The Nuclear Hype Cycle — Where It Stands After the Crash

The nuclear sector has experienced one of the most dramatic boom-bust cycles in recent market history. From September to October 2025, a constellation of nuclear stocks surged on the narrative that AI data centers would drive insatiable electricity demand, and nuclear was the only scalable clean baseload answer. Six months later, the speculative end of the sector has been destroyed — down 65-82% from peak — while companies with actual operating assets have held up far better. The question now is whether the crash has created opportunity or simply corrected a fantasy.

The Scorecard: Two Tiers

The nuclear trade has fractured into two distinct groups with dramatically different outcomes.

Pre-revenue / speculative tier (from 52-week high):

Revenue / operating tier (from 52-week high):

The divergence is stark. BWXT — which manufactures actual naval reactor components for the U.S. Navy — is within 3% of its all-time high. Meanwhile, NuScale, which has no firm construction contracts and is burning through cash, has lost 82% of its value. The market is no longer willing to pay for timelines measured in decades.

What Drove the Bubble

Three narratives converged in mid-2025 to create the nuclear mania:

1. AI data center power demand. The genuine insight was that AI workloads consume vastly more electricity than traditional computing. Data centers now account for ~4% of U.S. power consumption, projected to double by 2030. Nuclear is the only scalable 24/7 zero-carbon power source. This is real.

2. Landmark corporate PPAs. Microsoft signed a 20-year, 835MW power purchase agreement with Constellation Energy to restart Three Mile Island Unit 1 (renamed Crane Clean Energy Center, targeting 2027-2028 restart). Meta signed a 20-year PPA for Constellation's Clinton plant. Amazon announced a ~2GW deal with Talen Energy's Susquehanna plant. These deals validated the demand thesis with billions of real corporate dollars.

3. Policy tailwinds. The U.S.-Canada $80 billion reactor partnership, bipartisan support for nuclear in Congress, the NRC's ongoing review of SMR designs, and the global pivot back to nuclear (COP28 pledge to triple nuclear capacity by 2050) created a narrative of inevitability.

The problem: investors conflated "nuclear energy will be important" with "pre-revenue nuclear startups are worth billions today."

What Killed the Rally

Timeline reality. BloombergNEF's head nuclear analyst captured the contradiction: "There's a lot going on, and nothing is going on." The PPAs and policy announcements created activity, but no SMR company has broken ground on a commercial reactor. NuScale's first system isn't expected online until 2030 at the earliest. Oklo's first small reactor targets "late 2027" but still lacks NRC approval after its first application was denied in 2022. The revenue event horizon is 3-8 years away for every speculative name.

Fluor's exit from NuScale. Fluor Corp sold 71 million NuScale shares to Goldman Sachs at $19.05/share in February 2026 and is distributing another ~40 million shares through Bank of America and Citibank through Q2 2026. This structural selling pressure has been relentless — the stock has fallen from ~$19 to $10 since the sale began.

Valuation reality. At its October 2025 peak, Oklo had a $20 billion market cap with zero operating reactors, no commercial licenses, and no binding power contracts. NuScale traded at $18 billion on $31 million in engineering services revenue. These valuations required perfection on decade-long timelines — a setup that always corrects.

Broader risk-off. The KOSDAQ crash in early March 2026 and the early April selloff (likely tariff-related) hit speculative growth names hardest. High-beta nuclear stocks amplified every market drawdown.

The Demand Thesis Is Real — The Vehicle Was Wrong

This is the critical distinction. The underlying thesis — that the world needs significantly more nuclear power, and AI/data center growth accelerates that need — has not been invalidated. If anything, it has strengthened:

• Microsoft's Three Mile Island restart is proceeding (target 2027-2028)
• Meta and Amazon PPAs confirm sustained corporate demand
• The EIA projects U.S. electricity demand growth accelerating through 2030
• Global uranium prices remain elevated, supporting the mining tier

The mistake was in how investors expressed this thesis. Buying Constellation Energy (which owns 21GW of existing nuclear capacity and sells power today at premium prices) was a fundamentally different trade than buying Oklo (which has a PowerPoint and a Sam Altman board seat). The market has now violently repriced that distinction.

Where Are the Opportunities Now?

Operating tier — still expensive but fundamentally supported. CEG at 37x trailing P/E and 20x forward is not cheap, but it owns irreplaceable nuclear assets with 20+ year PPAs at premium rates. BWXT at 60x trailing has real Navy contracts and nuclear manufacturing revenue. These stocks have corrected 17-34% and may offer entry points if the broader market stabilizes, but they're not bargains.

Uranium miners — leveraged to the fuel cycle. CCJ at 116x trailing earnings prices in the uranium bull thesis. Uranium supply remains constrained and new reactor demand (including SMRs, eventually) will tighten the market further. UEC is earlier-stage. This tier depends on uranium spot and contract prices, which have held up relatively well.

Pre-revenue SMR names — binary outcomes. SMR at $10.15 has $3.93/share in cash and no debt. If NuScale lands a firm TVA contract or Romania FID in 2026-2027, the stock re-rates sharply higher. If catalysts continue to slip, it drifts toward cash value. OKLO at $48 still commands an $8.4 billion market cap with zero revenue — it has further to fall if execution disappoints. The risk/reward has improved from peak levels but these remain speculative bets on multi-year timelines.