Why did this AI doomer moment break the internet when so many others didn’t?

AI-safety concerns gained unusual political and public traction after former Anthropic employee Jacob Coxon’s resignation post exceeded 100 million views, with Anthropic researchers publicly supporting his warning of material extinction risk. The shift coincides with growing opposition to AI data-center expansion: 64% of Reuters/Ipsos respondents opposed rapid buildouts, 77% feared higher electricity prices, and at least 75 projects worth roughly $130 billion were blocked or delayed in Q1. Bipartisan calls for pauses, restrictions, and potential multinational rules raise regulatory and infrastructure-permitting risks for frontier AI developers and data-center investment.

Analysis

The investable transmission channel is not near-term model demand but permitting duration, power interconnection cost, and the required return on incremental AI infrastructure. A political reframing from “economic development” to “consumer electricity-cost risk” raises the probability that projects are delayed at the state/local level, reducing utilization visibility for EQIX and DLR and lengthening the revenue-conversion cycle for power-intensive campuses. The first-order impact on GOOG is modest given its balance sheet and geographic flexibility; the larger risk is that hyperscalers must pay more for firm power, community concessions, and distributed siting, depressing AI infrastructure ROI rather than halting capex.

Over the next 1-3 months, public safety rhetoric is likely to increase headline volatility in GOOG, MSFT, AMZN and META, but it is unlikely to alter earnings absent a concrete federal moratorium, export-control expansion, or state permitting action. The more consequential 6-18 month effect is competitive: incumbents with secured generation, transmission access, and operating data-center capacity gain share, while developers relying on speculative land/power queues face stranded-development risk. CEG, VST and GEV can remain relative beneficiaries if AI load is merely redistributed toward jurisdictions and sites with existing power availability.

Consensus may over-attribute any AI-equity weakness to existential-risk rhetoric. Policymakers have stronger incentives to regulate siting, energy pricing, disclosure, and liability than to impose a broad development pause; such rules could actually entrench GOOG, MSFT and AMZN by raising compliance and capital barriers for smaller labs. The thesis is falsified if permitting delays fail to appear in hyperscaler capex commentary, power procurement costs remain contained, and state restrictions are reversed or preempted federally.

Key Decisions for Investors

• Maintain a 1-3 month relative-value position: long GOOG / short a basket of higher-duration data-center REIT exposure (EQIX, DLR). The hedge isolates the likely cost-of-capital and permitting premium on asset-heavy capacity providers; exit if EQIX/DLR report unchanged development pipelines and leasing spreads despite local opposition.
• Add a 6-12 month long CEG or VST versus short DLR as a power-scarcity expression. Existing generation and contracted load become more valuable if new campuses are delayed, while data-center development returns compress; size modestly because a recession-driven power-demand reset would hurt both legs.
• Do not short GOOG outright on safety headlines. Instead, set an event alert around state-level project cancellations, utility rate-case outcomes, and hyperscaler commentary on interconnection delays; a confirmed capex deferral or AI monetization-guide cut would justify reassessing GOOG downside.
• For concentrated AI-beta portfolios, buy 3-6 month downside protection on QQQ rather than single-name puts around regulatory hearings. A broad policy shock would likely compress AI multiples across GOOG, MSFT, AMZN and META, while idiosyncratic enforcement risk remains too uncertain to price cleanly.