Here's Why Nvidia and Broadcom Are Still Leading the Pack for AI Investing

Nvidia posted 73% revenue growth to $68.1 billion in its latest quarter, while analysts expect growth to accelerate to 79% next quarter and 85% the following quarter. Broadcom's AI semiconductor division grew 106% year over year to $8.4 billion in fiscal Q1, and management expects its custom AI chip business could reach $100 billion or more in 2027. The article argues both stocks remain attractive long-term beneficiaries of multi-year AI data center spending, which Nvidia says could total $3 trillion to $4 trillion by 2030.

Analysis

The market is still underestimating how much of the AI capex cycle is becoming a two-engine oligopoly. NVDA remains the default beneficiary of generalized compute demand, but AVGO is the cleaner second-order winner because custom silicon usually follows the hyperscalers’ need to lower unit economics after they’ve saturated off-the-shelf GPU deployments. That creates a layered spend curve: first on GPU clusters, then on ASIC substitution, then on networking and replacement cycles, which extends the revenue runway beyond the initial build-out.

The key implication is that the strongest AI trade is no longer just “more capex,” but “capex intensity plus architecture shift.” If custom chips continue taking share, some incremental budget migrates away from pure GPU volume toward higher-margin integration, connectivity, and system-level optimization. That can actually support both names near term, but it increases dispersion beneath the surface: foundry capacity, HBM suppliers, advanced packaging, and high-speed interconnect vendors remain the bottlenecks, while software-first AI plays face monetization lag.

The main risk is not demand collapse but expectation compression. Both stocks are priced for a multi-year earnings compounding story, so any evidence that hyperscalers are moderating spend, extending depreciation lives, or internalizing more design work could trigger multiple contraction even if revenue keeps growing. The second-order bear case is that as custom ASICs become more prevalent, NVDA’s growth rate normalizes faster than the market expects, while AVGO’s design wins remain concentrated among a few customers, increasing headline risk.

Consensus is probably too focused on the durability of the spend cycle and not enough on who captures the margin pool as it matures. The better framing is that the AI trade shifts from “model training scarcity” to “infrastructure monetization efficiency,” which favors the best system architects and networking franchises. That keeps the bull case intact, but it argues for being selective on entry points and using volatility to express relative-value rather than outright chasing both names simultaneously.

Key Decisions for Investors

• Stay long NVDA, but prefer pullback entries after 5-10% drawdowns; the risk/reward is best when the market is pricing in a slower sequential growth rate and the trade can reaccelerate on hyperscaler order prints over the next 1-2 quarters.
• Overweight AVGO versus semis more broadly on a 6-12 month view; custom ASIC adoption is a multi-year theme, and the asymmetric upside is in continued share gains inside hyperscaler budgets rather than headline revenue growth alone.
• Pair trade: long AVGO / short a basket of lower-quality AI infrastructure names that rely on one-time buildout spend; if capex remains strong but shifts toward efficiency, the market should reward recurring design-win franchises over pure construction beneficiaries.
• Use call spreads instead of outright longs for event exposure in NVDA over the next 3-6 months; the trade is strong, but multiple risk is meaningful if guidance gets even slightly less explosive.
• Watch for signs of capex extension or internal ASIC announcements from hyperscalers; if those accelerate, trim NVDA on relative basis and rotate into AVGO and networking-related suppliers.