After the midterms, Big Tech’s AI data center problem is only getting bigger

Voters opposed data center development in 37 of 38 local ballot contests with complete counts, while Ohio voters in at least 14 municipalities approved bans or other restrictions. The results raise the risk of delays, tighter rules and higher costs for AI infrastructure, including potential added burdens for taxes, power and water; analysts say operators may shift projects or retrofit existing sites. Still, Amazon, Microsoft, Alphabet and Meta are projected to spend a combined $720 billion to $740 billion on capex in calendar 2026, reflecting continued demand for capacity.

Analysis

The key risk is not a blanket halt to AI capacity; it is a higher, less predictable cost and longer lead time to secure power, water, permits, and community consent. That shifts bargaining power toward utilities, owners of power-ready sites, and jurisdictions able to approve projects, while weakening the economics of marginal greenfield builds. Incumbents with multi-region portfolios and the ability to retrofit existing sites should be relatively advantaged; the cost may also flow through to AI-cloud customers, potentially tempering demand if pricing rises faster than workloads’ returns.

The exposure is asymmetric. Meta’s contested Ohio project is a local execution risk, not evidence that its national buildout is impaired. Oracle merits closer scrutiny: its disclosed debt burden and funding needs make delay or higher project costs more consequential than for better-diversified hyperscalers. Amazon and Microsoft have publicly offered to absorb more infrastructure and community costs, but those commitments could protect permitting access while pressuring project returns; the financial impact is unquantified.

Near term, watch for actual moratoriums, permitting rules, utility-rate treatment, and tax changes—not election rhetoric alone. Over 1–3 months, policy specificity in Texas and affected municipalities is the catalyst. Over 6–18 months, expect project relocation, phased builds, retrofits, and higher costs to become visible in capex efficiency and cloud margins. Contrarian angle: constraints may increase scarcity value of powered sites and strengthen incumbents’ advantage over smaller AI providers, even as headline capex plans face friction. Thesis weakens if proposed restrictions fail to become law, projects continue meeting schedules, or operators show no deterioration in cost per capacity added.

Key Decisions for Investors

• Consider a relative-value short ORCL / long MSFT, sized modestly and entered only as local or state rules become concrete. The thesis is greater financing and execution sensitivity at Oracle, not certainty of an absolute decline. Reassess if Oracle reports project milestones and funding on plan without worsening financing terms, or if Microsoft signals comparable site delays or cost escalation.
• Avoid treating the Ohio vote as a basis for a broad META short. Track permitting and construction milestones for the specific project; escalate the risk only if the dispute delays capacity or prompts material changes to company guidance.
• Add a watch item—not a trade yet—for power-ready data-center landlords, utilities, and grid-equipment suppliers: tighter site supply could support pricing, but the article provides no company-level evidence on contract exposure or who captures the economics. Verify lease rates, interconnection queues, and regulated cost recovery before positioning.
• Over the next 1–3 months, monitor Texas legislative proposals and municipal implementation, plus hyperscaler disclosures on schedule slippage, retrofit mix, and infrastructure spend. A clear rise in delays or cost per unit of capacity would support the relative-value thesis; rhetoric without enacted rules is not enough.