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

Polymarket assigns a 60% probability to a Democratic sweep in the November midterms, an outcome the article argues could trigger a short-term selloff in AI data-center stocks. Proposed and existing Democratic measures include a New York data-center construction pause of up to one year, potential removal of bonus-depreciation and Opportunity Zone tax benefits, and potential state taxes on large power users to fund grid infrastructure. Pure-play operators CoreWeave and Nebius are viewed as most exposed, while Nvidia, Amazon, and Alphabet could face indirect pressure because hyperscalers are projected to spend $1 trillion on AI infrastructure in 2025-26.

Analysis

The market mechanism is not a blanket AI-demand shock; it is a higher cost of incremental domestic capacity. Restrictions on tax treatment, interconnection obligations, and local power-cost recovery would raise project IRRs and delay new campuses, with the greatest valuation impact on externally financed operators whose equity cases assume rapid capacity additions. CRWV is more exposed than AMZN/GOOG because a higher cost of debt or equity simultaneously reduces build economics and pressures the multiple assigned to contracted growth. NBIS should be analyzed separately by geographic asset mix and power contracts rather than treated as a U.S.-policy beta.

The contrarian implication is that supply restrictions can increase the value of already energized, permitted capacity. If AI demand remains supply-constrained, incumbent operators with operational clusters may gain pricing power, offsetting slower expansion; an outright short of data-center equities immediately after an election result risks being caught in a scarcity rerating. NVDA faces a more delayed effect: installed-capacity and backlog conversion should cushion revenue for several quarters, while lower 2027-28 construction starts—not near-term GPU orders—would be the relevant earnings risk.

Election odds are not a tradeable regulatory outcome. Material federal tax changes are most plausible through a budget process after the new Congress is seated, while siting and utility-cost decisions remain predominantly state and local. Over the next 1-3 months, monitor hyperscaler capex guidance, utility interconnection timelines, and CRWV funding spreads; over 6-18 months, watch whether contracted capacity additions and lease pricing decelerate. The bearish thesis is falsified if operators retain build guidance while securing power and financing at unchanged economics.

Key Decisions for Investors

• Do not initiate a broad short on an election headline. Use any policy-driven selloff to assess long exposure to existing-capacity owners only after verifying lease duration, power pass-through provisions, and contracted-versus-planned MW; scarcity could make the first move directionally wrong.
• For a defined bearish expression, prefer a 6-12 month CRWV put spread over shorting NVDA: CRWV has more direct financing and expansion-IRR sensitivity, while the spread limits loss if capacity scarcity supports its valuation. Enter only if new debt/equity funding costs widen or management cuts capacity-addition guidance.
• Maintain AMZN/GOOG as relative hedges versus pure-play compute infrastructure. Their AI capex may be reprioritized, but cloud pricing, advertising, and retail/search cash flows make them materially less sensitive to a U.S. data-center permitting delay.
• Set an alert around NVDA supply-chain indicators and hyperscaler capex commentary: reduce NVDA exposure only if 2027 infrastructure commitments or GPU deployment schedules are revised down, not merely on legislative proposals. A sustained cut in aggregate capex guidance would be the catalyst for broader semiconductor downside.
• Watch state utility commission actions and federal tax-language release dates rather than polling alone. A final rule allowing utilities to recover incremental grid costs from large-load customers, or removal of accelerated tax benefits, would justify reassessing CRWV/NBIS estimates; absent implementation detail, the signal is weak.