Chip stocks just had their worst week in over a year. Is now the time to buy?

This looks more like a positioning reset than a fundamental air pocket. The key mechanism is that cheaper frontier models can actually expand total compute demand: if software becomes less expensive to deploy, inference volumes rise faster than training spend slows, which is structurally supportive for NVDA and AVGO rather than bearish. The market is still underestimating how much of the value pool migrates to the shovel sellers when model differentiation gets harder at the app layer.

The second-order loser set is not the chip complex; it is the high-multiple AI software cohort that relies on scarcity pricing and proprietary model access. If customers can switch to lower-cost open-weight options, spending shifts from software licenses toward throughput, memory bandwidth, networking, and power delivery. That favors semiconductor supply chains with pricing power and punishes names where growth has been justified by monopoly-like gross margins, especially if enterprise buyers use the recent selloff to renegotiate roadmaps.

The real risk is timing: if hyperscaler capex pauses for even one quarter, crowded semis can overshoot because the stock tape is ahead of the earnings tape. Over 1-3 months, a 10%-15% additional flush is plausible if management teams signal digestion; over 6-18 months, supply remains too constrained for a durable downcycle unless AI capex guidance rolls over. Contrarian takeaway: the market may be overpricing software disintermediation while underpricing hardware demand elasticity from cheaper models. Falsifier: any meaningful cut to 2026 AI capex plans or evidence that memory lead times shorten materially.