The story of AI power stocks is not a story about investors losing nerve. It is a story about the market discovering, one company at a time, that a financial clock and a physical clock run on different schedules — and that the gap between them is not closing.

The financial clock is simple. Moses Sutton of BNP Paribas named its mechanism explicitly: “Today what investors are willing to pay for is the pre-2030, pre-2032 narratives. They want to put less weight into what they believe will happen in the 2030s.” That sentence is not a prediction about the future. It is an operational rule the market is currently enforcing. Stocks whose revenue story stretches past 2030 are being discounted. Stocks whose revenue arrives sooner are not.

The physical clock operates on its own schedule and answers to no analyst. The nuclear permitting process runs on years-to-decade timelines that the Nuclear Regulatory Commission controls. Skilled labor — electricians, pipefitters, welders — is a finite pool that construction targets are already drawing down. Manufacturing capacity takes years to expand, as GE Vernova’s 2030 target demonstrated in the same week the company reported an 88% surge in orders. The physical clock cannot be accelerated by sentiment.

These two clocks are converging on the same set of companies. The mechanism is straightforward: the financial clock reprices the stock; the physical clock slows the construction that would produce the revenue that would validate the stock. Together they create a trap that repricing alone cannot break.

Oklo is the cleanest example because its trajectory is the most extreme. The advanced nuclear developer backed by Sam Altman lost roughly 75% of its value from a peak north of $25 billion last October, despite announcing deals to provide more than 14 gigawatts of power for Meta and others. One-point-two gigawatts of that total is a binding agreement with Meta. The rest consists of unsecured commitments and letters of intent. The company has not secured key permits from the NRC and has never built a power plant. Every month that passes with no permit application advancing, the discount investors apply to those 14 gigawatts grows. The selectivity loop — investors reallocating from distant-revenue to near-revenue names — has already squeezed Oklo from $25 billion to roughly $6 billion. And each successive squeeze tightens financing conditions, which makes it harder to fund the permitting work that would compress the timeline, which extends the timeline further. That is the vicious cycle. It is structurally self-reinforcing, and it is not finished.

Standard Nuclear faces the same liability from the opposite side of the value chain. It is a uranium fuel producer for advanced reactor designs — a market that does not yet exist because the reactors that would use its fuel are not commercially operating. The company went public last week after cutting its fundraising target from roughly $356 million to $150 million because demand was tepid. Shares lost 36% of their value in the first week of trading. The disconnect here is even starker than at Oklo: Oklo at least has a binding contract with a hyperscaler buyer. Standard Nuclear has a product with no market.

The selectivity mechanism is discriminating, not sweeping. GE Vernova reported second-quarter 2026 orders of $24.2 billion, up 88% year over year, and carries a backlog of $176 billion. Its shares declined in the same week. The reason was that the company set a new target to expand its manufacturing capacity by 2030 — pushing a portion of its growth narrative into the discount horizon investors are now penalizing. A company with proven technology, existing revenue, and surging orders still faced selectivity pressure when a piece of its growth story was back-loaded. Innio, a newly listed supplier of reciprocating engines competing with Caterpillar, has seen its shares slip in recent weeks despite operating in a conventional product category with no technology-readiness delay. The filter is consistent and independent of technology class. It looks at the calendar.

Fervo Energy proves the thesis better than any argument. Fervo adapts fracking techniques to generate geothermal power. Its first plant is due to start up late this year and reach 100 megawatts of capacity by early 2027. Jefferies analysts argued that grouping Fervo with out-of-favor advanced-nuclear companies “is unfair.” They are right. Fervo is not being swept down with Oklo and Standard Nuclear. The selectivity shift is looking at the calendar, not the technology class. Fervo has execution risk — its technology is still being proven, and any delay in the 2027 milestone would collapse its differentiation thesis — but it has not yet been re-categorized as a long-timeline story. The fact that Fervo is trading separately from Oklo proves the selectivity mechanism is real, active, and discriminating. If investors were merely panicking or rotating out of a theme, Fervo would be going down with the nuclear names. It isn’t.

The market’s frame is missing the parties who control the physical clock. The article that documents this repricing treats only market-visible parties as stakeholders: developers, analysts, buyers. But the physical ceiling is not a financial abstraction. It is a shortage of electricians, pipefitters, and welders — workers who are named as a bottleneck but given no voice in the reporting. It is NRC permitting timelines that operate independently of market sentiment. It is local communities near proposed reactor and geothermal sites who bear siting risk with no representation in capital allocation. It is ratepayers who will see transmission costs socialized and rates increased to subsidize data-center load. It is grid operators who will absorb the operational mismatch when power-source fragmentation produces capacity that does not align with load.

These parties are outside the investment-narrative frame, but their actions determine whether the buildout can happen at all. They are going to re-enter the story. The moment a skilled-labor shortage becomes a public controversy at a specific construction site, the moment a local community challenges a reactor siting decision, the moment a ratepayer group intervenes in a transmission cost allocation proceeding — that is when the physical ceiling stops being a background factor and becomes a named variable. The date is unpredictable. The structural logic says it happens within the next 24 months, because the buildout plans currently priced into stocks depend on physical conditions that the investor-sentiment frame has not acknowledged.

Three concrete tests will determine whether the structural thesis holds or breaks.

First: the day the next major delay is announced. If Oklo announces that NRC permitting has slipped by six months and the stock barely moves, the 75% decline may have been the full repricing. If the stock drops another 10–15% on the news, the selectivity shift is still accelerating and the convergence of the financial and physical clocks is confirmed. A second leg on delay news is the strongest signal that the market has not yet finished repricing long-timeline risk.

Second: Fervo’s 2027 milestone. If Fervo hits its 100-megawatt target by early 2027, the selectivity thesis is validated — near-revenue names are trading on fundamentals while long-timeline names remain discounted. If Fervo misses, the entire timeline-differentiation argument collapses and the company will be re-categorized as a long-timeline story facing the same selectivity pressure as Oklo. The market’s reaction to a Fervo delay will be a controlled experiment in whether the selectivity shift is discriminating by timeline.

Third: the entry of the absent parties. The date is unpredictable. The fact that it will happen is structural.

The fundamental thesis is that the repricing is structural and still underappreciated by anyone who thinks the worst is over. The investor selectivity loop and the physical-rate ceiling are not parallel mechanisms that happen to be active at the same time. They are converging on the same vulnerable set of companies. The companies with the longest delays — Oklo, Standard Nuclear — sit directly in the path of both forces. The companies with shorter timelines — Fervo, any gas turbine manufacturer already operating — are partially buffered because selectivity exerts less force on pre-2030 names and the physical constraint on their specific technology class is less severe. But even the buffered names are not safe from a second-order effect: if the selectivity lens widens further, it will capture more of the mid-duration pipeline.

Behind both clocks runs a self-reinforcing disadvantage cycle. Construction delays stretch the revenue horizon. That stretches the discount rate investors apply to distant cash flows. That compresses the stock price. That tightens financing conditions. That extends delays further. Oklo exemplifies it. Standard Nuclear exemplifies it. Any company whose revenue arrives after its financing needs will face the same structural vulnerability.

The AI power stock repricing is not over. It is not a correction. It is a structural adjustment to the discovery that the financial clock and the physical clock run on different schedules. Until the physical clock catches up — through faster permitting, more construction labor, or compressed manufacturing timelines — the financial clock will keep compressing the valuations of companies that cannot show revenue from operating assets. And the physical clock cannot be sped up by market sentiment. It can only be sped up by more labor, faster permits, and shorter supply chains — all of which are controlled by parties who are not yet in the story.

Analytical techniques used in this piece

This analysis applies the methods below. Each links to a short, plain-English explainer you can read and reuse.

Frame Audit
Surfaces the frame an argument adopts and what that framing quietly includes or excludes.
Stakeholder Mapping
Charts the parties to a situation — their interests, power, and alignments.
Systems Dynamics (Structural)
Maps a system’s structure — stocks, flows, and the architecture that shapes its behavior.