Is It Too Late to Buy Nvidia and Broadcom? Here's What History Says.

The key market issue is no longer demand durability; it is whether the supply chain can monetize that demand fast enough without margin compression. NVDA is increasingly exposed to a broader capex cycle because networking, systems, and software are becoming a larger share of the wallet, which is positive for revenue durability but can cap operating leverage if customers push back on total platform pricing. AVGO’s tighter visibility is a double-edged sword: contracted revenue lowers downside in the next few quarters, but it also invites valuation anchoring as investors start capitalizing a very long runway that can be disrupted by even a single large-customer design shift.

The second-order winner is likely the broader AI infrastructure ecosystem, not the obvious leaders. Every incremental dollar moving into custom silicon and high-speed networking raises demand for advanced packaging, HBM, photonics, foundry capacity, and power infrastructure; that is where the next leg of relative outperformance may emerge. Conversely, hyperscaler capex intensity should eventually create procurement discipline, which is a latent headwind for semiconductor multiple expansion even if reported earnings keep surprising.

The setup is vulnerable to a classic “good news is already owned” problem. Over the next 1-2 earnings cycles, both names can still outperform if bookings translate into guide raises, but the asymmetry worsens if management teams merely reaffirm instead of re-accelerate. The most plausible reversal catalyst is not AI demand fading; it is a digestion phase where customers slow order timing, lengthen depreciation assumptions, and force the market to re-rate growth from scarcity premium to industrial-like visibility.