Broadcom Inc. (AVGO) — Investment Tree v1
🔴 PRICE RE-ANCHOR (2026-05-30) — full Tier-A R2; verdict basis FLIPS from "undervalued" to "fully-valued." Scaffold anchored $235; live $446.77 (+90%; the scaffold's entire $140-345 scenario band now sits BELOW spot). The business CONFIRMED bigger than modeled — FY2025 revenue $63.89B (+24%), AI semi ~$20B (+65%), Q1 FY2026 AI $8.4B (+106%), CEO guide ">$100B AI chip revenue in 2027" + ~$73B backlog; VMware software margin 76.8% (Hock Tan playbook confirmed & exceeded). But at $446.77 it's ~94× trailing GAAP / ~40× forward — re-derived band Bear ~$280 / Base ~$470 / Bull ~$610, spot at fair-to-slightly-rich base; EV ~+4%, asymmetry ~1.0×. Verdict: Hold-with-sizing HOLDS, but the basis flips — no longer "undervalued +11.6% EV"; now "fully-valued; own for AI-compounding convexity, not cheapness; little margin of safety." 2-3% sizing OK; the K.4 ai-capex-high bucket cap binds. Primary cite: FY2025 10-K (accn 0001730168-25-000121) + Q1 FY2026 10-Q (accn 0001730168-26-000016). See
evidence_2026-05-30.jsonl. (Bucket-2 up-mover: +90% stale anchor flipped the verdict basis; the scenario tables + verdict in the body below are SUPERSEDED by this banner.)
⚠️ PARTIAL R2 VERIFICATION — HEADLINE FINDING: FY2025 REVENUE BEAT MODEL BY ~$5B (updated 2026-05-22) Original 2026-05-19 build was Tier-C throughout (sandbox 403s blocked all primary sources). R2 sweep 2026-05-22 (Claude Code, Windows local) upgraded 5 load-bearing claims to Tier-B via Wikipedia infobox (10-K-sourced; full Tier-A pending direct EDGAR access). - ✅ FY2025 revenue $63.89B (Tier-B; was M3 interpolation ~$57-60B† — actual is ~$5B above the upper bound of the interpolation range, implying ~24% YoY growth vs the ~12% the M3 model assumed) - ✅ FY2025 operating income $25.48B / margin 39.9% (Tier-B) - ✅ FY2025 net income $23.13B / margin 36.2% (Tier-B) - ✅ Segment mix 58% Semi / 42% Software (Tier-B — confirms original ~57-60%/40-43% range) - ✅ Total assets $171.1B; headcount 33,000 (down from 37,000 FY2024 — Hock-Tan margin-lift playbook on VMware continuing) (Tier-B) See
reports/AVGO/evidence_2026-05-22.jsonlfor the upgrade rows; originalevidence_2026-05-19.jsonlstays as historical M3 baseline. The big finding: the original Tier-C interpolation undershot revenue by ~10-13%. The "fairly-priced asymmetry" verdict was based on a too-conservative top-line. Recommend re-running Stage 4 scenarios with the corrected baseline — the actual growth trajectory may shift the prob-weighted asymmetry above "fairly-priced." Still M3 (deferred): Custom AI ASIC revenue (~$15-18B† estimate), AI ASIC FY2026 guidance ($25-30B+†), VMware operating-margin trajectory, Apple wireless revenue, free cash flow, Hyperscaler ASIC TAM estimates, Hock Tan acquisition history. These need direct 10-K + earnings-call commentary.
Stage 7 essay. Date: 2026-05-19 · Anchor: ~$235† · Forward PE: ~32× FY2027E · Dividend yield: ~2.5%
SOURCE QUALITY: Tier C throughout. Semi-design specifics M3.
0. Company Fundamentals — what Broadcom is and how it earns
Figures FY2025 (ended ~Nov 2025) unless noted.
What it is & how it earns. Broadcom is a diversified semiconductor + infrastructure-software business. FY2025 revenue was $63.9B (+24% YoY), split ~58% Semiconductor Solutions ($36.9B) — custom AI accelerators (XPUs/ASICs co-designed for hyperscalers: Google TPU, Meta MTIA, plus emerging Anthropic/OpenAI), networking silicon, wireless (Apple) and broadband — and ~42% Infrastructure Software ($27.0B) — VMware, CA and Symantec, run at a software operating margin near 77% on the Hock Tan integration playbook. AI semiconductor revenue was ~$20B (~+65%), making customer concentration in a handful of hyperscalers the central swing factor.
Cash-flow anatomy. Broadcom is a cash machine: capex is trivial (fabless), so operating cash flow converts almost fully to free cash flow.
| FY2023 | FY2024 | FY2025 | |
|---|---|---|---|
| Operating cash flow | $18.1B | $20.0B | $27.5B |
| Capex | ~$0.45B | ~$0.55B† | $0.62B |
| Free cash flow | $17.6B | ~$19.4B | $26.9B |
| FCF margin (FCF/revenue) | ~49% | ~38% | ~42% |
FCF margin remains exceptional; the FY2024 dip reflects a full year of lower-margin VMware revenue diluting the ratio, and reported GAAP net income (36% margin) is depressed by heavy VMware acquisition-related amortization, so FCF is the cleaner earnings proxy. (FY2024 capex/FCF marked † are derived from quarterly figures, not a direct 10-K line read.)
Balance sheet & capital allocation. Post-VMware, AVGO carries ~$66.1B total debt against ~$14.2B cash, i.e. net debt ~$51.9B — large but supported by ~$27B annual FCF and steadily deleveraging. R&D was ~$11.0B (~17% of revenue). The dividend was $0.59/quarter in FY2025 (raised to $0.65 in FY2026), ~$2.36–$2.60/yr, a sub-1% yield at recent prices; buybacks ran ~$8.5B in H1 FY2026 under a new $10B authorization, keeping total shareholder return in the tens of billions.
What drives it. The thesis turns on whether custom AI-silicon + networking can sustain 30%+ growth toward the CEO's ">$100B AI chip revenue in 2027" framework while the VMware software-margin engine holds — i.e. whether AVGO is a durable multi-engine AI compounder rather than a cyclically peak-valued capex play.
I. One-sentence verdict
Broadcom at ~$235† combines (a) the dominant custom AI silicon franchise (Google TPU + Meta MTIA + Microsoft Maia + emerging OpenAI), (b) VMware OPM expansion from 25% pre-acq to 50%+ via Hock Tan playbook, (c) Apple wireless revenue stability, (d) networking silicon AI-fabric pull, (e) capital-allocation discipline ($20B+/yr return), and (f) durability 21/25 Medium-High with 0 fatal flags — making this a Hold-with-sizing 2-3% initial; scale to 3-4% on K.4 sizing rule coordination with other ai-capex-high names.
II. Company snapshot
AVGO is a diversified semi + software company. FY2025 revenue ~$57-60B† at ~63-66% operating margin (non-GAAP); ~78-80% gross margin. Two segments: Semiconductor Solutions ~57-60% (custom AI ASIC + networking + wireless + broadband) + Infrastructure Software ~40-43% (VMware + Symantec + CA).
CEO Hock Tan since 2006 — legendary capital allocator (CA $19B, Brocade $5.5B, Symantec Enterprise $11B, VMware $69B). CFO Kirsten Spears.
Custom AI ASIC franchise: Google TPU dominant partnership; Meta MTIA production; Microsoft Maia involvement; emerging OpenAI + Apple AI silicon partnerships.
III. Five facts that drive everything
- Custom AI ASIC revenue ~$15-18B FY2025; growing 50%+ YoY. FY2026 guide $25-30B+. ✅C
- Hyperscaler diversification away from NVIDIA-only via AVGO custom ASIC designs. Top-3 customer set anchored; expanding. ✅C
- VMware OPM ~50%+ on Hock Tan playbook (vs ~25% pre-acq). Subscription mix ~65-70% (target ~85-90% by FY2027). ✅C
- AI-capex super-cycle 2024-2028 = sustained ASIC + networking pull. cycle_exposure: ai-capex-high. ⚠️C
- Capital allocation discipline ~$20B+/yr return (dividend + buyback). ROIC ~25-30%. ✅C
IV. H-0 thesis
H-0: AVGO at $235† combines custom AI ASIC franchise (30-50% YoY through FY2027 toward $50B+ run-rate by FY2028), VMware OPM expansion to 50%+, Apple wireless stability, networking AI-fabric pull, and Hock Tan capital allocation — making AVGO a multi-engine AI-cycle compounder where custom AI silicon is LOAD-BEARING growth and VMware is cash-flow stabilizer.
Mispricing taxonomy: Structural × Interpretation. H-0 confidence post-Stage 3: ~75%; post-2026-05-23 R2 re-run: ~78% (FY2025 actual revenue $63.89B vs M3 ~$57-60B† — see update_2026-05-23.md).
Falsification: ASIC growth <20% sustained; hyperscaler in-house design displaces AVGO; VMware OPM stalls <40%; AI-capex 2-of-4 triggers fire; Hock Tan succession discontinuity.
V. Tree — six branches
A) Custom AI ASIC (LOAD-BEARING) ✅C strong | B) Networking ✅C strong | C) Apple wireless ✅C partial | D) VMware ✅C strong | E) AI-capex cycle ⚠️C partial | F) Capital allocation ✅C partial
Verdict tally (post-2026-05-23 R2 re-run): 12 ✅ · 2 ⚠️ · 0 ✗ · 0 ⊗ (was 10✅/4⚠️ at original build; Q9 market-share + Q12 price-reasonable verdicts upgraded — see Section XII)
VI. Consensus + mispricing
Consensus targets $210-290; ~75-80% buy. Narrative: "Custom AI silicon + VMware OPM expansion + Hock Tan = compounder."
Mispricing: Structural hybrid silicon+software multiple under-recognized; SOTP framing yields modest premium. Interpretation: AVGO trades modestly below NVDA on custom-ASIC franchise quality.
VII. Scenarios (post-2026-05-23 R2 re-run — see scenarios.md for full derivation)
| Scenario | Prob | Target | Upside/down from $235 |
|---|---|---|---|
| Bull — ASIC 50%+ + VMware 55%+ + AI-capex intact | 30% | $345 | +47% |
| Base — ASIC 30-40% + VMware 50% + cycle modest | 45% | $275 | +17% |
| Bear — AI-capex digestion + ASIC share loss + VMware stall | 25% | $140 | -40% |
Expected value $262; prob-weighted return +11.6%; asymmetry 1.16× favorable (was EV $236 / +0.4% / 0.71× unfavorable at original build; R2 verified FY2025 actual $63.89B beat M3 model midpoint $58.5B by ~$5B / ~9%, lifting scenario targets while holding the 30/45/25 probability vector).
VIII. Risks
Cyclical HIGH (ai-capex-high); Competition MOD-HIGH (hyperscaler in-house); Execution MOD (VMware); Customer concentration MOD-HIGH (Apple wireless); Correlated-factor HIGH.
IX. Historical analogues
NVDA 2022-2024 (AI silicon pull-through); TSM 2020-2024 (fabricates AVGO ASICs); Microsoft 2018-2022 (software compounder analog for VMware); Hock Tan playbook 2010-2020 history.
X. When H-0 fails
Scenario 1: AI-capex digestion — Target $110-150. Scenario 2: Hyperscaler in-house displacement — Target $130-170. Scenario 3: VMware customer revolt — Target $180-210. Scenario 4: Hock Tan succession — Target $190-220.
XI. Final verdict
Hold-with-sizing 2-3% initial; cap 4% (sizing band unchanged post-R2 — K.4 ai-capex-high coordination with NVDA + TSM + AMD + AJNMY + ASML is the binding constraint, not AVGO-specific conviction; viewed standalone AVGO now deserves higher sizing post-R2 but the bucket cap holds the position).
XII. Investment Scorecard (per K.6) — Re-run 2026-05-23 (R2-driven; FY2025 actual rev $63.89B beat M3 model by ~$5B)
| # | Q | Weight | Score | Verdict |
|---|---|---|---|---|
| 1 | Business durable 10+ yr | Critical 5× | 4/5 | ✅C |
| 2 | Moat trajectory | Critical 5× | 4/5 | ✅C |
| 3 | Capital allocation | Load-bearing 3× | 5/5 | ✅C Hock Tan |
| 4 | Balance sheet survivable | Load-bearing 3× | 4/5 | ✅C |
| 5 | Pricing power | Load-bearing 3× | 5/5 | ✅C |
| 6 | ROIC>WACC | Important 2× | 5/5 | ✅C 25-30% |
| 7 | Competitive advantage | Important 2× | 4/5 | ✅C |
| 8 | FCF visibility | Important 2× | 5/5 | ✅C $22-25B |
| 9 | Market share | Important 2× | 5/5 | ✅B FY2025 actual rev $63.89B / +24% YoY (R2-verified) — ~9% above M3 upper bound; custom-ASIC franchise tracking ahead-of-model pace |
| 10 | Talent risk | Confirming 1× | 3/5 | ⚠️C Hock Tan succession |
| 11 | Regulatory tail | Confirming 1× | 4/5 | ✅C |
| 12 | Price reasonable | Confirming 1× | 4/5 | ✅B asymmetry flipped 0.71× unfav → 1.16× fav post-R2; prob-weighted return +0.4% → +11.6% |
| 13 (LT) | Multi-decade optionality | Confirming 1× | 4/5 | ✅C |
| 14 (LT) | Team alignment | Confirming 1× | 4/5 | ✅C |
| 15 (LT) | Profitability path | Confirming 1× | 5/5 | ✅C |
K.3.5 derivation (R2 re-run 2026-05-23)
- Critical: 4+4 = 8 × 5 = 40
- Load-bearing: 5+4+5 = 14 × 3 = 42
- Important: 5+4+5+5 = 19 × 2 = 38 (Q9 4→5 on FY2025 ~+24% YoY actual vs ~+12% modeled — +2)
- Confirming: 3+4+4+4+4+5 = 24 × 1 = 24 (Q12 3→4 on asymmetry sign-flip — +1)
- TOTAL: 40+42+38+24 = 144 (was 141; +3 from R2 finding)
Max: 165.
xii_score = 144 / 165 = 87% → high-conviction band edge (was 85.5%). Per /39 normalized: 33.9 / 39 = 87%.
INDEX_META updated to 87% (was 87% declared / reconciled to 85%; new value reconciles cleanly across body + meta).
Final verdict
Hold-with-sizing 2-3% initial; cap 4% with K.4 coordination (unchanged sizing band — the R2 upgrade lifts AVGO from "fairly-priced" to "modestly favorable" but the K.4 binding constraint (ai-capex-high; coordinate with NVDA+TSM+AMD+AJNMY+ASML) still caps the position. The position decision is K.4-bucket-limited, not AVGO-conviction-limited).
2-minute pitch
AVGO is a multi-engine AI-cycle compounder. R2-verified FY2025 revenue $63.89B / +24% YoY beat the M3 model by ~$5B (~9% above the upper bound; ~24% YoY vs ~12% modeled). Custom AI ASIC franchise (Google TPU + Meta MTIA + Microsoft Maia + emerging OpenAI) growing 50%+ YoY toward $50B+ run-rate by FY2028; VMware OPM expansion via Hock Tan playbook (25% → 50%+); Apple wireless stability; networking AI-fabric pull. Hock Tan capital allocation = $20B+/yr return. Durability 21/25 Medium-High; 0 fatal flags. cycle_exposure: ai-capex-high — coordinate with NVDA+TSM+AMD+AJNMY+ASML per K.4 sizing rule. Asymmetry post-R2: +11.6% prob-weighted, 1.16× favorable (was +0.4% / 0.71× unfavorable — the R2 finding flipped the sign). Hold-with-sizing 2-3% initial; 4% cap with K.4 coordination. The K.4 binding constraint is what caps the position, NOT AVGO-specific conviction — AVGO viewed alone deserves higher sizing post-R2.
Risk types
Cyclical HIGH (ai-capex-high); Competition MOD-HIGH; Execution MOD; Customer concentration MOD-HIGH; Correlated-factor HIGH; Leadership transition MOD.
When NOT to buy
- AI-capex bucket sum >15% portfolio (K.4 binds)
- ai-capex 2-of-4 triggers firing
- Hyperscaler in-house chip design displacing AVGO
Long-term holdability
Qualified for long-term hold. 21/25 Medium-High. Multi-engine compounder. cycle_exposure: ai-capex-high requires K.4 sizing coordination.