Egan-Jones Expects Both Parties to Keep Backing the AI Buildout

Egan-Jones says the four largest hyperscalers plan more than $700 billion in AI-related capital spending this year, while political support from both parties’ mainstream is likely to persist despite local opposition and tighter state-level rules. It views cost shifts and slower permits as positive for utilities but slightly negative for hyperscalers, negative for Oracle and lenders, and most concerning for smaller developers; $68 billion of projects were delayed or canceled last quarter.

Analysis

The political equilibrium looks more likely to change who pays for grid expansion than to stop AI construction. That reduces stranded-asset risk for utilities when projects fall through, but does not automatically raise utility earnings: the upside depends on regulators allowing cost recovery and on operators actually funding the required upgrades. Watch allowed-return decisions and rate cases, not just project announcements.

For developers, the risk compounds: delayed interconnection can push revenue out while construction and financing costs persist. Smaller, project-dependent operators are more exposed than diversified hyperscalers. Oracle is a relative vulnerability to monitor, but the article provides no company-level exposure, contract, or financing detail to quantify the effect; it does not support assuming the same impact across all of Oracle’s business.

Near term, permitting and grid-approval decisions—especially in Texas—and implementation of cost-allocation rules are the catalysts. Over 1–3 months, the key read-through is whether operators revise capex, delivery timelines, or power commitments. Over 6–18 months, persistent delays could shift workloads toward existing facilities and better-capitalized operators. The contrarian risk is that investors treat broad political support as protection from local bottlenecks: support for AI growth does not guarantee timely power access. Conversely, a cost shift alone is not evidence of demand destruction. Thesis weakens if approvals normalize without material schedule or capex revisions; it strengthens if project delays broaden or Oracle discloses a material deterioration in delivery economics.

Key Decisions for Investors

• Consider a modest ORCL underweight relative to a diversified hyperscaler peer basket, not an outright sector short. Reassess against Oracle’s disclosed data-center capex, financing, contracted capacity, and delivery timelines; those company-specific details are missing here.
• Favor regulated utilities selectively where filings show data-center customers bearing incremental grid costs and regulators permitting timely recovery. Avoid treating the policy shift as an earnings catalyst until rate-case evidence confirms it.
• Track Texas grid approvals and state cost-allocation implementation as 1–3 month catalysts. Escalating delays or capex/guidance revisions would support the cautious stance; normalized approvals and stable delivery economics would falsify it.
• Avoid a broad AI-infrastructure short based on this commentary alone: political support may preserve demand even as costs and schedules worsen. For smaller developers and related credit, require project-level liquidity, refinancing, and interconnection data before taking exposure.