Coming Fermi IPO Features ‘Donald Trump Nuclear Plant’ to Power AI Data Centers

Rendering of the Fermi America™ Private Grid Campus. (Courtesy Fermi America)

Investors will soon have a chance to buy into a Texas-size IPO that hopes to power the artificial-intelligence revolution.

The stock offering for the real estate investment trust Fermi is sure to grab attention, because it plays into some of the hottest themes in the stock market. Fermi says it will build massive power plants to provide electricity to AI data centers. By 2038, the company expects to be able to generate 11 gigawatts of power—more than five times as much as the Hoover Dam—from several natural gas plants, four large nuclear reactors, arrays of solar panels and batteries.

The nuclear component, which would be the largest of its kind built in the U.S. in decades, will be called the “Donald J. Trump Generating Plant.”

The company announced on Wednesday that it was starting its IPO roadshow and would be seeking to sell 25 million shares at prices between $18 and $22, for a capital raise of up to $550 million. The IPO could value Fermi at more than $13 billion. It plans to list on the Nasdaq and the London Stock Exchange with the ticker FRMI and could go public as soon as next week.

Fermi hasn’t yet generated revenue, and had spent $6.4 million through June 30, filings show**.** The company says it expects “to continue to incur operating losses for the foreseeable future, ” and probably won’t start receiving revenues from tenants until 2027.

Fermi will be organized as a REIT, a business structure that normally generates a lot of cash that is sent back to shareholders as dividends. But Fermi says that its large investment expenses mean that it won’t be issuing dividends for a while.

That may not matter to shareholders who believe in management’s vision to profit from the boom in AI data centers—which are expected to account for nearly half of all electricity demand growth through 2030.

Fermi was co-founded by former Texas Governor and Energy Secretary Rick Perry and longtime private-equity investor Toby Neugebauer. It’s one of dozens of planned projects in Texas that are designed to power data centers. But it’s got some attributes that could help it stand out from the pack.

The company has signed a 99-year lease on a parcel of land from Texas Tech University in Amarillo, Texas, measuring about nine square miles. The campus is close to a large natural gas field and major natural gas pipelines owned by Energy Transfer and ONEOK. Fermi has signed a memorandum of understanding for pipeline capacity from Energy Transfer, where Perry sits on the board of directors, and is negotiating for capacity with ONEOK. Fermi also has contracts for new turbine equipment from Siemens Energy and used equipment from GE Vernova , the company said.

Fermi says it expects to have up to one gigawatt of natural gas power online by the end of 2026, which would be a quick turnaround at a time when capacity constraints have made it difficult to add power plants.

Fermi says that it can power data centers “behind the meter,” essentially disconnecting from the broader power grid. It’s still aiming to have relationships with local utilities and will provide redundant power if necessary, but it says it can operate without those connections.

Fermi has also already raised some money, including $246 million worth of convertible debt, and financing for equipment like turbines. The IPO will help it buy even more equipment.

The financing, long-term lease, and pipeline access are real advantages, said one expert who helps companies finance power projects. “When you go through all those different things on the checklist, they’re positive,” said Patrick Johnson, a Houston-based partner at law firm White & Case, who advises companies on developing energy projects.

Fermi says it could potentially power all 11 gigawatts of its proposed capacity using natural gas, but it would rather do so using several different power sources. One of those power sources is nuclear, which has gaining renewed prominence as President Donald Trump aims to quadruple U.S. nuclear capacity by 2050. In a filing to the Nuclear Regulatory Commission, Fermi said it’s aiming to build four large-scale reactors to create the Donald J. Trump Generating Plant. The reactors would be built by Westinghouse, which has designed the most prominent large advanced reactor in the U.S. But the company’s two most recent reactors in Georgia came in more than $15 billion over budget and took years longer than expected. Fermi is targeting the first reactor to be operational by 2031.

Determining how to value Fermi is complicated. While many other companies are attempting to build power plants for data centers, there are few other stocks that look like Fermi—in fact, its business model may be one-of-a-kind in public markets. Companies such as Vistra and Constellation Energy also sell the electricity they generate at power plants—but they do so over a much broader operating base. And there are other companies that focus on making money from data centers like Coreweave , but they also have larger asset bases and closer relationships with tech companies.

Timm Schneider, founder of energy consultancy Schneider Capital Group, wrote in a note on Thursday that the company faces large execution risks, but that its valuation makes more sense if investors look out a few years.

“The initial implied valuation is rich for 2026 but looks more appealing beyond 2028 if the company delivers on its targets and at the midpoint of the expected price range,” Schneider wrote.

By 2029, the company’s projections imply earnings before interest taxes depreciation and amortization of $3.6 billion, Schneider calculated. At the midpoint of the IPO range, it would trade at 7.4 times its enterprise value to 2029 Ebitda—which would put it in a range familiar to energy investors.

The company’s operating model will be tricky to pull off. But once the stock starts trading, investors may be too focused on the letters AI to care.

Write to Avi Salzman at avi.salzman@barrons.com