For long-term investors, a 3-4% pullback in NVIDIA is not especially meaningful. The core thesis remains intact: hyperscaler AI spending is still growing, TSMC continues to report strong demand, and NVIDIA remains the dominant supplier of AI accelerators.

That said, the market is no longer pricing NVIDIA as a cyclical chip company. Expectations are extremely high. Broadcom's reaction shows that even strong results can trigger selloffs if guidance fails to exceed lofty forecasts.

My approach would be gradual accumulation rather than trying to time the bottom. Geopolitical headlines and profit-taking could create further volatility, but waiting for complete clarity often means buying at higher prices later.

The key question is not whether NVDA can grow, but whether growth can continue to outpace already elevated expectations. If AI capex remains strong into 2027, today's dip may look insignificant. If spending moderates, valuation could compress even with solid earnings.

I'm buying small on weakness, not going all-in.