NVDA is the cleanest read-through because this is no longer a simple beat/miss setup; it is a referendum on whether AI capex remains an industry-wide growth engine or starts to normalize. A strong guide would likely re-rate the entire semicap stack higher for 1-3 months, with the best second-order beneficiaries being the infrastructure suppliers that monetize every incremental rack: AVGO, ANET, MRVL, and memory/HBM exposure such as MU. The more crowded part of the market is the AI software layer; if spending looks disciplined, those names can compress faster than hardware because their valuation assumes a delayed monetization curve.

The Canada tariff escalation matters less for macro GDP than for margin fragmentation in cross-border manufacturing. The immediate losers are low-differentiation industrials, autos, metals, and logistics names with North American supply chains, where small tariff changes can force inventory pull-forward, route changes, and lower realized pricing. That makes the next 2-6 weeks noisy, but the bigger risk is a self-reinforcing escalation cycle that widens basis spreads and pushes buyers toward domestic substitution; if talks de-escalate, the pain should unwind quickly.

Contrarian view: the market may be underestimating how much of the AI trade is now about financing quality rather than demand quantity. If hyperscalers keep spending but free cash flow deteriorates, the winners shift from growth-at-any-price software to balance-sheet-stable infrastructure names, and eventually the whole complex can derate. The China/Iran angle is a slower-burn overhang for semis: it is not an earnings event today unless it broadens into fresh export controls or retaliation, but it increases the probability that China accelerates domestic substitution over the next 6-18 months.