Article published Jul 17, 2026. Prices below use latest available snapshots.
Four independent reads landed in the same week, all pointing the same direction: the electricity shortfall behind the AI build-out is getting worse, not better. And yet the market has spent three months selling every name that was supposed to win it. Both are right — because the fight has moved from whether power is scarce to who is allowed to get paid for the scarcity.
The evidence: the shortage is tightening
The grid's biggest market cleared short. PJM's latest capacity auction — the largest U.S. grid market — cleared at $325/MW-day even after thirteen governors imposed a price cap and floor, and still came in roughly 7,000 MW under its reliability target (Winston, @ChurchillWw). A market that clears high and short is telling you the shortfall is structural.
Demand is being rationed by politics. New York Governor Kathy Hochul went on Odd Lots to defend a one-year moratorium on new large data centers — the first state-level pause of its kind. Siting is no longer a formality: where the next gigawatt of demand is allowed to connect is now a political decision.
New supply is arriving on two clocks. Near term, four U.S. microreactor startups raced a DOE July 4 deadline to first criticality — the last, an Aalo Atomics sodium-cooled unit, went critical at Idaho National Laboratory (Winston, @ChurchillWw). Long term, General Fusion debuted on the Nasdaq via SPAC (Sherwood News), putting a private fusion moonshot on the public tape for the first time.
The tape: the trade already ran, and it's being repriced
Against that evidence, the complex looks like this (July 17 close):
| Ticker | Price | RSI | 30d | From 52-wk high | Note |
|---|---|---|---|---|---|
| CEG | $252.39 | 46.3 | −7.9% | −38.8% | merchant bellwether, still in a confirmed downtrend |
| VST | $155.44 | 48.5 | −4.9% | −29.3% | below its 200-day (~$166) |
| TLN | $372.37 | 46.3 | −14.7% | −17.5% | best of the merchants — a pullback, not a collapse |
| CCJ | $85.62 | 28.2 | −19.6% | −36.7% | fuel-cycle quality name, washed out |
| LEU | $156.05 | 44.2 | −18.5% | −66.4% | enrichment monopoly, full collapse regime |
| DUK | $125.01 | 48.0 | +0.9% | −7.1% | regulated utility — the only constructive tape |
The pre-revenue reactor developers are worse still: Oklo sits −79% and NuScale −87% from their 52-week highs (July 17 close). The 2025 "own anything with electrons" trade has been comprehensively unwound.
The tell: regulated held, merchant broke
The divergence is the message. The market is not ignoring the shortage — it is pricing intervention on the shortage's economics. Read the week's four datapoints again from that angle and each one is bearish for the merchant-scarcity trade even as it confirms the physical deficit: governors capping the very auction price that was the merchants' windfall; a state freezing the demand that was their growth story; new reactor supply arriving faster than the 2030s timelines the market had penciled in. Scarcity persists; the monetization of scarcity is being renegotiated in public. Regulated utilities — who earn on rate base, not on spot scarcity — are the structure the political system is comfortable rewarding, and the tape shows it.
Verdict
The bottleneck thesis survives; its expression changes.
The shortage is real, worsening, and now politically managed — which caps the blue-sky case for merchant power and pre-revenue reactor paper, but keeps the demand story fully intact for whoever builds and fuels the capacity. Don't chase the fallen headline names into confirmed downtrends. Watch the merchants for turn signatures (they now carry policy risk that wasn't priced in 2025), watch the washed-out fuel cycle for a base — the enrichment and uranium story is a supply-security question that outlives any one auction — and respect that the market's only working expression of this theme right now is the regulated compounder. That is itself information about where the cash flows are going to be allowed to land.
Desk Call
| Ticker | Call | Entry / condition | Invalidation | Review by |
|---|---|---|---|---|
| CEG | Watch | Base under the 50-day (~$269), then a reclaim that holds with a higher low | Revenue caps land on core markets, or new lows below the July shelf | 2026-09-15 |
| VST | Watch | Reclaim the 200-day (~$166) and hold it | Caps spread to its markets, or breakdown instead of a base | 2026-09-15 |
| TLN | Watch | Hold $360 and the 200-day ($371); clear the 20-day ($394) to confirm | A close below $360 | 2026-09-15 |
| CCJ | Watch | $84 shelf holds + first higher low; 50-day (~$105) reclaim confirms | The $84 shelf fails | 2026-09-15 |
| LEU | Watch | Full-quarter base, then a 50-day (~$177) reclaim | HALEU contract flow stalls, or lower lows continue | 2026-10-01 |
| DUK | Watch | A fresh 52-week high that holds | Loses the 200-day (~$123) | 2026-09-15 |
Sources
- PJM capacity auction ($325/MW-day clear, ~7,000 MW under the reliability target, thirteen-governor price collar): Winston, @ChurchillWw
- NY data-center moratorium: Odd Lots — NY Governor Kathy Hochul on her one-year data-center moratorium (Bloomberg)
- Microreactor first criticality (four units in five weeks; Aalo Atomics at Idaho National Laboratory): Winston, @ChurchillWw
- General Fusion Nasdaq debut: Sherwood News
- Price, RSI, moving-average, and drawdown figures: desk tape data as of the 2026-07-17 close (ai-power,
nuclear-fuel-cycle.json)