Where Will Broadcom Stock Be in 5 Years?

Broadcom's semiconductor-solutions revenue rose 88% in the first nine months of fiscal 2026 to approximately $48B, representing 68% of its $71B total revenue, as AI-chip demand remained the primary growth driver. Infrastructure software grew 13% to about $23B, providing a 32% revenue contribution that could cushion a potential semiconductor downturn and support integrated hardware-software offerings. The article expects Broadcom to outperform the market over five years, but flags valuation risk: the CAPE ratio of 41 is comparable to dot-com-era levels and raises the risk of an AI-led correction.

Analysis

The key underwriting issue is not whether AVGO has a software hedge, but whether VMware can convert from an acquisition-driven revenue base into durable, higher-margin recurring cash flow without accelerating customer defections. License changes and channel consolidation may lift near-term bookings and EBITDA, but they also create a 6-18 month migration risk for price-sensitive enterprise customers toward Nutanix (NTNX), Red Hat/IBM (IBM), and public-cloud alternatives. That makes reported infrastructure-software growth, renewal rates, and deferred-revenue trends more informative than headline segment revenue.

AVGO's custom-AI ASIC exposure is strategically differentiated from NVDA's merchant-GPU model: hyperscalers adopting internal accelerators can diversify procurement and improve AVGO's content per customer. The offset is concentration—large customers have substantial negotiating leverage, and a single capex pause can create a sharper revenue reset than the market expects. In the next 1-3 months, quarterly customer-capex commentary from MSFT, GOOGL, AMZN and META is the relevant read-through; over 6-18 months, the question is whether custom silicon expands the total accelerator market or merely displaces GPU spend.

Consensus appears too willing to treat software as a full-cycle stabilizer. Enterprise infrastructure budgets are cyclical, while AVGO's acquisition financing leaves equity duration sensitive to rates and to any multiple compression in AI beneficiaries. A broad AI de-rating would likely reduce AVGO's valuation despite resilient cash generation; the relative opportunity is therefore in AVGO's custom-silicon share gains versus more crowded GPU exposure, not an unhedged chase after a favorable retail-media narrative.

The thesis is falsified if AI semiconductor growth decelerates materially while VMware net retention or renewal metrics deteriorate, showing that one segment cannot offset weakness in the other. A sustained rise in long-end yields, hyperscaler reductions in 2027 capex guidance, or evidence that customers are migrating VMware workloads faster than expected would warrant reducing exposure.

Key Decisions for Investors

• Maintain AVGO as a core but do not add aggressively before the next earnings print; add only if management confirms AI backlog conversion and infrastructure-software renewal stability. Target a 6-12 month position sized below NVDA because customer concentration creates larger single-quarter gap risk.
• Implement a 3-6 month relative-value trade: long AVGO / short NVDA in equal beta-adjusted dollars after any post-earnings AVGO weakness. The trade monetizes custom-ASIC share gains and avoids paying for pure merchant-GPU duration; exit if NVDA regains accelerator-share momentum or AVGO reports an AI growth deceleration exceeding roughly 15 percentage points sequentially.
• Use NTNX and IBM as watch-list beneficiaries rather than immediate longs. Initiate only if disclosed VMware renewal friction, elevated migration activity, or weaker deferred revenue validates that pricing changes are generating customer churn; this is the cleanest second-order expression of software-integration risk.
• Hedge a long AVGO position with 6-9 month downside puts or a partial SOXX hedge around hyperscaler earnings. The hedge is warranted if the position is established ahead of AMZN, META, MSFT and GOOGL capex guidance, where an AI-spend reset would affect AVGO before software resilience can be demonstrated.