AI stocks hammered by midterm shock as data center buildout becomes a 'war of attrition'

After the U.S. midterm results, tech and AI-related stocks sold off sharply on Thursday: Nvidia fell 9.3%, AMD 17.3%, and data-center-focused Iren 17%, before U.S. tech stocks showed signs of recovery Friday morning. Investors are weighing potential Democratic-led regulatory hurdles and local opposition to data-center construction against Goldman Sachs' estimate of about $7.6 trillion in AI-related capital spending from 2026 through 2031. Dan Ives said the election could slow planned data-center construction and that Big Tech capex growth may peak in 2027, earlier than his prior 2028/2029 view.

Analysis

The more likely transmission is higher cost and slower deployment—not a nationwide stop to AI infrastructure. Local permitting, power availability and community opposition can push projects toward sites with existing grid access, lengthen payback periods and raise the value of scarce power and permitted capacity. That shifts advantage toward large balance sheets and established infrastructure operators, while leveraged neoclouds face greater utilization, refinancing and customer-concentration risk if delivery dates slip. Chip demand is exposed one step later: delays in commissioning can defer orders even if long-run compute demand remains intact.

The political signal is real but noisy: federal control does not itself resolve local siting decisions, and the article’s cited project data and forecasts do not establish company-level earnings impacts. Near term, positioning and rates can dominate; over 1–3 months, watch actual permitting decisions, power interconnection timelines and hyperscaler capex guidance. Over 6–18 months, sustained delays or weaker returns on AI investment would pressure infrastructure growth assumptions and related multiples. A reversal would be evidence that projects are relocating rather than canceled, with capex commitments and commissioning schedules holding. The contrarian opportunity is to distinguish financing and execution risk from compute demand: broad AI exposure may be oversold, but the weakest-funded buildout names are not automatically bargains.

Key Decisions for Investors

• Favor a conditional relative-value expression over a broad AI short: after volatility stabilizes, consider long NVDA versus short CRWV in risk-defined, modest sizing. The thesis is relative financing and deployment resilience, not immunity to a capex slowdown. Reassess if hyperscalers cut or defer capex materially, or if CRWV demonstrates durable funding and utilization; verify balance-sheet runway and customer commitments before entry.
• Avoid chasing a rebound in NBIS, IREN or CRWV until project-level power access, permitting status, funding runway and customer prepayments are clearer. A delay in commissioning or a worsening funding backdrop is a downside catalyst; confirmed schedules and secured power would weaken the caution.
• Track MSFT and META capex guidance alongside free-cash-flow conversion and project commissioning—not headline AI budgets alone. Broad capex reductions would challenge chip and infrastructure demand; stable commitments plus evidence of projects shifting to permitted sites would argue against extending the sector selloff.
• Treat power and permitting scarcity as a watch item for infrastructure exposure, including Brookfield Corporation (BN), rather than assuming its UK project offsets U.S. regulatory risk. Verify project financing, power arrangements and expected returns before attributing a material earnings catalyst.