Is It Too Late to Buy Nvidia and Broadcom? Here's What History Tells Us

Nvidia’s latest quarterly growth remained exceptionally strong at 73%, with management guiding to 77% growth next quarter, while Broadcom said AI semiconductor revenue rose 106% to $8.4 billion. Broadcom also projected its custom AI chips business could generate $100 billion or more in revenue by the end of 2027, reinforcing the scale of AI demand. The article argues both stocks still trade at attractive forward P/E discounts despite their sharp gains, making the setup constructive for long-term investors.

Analysis

The market is treating AI capex as a winner-take-most regime, but the more interesting read-through is that the value chain is bifurcating: NVDA still owns the general-purpose compute stack, while AVGO is increasingly the monetization layer for hyperscalers that have reached the scale where workload-specific silicon finally pencils. That is a healthier backdrop for both names than a pure substitution story, because custom silicon adoption usually expands total silicon spend by making previously uneconomic inference workloads viable at scale.

The second-order implication is pressure on everyone adjacent to AI infrastructure: network gear, memory, and power/cooling providers should keep seeing follow-on demand as accelerators proliferate, but smaller merchant semiconductor vendors are at risk of being squeezed by design wins moving in-house. If hyperscalers keep shifting from off-the-shelf accelerators to proprietary chips, the prize moves upstream into IP, EDA, packaging, and advanced interconnect rather than into commoditized compute alone.

Consensus is likely underestimating the duration of the capex cycle. A 1-2 quarter acceleration in near-term growth is already priced into sentiment, but the real catalyst is FY27 visibility: if AVGO’s custom silicon ramps as guided, it creates an annuity-like runway that can support multiple expansion even at a slower reported growth rate later. For NVDA, the key risk is not demand collapse but digestion: if customers are overbuying capacity into 2026, the stock can still de-rate on any sign of order normalization despite strong end-market seculars.

The contrarian angle is that both stocks may be ‘cheap’ only if investors believe 2027 guidance is executable without margin dilution from competition, export friction, or customer concentration. That makes the setup more attractive on pullbacks than on strength, because the next leg higher likely needs another round of capex confirmation rather than just headline growth prints.

Key Decisions for Investors

• Buy NVDA on any 5-8% pullback over the next 2-4 weeks; use a 3-6 month horizon. Risk/reward favors upside continuation if hyperscaler capex revisions keep landing high, but trim if customer inventories or order cadence flatten.
• Add AVGO on weakness with a 6-12 month horizon; the better asymmetric trade is the custom-silicon monetization curve into FY27. Stop-loss on a break below the market’s implied growth discount regime if forward multiple stops re-rating.
• Pair trade: long NVDA / short a basket of merchant AI semis with weaker hyperscaler exposure over 3-6 months. The thesis is that design-ins increasingly concentrate in the two platform winners while smaller names lose pricing power and socket share.
• Use call spreads rather than outright equity for NVDA into the next earnings cycle: 1-2 quarter upside is plausible, but options cap downside if the market punishes any digestion signal despite strong guidance.
• Watch for a relative-long AVGO vs short software-infrastructure names that monetize AI less directly; if capex remains strong, hardware-adjacent beneficiaries should outperform lagging AI software multiples over the next 1-2 quarters.