It’s Donald Trump Versus MAGA on Data Centers

More than 60% of Americans, including 47% of Republicans, oppose new data-center construction, creating a growing political risk for AI infrastructure expansion ahead of the midterms. Republican campaign strategists have warned candidates that unqualified support for AI data centers is a losing position, even as President Trump continues to promote the sector on competitiveness grounds versus China. The backlash could complicate permitting and development, with the Interior Department considering at least 12 proposals across 17,600 acres of public land and Virginia weighing potential limits on new facilities.

Analysis

The investable signal is a shift from federal AI enthusiasm to local permitting, power procurement, water, noise, and land-use friction. This raises the probability that hyperscaler capex converts into commissioned capacity more slowly than consensus, creating a 6-18 month mismatch between equipment orders and revenue-producing compute. The relative beneficiaries are scarce powered-land and grid-enablement assets—EQIX, DLR, VRT, ETN, CEG and VST—rather than developers reliant on greenfield approvals.

For AMZN, the near-term issue is less headline reputational damage than whether AWS must pay higher mitigation and resiliency costs to preserve its expansion pipeline. A localized customer-data-loss incident is unlikely to move consolidated earnings, but it gives enterprise buyers and regulators a tangible argument for multi-region redundancy, favoring interconnection-rich incumbents such as EQIX and potentially increasing AWS network and backup-power spend. The key watch item is whether AWS, MSFT, GOOGL or META shift from aggressive capacity targets to more conditional language around utility availability and permits in upcoming earnings calls.

Consensus remains focused on AI demand and aggregate capex, while underpricing geographic concentration risk: opposition in one or two critical power markets can redirect projects, not eliminate spend. That is constructive for electrical equipment and existing data-center landlords, but negative for merchant developers with uncontracted land banks and for hyperscaler margins if power costs are bid up. The thesis is falsified if states adopt expedited permitting or if utilities demonstrate sufficient incremental generation and transmission commitments without materially higher connection costs over the next 1-3 months.

Key Decisions for Investors

• Prefer a 6-12 month long VRT and ETN basket over a broad hyperscaler basket: grid, cooling and power-quality content rises when projects require mitigation and backup infrastructure. Use a 10-15% drawdown or a material reduction in hyperscaler capex guidance as thesis stops.
• Pair long EQIX / short DLR over 3-6 months if permitting friction intensifies: EQIX's interconnection density and enterprise colocation mix should be more defensible than incremental wholesale development economics. Avoid entry if DLR demonstrates accelerating signed leases with fully secured utility capacity.
• Do not add to AMZN solely on this development; monitor AWS commentary on availability-zone resilience, insurance/recovery costs, and regional capacity constraints at the next earnings release. A disclosed rise in remediation costs or cloud-growth deceleration would justify reassessing relative exposure versus MSFT and GOOGL.
• Maintain a watchlist on CEG and VST for pullbacks rather than chase: delayed data-center construction can defer load realization, but contracted or highly probable hyperscaler demand should support power-price and capacity-value upside over 12-24 months. The risk is regulatory restrictions on large-load interconnection or power-price intervention.