Six months ago, nuclear AI startup Fermi looked unstoppable. Fresh off a $450 million Series B that valued the company at $3.2 billion, CEO Sarah Martinez was telling investors that nuclear power would solve AI's energy crisis. This week, both Martinez and CFO David Chen resigned within 48 hours of each other, leaving the Rick Perry-backed company scrambling to save its $2.8 billion Texas project.
Something fundamental just shifted in the nuclear AI gold rush.
Key Takeaways
- Fermi's CEO and CFO both resigned within 48 hours amid Texas project delays
- Nuclear Regulatory Commission oversight and cost overruns pushed completion to Q2 2028
- Nuclear AI startups raised $8.4 billion in 2026 despite mounting sector challenges
The 48-Hour Collapse
Martinez cited "strategic differences with the board" in her resignation letter — the kind of diplomatic language that usually means something uglier happened behind closed doors. She'd been CEO for barely 11 months since joining in January 2025. CFO David Chen, who had shepherded the company through its massive Series B round just three months earlier, followed her out the door before the week ended.
Board Chairman Rick Perry — yes, the former Energy Secretary who once forgot the name of his own department — will serve as interim CEO while the company searches for new leadership. Perry co-founded Fermi in 2024 betting that AI's voracious energy appetite would finally make small nuclear reactors economically viable.
The timing tells you everything. According to sources familiar with the company's finances, Fermi has already burned through $180 million of its latest funding while facing cascading delays in Texas. When your CEO and CFO both bail within two days, they've seen numbers the rest of us haven't.
Texas-Sized Problems
Fermi's 400-megawatt nuclear facility in East Texas was supposed to be the proof of concept — a dedicated power plant serving AI data centers with clean, reliable energy. Instead, it's become a case study in regulatory reality. The Nuclear Regulatory Commission placed the project under enhanced oversight in February after Texas environmental officials raised safety concerns.
Those delays have real consequences. The facility's completion date slipped from Q3 2027 to Q2 2028, adding an estimated $400 million to the project cost. More painful: several hyperscale cloud providers with preliminary agreements for 300 megawatts of capacity are reportedly reconsidering their commitments.
"The nuclear AI power model requires flawless execution on both the technology and regulatory fronts. When either falters, investor confidence evaporates quickly." — Michael Zhang, Partner at Andreessen Horowitz
Here's what most coverage misses: this isn't just about construction delays. It's about the fundamental mismatch between nuclear timelines and AI urgency. Nuclear facilities take 7-12 years to build and decades to pay off. AI companies need power next quarter, not next decade.
The Nuclear AI Paradox
Fermi operates in a sector that raised $8.4 billion in 2026 — more money than sense, arguably. Companies like Helion Energy, Commonwealth Fusion Systems, and NuScale Power are all chasing the same dream: capturing AI's energy demand with nuclear baseload power. The problem is that dream keeps getting more expensive while alternatives keep getting cheaper.
Bloomberg New Energy Finance projects renewable costs will drop another 30% by 2028. Meanwhile, every nuclear AI project faces the same regulatory gauntlet that's strangling Fermi. Microsoft and Google have already started exploring geothermal and advanced battery storage — solutions that don't require a decade of federal approvals.
The math is getting uncomfortable. Energy represents up to 40% of operating costs for large AI deployments. When your nuclear plant takes twice as long and costs twice as much as projected, you're not solving the energy crisis — you're creating a capital allocation crisis.
Money Talks, Executives Walk
Those executive departures have immediate financial implications. Fermi was reportedly preparing an $800 million Series C for Q3 2026 — funding that's now in serious jeopardy. Lead investor Breakthrough Energy Ventures declined to comment on the round's status, which is never a good sign.
The company's cash runway extends through Q2 2027 based on current burn rates, but those Texas delays could accelerate spending dramatically. With 340 employees across Austin, Houston, and D.C., Fermi faces the classic startup death spiral: delays increase costs, increasing the funding needed, making funding harder to secure.
Competitor NuScale Power offers a cautionary tale. The publicly-traded nuclear company has lost 60% of its value over the past year amid similar project delays and cost overruns. When even public nuclear companies can't maintain investor confidence, private startups face an uphill battle.
Perry's Nuclear Gambit
Rick Perry now owns this mess. The former presidential candidate who once advocated for eliminating the Department of Energy must now navigate the same regulatory apparatus he once criticized. His immediate challenge: stabilizing operations while addressing the mounting regulatory issues that spooked his top executives.
The broader industry is watching closely. Several other proposed nuclear AI facilities face opposition from environmental groups and local communities. If Fermi — with its high-profile backing and substantial funding — can't make the model work, it raises fundamental questions about whether nuclear-powered AI infrastructure is viable at all.
The next six months will reveal whether this executive exodus represents a temporary setback or the beginning of a broader reckoning in nuclear AI. Either way, the easy money phase of this sector just ended.
