Own Broadcom by Sept. 21 to Qualify for Its Sept. 30 Dividend. Here's How Many Shares You'd Need for $5,000 in Yearly Dividends.

Broadcom guided for AI revenue of $115 billion in fiscal 2027 and $230 billion in fiscal 2028, up from its $100 billion fiscal-2027 outlook six months earlier, driven by custom XPU accelerators and AI networking demand. The company generated $31.94 billion of free cash flow in the first nine months of fiscal 2026 versus $9.28 billion of dividends, while also allocating $8.45 billion to buybacks and $10.53 billion to debt repayment. Its annual dividend is $2.60 per share, following a 15th straight annual increase, but the 0.8% current yield and 29.1x forward P/E leave valuation and customer concentration as key risks. OpenAI and Anthropic could become Broadcom's largest AI customers, increasing growth potential but also exposure to any slowdown in AI capex.

Analysis

AVGO’s upside is increasingly governed by a small number of bespoke accelerator programs rather than diversified semiconductor demand. Custom silicon creates materially higher switching costs once a customer commits its software stack, networking topology, and power design, but it also converts backlog into a binary execution risk: a delayed model launch, financing constraint, or lower utilization rate at one frontier lab can remove several quarters of expected revenue. The key near-term datapoint is not dividend capacity but whether customer-specific XPU ramps translate into networking attach and gross-margin stability; weak attach would imply customers are optimizing total cluster cost more aggressively than expected.

The market is likely underpricing the difference between hyperscaler-funded demand and lab-funded demand. GOOG and META can sustain capex through internal cash generation, while OpenAI and Anthropic depend more heavily on external capital, cloud partners, and commercialization trajectories; their orders should therefore carry a lower valuation multiple until take-or-pay terms, deposits, and deployment schedules are independently verified. Over 6-18 months, successful ASIC adoption is a share loss for NVDA at the largest customers, but not necessarily a demand loss for NVDA: lower accelerator unit cost can expand total compute budgets and preserve demand for Nvidia in less customized workloads. A capex pause would hit AVGO’s incremental earnings harder than its broader software franchise suggests, creating asymmetric multiple-compression risk if AI mix rises faster than recurring software cash flow.

Key Decisions for Investors

• Maintain a tactical long AVGO / short NVDA pair over the next 1-3 months only if AVGO’s AI-revenue guide is reaffirmed and customer deployment milestones remain on schedule. The thesis is ASIC share gains and networking attach; exit if AVGO reduces AI guidance or if NVDA reports renewed concentration-driven accelerator demand that signals custom programs are slipping.
• Do not underwrite the dividend-growth narrative as a primary catalyst at current valuation. Treat capital returns as downside support, not rerating fuel; require evidence that post-buyback free cash flow remains sufficient to reduce leverage while funding AI-related working-capital needs.
• Set an event-driven alert around OpenAI/Anthropic financing, cloud-capacity contracts, and reported inference monetization over the next 90 days. Any funding delay, cloud-partner renegotiation, or deployment deferral is a reason to reduce AVGO exposure before consensus estimates incorporate lower custom-chip volumes.
• For portfolios already long AI semis, hedge 6-12 month AVGO single-name concentration with modest put spreads rather than adding outright exposure after guidance-driven strength. The adverse scenario is a customer capex reset that compresses AVGO toward a diversified-semiconductor multiple; the hedge should be removed if customer commitments become contractually visible and networking margins expand.