1. What Does Broadcom Actually Do?
Two segments, roughly 70/30 of revenue today:
- Semiconductor Solutions (70% of Q3 revenue, $20.8B): custom AI accelerators ("XPUs") co-designed with hyperscalers, AI networking silicon (Tomahawk/Jericho switches), broadband, wireless (still a large Apple RF supplier), storage, and industrial chips.
- Infrastructure Software (30%, $8.8B): VMware virtualization, mainframe software (CA), and cybersecurity (Symantec enterprise).
Q3 AI semiconductor revenue of $16.7 billion grew 221% year-over-year and 54% quarter-over-quarter — meaning AI chips alone are now more than half of total company revenue, and non-AI semiconductor + software is effectively the "stable base" underneath a much faster-growing AI layer. The Semiconductor Solutions segment produced record revenue of $20.8 billion, up 127% year over year, accounting for 70% of consolidated revenue.
Durability of the moat: custom silicon co-design cycles run multiple years and get embedded deep in a customer's infrastructure roadmap — high switching costs once a hyperscaler commits. The risk is customer concentration: a handful of hyperscale/AI-lab relationships drive nearly all AI growth.
2. How Does Revenue Grow?
| Fiscal Year | Revenue | YoY Growth | Driver |
|---|---|---|---|
| FY2019 | $22.6B | — | Broadcom (semis) + CA/Symantec |
| FY2022 | $33.2B | +21% | Core semis, early software |
| FY2023 | $35.8B | +8% | Semi cyclical slowdown |
| FY2024 | $51.6B | +44% | VMware consolidation |
| FY2025 | $63.9B | +24% | AI accelerator ramp begins |
| FY2026E | ~$105.9B | +66% | AI accelerator hyper-growth |
Broadcom closed fiscal year 2025 with revenue of $63.9 billion, up 23.9% year over year. Q3 FY26 alone saw consolidated revenue of $29.6 billion, up 86% year-on-year, driven by a significant increase in AI semiconductor revenue, and Q4 consolidated revenue growth is forecast to increase 93% year-over-year to $34.8 billion.
3. How Profitable Is That Revenue?
| Metric | FY2024 | FY2025 | Q3 FY26 |
|---|---|---|---|
| Gross margin (GAAP) | 63.0% | 67.8% | 69.1% |
| Operating margin (GAAP) | 26.1% | 39.9% | 53.9% |
| Operating margin (Non-GAAP) | — | — | 68% |
FY2025 revenue was $63,887M with a gross margin of 67.8%, up 4.8 points, and operating margin of 39.9%, up 13.8 points from FY2024. The margin story is the AI mix shift plus VMware integration maturing — software carries much higher margins than the FY2024 acquisition costs implied. Guidance keeps this elevated: Q4 non-GAAP operating margin guidance of approximately 66 percent, flat with a year ago.
4. How Much Becomes Cash?
This is where Broadcom's fabless, capital-light model shows up starkly:
- Q3 FY26: cash from operations of $14.2 billion, less capital expenditures of $0.5 billion, resulted in $13.7 billion of free cash flow, or 46 percent of revenue.
- Nine-month FY26 FCF: $31.9B on $71.1B revenue (~45% FCF margin).
- CapEx is tiny relative to revenue (~1.4% of nine-month revenue) — Broadcom doesn't build fabs or data centers; TSMC and its customers carry that capital burden.
5. Reinvestment and Returns
| Net debt (Aug 2026) | ~$35.4B |
| Stockholders' equity | $99.7B |
| Goodwill (legacy M&A) | $97.8B |
| Approx. invested capital (debt + equity − cash) | ~$135B |
| TTM GAAP operating income | ~$42.8B |
| Illustrative NOPAT (~15% tax rate) | ~$36.4B |
| Illustrative consolidated ROIC | ~27% |
That ~27% is dragged down by ~$98B of purchased goodwill sitting on the balance sheet from the CA, Symantec, and VMware deals. On an incremental basis — the return on each new dollar spent on AI capex and custom-silicon R&D — returns are almost certainly much higher, because capex is running at roughly 1–2% of revenue while operating profit is compounding at 60–90%. That's the mechanical reason FCF margins keep expanding.
6–7. How Long Can This Last, and What's the Terminal State?
Bull case for duration: switching costs in multi-year custom-silicon programs, an addressable market (AI compute) still expanding, and Broadcom's position as the leading alternative to building AI silicon in-house or buying merchant GPUs.
Bear case for duration: heavy dependence on a small number of hyperscale/AI-lab customers, competitive pressure from Nvidia's merchant GPUs and Marvell's custom-silicon push, and the residual-value-guarantee financing noted above, which raises the question of how much future "demand" is being pulled forward or subsidized.
8. Balance Sheet: EV to Equity
| Cash & equivalents | $24.0B |
| Total debt | $59.4B |
| Net debt | $35.4B |
| Diluted shares (Q3 FY26 weighted avg) | 4.89B |
9. What Does the Current Price Already Assume?
As of the last close before earnings, AVGO traded at $370.34 with a market cap of $1.76 trillion, a trailing P/E of 61.65, trailing EPS of $6.01, and a 52-week range of $287.17–$495.00. On Non-GAAP TTM EPS (~$9.75), the multiple is closer to ~38x — the GAAP/non-GAAP gap is almost entirely acquisition-intangible amortization from CA/Symantec/VMware, which is a real accounting charge but not a cash cost going forward.
Reverse-DCF Read
Running a simplified reverse-DCF — enterprise value ≈ $1.80T (market cap + net debt), trailing FCF ≈ $39.4B, 9% discount rate, 4% terminal growth after a 5-year high-growth window — the price is consistent with roughly 23% annual FCF growth for five years before fading to a mature growth rate. That's notably lower than what management is currently guiding for the next two years (AI revenue alone roughly doubling in FY27 and again in FY28).
Two readings are possible: either the market doesn't fully believe the multi-year guidance will convert to FCF at this rate, or it's pricing meaningful deceleration/multiple compression once the current AI buildout matures.
Notably, Broadcom's own outlook is for AI revenue to reach approximately $230 billion by fiscal 2028, supported by secured supply, and management said the company is on target for over $30 in EPS by fiscal 2028, well above the LSEG consensus of $25.86 at the time — so the company itself is guiding above what analysts had modeled, which is part of why the stock reacted positively even though near-term Q4 revenue guidance modestly missed expectations.
10. Stress Test
| Bear | Base (≈ management guidance) | Bull | |
|---|---|---|---|
| FY27 AI revenue | ~$85B | $115B | $140B+ |
| FY28 AI revenue | ~$150B | $230B | $280B+ |
| Non-GAAP op margin | 60% | 66–68% | 70%+ |
| Key risk/driver | Hyperscaler capex pause, ASIC share loss to Nvidia/Marvell | Guidance plays out as stated | Broadcom wins more custom-silicon programs, software cross-sell accelerates |
Comments
Post a Comment