Democrats may go after data centers. That's not a reason to dump all your AI stocks

Analysts argued that Democratic pushback against data-center expansion could slow AI investment and help deflate a potential bubble gradually, making the boom more sustainable but increasing near- to medium-term volatility. AI-related capex has grown at a 26% annualized pace in real terms since the beginning of last year; on Wednesday, the S&P 500 rose 2% by noon ET, while hyperscalers gained more than 4% and GE Vernova fell about 10%. AI-dedicated neoclouds also dropped sharply, including IREN down 11%, amid concerns about funding and profitability.

Analysis

The key distinction is not “AI winners versus losers,” but who bears the financing and utilization risk if projects slip. Large platforms can defer discretionary capex and retain strategic flexibility; compute lessors and neoclouds may still owe for capacity, power, or equipment before customer demand supports utilization. That asymmetry favors cash-generative platforms over intermediaries, but only if hyperscalers can actually renegotiate or postpone commitments. If obligations are sunk, a slower buildout may delay revenue without materially relieving cash needs.

Power and equipment exposures are not one-way shorts. Delayed projects could defer orders for GE Vernova (GEV) and incremental load for Constellation Energy (CEG), while persistent grid constraints may preserve the value of existing capacity and make eventual buildout more expensive. Texas permitting and interconnection decisions are the near-term swing factor; the longer-run test is whether enterprise AI revenue catches up with infrastructure commitments. The contrarian risk is treating political friction as a clean, benign “bubble deflation”: it could instead strand highly leveraged capacity and tighten credit before capital is efficiently redeployed.

Key Decisions for Investors

• Consider a staged relative-value position: long a basket of AMZN, MSFT, and GOOG against a smaller, risk-capped short in CRWV, NBIS, and IREN. Enter on policy or financing confirmation rather than extrapolating one-day moves. Thesis: platforms have more capex flexibility, while specialist operators are more exposed to utilization and refinancing. Falsify if neocloud customer commitments/utilization strengthen and credit spreads tighten while hyperscaler capex guidance remains elevated.
• Keep GEV and CEG on a watchlist rather than shorting solely on permitting headlines. Reassess on Texas policy detail, project cancellations or delays, and company order/load commentary; a delay in new projects is not evidence that existing generation or grid capacity has lost scarcity value.
• Over the next 1–3 months, monitor hyperscaler capex guidance against AI-related revenue realization, neocloud financing terms and credit spreads, and state-level permitting/interconnection actions. A widening funding gap alongside weaker utilization would support the relative-value trade; durable contracted demand and easier financing would invalidate it.