Broadcom is often described as a chip company.
That description is too small.
Broadcom designs custom AI accelerators, networking silicon, high-speed connectivity, and optical components. It also owns infrastructure software that produces recurring cash flow.
The combination is powerful: semiconductor growth supplies the acceleration, while software helps fund it.
But at $353.70, investors are not discovering that model. They are paying for it.
What Broadcom actually controls
Broadcom sits where hyperscaler ambition becomes physical infrastructure.
Its custom accelerators help large customers build specialized AI systems. Its switch silicon and SerDes move data between processors. Its optical technology helps those systems scale beyond a rack. VMware and the rest of the software portfolio add a separate stream of infrastructure revenue.
In fiscal Q3 2026, revenue reached $29.59 billion, up 86% from a year earlier. AI semiconductor revenue was $16.7 billion, up 221% year over year and 54% sequentially.
Why does that matter?
Because the AI thesis is no longer a small option inside Broadcom. It is the main growth engine.
Infrastructure software contributed $8.75 billion of revenue, up 29%. Free cash flow reached $13.67 billion, or 46% of revenue. That is the quality of the model: Broadcom converts enormous scale into cash.
The strength and the concentration are the same fact
The market loves Broadcom because a handful of hyperscalers are racing to build custom AI infrastructure.
That is also the risk.
Custom silicon programs are large, customer-specific, and concentrated. A delayed deployment or a lost design can matter. Hyperscalers also negotiate from strength.
Software brings diversification, but the balance sheet carries the history of acquisition. Broadcom ended the quarter with roughly $59.4 billion of short- and long-term debt against about $24.0 billion of cash.
The company can service that debt because cash flow is extraordinary. The debt still reduces the room for error.
What $353.70 requires
I use fiscal 2030 as the common checkpoint and an 11% required annual return. The hurdle begins with the current risk-free rate and equity risk premium, then reflects Broadcom's leverage, customer concentration, and unusually strong cash generation.
At a 28 times terminal earnings multiple, the current price requires about $19.20 of fiscal 2030 EPS.
Fiscal Q3 non-GAAP EPS was $3.32. Annualizing one quarter gives $13.28, but Broadcom's Q4 revenue guide of $34.8 billion tells us the business is still moving quickly.
The market is asking Broadcom to grow into the valuation. It is not asking for fantasy. It is asking for sustained excellence.
Three possible outcomes
Conservative: about $217 today
Broadcom earns $15 per share in fiscal 2030 and receives a 22 times terminal multiple.
AI revenue keeps growing, but custom-silicon growth slows, customer concentration becomes more visible, software integration matures, and the market assigns a lower multiple to a leveraged infrastructure company.
Base case: about $406 today
Broadcom earns $22 per share in fiscal 2030 and receives a 28 times terminal multiple.
This assumes custom accelerators and networking continue gaining share of hyperscaler budgets, software remains durable, and cash flow steadily reduces balance-sheet risk.
Upside case: about $632 today
Broadcom earns $30 per share in fiscal 2030 and receives a 32 times terminal multiple.
This requires custom AI silicon to become a lasting second architecture beside merchant GPUs, Ethernet networking to capture more scale-up and scale-out spending, and software margins to remain exceptional.
At $353.70, AVGO is below my base case, but the discount is not wide enough to ignore execution risk.
Debt, dilution, and the quality of earnings
Broadcom's fiscal Q3 Form 10-Q showed roughly 4.77 billion shares outstanding and 4.89 billion diluted weighted-average shares for GAAP EPS. Stock-based compensation was about $2.0 billion in the quarter.
Why does that matter?
Because per-share value is what shareholders own. Revenue growth does not automatically create value if debt and share issuance claim too much of the result.
I want free cash flow to keep covering interest, debt reduction, dividends, and repurchases without requiring the AI boom to remain at its fastest point forever.
What to watch out for
- Quarterly AI semiconductor revenue - because the market's expectations depend on that engine remaining durable.
- Custom accelerator customer wins and concentration - because a small number of programs can move billions of dollars.
- Ethernet switching, SerDes, and optical execution - because accelerators create value only when data can move efficiently.
- VMware revenue, bookings, and customer retention - because software is the stabilizer in the model.
- Debt, share count, and stock compensation - because cash flow must reach each share, not disappear into the capital structure.
What could change the thesis
The thesis would strengthen if Broadcom expands its custom-silicon customer base, sustains networking leadership, and uses free cash flow to reduce net debt faster than expected.
The thesis would weaken if AI revenue becomes dependent on one or two deployments, VMware economics weaken after the initial repricing, or stock compensation prevents per-share cash flow from keeping pace with reported growth.
The most important disconfirming evidence would be a widening gap between AI revenue growth and free cash flow per share.
Bottom line
Broadcom may be the cleanest way to own the part of AI infrastructure that Nvidia does not control by itself.
Business quality: exceptional.
Financial engine: exceptional, but leveraged.
Valuation: supportable if earnings reach roughly $19 per share by fiscal 2030 and the premium multiple survives.
At $353.70, the stock offers upside in my base case, but not enough to stop asking hard questions about concentration and debt.
The practical principle is simple: when growth is extraordinary, follow the cash per share and the obligations attached to it.
Sources
- Broadcom fiscal Q3 2026 earnings release
- Broadcom fiscal Q3 2026 Form 10-Q
- Federal Reserve H.15 interest rates
- Damodaran implied equity risk premium
Scenario values are analytical estimates, not company guidance or price targets. Educational only - not financial advice.