‘People really effing hate data centers’: How the AI buildout became the midterms’ affordability issue

PJM capacity prices rose from $28.92 per megawatt-day for 2024-25 to $329.17 for 2026-27; the grid monitor attributed 63% of the increase in the 2025-26 auction to data centers, while IEEFA estimated added costs of about $16 per month for Ohio households and $18 in western Maryland. Data center energy costs became a campaign issue, with winning candidates backing measures such as rate protections, eliminating tax breaks, or moratoriums, and at least 14 Ohio municipalities approving bans or restrictions. Industry groups dispute the cost link, while data center capacity under construction is up 48% nationwide; Goldman Sachs strategists see possible curbs to tax incentives but consider a nationwide moratorium unlikely.

Analysis

The investable shift is from “AI demand” to who bears the marginal cost of connecting it. If regulators require dedicated generation, grid upgrades, or separate tariffs, projects may remain viable but face higher upfront capex, slower returns, and longer siting timelines; utilities could also lose the political latitude to socialize costs across households. That makes local permits and utility rate cases more important near-term indicators than national AI rhetoric.

For Amazon.com (AMZN) and Oracle (ORCL), treat this as a project-level execution risk, not evidence that aggregate cloud demand is weakening. Oracle has a specifically identified Saline Township exposure; Amazon faces local restrictions in Ohio, but the article does not establish that any particular ballot measure blocks construction. The burden of proof is on proponents to show whether projects can secure power and permits on acceptable economics. Power generators and developers able to contract directly with data centers may benefit if customers fund incremental supply, though policy could limit that upside.

Contrarian read: opposition is politically potent, but local resistance does not equal a national moratorium, and projects can adapt through self-generation, long-term power contracts, or relocation. Near-term risk is headline-driven multiple volatility; over 1–3 months, watch legislation, municipal enforcement, and utility tariff/rate-case decisions. Over 6–18 months, the key test is whether these change commissioning dates or capital returns. A thesis of broad AI-capex impairment is falsified if projects continue to obtain permits and power agreements without material schedule or economics changes.

Key Decisions for Investors

• No immediate broad short of AI infrastructure on this signal alone. Track AMZN and ORCL separately; do not extrapolate the Saline project’s exposure to Oracle’s consolidated business or assume Ohio measures halt Amazon projects.
• Use a relative-risk watch rather than an unconditional pair: consider underweighting ORCL versus AMZN only if Saline permitting, power access, or project timing deteriorates while Amazon’s affected Ohio sites remain on schedule. Reassess against company disclosures and actual project-level updates.
• Over the next 1–3 months, monitor state legislation, municipal implementation or legal challenges, and PJM utility tariff proposals. Escalate the risk if rules shift incremental grid or generation costs to developers; de-escalate if separate tariffs or customer-funded power arrangements preserve project economics.
• Avoid treating higher capacity prices as an automatic utility or generator long: political pressure may constrain cost recovery or reshape procurement. Verify rate-case treatment and contracted data-center demand before positioning.