Remember When Nuclear Was Going to Save AI?
Remember when every AI company was announcing a nuclear power deal and it felt like the energy problem was basically solved? Microsoft, Google, Amazon — they were all signing agreements, dusting off old reactor sites, and telling investors that clean, always-on nuclear power was the obvious answer to running data centers that eat electricity like it’s free. Fermi Inc. was supposed to be the pure-play version of that bet. Not a tech giant dabbling in energy. An actual startup built around the idea of powering AI infrastructure with nuclear from day one.
That story got a lot harder to tell this week.
Two Exits, One Very Bad Week
On April 17, 2026, Fermi co-founder and CEO Toby Neugebauer stepped down. CFO Miles Everson followed shortly after. The market did not take it well — shares dropped 22% on the news, with some reports tracking the decline closer to 17-18% depending on the session. Either way, that’s a significant chunk of market value gone in a matter of days.
Neugebauer isn’t completely out of the picture. He’s staying on as chairman of the board, which is the kind of move that’s meant to signal continuity but often reads as a soft landing after things didn’t go as planned. When a founder steps back from the CEO role but keeps a board seat, it usually means the company needs a different kind of operator — someone who can execute where the visionary couldn’t.
What Fermi Was Actually Trying to Do
For anyone who hasn’t been following this one, Fermi was co-founded with former U.S. Energy Secretary Rick Perry attached to the project — which gave it a certain political credibility in the energy space. The core idea was building an AI campus in Texas powered by nuclear energy. Not buying offsets. Not signing a future delivery contract. Actually building the infrastructure to run AI workloads on nuclear power.
That’s an ambitious plan under any circumstances. Nuclear projects in the U.S. are notoriously slow, expensive, and politically complicated. Pairing that with the speed expectations of the AI industry — where investors want to see compute capacity come online fast — creates a tension that’s genuinely hard to manage. You’re essentially asking two very different industries to move at the same pace, and they don’t.
What This Looks Like From a Toolkit Reviewer’s Seat
I spend most of my time here at agntbox.com looking at AI tools — what works, what overpromises, what quietly disappears after a funding round. Fermi isn’t a tool I’d review in the traditional sense, but the pattern here is familiar. A company builds a compelling narrative around a real problem (AI needs a lot of power, nuclear is clean and reliable), attracts serious attention, and then runs into the part where the narrative has to become an actual operating business.
That transition is where most things break. And losing both your CEO and CFO simultaneously — the two people most responsible for strategy and financial execution — is not a sign that the transition is going smoothly.
The Texas Campus Question
Fermi’s AI campus in Texas was the centerpiece of its pitch. Texas has a deregulated energy market, land, and political appetite for big infrastructure projects. On paper, it makes sense. But the departures suggest the company has been hitting serious obstacles in turning that vision into something operational. We don’t have specifics on what those obstacles are — permitting, financing, technical delays — but the timing of two senior exits points to something more than routine leadership transition.
What Happens Next
Fermi will need to move quickly on naming a new CEO who can actually close the gap between the company’s ambitions and its current position. Investors are already spooked. A credible operator with energy infrastructure experience — not just AI hype credentials — would go a long way toward stabilizing confidence.
The underlying thesis isn’t necessarily wrong. AI does need power. Nuclear does offer things that solar and wind can’t match for always-on compute workloads. But a good thesis and a functioning company are two different things, and right now Fermi has to prove it can be the latter.
From where I sit, this is a company that sold a very good story. The next chapter depends entirely on who’s left to write it — and whether they can actually build the thing.
🕒 Published: