AI Capex Jitters Hit Semiconductor Valuations

Semiconductor stocks tied to data center and AI demand are currently facing valuation pressure as investors question the durability of AI spending. Concerns around Anthropic’s annual recurring revenue are being treated as a signal about broader AI monetization and the sustainability of hyperscaler infrastructure budgets.

Free cash flow for leading AI chipmakers and foundry names is now closely aligned on a near dollar‑for‑dollar basis with hyperscaler capital expenditure. When that capex trajectory is viewed as outpacing underlying AI service revenues, equity markets typically discount future semiconductor earnings, even while reported results and cash generation remain strong.

Historically, similar dynamics have appeared when a concentrated customer group drove incremental semiconductor demand and then faced its own growth or ROI doubts. Episodes linked to cloud data center digestion in 2018‑2019 and the 2020‑2021 crypto‑mining cycle showed that once customer capex looks ahead of sustainable end demand, semiconductor multiples tend to compress for 6‑12 months.

In the current AI build‑out, free cash flow at hyperscalers is already under pressure while expectations for the chip complex remain elevated. Names such as NVIDIA (NVDA), Broadcom (AVGO), Marvell Technology (MRVL) and Taiwan Semiconductor (TSM) sit at the center of this linkage, with a large share of incremental profits tied to AI accelerators, custom silicon and advanced foundry capacity.

With Anthropic‑related ARR uncertainty amplifying existing worries about AI return on investment, markets are increasingly focused on the gap between slowing capex growth and optimistic semiconductor revenue forecasts. As long as that gap persists, historical patterns suggest elevated volatility and ongoing re‑rating risk across the AI‑exposed semiconductor supply chain and, by extension, the hyperscaler and AI infrastructure ecosystem.