting them for land-based use raises engineering, licensing and regulatory questions that no company has fully answered. The Nuclear Regulatory Commission’s process for licensing a commercial plant is measured in years, and the novel provenance of HGP’s reactors would make the review anything but routine. Whether retired naval reactors can be certified for commercial operation is an open question at the center of the company’s entire business plan.
The company’s flexibility pitch is designed for the AI market’s actual needs. Data centers have historically been treated as fixed loads, but the AI workload is more variable, with training runs and demand surges that can be shifted. HGP says its facilities can modulate output to follow grid conditions and computing demand, a claim that, if it holds, would make its plants more valuable than conventional baseload reactors, which prefer to run flat.
The SPAC route carries its own signals. Companies that list through a merger with a blank-check firm typically do so when they need capital quickly and want to avoid the scrutiny of a traditional IPO process. For a nuclear startup with regulatory questions ahead, the speed of a SPAC deal has obvious appeal, even if it means accepting the skepticism that has attached to the vehicle since the last wave of SPAC listings ended in disappointment for many investors.
The timing also reflects a broader shift in how energy markets are being valued. Utilities and power developers that once traded on predictable dividends now carry premiums tied to their ability to serve data centers, and companies with no operating plants have raised billions on the strength of contracts with technology firms. The reordering has drawn comparisons to the early days of other infrastructure booms, when capital flowed to concepts before the engineering was proven.
HGP’s talks are at an early stage, and Bloomberg reported that terms could change or the deal could fall apart. What the report captures is the state of the market: artificial intelligence has grown so hungry for power that investors are willing to price a company based on second-hand naval reactors at more than a billion dollars, before a single such reactor has been licensed for commercial use. The energy industry’s valuation system has been redrawn by the AI boom, and stories like HGP’s are the evidence.
