Fermi’s Nuclear AI Dream Loses Its Architects

Toby Neugebauer co-founded Fermi on a bold premise — that nuclear power could solve the energy crisis threatening to choke AI’s growth. Now he’s gone, stepping down as CEO effective April 17, 2026, with CFO Miles Everson following close behind. For a company built around one of the most technically demanding intersections in modern infrastructure, losing both its chief executive and financial officer in rapid succession is not a routine leadership shuffle. It’s a signal worth reading carefully.

What Fermi Was Actually Trying to Do

Fermi’s pitch was architecturally interesting from a systems perspective. AI training and inference workloads are extraordinarily power-dense. A single large-scale GPU cluster can draw hundreds of megawatts continuously, and the grid in most regions simply wasn’t designed for that kind of sustained, concentrated demand. Fermi’s answer was to pair AI compute campuses directly with nuclear generation — stable, carbon-free baseload power that doesn’t fluctuate with weather or time of day.

The company, co-founded alongside former U.S. Energy Secretary Rick Perry, was developing an AI campus in Texas. The geographic choice made sense on paper: Texas has a deregulated energy market, available land, and political appetite for large industrial projects. But the technical and regulatory path from concept to operational nuclear-adjacent AI campus is extraordinarily long, expensive, and fragile to leadership instability.

Why These Departures Hit Differently

Markets reacted immediately. Fermi’s shares dropped 22% following the announcements, with some reports tracking a decline closer to 17.56% on a single trading day. That kind of reaction tells you investors weren’t just surprised — they were recalibrating their entire thesis on the company.

Neugebauer transitioning to board chairman rather than exiting entirely is the one detail that softens the narrative slightly. It suggests this wasn’t a falling-out or a scandal-driven exit. But it also doesn’t explain why both the CEO and CFO departed at the same time, and that simultaneity is what makes this feel less like a planned transition and more like a structural problem surfacing.

From a technical program management standpoint, nuclear infrastructure projects require an almost obsessive continuity of leadership. Regulatory relationships, utility negotiations, site permitting, and reactor procurement timelines are all deeply personal — they depend on specific individuals who have built trust with counterparts over years. When those people leave, the institutional knowledge and relationship capital they carry doesn’t automatically transfer to a successor. It has to be rebuilt, and that takes time the company may not have.

The Harder Problem Underneath

There’s a deeper architectural tension in Fermi’s model that these departures bring into focus. Nuclear power plants — even small modular reactors, which are still largely pre-commercial — operate on decade-scale development timelines. AI infrastructure, by contrast, moves on 18-to-36-month cycles. The GPU generation you’re designing a data center around today may be two generations obsolete by the time a new nuclear facility comes online.

This mismatch isn’t fatal on its own. There are ways to structure the problem — using existing or near-term nuclear capacity rather than building new, or designing facilities that can absorb power from multiple sources. But it requires a leadership team that deeply understands both the energy side and the compute side, and can hold that tension productively over a long horizon. That’s a rare combination, and Fermi just lost two of the people who were supposed to be holding it together.

What This Means for the AI Energy Space

Fermi is not alone in chasing nuclear power for AI. Microsoft has explored restarting Three Mile Island’s Unit 1 reactor. Google has signed agreements around small modular reactor development. Amazon has made similar moves. The thesis — that AI’s power appetite will outpace what renewables and the existing grid can reliably deliver — is broadly shared across the industry.

But most of those efforts are backed by companies with enormous balance sheets and the ability to absorb years of slow progress. A startup operating in this space faces a fundamentally different risk profile. Capital markets are impatient, and a 22% single-day drop is a reminder that investor confidence in early-stage nuclear-AI ventures is thin and easily shaken.

Reading the Architecture of a Company

As someone who spends most of my time thinking about system architecture, I find leadership structures genuinely diagnostic. The way a company organizes its decision-making, and who it puts at the center of that structure, tells you a lot about what it believes its hardest problems are. Fermi just lost the two people most responsible for its strategic and financial architecture simultaneously.

That doesn’t mean the company is finished. New leadership could bring fresh energy and clearer execution focus. But the clock is running, the technical challenges haven’t gotten easier, and the Texas AI campus still needs to be built. Whoever steps into these roles inherits a genuinely difficult problem — and not much runway to figure it out.

🕒 Published: April 21, 2026