The Biggest Trade If the Democrats Win Back the House and Senate in November's Mid-Term Elections: Short Artificial Intelligence (AI) Data Centers

With Polymarket assigning a 60% probability to a Democratic congressional sweep, the article argues that AI data-center stocks could face selling pressure from stricter regulation, permitting limits and reduced tax incentives. Proposed measures include ending data-center access to bonus depreciation and Opportunity Zone benefits, while legislation from Virginia lawmakers would require large power users to fund infrastructure through new taxes. Pure-play operators CoreWeave and Nebius are presented as most exposed, while hyperscalers Amazon and Alphabet—expected to spend $1 trillion on AI infrastructure in 2025-26—could also decline initially despite more diversified businesses.

Analysis

The relevant transmission mechanism is not an outright reduction in AI demand but a higher cost of delivered compute: slower permitting, utility interconnection charges, and reduced tax shielding raise the all-in cost of capacity. CRWV and NBIS are more exposed than hyperscalers because their equity cases require rapid capacity commissioning and high asset utilization to amortize fixed power, GPU, and financing costs. A delay of even one or two quarters can impair IRR and refinancing flexibility for leveraged GPU-cloud operators, while AMZN and GOOG can redirect capex across regions and fund infrastructure from operating cash flow.

The market may be over-attributing the risk to NVDA. A permitting delay shifts accelerator deliveries and can create a near-term digestion period among GPU lessors, but does not necessarily eliminate compute demand; constrained domestic supply could ultimately support cloud pricing and utilization. The more immediate second-order losers are data-center construction and electrical-equipment orders tied to speculative campuses, while regulated utilities and grid vendors can benefit if legislation formalizes cost recovery from large-load customers. PWR, ETN and VRT merit monitoring, but the outcome depends on state-level rate-case treatment rather than federal election results alone.

A Democratic sweep is already partly reflected in prediction-market odds, and federal legislation would face drafting, committee, and implementation delays. The actionable catalyst is therefore post-election evidence of tax-law inclusion or state utility commissions imposing incremental interconnection contributions, not the election result in isolation. Falsify the bearish pure-play thesis if CRWV/NBIS report backlog conversion and energized capacity above plan without rising power-cost commitments, or if hyperscaler capex guidance remains intact while GPU-cloud pricing strengthens over the next two quarters.

Key Decisions for Investors

• Establish a 3-6 month relative-value position: short CRWV versus long AMZN and GOOG in equal beta-adjusted weights. The thesis is financing and commissioning-risk divergence, not broad AI weakness; cover the short if CRWV demonstrates two consecutive quarters of capacity delivery ahead of guidance or materially lower funding costs.
• Use December 2026 or January 2027 CRWV put spreads rather than an unhedged directional short ahead of the election, sized to a defined premium loss. Target a 2:1 payoff structure; avoid implementation until implied volatility is checked against its post-earnings range.
• Maintain NVDA exposure but reduce incremental purchases into election-related headlines. Re-add only if supplier commentary confirms that any campus delays are being offset by demand from sovereign, enterprise, or non-U.S. deployments; a broad order deferral from CRWV and other GPU lessors would be the warning signal.
• Create an alert basket of PWR, ETN, VRT and select regulated utilities with large-load exposure. Initiate only after specific state rate orders establish that data centers bear incremental transmission/distribution costs with permitted utility returns; absent that evidence, the regulatory outcome may simply delay projects rather than improve supplier economics.