AVGO — Broadcom Inc.

The tape and the reported numbers are pulling in opposite directions. Anchored to 2026-10-16: shares were down 8.5% in one week, 10.8% over three months and 18.6% over six months, RSI(14) was 35.54, and price sat 6.94%/9.56%/10.15% below its 20-, 50- and 200-day averages. The negative side is mostly narrative, not reported results: CEO Hock Tan's mid-October 2026 comment that access to capital — not wafers or memory — is becoming the binding constraint on data-center build-outs, plus a run of insider sales by Tan and CFO Kirsten Spears. Against that, Q3 FY26 revenue was $29.59B (+85.5% YoY), adjusted EPS $3.32, AI revenue $16.7B (+54% sequentially), Q4 AI revenue guided to ~$21.7B, and the sell-side remains lopsidedly positive (54 Buy, 6 Hold, 0 Sell; consensus target ~$510). Net: a genuinely divided market, with price action bearish and fundamentals bullish.

Three-year arc: a VMware-assisted step-up, then a real AI inflection. Revenue ran $9.30B (Q4 FY23) into $11.96B, $12.49B, $13.07B, $14.05B through FY24 at low-single-digit sequential growth, then $14.92B, $15.00B, $15.95B, $18.02B through FY25 — still only ~4–8% per quarter. FY26 breaks the pattern: $19.31B → $22.19B → $29.59B, i.e. +7%, +15%, +33% sequentially, with the latest quarter up 85.5% YoY. This is acceleration, not a one-quarter blip — management guided Q4 AI revenue to ~$21.7B after $16.7B in Q3. Operating margin tells the same story: 36.9% (Q3 FY25) → 41.7% → 44.9% → 48.9% → 54.3% (Q3 FY26), a 17.4-point YoY expansion, with gross profit of $20.46B on $29.59B of revenue. Diluted EPS went $0.85 → $1.74 → $1.50 → $1.91 → $2.68; the odd dip in the Feb-2026 quarter (net income fell to $7.35B from $8.52B on higher revenue) looks like a non-operating/tax item rather than an operating break, since operating income rose in that quarter. Free cash flow is compounding faster still: $7.02B (Aug-2025) → $7.47B → $8.01B → $10.26B → $13.67B, or ~$39.4B TTM. Caveat: the Q4 FY23 and Q1 FY24 prints straddle the VMware close and a fiscal-year change, and the Aug-2024 quarter carried a -$1.88B net loss from acquisition accounting, so the oldest year-over-year comparisons are noisy.

As of 2026-10-16 the stock traded at $331.14 on 41.1x TTM EPS of $8.03 versus its own 5-year average of 56.4x — optically cheap on earnings, but only because trailing EPS is still catching up to a price that has already de-rated. On sales the picture flips: 17.7x P/S vs a 5-year average of 13.1x, a clear premium to its own history. PEG of 0.409 flatters because the trailing growth rate is extraordinary and not repeatable. Applying the supplied next-FY consensus EPS of $11.676 gives ~28.4x forward — though that estimate looks stale: Q3 FY26 GAAP EPS alone was $2.68 (~$10.7 annualized and rising), and next-FY consensus revenue of $106B sits below the current annualized quarterly run-rate of ~$118B. The research brief cites ~19x forward earnings, implying the Street's live numbers are well above the stale feed. Net: not a value stock on any absolute measure — 31.6x EV/EBITDA, 15.8x book, 0.72% dividend yield — but not expensive for a business growing revenue >85% with a 54% operating margin, especially against a 5-year average P/E of 56x. The valuation argument here is entirely about the duration of the AI cycle, not about the current multiple.

The setup is unusual: earnings are beating and guidance is rising, yet the stock keeps falling, down 18.6% over six months as of 2026-10-16. That means the market is not arguing with the reported numbers; it is discounting their durability. The shares are effectively pricing a future deceleration — capex digestion, customer concentration, vendor-financing strain — rather than any current shortfall. Two implications. First, the next print (2026-12-10, estimated) is a low bar on the printed numbers and a high bar on guidance: the FY27 AI revenue framing matters far more than the Q4 beat. Second, if AI revenue actually tracks toward the reiterated $115B FY27 target, the current price is discounting a shortfall that is not visible in orders. Watch hyperscaler capex commentary and any change to the multi-year AI targets — that is the variable the multiple is keyed to. Do not anchor on the trailing 41x P/E; it is a lagging artifact of an earnings ramp that has not finished.

Anchored to 2026-10-16: RSI(14) was 35.54 — weak, but not yet oversold (30 would be the washout signal). Price was 6.94% below the 20-day SMA, 9.56% below the 50-day and 10.15% below the 200-day, which is the textbook definition of a downtrend with no base formed. The prior close was $344.33 and the stock lost another 3.8%, so momentum is still negative. Against a 52-week range of $289.96–$495.00, the shares sit roughly 14% above the 52-week low and about 33% below the high. Entry-timing implication: there is no technical confirmation of a bottom here. An RSI in the mid-30s inside a confirmed downtrend is not, on its own, a reliable dip-buy trigger; the more meaningful references are the $290 area (52-week low) as support, and the 50-day average as the first level that would have to be reclaimed to suggest the trend has actually turned.

• Hyperscaler AI capex cycle — Broadcom's custom ASIC and AI networking revenue is directly levered to a handful of cloud customers' infrastructure budgets, so a digestion pause hits orders before it shows up in reported revenue.
• AI financing and capital availability — CEO Hock Tan's mid-October 2026 remark that capital access, not wafer or memory supply, is becoming the binding constraint raises the question of how much of the build-out is being vendor- or customer-financed, a structural risk to the order book.
• Memory/HBM and wafer input cost inflation is pressuring gross margin; the 68.8% TTM gross margin has held up so far mainly because operating leverage has outrun input costs.
• Interest rates and leverage — Broadcom carries $57.2B of long-term debt ($33.2B net of $24.0B cash) with a 1.476 beta, so both refinancing cost and the discount rate applied to long-duration AI growth are rate-sensitive.
• Sector rotation and high-multiple de-rating — the AI complex has been volatile on monetization concerns, and a 17.7x sales multiple leaves little insulation from a sentiment-driven repricing even though AVGO screens cheaper than pure-play GPU names.

• Q4 FY26 earnings on 2026-12-10 (estimated) — the printed quarter matters less than the FY27 AI revenue framing and any update to the $115B (FY27) / $230B (FY28) targets.
• Hyperscaler capex announcements and custom accelerator deployment updates with customers including Anthropic and OpenAI — the orders that either validate or break the multi-year targets.
• Gross margin trajectory as HBM and memory costs flow through — evidence that operating leverage can keep absorbing input inflation.
• Further debt reduction and capital return funded by ~$39.4B TTM free cash flow (dividend yield is only 0.72%), which could carry more of the story if growth sentiment cools.
• VMware/software integration and networking attach rates inside AI clusters — the less cyclical, higher-visibility half of the business.

• AI capex digestion: if hyperscalers slow spend, the $115B FY27 AI revenue target becomes unreachable and a 17.7x sales multiple has a long way to fall.
• Customer concentration plus vendor-financing/counterparty risk — the CEO himself flagged capital availability as the emerging constraint in mid-October 2026.
• Insider selling by CEO Hock Tan and CFO Kirsten Spears through recent weeks, a poor signal at a moment when the shares are already under pressure.
• Gross margin risk from HBM/memory price inflation; gross margin is the shock absorber and has limited room if input costs keep climbing.
• Valuation/duration risk: 31.6x EV/EBITDA, 15.8x book and a 1.476 beta leave no margin of safety if the AI narrative cools, regardless of reported earnings.
• Balance-sheet leverage — $57.2B of long-term debt is manageable against $39.4B TTM free cash flow but constrains flexibility if rates stay high.

The business is executing better than at any point in its history — Q3 FY26 revenue +85.5% YoY, operating margin 54.3%, $39.4B TTM free cash flow, and a reiterated path to $115B of AI revenue in FY27 — and the stock has de-rated 18.6% over six months anyway, sitting below every major moving average as of 2026-10-16. That gap is the opportunity if the AI targets hold and the trap if they don't. On the deterministic score's own terms this is Strong Fundamentals at a Neutral-versus-history valuation, which supports a constructive lean; the honest caveat is that the entire thesis rests on a handful of hyperscaler customers continuing to spend, and the market has started pricing that risk rather than ignoring it. Reasonable for a long-horizon investor comfortable with that concentration and with buying into a downtrend; not a trade for someone who needs the tape to confirm first.

Applied ~31x to my own FY2027 GAAP EPS estimate of ~$14.00, giving ~$434. The multiple is deliberately below the stock's 5-year average P/E of 56.4x and roughly a PEG of 1.0–1.1 on a normalized ~28–30% forward growth rate — a premium to a mature semiconductor multiple but a discount to where the Street's targets imply the stock should trade. The EPS input is mine, not the feed's: the supplied next-FY consensus of $11.676 is stale against a Q3 FY26 quarterly GAAP EPS of $2.68 (~$10.7 annualized and still compounding sequentially), so I blended consensus with the visible run-rate. Cross-check: ~30x on ~$65B of FY2027 free cash flow (TTM $39.4B, up from $20.3B two years earlier) implies roughly $400/share after net debt — the two methods bracket $400–435, and I take the upper end at $435.