Altair
Skip to content

Trading Analysis Report: AVGO

Analysis date: 2026-10-21 · Model: claude-opus-5-5 · AI-generated, not financial advice.

  • Analysis date: 2026-10-21
  • Rating: Hold
  • Generated: 2026-10-21 11:28:42
  • TradingAgents 0.6.0: anthropic, deep claude-opus-5-5, quick claude-sonnet-5-5
  • Analysts: market, sentiment, news, fundamentals; research debate rounds 5, risk debate rounds 5
  • Data vendors: core_stock_apis yfinance, technical_indicators yfinance, fundamental_data sec_edgar,yfinance, news_data yfinance, macro_data fred, prediction_markets polymarket

V. Portfolio Manager Decision

Rating: Hold

Executive Summary: Hold AVGO at $339.07. Do not buy at market. Holders at or below 1.0x of a standard allocation keep their position, and anyone above 1.0x should offer the excess at $344-352 and sell back to 1.0x on a daily close below $328.58. If a Q4 guide lookup clears the ~$31.1B veto, build toward 0.5x total AVGO exposure: 0.2x at $321 (live after Meta reports, only if price is still above it), then 0.2x at $317 and 0.1x at $309 (live after the later of the Oct 28-29 FOMC and the last big hyperscaler capex print). Size every tranche to survive a 13% gap, exit all newly added exposure on any daily close below ~$292, add above $368.78 only in 0.25x steps to a 0.75x ceiling until the December print, and treat $351 as the 2-3 month target.

Current Price: 339.07

Price Target: 351.0

Confidence: Medium

Time Horizon: 2-3 months

IV. Risk Management Team Decision

Aggressive Analyst

Aggressive Analyst: Neither of you has spoken yet, so I'll lay out my case and then get ahead of where I expect you to push back.

People hear "Hold" and assume it's the timid option. It isn't. A Hold at 339 with scaled bids at 321, 317 and 309, a market starter if the guide comes in at or above about $32.6B, and an extension to 1.0x on a high-volume close above 368.78 is a loaded spring. The plan says we want more of this business, but not at a price where the edge is thin and its sign depends on one uncertain give-back probability. The scenario grid only clears the 9% hurdle around 307 to 330, and the bids sit in that zone.

The business is the reason to want more. Revenue growth accelerated from 29.5% to 47.9% to 85.5% over three quarters. Operating expenses have stayed near $4.5B a quarter, so operating margin hit 53.9% with incremental margins around 70 to 74%. Free cash flow was $13.7B in Q3 alone. Cash is up 124% year over year, the current ratio is 2.5, and net debt is roughly 0.7x EBITDA by my derivation. The price has moved the other way, down about 29% from the 479.94 close on June 2. On Q2 run-rate EPS of about $7.64, the stock was near 63x at the peak. On Q3 run-rate EPS of $10.72 it's 31.6x. The multiple has roughly halved while earnings power rose about 40% in a quarter. I'm not calling that cheap, which is why we aren't buying at market. But it is a very different setup from June.

To the conservative side, I'll concede what's true. The weekly SuperTrend is down, price is below the 10 EMA, 50 SMA and 200 SMA, the 50 is under the 200, and closing lows have stepped down to 328.75. RSI never hit 30 and every z-score is inside ±2, so a washout lower is entirely possible. But that is the scenario the plan is built to profit from. If the stock flushes to 321, 317 or 309, we're buyers at prices where the grid says the edge is positive. If it gaps up on a strong guide, the starter covers us. The only way to lose the argument is to be fully loaded at 339, and the plan explicitly doesn't do that.

I'd also resist treating the trend indicators as a reason to stay away. They lag. The weekly flip level is 429.30, about 27% above price, so waiting for confirmation there means paying 27% more. The monthly SuperTrend is still up with its stop 28% below. Meanwhile the daily tape is showing early cracks in the selling. The TD-9 is complete, though provisional. RSI is 41.98 against 32.27 when price last sat at 339 in September, OBV is higher too, the MACD histogram has narrowed from -2.48 to -2.08, and the last two sessions drew 15.8 to 17.5M shares against 23 to 30M on the heavy down days. I'll be candid that today's bar is partial at 8.2M shares, so none of this is proof. But it's the right shape for a bounce setup.

The risk/reward from the plan's 317 entry with a 292 stop is about 7.9% at risk against 16% to the 368.78 resistance cluster and roughly 35% to the weekly flip. That's about 2 to 1 on the near target and 4.5 to 1 on the far one. The 13% gap risk and the 1.48 beta are sizing issues, and the plan handles them by capping total exposure at 1.0x and sizing tranches to survive the gap.

On insiders, $674M of sales sounds ominous. It's about 0.04% of a $1.6T market cap, and it clusters after vest dates, which looks scheduled, though that's my inference. The CEO hasn't sold since January. I'll grant that insider buying is negligible, since one director put in about $0.7M. It's a mild negative, not a thesis breaker.

On the AI bubble, Griffin's comment and the Polymarket odds of a 2026 burst at 10%, up 3.5 points, are real tail risks. But that still leaves 90% on the other side, recession odds are 8%, and the hike odds that were weighing on multiples fell from about 33% to 14% in a week. I'll flag that the FRED data failed, so I have no verified yields or VIX. Sentiment is low-confidence, with only 12 tagged posts, and I wouldn't build a thesis on it. But it shows frustration rather than euphoria, and nobody is crowding into this name.

To the neutral side, you'll probably say the plan is balanced and we should wait for data. I agree with the Hold, but waiting can't mean passivity. The gates are the Q4 guide, a clean 10-Q, no hyperscaler capex cut and no FOMC hike. If they clear, we should deploy at once, not ask for another round of confirmation, because the cost of over-confirming is buying near 369 or higher instead of 317 to 321.

For existing holders, selling a compounder like this after a 29% drawdown, ahead of catalysts that could reverse the tape, gives away the bounce setup for no gain. Without portfolio context I can only give sizing relative to a standard allocation: holders at or below 1.0x keep their position, and holders above it shouldn't add. My position is Hold, with the buy orders staged and the triggers set. Aggressive Analyst: Several of your corrections are right, so I'll fix those first. The monthly SuperTrend cushion is about 22% from price, not 28%. The 429 weekly flip is a trend-change signal, not a price objective, so my 4.5-to-1 was decoration. The RSI comparison should be low to low, 34.89 against 32.27, which is a mild 2.6-point improvement. "Nobody is crowding in" can't be drawn from 30 messages in a two-hour window, and the ETF count is mechanical. I withdraw all four. Conservative is also right that I was loose about the bids. They sit behind the gate. A bid that is live ahead of Meta only fills if Meta goes badly, which is adverse selection. So nothing is live tonight.

None of that touches the core of my case, which is price against earnings power. On the technicals, I only need them to tell me not to chase at market.

Conservative, you call 43x trailing the honest multiple. Neutral showed it includes an 85-cent quarter. It also only works as a forward multiple if Q4 EPS comes in around $1.96, which is 27% below the $2.68 just reported. Operating expenses are flat near $4.5B and incremental margins are 70%. That is a demand-collapse scenario, not a lumpy-quarter one. The lumpiness question is worth the 29x to 35x range Neutral gave.

You also say the market paid less for better earnings. I don't need the market to be wrong. I need the expectations embedded in $339 to be lower than in June, and they plainly are. The multiple halved while the earnings it applies to rose 40%.

The filing flags are smaller than they look. Gross margin slipped 0.4 points in a quarter where revenue rose $7.4B, and it's still 2 points up year over year, with operating margin up 5 points sequentially. The 89% cash conversion compares after-tax, after-interest cash flow with pre-tax operating income. Q3 net income is roughly $13B by my math from EPS and share count, so operating cash flow is about 110% of that. Amortization and stock comp flatter that, so I won't call it clean. It's a 10-Q lookup, not a thesis. The net debt ratio is a vendor derivation, but the statements themselves show liabilities down $3B in a quarter while assets rose $9B.

On risk and reward, you're right that a gap turns my 7.9% loss into about 13%. But a gap hits every entry equally. A 13% gap from 345 is also a 13% loss. That argues for sizing, which the plan already does, not for a different entry price. Run the break-even odds both ways. From 317 to 368.78, I need about a third of trades to work if the stop fills, and about 45% if every loss is a 13% gap. From Conservative's 345 entry with a stop under 328.58, the hurdle is about 41% without a gap and about 65% with one. The lower entry wins either way, and 368.78 is a level the stock closed near just two weeks ago.

Neutral's point about the gaps fits the dates. They look like earnings gaps, and the next earnings report is in December, so the build has seven or eight weeks before that window. The plan's exit is also a close below 292 with a tripped tripwire, not a resting stop order that sweeps into the March-low cluster.

On the cost of waiting, you say a few percent. But your extension rule needs a weekly close above the 200 SMA with the 50 SMA flattening. The 50 has fallen about 18 points in a month, so that means paying well into the 380s, over 20% above 317. That is the real price of your confirmation.

The AI-bubble contract is $2.5M of volume with no definition of "burst." Even taken at face value, a 30% hit on a 0.5x build costs 15% of a standard allocation before any exit. The plan's review and exit levels cut that off well short. A stock already 29% off its peak is a different exposure from one at 480. The FRED outage is a vendor problem. VIX is on any quote screen, so we aren't trading blind.

On insiders, I'll verify the CFO label, but it's one lookup. The CEO did not sell into the 480 to 493 peak, and nobody has sold since July 10, so no one is selling at 339.

Neutral, you say the guide is a lookup, and I agree. But then "nothing at market today" is a decision made before the lookup you asked for. If the guide is at or above $32.6B and the 10-Q is clean, the plan's starter is live. I'd fire it once Meta clears.

I disagree with holding the 317 and 309 tranches until the 29th. The hold is 86% priced, and a hike tail of 14% against 0.2x and 0.1x tranches is a point or two of a standard allocation. If we wait, we chase a relief rally in the likely case. I'll take your 0.2/0.2/0.1 split, whose weighted average is exactly 317, and your upside leg of 0.15x after two closes above 343.86. That fixes the hole in my plan where it grinds up without filling anything. I'll take your extension ladder over Conservative's weekly condition. For holders above 1.0x, a limit sale into 344 to 352 is fine, but a forced sale below 328.58 sells the excess in the flush where the plan's own edge turns positive.

So it stays a Hold, with the guide and 10-Q checked today. The starter goes in after Meta if both pass, and the three bids go live after Meta. The upside leg and the extension ladder stay in. I'm not cutting the build to 0.25x, because that is a preference, not a finding. Aggressive Analyst: I'll start with what I'm giving back, because it's more than usual this round.

You're both right about the market starter. My case was that a strong guide would gap the stock and the starter would cover us. But the guide is a seven-week-old fact, not a catalyst, and that mistake came from the plan. With the stock sitting at its September 15 close and the guide already public, a starter at 339 is just buying valuation where the grid says the edge is thin. It's dropped. The 7.9% risk is also a floor with a close-based exit, so every tranche gets sized to the gapped price.

I'll take the 0.2/0.2/0.1 split with 321 live after Meta, and I'll stop arguing the 317 and 309 timing. Once Meta is read, one of the two tails is gone, so the "same risk factor" argument loses most of its force. But Neutral is right that this is a few tenths of a percent of a standard allocation. The 309 cancel rule is fine too.

Neutral's reductio on holders above 1.0x also works on me. If a holder at 1.0x can't add in the flush, a holder at 1.3x has no claim to keep the excess there. Offer it at 344 to 352 and sell it back to 1.0x, no further, on a close below 328.58. Nobody is dumping a compounder. They're coming back to budget.

Here is where I'm not moving.

Conservative, you called 63x a strange anchor, so use a better one. Tan sold at about 347 on January 6, and You bought at 325 on December 18. The last time this stock traded where it does now, the latest reported quarter was Q4 FY25, whose EPS the filings imply at about $1.74, or roughly $7 annualized. That's about 50x. Today the same price buys $10.72 of run-rate, 31.6x. It's the same price and the same stock, with three more prints behind it, and the multiple is down by more than a third. The 38% run to 480 overshot, and I'm not asking for a repeat. But you can't call that setup a hope that this time is different.

On the market paying less for better numbers, Neutral is right that September isn't established, since nobody has read the September 3 row. What the data does show is the path after it: 338.65 on September 15, 376.51 on October 7, an 11% rally, then all of it gone by October 16 with no company event in our data. A round trip on no news, with Griffin's bubble comment, Credo down 15% and the "too dependent on AI spending" headline in the tape, looks like narrative and positioning to me. I can't prove that, and you can't prove the durability reading. When neither side can prove it, the price we pay is the one thing we control, and staged bids are the cheaper way to hold that uncertainty.

On break-even, you're right that a 317 fill is conditional on support breaking and that equal win rates is an assumption. But the resistance ladder at 351.69, 360.20, 366.94 and 368.78 is the same ladder a buyer at 345 has to climb. The 317 buyer is paid 16% for it and the 345 buyer is paid 7%. The conditioning also cuts both ways. The technical report's own continuation scenario has RSI under 30 and z-scores heading toward minus 2, and at 317 the run-rate multiple is about 29.6x. That's the washout the bids are built for. We can't measure the win rate there, and we agreed that argues for sizing, not for a different entry.

The real cost of the unfilled path is bigger than "a few points." If the stock grinds up without touching a bid, the plan holds 0.15x after two closes above 343.86. The rest of the position only arrives through the extension ladder at 369 and above, about 15% over the 321 bid. So the extension rule matters more than anyone has treated it. Neutral moved the 1.0x step from a second close to a Friday weekly close above 369. On a stock with a 3.3% ATR, that can cost four sessions. I'd go back to the earlier version: a second daily close above 368.78 on volume, or a held retest of 360, whichever comes first.

On the AI-semi cap, I agree Meta's capex moves NVIDIA and Micron too, and anyone holding all three should size smaller. But "count it toward the cap" has to be a correlation haircut, not one-for-one. Otherwise the rule says a normal tech book can never take the 0.5x build, which is a sector-wide sell signal nobody here is arguing for. Without portfolio context, the more AI semis you already own, the closer you stay to the low end of the build.

On December, I accept the rule: hold the base build into the print, and trim an extended position to 0.75x. The one gap whose setup we can see, June 4, hit a stock that had run roughly 38% in five months. A December print landing on a position built near 30x run-rate after a 29% drawdown is a different bet. I'd still size it to 13%.

So the Hold stands. Do the lookups today: the guide and 10-Q, the September 3 row, the CFO status, VIX, the 10-year and Meta's date. A failed gate cancels the build. If the gate passes, 321 goes live after Meta, and 317 and 309 go live after the FOMC. The 0.15x upside leg sits inside the 0.5x total. The extension goes to 0.75x on a high-volume close above 368.78, and to 1.0x on a second close or a held 360 retest.

The worst case on the base build is about 6.5% of a standard allocation if a 13% gap hits it. The cost of sitting out is not owning a business growing 85% a year through a 16% move from the bid zone to resistance. That's an asymmetry worth taking. Aggressive Analyst: Neutral, your catch on the extension is the best point of the round, and it's a hole in my own plan. I wrote the ladder as totals, 0.75x and then 1.0x, and never ran it from the path I keep saying is the real cost of sitting out. From the 0.15x upside leg, getting to 0.75x means buying 0.6x at about 369, right under the 200 SMA, the daily flip level and the October 7 close, in a weekly downtrend. That's bigger than the whole bid build, and a 13% gap on that piece alone is almost 8% of a standard allocation. So I take +0.25x per confirmation, two closes above 368.78 for the first step, a held 360 retest or a weekly close above about 369 for the second, and a 0.75x ceiling until the December print.

The ceiling does cost something. If the stock runs to the 429 weekly flip before December, the 0.25x we left unbought is worth about 4% of a standard allocation. The gap it protects against is about 3.3%. Those are two conditional tail outcomes of similar size, so I'll take the cap and not pretend it's free.

You're also right about the framing. The business accelerated for three quarters while the stock fell 29%, so owning the growth was never the issue. The cost of sitting out is the 16% from the bid zone to the ceiling. The December point is fair too. Until someone reads the September 3 row, I'm dropping any claim that a de-rated stock is safer going into a print, and I'm sizing to 13%. The half-weight rule for other AI semis is the number I was asking for. A partly invested holder fills the tranches from the top, 321 first, until total AVGO reaches 0.5x.

Conservative, I accept three things. The bid rule is cheap: a bid at 321 only counts if price is above it, and a gap through costs about 2.4% on a 0.2x tranche, roughly half a percent of a standard allocation. The two closes on the first extension step are right. A single close on October 7 would have bought into 376.51 and then eaten a 12.7% fall, and I can't see whether that day cleared the volume bar, so I'm leaning on the arithmetic, not the anecdote. And I take your 10% stress number as the ceiling. It needs all three tranches filled, a flush through 309, a close under 292, and then a 13% gap on the exit. That puts the stock near 252, below the monthly stop at 264.68 and about 47% under the June close. That's a broken-regime outcome, which is what a ceiling is for.

Your best remaining argument is that semis multiples look lowest at peak earnings. I can't refute that from this data, but it has a test, and we're running it today. If Q4 guidance shows a stall, that's the high-water mark you're worried about. The plan's veto line is a guide near $31.1B, about 5% above Q3 revenue of $29.6B. Anything under that kills the build. So the cyclical worry is answered by the lookup, not by pulling the bids back. If the guide clears, the bids sit at 29 to 30x run-rate, against roughly 50x at this same price in January.

You also called passing the gate the absence of a veto, not a green light, and I agree. The gate is a veto and price is the trigger. The bids only fill on weakness, the 309 tranche is the smallest and has a cancel rule, and nothing is bought at 339. You've signed off on 0.5x, the 0.75x cap, two closes, the bid rule and the hard limits for holders above 1.0x. That isn't a comfortable plan. It's a plan built around being wrong, with a small, staged position that lets us own the bounce if it comes.

So here's the Hold as I'd sign it. Nothing at market today. Run the lookups: the guide against the $31.1B veto line, the 10-Q receivables and inventory, the September 3 reaction, the CFO status, VIX and the 10-year, and the dates for Meta and the December print. A failed gate cancels the build. After Meta, 0.2x at 321 goes live if price is above it, and if price gapped through, we wait for a close back above 328.58. After the FOMC, 0.2x at 317 and 0.1x at 309 go live, and the 309 bid is cancelled if closes slip under 317. The 0.15x upside leg after two closes above 343.86 sits inside the 0.5x total. Extensions come in 0.25x steps, capped at 0.75x until the December print, and the base build is held through it. Total AVGO exposure is 0.5x, with other AI semis counted at half weight against the 1.0x cap. Anyone above 1.0x offers the excess at 344 to 352 and sells back to 1.0x on a close below 328.58. Every tranche is sized to the gapped price, not to the 292 stop.

That keeps a real path to size on both the flush and the grind-up, and it caps the worst case at about 10% of a standard allocation instead of leaving it open. We don't have to call the bottom. We have to be positioned to own the move from roughly 317 to 369 if it comes, without being the person fully loaded at 339 when it doesn't. Aggressive Analyst: Neutral, your calendar point changes what I said last round, so I'll take the hit first. I claimed the guide lookup answers Conservative's peak-earnings worry. If fiscal Q4 ends around November 1, using the same thirteen-week spacing as the May 3 and August 2 quarter ends, the guide covers a quarter that's almost nine-tenths over. The lookup can tell us whether this quarter stalled. It can't tell us whether $10.72 is the high-water mark. The Q1 guide in December answers that. So for the next seven weeks we're taking a modest position ahead of that test, at a price that already discounts a lot of doubt, sized so a bad answer is survivable.

The same calendar fact helps the valuation case, and neither of you has used it. The 31.6x is on Q3 EPS, and Q4 is nearly in the books. Take the veto-line guide of $31.1B, hold opex near $4.5B and gross margin near 69%, and Q4 operating income comes out about 6% above Q3. That's EPS near $2.85, or roughly $11.40 annualized. At 339 that's under 30x, and at 317 it's under 28x. This is the weak end of the passing range, and it's my arithmetic, not guidance, so hold it loosely. That $31.1B line is also still about 73% growth over the roughly $18.0B Q4 FY25 base, and $32.6B is about 81%. The middle band is growth easing from 85%, not a stall.

On the dial, I side with Neutral against cutting the build to 0.2x on a seven-week-old number with no consensus to measure it against. I'll accept dropping only the 309 tranche in the middle band. It costs 0.1x, and if the guide is merely adequate, I'd rather not own the tranche that fills last.

Conservative, the extension failure line is your best point of the round, and I'm taking it at a fixed close near 352, as Neutral wrote it. The 292 exit was drawn for entries in the low 300s and makes no sense for a step bought near 370. One consistency note. Since the exit is on a close, your 1.2% is a floor, not a cap, which is the argument we all applied to my 292 stop. The stock has had 4.4% days twice this month, so I'd write a failed step as 1.2% to 2%. That's still far better than 5% to 8%.

I'll also give you the upside leg waiting for the FOMC. Exactly one live order before the 29th is clean, and the cost is a few days on 0.15x.

Where I'm staying with Neutral is the mandatory trim to 0.75x a week before the print. Your arithmetic is right: 13% of 1.0x is more than 13% of 0.75x. But the gap is symmetric. Nobody here has shown which way it leans, and we haven't even read the September row. The 0.75x ceiling on new money isn't about unpaid gap risk. It's about not buying the last 0.25x at the top of a resistance stack. A standard allocation already includes the stock's gap risk by definition. A calendar sale for a book we've never seen, with unknown basis and taxes, is price-blind, and we argued against exactly that at 339. The checkpoint a week before the print is the right tool. A holder who wants 0.75x gets the plan's blessing, and nobody is forced.

I'd add one lookup. The gate says "no hyperscaler capex cut," plural, but our sequencing names only Meta. From memory, not the tools, the other big hyperscalers usually report in the last days of October, and Broadcom's custom-accelerator business leans on several of them. Pull those dates today. If one lands after the 29th, I'd rather know before the 317 and 309 bids go live than after.

So here's the plan as I'd sign it. Nothing at market today. Run the lookups: the guide, the 10-Q working-capital lines, the September 3 reaction, the CFO status, VIX, the 10-year, and the Meta, other hyperscaler and December dates. A guide under about $31.1B cancels the build, and the middle band drops 309. After Meta, 0.2x at 321 goes live only if price is above it. After the FOMC, 0.2x at 317 and 0.1x at 309 go live, the 0.15x upside leg arms, and everything sits inside 0.5x of total AVGO exposure, with other AI semis at half weight. Extensions come in 0.25x steps up to 0.75x, each abandoned on a close under about 352. The base build is held into December with a checkpoint a week before. Excess above 1.0x is offered at 344 to 352 and sold back to 1.0x on a close below 328.58. Every tranche is sized to the gap, not the stop.

Conservative, you've now signed off on a plan where every order is gated, every step has an exit, and every size is capped. If that's still a notch too comfortable, the only plan you'd call safe is owning nothing. That has a cost too: the 16% from the bid zone to the ceiling, and a seat at the Q1 guide with no position if the stock re-rates. This plan gets us into either the flush or the grind with a bounded downside, and we don't have to guess which one comes.