Broadcom Has Trailed the S&P 500 This Year. That Shouldn't Last Much Longer.

This setup is less about whether AI demand is real and more about whether hyperscaler capex is broadening from GPUs into custom silicon without a pause. Broadcom’s edge is not unit growth alone; it is design-in stickiness and the ability to monetize a few large programs with outsized operating leverage. If Nvidia’s print is a clean read-through, it argues that the spending cycle is still intact, which raises the odds Broadcom can beat and, more importantly, lift forward AI revenue guidance again.

The market is probably overreading the Marvell-Alphabet angle as a direct displacement story. In practice, the hyperscalers want supplier diversity, pricing leverage, and redundancy, so a second ASIC vendor often expands the pie before it takes share. That means MRVL can get a narrative bid while the cash-flow inflection remains with AVGO, whose scale and customer penetration make near-term revenue visibility better than the headline suggests. AMD is a weaker relative beneficiary here because custom silicon spend is not a zero-sum GPU substitute; it is an incremental workload split.

The key risk is timing: the next 1-3 weeks are about earnings and guidance, but the real failure mode is a capex air pocket in 1-2 quarters if enterprise AI spend stalls or a major customer slows deployments. The thesis breaks if Broadcom misses the implied AI revenue run-rate or signals that sequential growth is normalizing faster than expected. Over 6-18 months, the bigger question is whether the market starts valuing AVGO like a cyclical chip supplier rather than a structural AI infrastructure compounder; if so, multiple expansion can outrun fundamentals both ways.