Analysis
Holtec priced its IPO this week at up to $10.2 billion, and I think it's underpriced relative to the actual thesis it's selling, which has nothing to do with spent-fuel storage or decommissioning -- Holtec's slower, steadier existing businesses -- and everything to do with one line in its prospectus: it owns and is restarting an 800-megawatt reactor under long-term power purchase agreements, in a year when every hyperscaler building an AI data center has discovered that GPUs are not the constraint. Power is.
The same thesis, priced twice
That is the same thesis behind SB Energy's IPO, which Pulse has covered carrying a $439 billion contract backlog against $269 million of current revenue -- a company effectively pre-selling decades of power capacity to data-center operators before it has built most of it. The two companies sell different generation technology, solar and storage versus nuclear restart and SMRs, but they are pricing the identical bet: that the AI buildout has turned firm, contracted electricity into a scarcer and more valuable asset than the compute it powers.
“A generalist growth investor comparing Holtec to a normal utility IPO will anchor on its historically modest storage-and-decommissioning growth rate.”
Why the market hasn't priced it yet
I don't think public markets have fully priced that yet, because energy investors and AI investors are still mostly different people reading different filings. A generalist growth investor comparing Holtec to a normal utility IPO will anchor on its historically modest storage-and-decommissioning growth rate. An AI infrastructure investor who has watched Crusoe, Nscale and Fluidstack all cite power access as the hard limit on growth will read the exact same prospectus and see a scarce-asset story. That mispricing gap is where the opportunity sits, at least until the comp set catches up.
Room for disagreement: The strongest counter is that nuclear restart and SMR execution risk is real and historically brutal -- SMR programs across the industry have missed timelines for a decade, Palisades is the first restart of its kind with no direct regulatory precedent, and a single serious delay or safety incident resets the entire thesis overnight in a way solar and storage delays do not. Nuclear's premium may be exactly right, not underpriced, given that tail risk. If you believe execution risk in nuclear restarts is systematically underestimated by AI-infrastructure investors chasing a power-scarcity narrative, Holtec at $10.2 billion is fully priced or worse.
What would actually re-rate this
What I'd actually underwrite: whether Holtec's SMR order book, not its Palisades restart, is what re-rates the stock in year two. Restart execution is a known, if hard, engineering problem. A real SMR customer contract from a hyperscaler would be the first proof that the power-scarcity thesis extends past legacy nuclear assets into genuinely new capacity.
There is a second-order trade underneath this one worth flagging for anyone reading the S-1 line by line: Holtec's decommissioning business, the least glamorous of its three segments, is the one with the least correlation to whether the AI-power thesis pans out at all. A fund that wants exposure to the nuclear-restart story without betting the entire position on hyperscaler demand materializing on schedule could reasonably treat Holtec as a blended instrument -- a defensive, cash-generating base with a real option attached, priced as if it were the option alone.
