There's a Part of AI That Nvidia Can't Monopolize -- and This Stock Owns It

Broadcom is positioned to benefit from multiple AI infrastructure growth drivers, including custom chips, Ethernet networking, and VMware-driven enterprise AI software. Management expects more than $100 billion in AI chip revenue in 2027, with Q2 fiscal 2026 revenue projected at about $22 billion and AI semiconductor revenue of $10.7 billion versus $8.4 billion in Q1. The article is constructive on Broadcom’s long-term outlook, but notes valuation at 80.5x trailing earnings and customer concentration risk across six major AI customers.

Analysis

Broadcom’s edge is not simply “AI exposure”; it is becoming the default toll-collector for the non-GPU layers hyperscalers need once they move from model training prototypes to industrial-scale deployment. The second-order effect is that custom silicon tends to fragment Nvidia’s wallet share, but it expands the total spend on interconnect, switch silicon, PHYs, and design services — areas where Broadcom can monetize both complexity and customer lock-in. That makes AVGO’s growth more durable than a pure accelerator vendor, because every incremental custom chip program creates follow-on networking and software attach.

The more interesting read-through is to suppliers and competitors. If hyperscalers keep internalizing accelerator design, the marginal beneficiary may be less about who wins the compute chip and more about who owns the “glue” between heterogeneous chips, racks, and clouds. That is structurally supportive for AVGO and, to a lesser extent, for other high-end networking and optical vendors; it is relatively negative for Nvidia’s long-run mix if custom ASICs keep eating into full-stack GPU deployments. The risk is that this thesis depends on sustained capex intensity from a small number of customers, so any pause in AI buildouts would show up quickly in order cadence before it hits reported revenue.

VMware adds a different option value: enterprise AI infrastructure spending should be less cyclical than hyperscaler capex, but it will ramp slower and be more execution-sensitive. The market may be underestimating how sticky private/hybrid AI deployments become in regulated verticals, which could offset some concentration risk over time. Near term, though, the stock is priced like a near-flawless compounder, so any guide-down, customer delay, or margin compression would likely re-rate it sharply over a 1-3 month horizon.

Key Decisions for Investors

• Maintain long AVGO exposure, but size it as a quality-growth compounder rather than a momentum trade; best entry is on post-print consolidation or any 5-8% pullback from hyperscaler capex fears, with a 6-12 month upside case driven by custom chip + networking attach.
• Pair trade: long AVGO / short NVDA on a 3-6 month horizon as a relative-value expression of custom silicon share shift; risk is NVDA re-accelerating enterprise demand or AVGO concentration risk derating the long leg.
• Use META and GOOGL as higher-beta indirect beneficiaries only on weakness; both have leverage to custom AI infrastructure, but AVGO is the cleaner way to own the infrastructure monetization layer.
• Avoid chasing AVGO into strength at peak optimism; consider selling cash-secured puts or put spreads 10-15% below spot to express a bullish view with better entry and defined downside.
• Watch for any evidence of hyperscaler capex deceleration over the next 1-2 quarters; that would be the main trigger to trim AVGO and cover any AVGO/NVDA relative-value position.