The pattern exposes a structural flaw in global AI infrastructure investing. Markets worldwide have priced AI chip demand at maximum growth velocity. Guidance that meets — but does not beat — consensus now triggers double-digit drawdowns. The gap between "good" and "good enough" has collapsed.
This is not a demand story. Hyperscalers across the US, Europe, and Asia are still building data centres. Chip orders remain robust. The problem is valuation: semiconductor stocks have absorbed so much anticipated future growth that any shortfall in the guidance narrative produces asymmetric selling.
Broadcom's two-day drop illustrates the dynamic directly.1 Strong results no longer protect a stock if the guidance fails to accelerate. Earnings have shifted from fundamentals-driven events to binary guidance moments, with outsized downside as the default outcome for in-line prints.
The same risk applies to NVIDIA, AMD, and Marvell as Q2 guidance season approaches.1 Options markets are expected to price elevated implied volatility into each name ahead of their releases. In-line guidance is now a negative scenario in every time zone where these stocks trade.
For global AI infrastructure as a sector, the volatility raises a harder question. Semiconductor stocks have functioned as international proxies for AI spending confidence — influencing capital flows from sovereign wealth funds in the Gulf to pension allocators in Northern Europe. Swings of nearly 20% on guidance language signal fragility in a consensus that stretched well beyond any single market.
The underlying spend is real. But sustaining current valuations demands continuous upside surprises across multiple guidance cycles — a difficult standard as hyperscaler commentary grows more nuanced about deployment timelines and return on AI investment.
Q2 earnings season now tests whether the global AI infrastructure trade holds at current expectations, or whether Broadcom's move marks the first visible crack in a consensus that ran ahead of delivery.
