Federal regulators back Trump's plan to speed power to energy-hungry AI data centers

FERC unanimously voted to speed interconnection for AI data centers and other large power users, with the order requiring them to pay the full cost of any grid upgrades. The move should support faster AI infrastructure buildouts, but it does little to relieve tightening power supply, rising electricity bills, and grid strain already linked to data-center growth. Data centers now consume about 5% of U.S. electricity demand and could triple by 2035, with Virginia already above 25% of load.

Analysis

The policy shift lowers a key bottleneck for AI load growth, but the biggest economic winner is not the hyperscalers per se — it is the “picks and shovels” ecosystem that monetizes accelerated interconnection: transformers, switchgear, gas turbines, substation engineering, and grid software. If large loads can jump the queue, the value migrates toward companies with existing power contracts, on-site generation, or modular deployment capability, while purely speculative data-center developers face a higher bar to prove they can secure electrons before competitors do.

Second-order, this is likely inflationary for power infrastructure capex rather than immediately bullish for utility earnings. Utilities and grid operators may protect returns by socializing more of the transmission buildout through rate base expansion, which is constructive for regulated names with constructive state PUC relationships, but negative for regions where backlog is already severe and political resistance to higher bills is rising. The real constraint remains physical: interconnection reform can shorten paperwork, not manufacture turbines, copper, skilled labor, or cooling water, so the earnings upside for AI beneficiaries is deferred and increasingly gated by supply-chain lead times.

The consensus is underpricing the possibility of a bifurcation: the biggest AI platforms can self-fund power and lock in supply, while smaller second-tier cloud/AI users get squeezed by longer waits and higher all-in utility costs. That favors scale leaders with balance sheet flexibility and penalizes marginal inference/training operators whose unit economics depend on cheap, reliable grid access. Near term, sentiment may stay positive, but over 3-12 months the market should start differentiating between firms with secured megawatts and those merely announcing demand.

Key Decisions for Investors

• Long GOOGL/MSFT/META vs short a basket of smaller AI infrastructure-exposed names on any post-news strength; thesis is that scale and balance sheet will convert faster interconnection into capacity, while marginal players face execution slippage over 6-12 months.
• Add to infrastructure-enablers with supply scarcity exposure: LECO, ETN, PWR, POWL on pullbacks; these names have the cleanest leverage to accelerated grid buildout and backlog repricing over the next 4-8 quarters.
• Pair long regulated utilities with strong transmission growth visibility (e.g., NEE, DUK) against short utilities in regions with acute data-center load risk and political backlash; look for 6-18 month rate-base upside versus margin pressure from customer pushback.
• For event-driven traders, buy 3-6 month call spreads on MSFT or AMZN only on weakness after initial enthusiasm fades; the policy reduces strategic risk, but the operating benefit remains delayed until physical power is secured.
• Hedge the theme with shorts in water-constraint and high-capex AI-adjacent developers where the interconnection queue is the core asset; if permitting/turbine shortages persist, these names are most vulnerable to multiple compression.