Chip Stocks Record Steepest Losses Since 2023 as AI Spending Scrutiny Intensifies

MarketWatch reported growing Wall Street impatience ahead of Anthropic’s investor day, with investors demanding clearer paths to monetization after repeated promises of transformative AI productivity gains. Semiconductor order books remain elevated, yet forward multiples have compressed as sell-side models now embed lower terminal growth assumptions for data-center accelerators. Primary earnings transcripts from Nvidia’s July quarter already showed sequential deceleration in data-center revenue growth rates compared with the prior three quarters.

The pattern mirrors earlier capital-expenditure cycles in which initial hardware demand outran verifiable enterprise ROI, prompting re-pricing before the next platform wave. Central-bank liquidity data and corporate capex guidance from Microsoft and Google parent Alphabet indicate sustained dollar outlays, yet allocation scrutiny has shifted from volume to utilization metrics. This repricing transmits directly to foundry utilization forecasts at TSMC, where advanced-node capacity bookings for 2025 are being stress-tested against lower AI attach rates.

What follows is a narrower set of winners: only those vendors able to demonstrate measurable inference-cost reduction or new sovereign-AI contract wins will maintain premium valuations. Secondary suppliers without direct exposure to frontier-model training clusters face accelerated multiple contraction. The ledger shows hyperscalers gaining optionality on future architectures while absorbing margin pressure in the near term; chip makers absorb valuation resets in exchange for continued volume visibility.