U.S. equities finished the week lower, with AI-exposed semiconductor names at the center of the decline. NVDA and AMD were among the hardest-hit tickers as investor appetite for AI infrastructure plays cooled, extending a stretch of weakness that has weighed on the broader tech complex. The headline doesn't specify a single catalyst, pointing instead to a confluence of valuation concerns, geopolitical chip restrictions, and mixed signals from hyperscaler capex commentary.

NVDA's fundamentals remain exceptional on paper — $215.9B in revenue, +65.5% YoY growth, 71.1% gross margins, and $4.90 diluted EPS for FY2026. AMD is growing fast too at +34.3% YoY but operates on thinner margins (49.5% gross, 12.5% net), leaving it more exposed if AI spending narratives soften. Both names have run hard into these numbers, meaning any growth-rate deceleration or demand-guidance miss carries outsized downside risk.

The bull case for owning the dip rests on NVDA's near-monopoly in training infrastructure and AMD's credible challenger position in inference, with both benefiting from a multi-year datacenter buildout cycle that hasn't shown structural cracks. But the bear case is real: at elevated multiples, even slight deceleration in hyperscaler AI capex — or new export restrictions — can reprice these stocks sharply, as this week demonstrated.

What to watch: any commentary from major cloud providers (MSFT, GOOGL, AMZN) on capex trajectory, NVDA's next earnings print for H100/B100 demand signals, and the regulatory drumbeat around chip export controls. Until those catalysts resolve, the risk/reward on a blind dip-buy is asymmetric in the wrong direction.