Broadcom's AI chip business is growing fast, and it's already showing up in results, not just future promises: AI-related revenue reached $10.8 billion last quarter, up 143% from a year earlier, and now makes up about half of total revenue.
Debt taken on to buy VMware in 2023 is being paid down on schedule, and a major credit-rating agency just upgraded the company as a result.
At today's price, the stock is priced for the company's own multi-year AI growth targets ($56 billion in AI revenue this fiscal year, more than $100 billion the year after) to arrive largely on schedule; real upside exists if those targets are beaten, but a handful of large customers make up around 40% of revenue, and chip demand has a long history of booms followed by busts.
Fairly Valued, the price already reflects a lot of good news that still has to show up.
Broadcom sells custom AI chips and high-speed networking semiconductors to large cloud computing companies, its fastest-growing business, plus infrastructure software (VMware) sold mainly to large enterprises running their own data centers.
We'd revisit this call if the price falls below $220.00 or rises above $560.00.
These are review triggers for us to take another look, not price targets or a promise the stock will reach them.
| Metric | AVGO (subject) | Peer median |
|---|---|---|
| EV/Sales | 29.90x | 19.39x |
| EV/EBIT | 74.96x | 79.08x |
| EV/EBITDA | 74.96x | 56.83x |
| P/B | 23.24x | 13.18x |
| Gross Margin | 67.8% | 53.2% |
| ROIC | 19.3% | 12.4% |
| ROE | 28.4% | 18.7% |
| Rule of 40 | 54.7% | 19.8% |
TTM revenue (through 5/3/2026) $75.465B, built from Q3FY25 $15.952B + Q4FY25 $18.015B + Q1FY26 $19.311B + Q2FY26 $22.187B (company press releases). Quarterly YoY growth has accelerated every quarter disclosed (22% -> 28% -> 29% -> 48%), driven almost entirely by AI semiconductors ($10.8B Q2 FY26, +143% YoY); non-AI semiconductor revenue (~$4.2B/quarter) is roughly flat. The 12-month debate: is $100B+ FY2027 AI revenue (vs. $56B FY2026 guide) achievable, or does it require flawless multi-year execution across a handful of customers.
GAAP operating income $10.788B in Q2 FY26 (48.6% margin); non-GAAP operating margin ~67-68% every quarter, a wide and widening GAAP/non-GAAP gap driven mainly by VMware purchase-accounting intangible amortization (~$2.05B/quarter) - a real, recurring cost of Broadcom's serial-acquirer model, not a one-time add-back (B0.7 Q4 serial-adjuster flag). Infrastructure Software operating margin reached ~79% (+310bps YoY) in Q2 FY26, showing real VMware-integration cost discipline.
TTM OCF $33.622B, capex only $0.860B (fabless model - no AMZN/GOOGL-style AI-capex FCF crush), SBC $8.785B. Plan-formula FCF (OCF-capex-SBC) = $23.977B, a 31.8% margin on TTM revenue - genuinely strong, undistorted cash conversion (company-defined FCF, OCF-capex only, is $32.762B / 43.4%). TTM GAAP net income $29.317B (10-Qs/8-Ks, 2025-2026).
Total debt $64.907B, cash $19.628B, net debt ~$45.3B as of 5/3/2026. Despite the large absolute VMware-era debt load, it is now only ~3.5% of enterprise value (~$1.84T) given the market-cap run-up. Moody's upgraded senior unsecured to A3 (from Baa1, positive outlook); total debt/EBITDA is guided below 1.5x by FYE2026 versus ~3x at FYE2024 (S&P Global Ratings commentary). Dilution is minimal (diluted shares ~4.876B). Buyback capacity is modest: the $10B authorization (approved ~Dec 2025) was ~93% utilized by 2/1/2026 (only $700M left) - well under the 3% buyback support threshold; the dividend, not buybacks, is the primary capital return ($11.142B paid FY2025 vs. $2.45B repurchased; 15th consecutive annual increase targeted at $2.60/share, yield ~0.7%).
FCF yield ~1.3% of market cap ($23.977B/$1.79T); TTM EV/adjusted-EBITDA ~35.9x (EV $1.84T / TTM adj. EBITDA $51.29B) but ~26x on FY2026 guided-run-rate EBITDA - the gap itself signals how much of the multiple already assumes forward growth. Trailing P/E ~60-62x; forward P/E ~20-23x; PEG ~0.5.
AVGO's ~80% share of high-end Ethernet switching (Tomahawk 6, Jericho4) is a genuine moat as hyperscalers shift AI-cluster fabric from InfiniBand to open Ethernet. Versus AI-chip peers: NVDA data-center revenue +56% YoY (forward EV/Sales ~17.6x); AMD data-center +57% (forward P/E ~84x, richer than NVDA on a much smaller base); MRVL data-center +69% YoY (forward EV/Sales ~8.4x). AVGO's blended growth (semis +79% YoY) sits between MRVL and NVDA in rate but carries a software-margin cushion (93% gross margin Infrastructure Software) neither pure-play peer has - a stock-specific quality premium, not purely a sector bet.
Sensitivity channel = AI/hyperscaler capex cycle. 10Y UST ~4.5-4.5% (July 2026) raises financing costs broadly; the more direct transmission is hyperscaler capex intentions (Google, Meta, Microsoft, Amazon) - any pullback in disclosed AI infrastructure budgets (as flagged in AMZN's own FY2026 capex guide) would hit AVGO's order book faster than rates would.
CEO Hock Tan sold ~300K shares (~$101.3M) in January 2026 near $347-348; Chairman Henry Samueli sold >$651M over the trailing three months plus filed to sell a further $71M (2026-06-29) - large in dollar terms but ~0.04% of market cap, below the S1 net-selling threshold; reads as diversification against concentrated legacy stakes, not a fresh signal, though the scale of Samueli's selling warrants a DEGRADED watch note.
Catalysts - Q3 FY2026 print (guided $29.4B revenue, $16B AI revenue, ~67% non-GAAP op margin); new Apple RF/ASIC agreement (>$30B through 2031, disclosed 2026-07-06, +4-5% stock reaction); further named hyperscaler XPU wins (OpenAI, Anthropic already disclosed as supply-deal counterparties). Risks - top-5-customer concentration at ~40% of revenue; AI order book depends on continued hyperscaler capex intensity that peers (AMZN, GOOGL) are already flagging as elevated; semiconductor-sector cyclicality precedent (2022-2023 correction) if AI-server demand digests.
(Q6) AI revenue beat/matched guidance every disclosed quarter and bookings backlog exceeds $30B, arguing against a pure narrative story; Infrastructure Software margin genuinely expanded post-VMware, evidencing real integration execution, not just cost-cutting theater; debt is being paid down on schedule with rating upgrades, undercutting the 'over-levered acquirer' framing. (Q7) Falsification: if quarterly AI semiconductor revenue growth decelerates below 100% YoY for two consecutive quarters, or FY2027 AI revenue guidance is cut from >$100B, the FAIR call should shift toward OVERVALUED.
reassess_low ~$209 (0.55x Base $380) - approach toward Bear. reassess_high ~$418 (1.10x Base) - approach toward Bull materializing.
Mechanical model output, a screen, not a fair value.conservative screen, understates growth/AI-capex names
A scenario EV/EBITDA sanity-check anchored on management's own multi-year AI-semiconductor guidance (FY2026 $56B, FY2027 >$100B) puts Base fair value near $380/share against a $367.97 price (~3% below Base, inside the FAIR band); Bear (~$220, AI-capex digestion/deceleration) and Bull (~$560, targets beaten, multiple re-rates further) show the price already requires the currently-guided ramp to land largely on schedule, without yet demanding upside beyond it.
The mechanical screen and the research view reach the same verdict.
Confidence = 0.7 × research sub-score + 0.3 × mechanical sub-score. Mechanical sub-score rewards the Layer 1 screen for agreeing with Layer 2's direction (corroboration), not just for being conviction=NORMAL in isolation.