A startup that promised to revolutionize AI infrastructure by building dedicated nuclear power plants for data centers has imploded spectacularly, raising serious questions about investor due diligence in the AI boom. Fermi Energy, which went public in October 2025 with a market value exceeding $19 billion despite having zero revenue and no signed customers, fired its chief executive in May 2026 after months of failed negotiations to land even a single client.

What Happened to Fermi Energy's Nuclear AI Vision?

Fermi's pitch seemed almost irresistible on paper. The company proposed building power plants capable of generating 17 gigawatts of electricity, roughly three times the amount typically consumed by New York City. The plan involved installing hyperscalers' data centers directly on-site, allowing them to tap into dedicated power from natural gas turbines initially, with a transition to nuclear reactors later. The combination of artificial intelligence, nuclear energy, and political connections attracted investors who saw the future of AI infrastructure taking shape.

The company's flagship project, dubbed Project Matador or the President Donald J. Trump Advanced Energy and Intelligence Campus, sprawled across more than 5,000 acres in the Texas panhandle. Yet despite the grand vision and substantial funding, the site remains mostly unfinished. In May 2026, Fermi's board fired Chief Executive Toby Neugebauer after extended negotiations failed to produce a single binding customer agreement. Chief Financial Officer Miles Everson also departed. The company's stock has since tumbled 84 percent from its peak, leaving investors with a cautionary tale about hype outpacing reality.

Why Did Investors Back a Company With No Customers?

The collapse of Fermi reveals a fundamental disconnect between investor enthusiasm for AI infrastructure and the practical realities of energy markets. Jigar Shah, a former official at the U.S. Department of Energy who ran the Loan Programs Office during the Biden administration, offers a blunt assessment of what went wrong.

"We're allowing these types of projects to continue to be viewed as viable when they most certainly are not," Shah stated, calling Fermi a failure "of monumental proportions."

Jigar Shah, former Director of the Loan Programs Office, U.S. Department of Energy

Shah explained that the fundamental premise of off-grid data center power is flawed from a financing perspective. Banks and major hyperscalers prefer the reliability of grid power, which draws from multiple sources and has proven infrastructure. A handful of expensive, on-site nuclear plants cannot compete with that redundancy and stability. One investor who visited the Texas site in February described it bluntly to short sellers: "A piece of dirt with a dream".

Shah

How to Evaluate AI Infrastructure Investments More Carefully

• Demand Signed Contracts: Investors should require evidence of binding customer agreements before valuing a company, not just theoretical demand or political enthusiasm. Fermi's complete lack of signed clients should have been a red flag from the start.
• Assess Regulatory Feasibility: Nuclear power plants take decades to build and face extensive regulatory scrutiny. Projects promising rapid deployment of nuclear capacity should be viewed with skepticism, particularly when timelines don't align with regulatory realities.
• Verify Market Demand From Actual Customers: Hyperscalers have existing relationships with grid operators and established power procurement strategies. New entrants claiming to disrupt this market need documented interest from major customers, not just investor enthusiasm.
• Compare Against Proven Alternatives: Grid power, despite its challenges, has proven reliability and financing mechanisms. Any alternative must demonstrate clear advantages, not just theoretical benefits, to justify the additional risk.

The Fermi collapse echoes earlier cautionary tales in clean energy. The bankruptcy of Solyndra, a renewable-energy firm that received Department of Energy loan guarantees, became a symbol of government investment failures. However, that program ultimately returned profits of $5 billion or more across all investments. The difference is that Solyndra at least had a real product and actual market competition; Fermi never progressed beyond a business plan.

Shah's warning deserves particular attention given the current AI infrastructure frenzy. The market's enthusiasm for AI-related ventures has created an environment where companies can attract billions in funding based on grand visions rather than demonstrated customer demand or operational feasibility. Fermi's $19 billion valuation despite zero revenue represents an extreme case, but it reflects a broader pattern of investor optimism outpacing business fundamentals.

The implications extend beyond Fermi itself. Similar off-grid data center projects are being developed elsewhere, and many deserve more skepticism than they have received from investors and regulators. The nuclear-AI energy sector has attracted legitimate interest from major technology companies and serious energy firms, but the Fermi failure demonstrates that not every venture combining these buzzwords represents a viable business opportunity. As AI infrastructure becomes increasingly critical to the technology industry, distinguishing between transformative projects and speculative ventures will become more important, not less.