AI's insatiable appetite for electricity could revive a forsaken energy source

The article argues that surging data-center power demand could benefit Babcock & Wilcox, which has a $2.7 billion backlog including a $2.4 billion Base Electron deal tied to Applied Digital. It highlights B&W's 244% year-to-date stock gain, Applied Digital's move from $1.5 billion to $12 billion in market value, and the possibility that coal and natural-gas plant demand could revive. The piece is bullish on coal-linked and power-infrastructure names, but notes significant short interest, B. Riley-related controversy, and execution risk.

Analysis

The market is starting to price the data-center buildout as a power-scarcity trade, not just a compute trade. That shifts the center of gravity from semiconductors to whoever can deliver megawatts fast: gas-turbine bottlenecks, grid interconnection delays, and retrofit/maintenance spend become the real choke points. In that framework, coal is less a long-duration secular bet than a short- to medium-term optionality trade on permitting, plant life-extension, and emergency reliability policy.

The key second-order effect is that a sustained “keep-the-lights-on” policy makes stranded-asset math less one-way. If coal plants get deferred instead of retired, the beneficiaries are not only miners but also service and equipment names tied to boiler life extension, emissions controls, and outage repairs. That also reduces the urgency premium in gas and renewable adjacencies: every extra year of coal uptime is a year of fewer new-build megawatts, which can flatten the near-term order book for higher-quality power infrastructure names.

The biggest risk is that the political thesis outruns the legal and utility capex reality. Federal pressure can slow retirements, but it cannot instantly create transmission, fuel logistics, or utility willingness to re-rate depreciating assets; if courts or state regulators push back, the coal trade can unwind quickly. For APLD/CRWV, the more important catalyst is not contract announcements but financing durability: if equity markets stop rewarding backlog leverage, highly shorted data-center developers can de-rate 30-50% even with intact demand.

Consensus is probably overestimating how much of the upside accrues to the obvious hyperscaler names and underestimating the equipment and retrofit layer. The cleaner expression is to own the bottlenecks with pricing power and avoid the most balance-sheet-stretched developers. Coal can work tactically, but only as a policy beta trade with tight risk controls; the higher-conviction structural winners remain electrification, grid hardware, and gas-fired capacity expansion.

Key Decisions for Investors

• Long BTU/CNR/ARLP basket for 3-6 months as a policy-beta trade; use equal-weight sizing and trail stops tightly because the upside is large if plant-life extensions continue, but drawdown risk is severe if courts or regulators reverse the narrative.
• Pair long ETN or CAT against short GEV on a 1-3 month horizon: if the market is pricing more megawatts but supply is constrained, the scarce-grid-and-implementation names should outperform the already-loved turbine installer. Risk/reward improves if power backlog growth broadens beyond headline developers.
• Buy APLD on weakness only, but hedge with long-dated puts or a CRWV short overlay; the trade is attractive for 2-4 weeks into financing/news flow, yet highly vulnerable to any loss of credibility around lease monetization or equity issuance.
• Initiate a relative-value long BTU / short GEV or long BTU / short AEP basket if the objective is to express policy-driven coal extension rather than pure utility growth. This captures the highest beta to the thesis with less direct exposure to rate sensitivity.
• Avoid chasing NVDA here; the marginal trade is in power delivery, not compute hardware. If you want AI exposure, prefer the picks-and-shovels of grid capex over the headline beneficiaries until power availability, not chip demand, becomes the binding constraint.