Trading Analysis Report: AMD¶

Analysis date: 2026-10-21
Rating: Underweight
Generated: 2026-10-21 11:26:32
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

I. Analyst Team Reports¶

Market Analyst¶

Current Price: 555.60 Price As Of: 2026-10-21

AMD (Advanced Micro Devices, Inc., NMS) Technical Report¶

1. Data integrity notes¶

Bar discrepancy. The raw get_stock_data CSV shows the 2026-10-21 bar as close 555.53 and volume 10,101,194. The verified snapshot shows close 555.60 and volume 10,101,395. I use the snapshot figures and have not reconciled the two.
Possible partial session. Volume of about 10.1M is the lowest of the last 30 sessions I can see (the next lowest is 13.7M on 10-05 and 10-09). That is consistent with an unfinished bar, but the tools don't say so. If so, the 10-21 values of OBV, TD-9, RSI, MACD and the supertrend/z-score tiers could still change.
Percentages. Percentages below are my arithmetic on verified prices, except the supertrend "distance" figures, which come from the tool.
Scope. These tools cover price and technicals only. They give no news, earnings or fundamental information, so I can't explain the sell-off or the earlier gaps.

2. Indicators selected and why¶

I used eight indicators, one or two per category and none redundant: - Trend: close_50_sma and supertrend (weekly, monthly and daily tiers). - Momentum: macd and rsi. - Volatility: atr. - Volume: obv. - Exhaustion: td_9. - Stretch: z_score.

The snapshot also gave the 10 EMA, 200 SMA, Bollinger bands and MACD signal/histogram, which I use as extra context.

3. Big picture: a large uptrend, now in a sharp pullback¶

From the CSV: - AMD closed at 190.95 on 2026-03-03, the low close of the 12-month window. It reached a high close of 649.42 on 2026-10-06 (intraday high 658.52). That is roughly a 240% advance. - The advance came in steps. Gaps up came on 04-24 (close 347.81 on 81.6M shares) and 05-06 (close 421.39 on 87.7M). A consolidation ran from about mid-May to early September, with a sharp drop to a 429.56 close on 07-29. A breakout followed on 09-21 (close 615.52 on 44.5M, the heaviest volume in the recent window). - Since the 10-06 peak, the stock fell to a 528.75 close on 10-16, about -18.6% from the peak close. It now sits at 555.60, about -14.4% below the peak. - The 10-15 session opened at 540.00 against a prior close of 570.82, a gap down. Its volume of 28.8M was the highest since 09-21. The tools don't say why. - Today's bar (O 545.69, H 556.48, L 537.30, C 555.60) is a +3.9% bounce from yesterday's 534.74. It closed near the day's high, but on light volume.

4. Trend analysis¶

Moving averages (verified) - 10 EMA: 566.57. Price is about 1.9% below it, so short-term trend is still pointing down. - 50 SMA: 541.16, rising every day in the 30-day window (from 497.00 on 09-21). Price is about 2.7% above it. - Closes on 10-16 (528.75), 10-19 (538.05) and 10-20 (534.74) were below the 50 SMA. Today's close reclaims it. - Medium-term trend is intact, but the line was only just reclaimed. - 200 SMA: 399.10. Price is about 39% above it, which is a very large gap and a reminder of how far the stock has run.

SuperTrend (14, 3x ATR). Weekly outranks monthly, which outranks daily. - Weekly (primary): UP. Trailing stop 453.83, with the close 22.36% above it. - Monthly (regime): UP. Stop 336.14, with the close 65.20% above it. - Daily (timing): DOWN. Stop 617.04, with the close 10.00% below it. - The higher tiers are bullish and the daily tier is bearish, so this reads as a pullback inside an uptrend rather than a confirmed trend reversal. - A daily flip back to up would need a close above 617.04, about 11% higher. That is close to the 10-01 close of 615.73 and well below the 10-06 high.

5. Momentum¶

MACD: -1.94, signal 11.16, histogram -13.11. All are bearish readings.
The MACD line fell from 36.55 on 10-07 to -1.94, and crossed below zero between 10-19 (+3.28) and 10-20 (-0.59).
The pace of decline is slowing: the one-day changes were -4.55 (10-16 to 10-19), -3.87 (10-19 to 10-20) and -1.36 (10-20 to 10-21). That is an early sign of stabilization, not a buy signal. MACD is a lagging measure and was still falling on today's bounce.
RSI: 45.83, up from 39.81 yesterday.
The window low was 37.77 on 10-16, and RSI never reached the 30 oversold zone.
It had been in the 65–74 range from 09-21 to 10-07, peaking at 73.88 on 09-22. It has dropped about 26 points from the 72.16 reading on 10-06.
The pullback has erased the overbought condition without producing an oversold one.

6. Volatility and risk sizing¶

ATR: 26.53, about 4.8% of price. It rose from 23.4 (10-07) to 27.27 (10-16) and has eased since. A typical daily range is about ±$26, which is wide.
Bollinger bands: lower 523.78, middle 599.13, upper 674.47. Price is about 21% of the way from the lower band to the upper (my calculation), roughly 7% below the middle band. It sits in the lower half but is not at the lower band.
Stop distances in ATR terms:
The daily supertrend stop (617.04) is about 2.3 ATR above price.
The weekly stop (453.83) is about 3.8 ATR below price, a wide risk distance of about $102.
2 ATR is about $53.
The 10-16 intraday low of 526.83 is about $29 (about 1.1 ATR) below the close. That is a tight, observable level but small relative to normal daily noise.

7. Volume (OBV)¶

OBV peaked at 2.272B on 10-06, the same day as the price peak, and is now 2.161B. That is a decline of about 112M.
Its low in the window was 2.145B on 10-16. It has risen slightly since, consistent with the bounce.
OBV is also below its 09-21 level of 2.192B, when price was 615.52.
OBV and price are moving together, so there is no divergence in either direction. The decline has had moderate distribution, not a collapse. Today's bounce is on light volume, which gives it weak confirmation so far.

8. Exhaustion and stretch¶

TD-9:
Daily: +9, buy setup complete. This is a reversal watch for a bounce or stabilization.
Weekly: +2 of 9.
Monthly: +1 of 9.
The daily count is exhausted, but the weekly and monthly counts are early and show no higher-timeframe exhaustion. Weight weekly over daily, so this is a short-term bounce signal, not a major bottom.
Count check from the CSV: by my read the 9^th qualifying close was 10-20. The 10-21 close (555.60) is still below the close four bars earlier (562.23). The tool shows +9, so the count may be capped at 9, or the unfinished bar may be treated differently. This is my inference and not tool-verified.
Z-score (20-period):
Weekly: +0.52 (near the mean).
Monthly: +1.50 (above the mean).
Daily: -1.16 (below the mean).
Nothing reaches |2|, so none of the three timeframes is statistically stretched.
The daily reading is mildly negative and would support a modest mean-reversion bounce. The weekly reading is near fair value, so the pullback has mostly normalized the earlier stretch.

9. Synthesis¶

Bullish evidence - Weekly and monthly supertrends are up, with stops far below price. - Price is above the rising 50 SMA and the 200 SMA. - Daily TD-9 buy setup is complete. - RSI is turning up from 37.77. - The MACD decline is slowing. - Today's close is near the day's high. - OBV shows no divergence.

Bearish evidence - The daily supertrend is down. - Price is below the 10 EMA. - MACD is below both zero and its signal line, with a histogram of -13.11. - The drop from 10-06 included a high-volume gap down on 10-15. - OBV is below its early-October level. - Today's bounce comes on the lightest volume of the window (possibly a partial bar). - Price is about 39% above the 200 SMA, so there is a long way to fall if the trend breaks.

Net read: The medium and long-term trend is still up, and the short-term trend is down and in repair. The most likely near-term picture is a consolidation or relief bounce, but the tools show no confirmed reversal. This is a moderate-confidence reading. It rests on mixed signals and a possibly incomplete final bar.

10. Levels to watch (from tool outputs only)¶

Role	Level	Source
Resistance	566.57	10 EMA
Resistance	599.13	Bollinger middle
Resistance	617.04	Daily supertrend stop (flip level)
Resistance	649.42 / 658.52	10-06 high close / intraday high
Support	541.16	50 SMA (just reclaimed)
Support	523.78 / 526.83	Lower Bollinger band / 10-16 intraday low
Support	453.83	Weekly supertrend stop
Support	399.10	200 SMA

I have not verified these as past support or resistance with tested bounces. They are reference levels, not validated zones.

Conditional scenarios - Constructive: A hold above the 50 SMA (541.16) on rising volume, a daily close above the 10 EMA, and the MACD histogram narrowing would support a recovery toward 599 and the 617 flip level. - Cautionary: A close back below the 50 SMA and the 10-16 low (526.83), especially with the MACD histogram widening and OBV making new lows, would put the lower Bollinger band (523.78) in play. After that the next reference is the weekly supertrend at 453.83.

11. Summary table¶

Item	Reading	Interpretation
Current price (2026-10-21)	555.60	Snapshot close. The CSV shows 555.53, a discrepancy to flag
Day's range	537.30–556.48	+3.9% bounce, closing near the high
Volume	~10.1M	Lowest in the 30-session window, possibly a partial bar
10 EMA	566.57	Price below it: short-term bearish
50 SMA	541.16	Price above it and rising: medium-term support
200 SMA	399.10	Price about 39% above it: long-term uptrend, large cushion
Supertrend weekly	UP, stop 453.83	Primary trend bullish
Supertrend monthly	UP, stop 336.14	Regime bullish
Supertrend daily	DOWN, stop 617.04	Short-term bearish, flip needs a close above 617.04
MACD / signal / hist	-1.94 / 11.16 / -13.11	Bearish, but the decline is slowing
RSI	45.83	Neutral. 10-16 low was 37.77, peak 72.16 on 10-06
Bollinger (L / M / U)	523.78 / 599.13 / 674.47	Price in the lower half of the band range
ATR	26.53 (~4.8% of price)	High volatility, size positions accordingly
OBV	2.161B (peak 2.272B on 10-06)	Moves with price, no divergence
TD-9 (D / W / M)	+9 / +2 / +1	Daily exhaustion, higher timeframes early
Z-score (D / W / M)	-1.16 / +0.52 / +1.50	No stretch above
Net read	Uptrend pullback with a tentative bounce	Reversal unconfirmed, watch 541 and 527 on the downside and 567, 599 and 617 on the upside

Sentiment Analyst¶

Overall Sentiment: Mildly Bullish (Score: 6.3/10) Confidence: Medium

SOURCE-BY-SOURCE BREAKDOWN (AMD, 2026-10-14 to 2026-10-21). Data limits first: Reddit was disabled by config, so there is no Reddit read. News is 20 headlines with titles only and no article bodies. StockTwits is a 30-message sample. Confidence is therefore medium, not high.

News (Yahoo Finance feed, 20 headlines): about 9 are AMD-focused. The rest are peripheral (NVDA, INTC, AVGO, SpaceX/Tesla, ETF flow notes) or irrelevant (a MindWalk biotech press release). Constructive items: (a) Susquehanna raised its AMD price target to $620 from $550, and a Stocktwits item says it sees 'modest upside' for chip earnings, with AMD and QCOM raised and INTC and COHR cut; (b) iShares SOXX added AMD on Oct. 19, and a separate headline cites $332.1M of net ETF buying in AMD that day, which is a tangible passive and flow tailwind; © an Insider Monkey piece says the data center business is 'finally delivering profits'; (d) a 24/7 Wall St. piece says Anthropic has locked in over $600B of compute and still wants more, framed as 'big news for AMD'; (e) Supermicro lifted its revenue forecast, a read-through for AI server demand; (f) tech led the market higher. Cautionary items: (a) a Motley Fool headline says AMD 'plummeted 13% last week' and asks whether to buy the dip before Nov. 3, which points to the next earnings date and recent volatility; (b) a 24/7 Wall St. headline asks whether AMD is 'priced for perfection before earnings' after a 137% run; © several headlines favor NVDA ('cheapest NVIDIA has looked all month', 'road to $1 trillion in annual revenue', 'AMD vs. Nvidia, 1 metric proves which is best'). I only have the titles, so I cannot say what the AMD vs. NVDA piece concludes. Net news tone: moderately positive on fundamentals, analyst support and flows, tempered by valuation and pre-earnings risk.

StockTwits (30 most recent messages, timestamped Oct. 21 15:59-17:18Z): 14 Bullish (47%), 0 Bearish (0%), 16 unlabeled. Among labeled posts the split is 100% bullish, but n=14 is small and quality is uneven. Roughly half of the bullish tags sit on posts where AMD is incidental: cashtag lists, $BB/$ONDS/$PLTR promotion, and link-only posts. Only about 5-8 bullish posts carry an AMD-specific view: 'back to $1T and beyond' (@BlueTegu), 'Number 1 on my scoring model, will be adding more' (@shoc24), 'short squeeze after hours when meta reports. We opening 575+' (@shortsqueezing18), 'AMD soon' (@iBlezt), and two @Maribor posts arguing that fund managers calling AI a 'bubble' are just trying to cool sentiment. One user cites AMD at about 555.55 ('sakes alive!'), and another says 'back to $600'. I cannot verify these prices with the evidence provided. Unlabeled posts add context: @MaintenanceRequirement bought Dec '28 $870 calls (long-dated, very bullish positioning), several posts discuss 550 calls possibly 'printing', and @TopOptionAlert (a commentary-style account) discusses a Friday 525 put hedge and warns that Meta spending concentration 'cuts both ways'. One unlabeled post claims AMD was '40% off the highs' with a $5 one-candle move, which I cannot reconcile with the 137% run headline and cannot verify. Another poster says institutions are still rotating out of other sectors and trying to 'hold this dip'. Zero bearish labels is unusual and may reflect a sample or labeling skew rather than an absence of bears, since the hedging chatter shows caution exists but is not tagged.

CROSS-SOURCE DIVERGENCES AND ALIGNMENTS. Aligned: both news and retail lean positive on AMD's AI and data center trajectory, and the Anthropic compute theme, Susquehanna's $620 target (roughly 12% above the $555.55 print, if accurate) and ETF inflows echo retail's $575-$600 aspirations. Divergent: retail is exuberant and short-squeeze oriented, with no bearish tags, while the news flow carries explicit valuation and pre-earnings caution and notes a 13% drawdown the prior week. Retail also shows tactical anxiety through put hedges and the Meta-concentration worry, which an all-bullish tag count hides. Retail is dismissive of Ken Griffin's reported 'AI bubble' comment, treating it as a fund manager talking down prices, while the news feed contains no direct counter-narrative. That dismissiveness is a mild contrarian caution flag. A 0% bearish read with a small, noisy sample carries over-extension risk.

DOMINANT NARRATIVE THEMES. (1) Rebound after a sharp pullback: dip-buying into a recovery toward $550-$600. (2) AI infrastructure demand: Anthropic's $600B+ compute commitments, Supermicro's raised forecast, Meta spending. (3) Pre-earnings positioning ahead of the Nov. 3 report (per the Motley Fool headline), including options activity around the 525/550 strikes. (4) Relative competition: AMD vs. NVDA framing, with NVDA favored in several headlines, and Intel foundry chatter (Musk/Terafab, 18A customer gaps). (5) AI-bubble debate, with retail rejecting the bubble thesis. (6) Passive flows: SOXX inclusion and ETF buying.

CATALYSTS AND RISKS. Catalysts: Meta's earnings (a retail-cited after-hours catalyst; timing and content not verified here), Tesla earnings and SpaceX/Tesla merger headlines affecting risk appetite, AMD's Nov. 3 earnings, further AI customer commitments (Anthropic and others), and continued ETF and index inflows. Risks: valuation after a 137% run ('priced for perfection'), a 13% one-week drop showing high volatility, customer concentration in hyperscaler capex (a Meta pause could hit AMD first, per commentary), AI-bubble rhetoric from prominent investors, NVDA's relative dominance, possible sector rotation or liquidity strain, and crowded bullish retail positioning including call-heavy and squeeze-hopeful flow. Not available: Reddit data, article bodies, actual price and volume series, and any confirmation of earnings expectations. This is sentiment signal only, not a price call.

SUMMARY TABLE | Signal | Direction | Source | Evidence | |---|---|---|---| | Analyst target raise | Bullish | News (MT Newswires/Stocktwits) | Susquehanna PT to $620 from $550; sees modest upside for chip earnings | | ETF/index inflows | Bullish | News (GuruFocus) | SOXX adds AMD Oct. 19; $332.1M net ETF buying | | AI demand read-through | Bullish | News (24/7 Wall St., Yahoo) | Anthropic $600B+ compute commitments; Supermicro lifts revenue forecast | | Data center profitability | Bullish | News (Insider Monkey) | Data center 'finally delivering profits' | | Valuation / pre-earnings risk | Bearish/Cautious | News (24/7 Wall St.) | 137% run, 'priced for perfection'; earnings Nov. 3 per Motley Fool | | Recent drawdown | Bearish/Cautious | News (Motley Fool) | Stock 'plummeted 13% last week'; buy-the-dip framing | | Relative strength vs NVDA | Mildly Bearish | News (multiple) | Several NVDA-favoring headlines; AMD vs. NVDA comparison (conclusion unknown) | | Retail tag ratio | Bullish | StockTwits | 14 Bullish / 0 Bearish / 16 unlabeled (n=30); many tags on multi-ticker or off-topic posts | | Options positioning | Bullish with hedging | StockTwits | Dec '28 $870 calls; 550 calls 'printing'; Friday 525 put hedge discussion | | Squeeze/price-target chatter | Bullish (speculative) | StockTwits | '575+ open', 'back to $600', 'back to $1T' (prices unverified) | | Meta capex concentration | Cautious | StockTwits (@TopOptionAlert) | One Meta headline can lift or hit the name | | AI-bubble dismissal | Contrarian caution | StockTwits | Multiple posts dismiss Ken Griffin's bubble comment | | Reddit | Unavailable | Reddit | Disabled by config; no read |

News Analyst¶

AMD (Advanced Micro Devices, Inc.) News and Macro Report, week ending 2026-10-21¶

0. Data coverage and limits¶

FRED macro data was unavailable. Every get_macro_indicators call failed: fed funds, 10Y Treasury, CPI, unemployment, yield curve and VIX, plus the raw series IDs DFF and DGS10. The first error was a parse failure and the retries returned HTTP 403 from the FRED API. I have no verified readings for rates, inflation, labor, curve shape or volatility, and I have not estimated any.
The news tools returned headlines and links only, with no article bodies or prices. I have no AMD price level, so I can't say how far the 13% drop or the 137% run has taken the stock. Claims below come from headlines, so treat them as leads to verify.
Prediction-market liquidity varies a lot. Fed markets trade around $1M or more, but the AMD-specific markets trade in the hundreds of dollars. I flag this where it matters.

1. AMD company developments¶

The setup is a big run, then a sharp pullback, with earnings close. - Motley Fool: "AMD Stock Plummeted 13% Last Week. Should You Buy the Dip Before Nov. 3?" This points to Q3 earnings around Nov 3, about 13 days away. Confirm the date against AMD's IR calendar. - 24/7 Wall St.: "After a 137% Run, Is AMD Priced for Perfection Before Earnings?" The stock has had a very large run, so expectations are high.

Sell-side and flows are supportive. - Susquehanna raised its AMD price target to $620 from $550 (MT Newswires). It also raised QCOM and cut INTC and COHR, and it sees "modest upside" for chip earnings overall (Stocktwits). - GuruFocus reported $332.1M of net ETF buying in AMD on Oct 19. It also reported that SOXX added AMD and INTC on Oct 19. This reads as ETF-driven demand during the dip, but the headline gives no detail on what the "add" means.

AI demand headlines are a tailwind. - 24/7 Wall St. says Anthropic has locked in over $600B of compute and wants more, and calls that "big news for AMD." Which vendors that spending goes to is a headline interpretation, not confirmed AMD contract wins. - Supermicro raised its revenue forecast and the stock jumped (Yahoo Finance). Supermicro is a server OEM that sells AMD-based systems, so this is a read-through for AI server demand. - Insider Monkey says AMD's data center business is "finally delivering profits." That is a positive tone on the segment's margins heading into earnings.

Competitive and sector context. - Intel rose after Musk signaled a "Terafab" breakthrough (Barron's). This is a possible foundry or competitive narrative shift, and it is unquantified. - Nvidia coverage is heavy: "cheapest it's looked all month," a "$1 trillion annual revenue" prediction, and Nvidia versus AMD comparisons. Relative-value debate between the two names is active. - Tech led the market higher during the week, with SpaceX shares surging (Yahoo Finance). Meta and Tesla earnings were due, and IBM and Palantir were in the earnings conversation.

2. Prediction-market signals¶

Fed policy (the markets are deep, roughly $1M to $9M in volume, so this is the most reliable data I have): - "No Fed cuts in 2026": 97%. Every market for 6 or more cuts is at 0%. - December 2026 FOMC (resolves Dec 9): - Hike 25 bps: 78% - Hold: 18% - Cut 25 bps: 1% - Hike 50+ bps: 2% - Markets are pricing a hawkish bias, with a hike more likely than a cut or a hold. I couldn't confirm this against FRED or the current fed funds level. If it's accurate, it matters a lot for long-duration, high-multiple growth stocks like AMD. Check it against fed funds futures or CME FedWatch before relying on it.

Recession: - US recession by end of 2026: 8%, which is low. UK is 6% and Japan 4%, on very thin volume. The market is not pricing a growth shock, so the macro risk is rates rather than demand.

AMD Q3 data center revenue (Polymarket, resolves around Nov 3):

Threshold	Implied probability
> $7.6B	78%
> $8.0B	62%
> $8.4B	28%
> $8.8B	16%
> $9.2B	10%

Volume is only about $200 to $500 per market, so this is not reliable. It does imply a crowd median near $8.1B to $8.3B. Compare that with the sell-side consensus figure, which I don't have.

Policy and regulation (thin volume): - "AI-chip export licensing becomes law this year": 14%, down 13 points in a week. - "Export-control chip security becomes law this year": 35%, up 3.5 points. - Low odds for new China-related chip legislation is mildly positive for AMD. The volume is only a few thousand dollars, so treat it as weak evidence.

3. Macro and global backdrop¶

The global news feed was mostly sector and corporate items: AI software partnerships, quantum computing stocks (IonQ and Rigetti down sharply), IBM earnings previews, and a silver note saying silver is "more resilient than gold." It had no major macro, geopolitical or central-bank headlines for this window, so I can't characterize CPI, jobs, tariffs or geopolitics from these tools.
Equity sentiment looks risk-on for tech. Stocks were rising pre-bell ahead of Meta and Tesla earnings, and tech led the market.
Quantum and speculative-tech names are weak, which may suggest selectivity within speculative technology.

4. Trading implications¶

The Nov 3 earnings event dominates AMD's near-term risk. After a 137% run, a 13% drop and a PT raise to $620, the stock is volatile and expectations are high. Consider the option-implied move and position sizing around the print. I don't have the implied move.
The key numbers to watch are data center revenue, the MI-series GPU ramp and Q4 guidance. A beat on data center revenue, rather than the headline number, is the likely driver.
The Fed is the main macro risk. A priced-in December hike (78%) is a headwind for high-multiple semiconductors. A hot data print or hawkish Fed commentary before earnings could extend the pullback.
AI capex news is the main tailwind. The Anthropic compute commitments, Supermicro's raised guide and Meta's earnings (capex commentary) are positive read-throughs.
Intel/Terafab and Nvidia relative valuation are narrative cross-currents. Watch for rotation between AMD, Nvidia and Intel.
Dip-buyers are visible (ETF inflows and the SOXX addition), but I can't tell whether that is durable demand or mechanical flow.

Caveats: I couldn't verify rates, inflation or the VIX. The AMD price and technical levels are missing. Everything else rests on headline-level information.

Summary table¶

Category	Key point	Evidence / source	Reliability	Implication for AMD
Earnings catalyst	Q3 report around Nov 3	Motley Fool headline	Medium (confirm date)	Binary event risk in about 13 days
Price action	Down 13% last week after a 137% run	Motley Fool, 24/7 Wall St.	Medium (no price data)	High expectations, volatile
Analyst view	Susquehanna PT $620 (from $550)	MT Newswires, Stocktwits	High	Supportive sentiment
Flows	$332.1M net ETF buying Oct 19; SOXX adds AMD	GuruFocus	Medium	Dip demand or mechanical flow
AI demand	Anthropic has $600B+ of compute commitments	24/7 Wall St.	Low to medium (vendor split unconfirmed)	Potential tailwind
Server read-through	Supermicro raised revenue forecast	Yahoo Finance	Medium	Positive for AI server demand
Data center	Segment turning profitable	Insider Monkey	Medium	Key earnings driver
Competition	Intel/Musk Terafab; Nvidia valuation debate	Barron's, 24/7 Wall St., Motley Fool	Low	Narrative rotation risk
Fed (prediction mkt)	No 2026 cuts 97%; Dec hike 78%, hold 18%, cut 1%	Polymarket (deep volume)	Medium to high, unverified vs FRED	Headwind for high-multiple tech
Recession odds	US recession by end 2026: 8%	Polymarket	Medium	Growth shock not priced
AMD Q3 DC revenue	Crowd median about $8.1B to $8.3B	Polymarket (volume under $1k)	Low	Compare with consensus
Export controls	AI-chip licensing law 14%; chip security law 35%	Polymarket (thin)	Low	Limited near-term regulatory shock
Macro data	Rates, CPI, jobs, curve, VIX	FRED, 403/parse errors	Unavailable	Not assessed

Fundamentals Analyst¶

AMD (Advanced Micro Devices, Inc.) — Fundamental Report as of 2026-10-21¶

0. Scope and data freshness¶

Nothing new has been filed in the past week. The newest financial statements cover the fiscal quarter ended 2026-06-27 (Q2 FY26). The Q3 FY26 quarter (ended about late September) is not in any tool output. All financial analysis below is roughly 3.5 months old.
One insider transaction falls inside the past week: Forrest Norrod's sale on 2026-10-15 (Section 6).
The fundamentals snapshot is current. The implied share price is about $555.7 (P/B × book value and P/E × EPS both give this).
Figures marked (derived) are my own arithmetic on tool data. The tools don't report them directly. Examples are Q4 figures and TTM totals built from SEC data.

1. Company snapshot and valuation¶

Metric	Value	Comment
Market cap	$907.2B	Implied price about $555.7
P/E (TTM, GAAP)	141.8x	EPS TTM $3.92
Forward P/E	35.3x	Forward EPS $15.76 (basis unspecified)
PEG	0.53	Implies about 67% growth if based on forward P/E (derived)
Price/Book	13.5x	Book value/share $41.19
Price/Sales	about 22x (derived)	Revenue TTM $41.3B
EV/EBITDA	about 95x (derived)	EBITDA $9.56B, EV about $906B
Price/FCF	about 103x on provider FCF ($8.84B); about 108x on SEC-derived FCF ($8.40B)	FCF yield about 0.9–1.0%
Dividend yield	0%	No dividend
Beta	2.45	Very high volatility
52-week range	$188.22 – $658.52	Price is 15.6% below the high and 195% above the low
50-day / 200-day average	$539.54 / $397.42	Price is about 3% above the 50-day and about 40% above the 200-day
ROE / ROA	10.2% / 5.1%	Equity base is inflated by the 2022 Xilinx deal (Section 3)

Valuation read: - The forward EPS of $15.76 is about 4.0x TTM GAAP EPS. At about 1.66B diluted shares, that means roughly $26B of net income, against $6.4B over the last twelve months (derived). - It is also about 2.9x the Q2 FY26 GAAP run-rate (4 × $1.38 = $5.52). - Forward EPS is probably a non-GAAP consensus figure, so the comparison with GAAP trailing EPS is not like-for-like. - Either way, the price embeds a very large earnings step-up that has not yet shown up in reported results.

2. Income statement¶

Quarterly trend (USD millions)¶

Quarter	Revenue	YoY	Gross profit	GM	Op. income	Op. margin	Net income	Dil. EPS
Q1 FY25 (Mar-25)	7,438	+35.9%	3,736	50.2%	806	10.8%	709	0.44
Q2 FY25 (Jun-25)	7,685	+31.7%	3,059	39.8%	(134)	-1.7%	872	0.54
Q3 FY25 (Sep-25)	9,246	+35.6%	4,780	51.7%	1,270	13.7%	1,243	0.75
Q4 FY25 (derived)	10,270	+34.1%	5,577	54.3%	1,752	17.1%	1,511	0.92
Q1 FY26 (Mar-26)	10,253	+37.8%	5,416	52.8%	1,476	14.4%	1,383	0.84
Q2 FY26 (Jun-26)	11,536	+50.1%	6,203	53.8%	1,990	17.25%	2,297	1.38

Q4 FY25 = FY25 annual minus the first three quarters. It is not reported as a standalone quarter. The derived Q4 revenue reconciles exactly to the provider's $41,305M TTM revenue.

Annual trend (USD millions)¶

FY	Revenue	Growth	Gross margin	Op. income	Op. margin	Net income	EPS
2020	9,763	–	44.5%	1,369	14.0%	2,490	2.06
2021	16,434	+68%	48.2%	3,648	22.2%	3,162	2.57
2022	23,601	+44%	44.9%	1,264	5.4%	1,320	0.84
2023	22,680	-4%	46.1%	401	1.8%	854	0.53
2024	25,785	+14%	49.4%	1,900	7.4%	1,641	1.00
2025	34,639	+34%	49.5%	3,694	10.7%	4,335	2.65

Key observations¶

Growth is accelerating. YoY revenue growth went from about 32–36% through FY25 to +50.1% in Q2 FY26. Sequential growth was +12.5% in Q2 FY26 after a flat Q1 (-0.2%). H1 FY26 revenue of $21.8B is up 44% on H1 FY25. The FY2020–FY2025 revenue CAGR is about 28.8% (derived).
Margins are expanding.
Gross margin rose from about 50% to 53.8%.
Operating expenses (gross profit minus operating income) were about 36.5% of revenue in Q2 FY26. That compares with 41.5% in Q2 FY25 and 38.4% in Q1 FY26.
Opex grew 32% YoY against revenue growth of 50%.
Incremental operating margin in Q2 FY26 was about 40% (+$514M operating income on +$1,283M revenue) (derived).
Earnings quality is a flag. Net income exceeds operating income in several periods:

Period	Operating income	Net income
Q2 FY25	-$134M	+$872M
FY25	$3,694M	$4,335M
Q2 FY26	$1,990M	$2,297M

In the other recent quarters, net income ran at about 86–98% of operating income. If Q2 FY26 had followed the roughly 93% average of Q3 FY25 to Q1 FY26, net income would be about $1.84B and EPS about $1.11, not $1.38. That leaves about $0.27/share unexplained (derived estimate).
These tools don't break out tax, interest or investment items. Check the 10-Q for a tax benefit or investment gains before treating $1.38 as run-rate EPS.
Q2 FY25 was a one-off trough. Gross margin fell to 39.8% and operating income turned negative. The tools don't label the cause. Year-over-year comparisons against that quarter are flattered, so use Q3 FY25 as the cleaner base.
Diluted share count is roughly flat at about 1.64–1.66B (derived from net income ÷ EPS). There is no heavy buyback shrinkage. Financing cash flow is only slightly negative.

3. Balance sheet (USD millions)¶

Date	Total assets	Current assets	Cash	Current liabilities	Equity
FY24 (Dec-24)	69,226	19,049	3,787	7,281	57,568
FY25 (Dec-25)	76,926	26,947	5,539	9,455	62,999
Mar-26	79,642	28,628	5,585	10,506	64,462
Jun-26	84,464	31,522	5,086	12,081	67,224

Liquidity is strong. The current ratio is 2.61 (31,522 ÷ 12,081, which matches the provider).
Leverage is minimal. Debt/equity is 6.4%, implying about $4.3B of debt (derived). Cash of $5.1B is roughly equal to that debt. Short-term investments are not in the tool output, so true net cash is probably higher.
Total liabilities are not tagged in the data. Total assets minus equity gives about $17.2B, up from $15.2B in Q1 (derived).
Cash fell and working capital grew.
In Q2, cash fell $0.5B and total assets rose $4.8B.
Current assets rose $2.9B, so non-cash current assets (receivables, inventory and similar) rose about $3.4B. Current liabilities rose $1.6B.
The tools don't split receivables from inventory. Check the 10-Q: a receivables-driven build is acceptable, but an inventory-driven build ahead of a new product ramp is a risk.
The equity base is acquisition-driven. Assets jumped from $12.4B to $66.9B in Q1 FY22 (the Xilinx deal). That inflates book value, depresses ROE (10.2%) and makes P/B (13.5x) look less meaningful. Tangible book is far lower.

4. Cash flow (USD millions)¶

Quarterly cash-flow figures are cumulative year-to-date, so I differenced them (derived).

Period	Operating CF	Capex	Free cash flow	FCF margin
FY23	1,667	546	1,121	4.9%
FY24	3,041	636	2,405	9.3%
FY25	7,709	974	6,735	19.4%
Q3 FY25	2,159	258	1,901	20.6%
Q4 FY25	2,600	222	2,378	23.2%
Q1 FY26	2,955	389	2,566	25.0%
Q2 FY26	2,366	808	1,558	13.5%
H1 FY26	5,321	1,197	4,124	18.9%
TTM	10,080	1,677	8,403	20.3%

Observations: - Cash generation has inflected. FY25 FCF was 2.8x FY24. H1 FY26 FCF is up 68% on H1 FY25 ($4.1B vs $2.5B), and H1 operating cash flow is 1.45x net income. - Q2 FY26 weakened. - FCF fell 39% sequentially. - Operating cash flow slipped even as revenue grew 12.5%, consistent with the working-capital build. - Capex doubled sequentially to $808M, 7.0% of revenue versus about 2–4% in prior quarters. It is up about 2.9x YoY. - Investing outflows are large. H1 FY26 investing outflow was $5.4B, of which only $1.2B was capex. About $4.2B went to other uses (securities purchases, acquisitions or strategic investments). FY25 showed a similar pattern ($5.5B investing vs $1.0B capex). The tools don't itemize this. - Shareholder returns are small. Net financing was only -$431M in FY25 and -$365M in H1 FY26.

5. Data-consistency flags¶

TTM gross profit. The provider says $23.0B (55.7% margin). SEC-derived TTM is $21.98B (53.2%). The provider's figure would require about a 64% gross margin in Q4 FY25, against a derived 54.3%. I'd trust the SEC-derived figure.
TTM FCF. The provider says $8.84B. SEC-derived is $8.40B, a difference of about 5%.
Operating margin of 17.25% from the provider is the latest quarter, not TTM. TTM operating margin is about 15.7% (derived).
Insider data quirks.
Some Lisa Su rows from 2024-12-04 appear duplicated.
Mark Papermaster's 2026-08-14 exercise of 129,345 options shows a value of $625K, which doesn't reconcile with the $84.85 strike price. The value field is unreliable on that row.
The "Transaction" column is blank throughout.
There is no 10b5-1 flag.

6. Insider activity¶

Past week¶

2026-10-15, Forrest Norrod (Officer):
Exercised 7,262 options at $84.85.
Sold 17,262 shares at $535.15–$543.03, about $9.33M.
This is his third consecutive monthly sale of 17,261–17,262 shares: 8/24, 9/15 and 10/15.
No other insider transactions appear in the past seven days.

2026 year-to-date open-market sales (derived from tool values)¶

Insider	Role	Shares sold	Proceeds	Notes
Lisa Su	CEO	570,000	$212.3M	6 sales between Feb and Sep. Gifted 65,000 shares in 2026.
Mark Papermaster	CTO	118,936	$45.2M	Monthly option exercises at $84.85, plus larger April sales. Gifted 440,614 shares (including 413,212 on 2/24).
Forrest Norrod	Officer	90,721	$38.6M	Monthly cadence since August
Paul Grasby	Officer	46,876	$20.1M
Jean Hu	CFO	34,956	$11.1M
Nora Denzel	Director	10,447	$5.5M
Ava Hahn	General Counsel	5,864	$2.2M
Total			about $335M

Trailing about 90 days (since 7/23): about $115M sold across six insiders.
Trailing 30 days: only Norrod's $9.3M.
Open-market purchases: only Philip Guido bought, 4,645 shares at $107.56 on 2025-02-07 and 8,800 shares at $113.56 on 2025-05-20, about $1.5M in total. There have been no insider buys in the last 17 months.
CEO selling has slowed in share terms. Su sold 250,000 shares in May–June but only 110,000 in August–September.
Unpriced entries on 2/13, 2/27, 8/7 and 8/14 across many officers have no price or text. They look like equity-award vesting or tax-withholding events rather than discretionary trades. That is my inference, since the tool doesn't label them.

Interpretation: - The selling is broad-based and regular: fixed monthly share counts, recurring option exercises at the same strike, and sales around the 10^th–25^th of each month. The cadence is consistent with pre-arranged plans, though the tool doesn't confirm that. - The absence of any buying, at a price near 3x the early-2026 level, offers no insider-conviction support at current valuation. It isn't an outright negative signal on its own.

7. Takeaways for traders¶

Supportive evidence - Revenue growth is accelerating (+50% YoY), with gross and operating margins both expanding. - FCF is $8.4B TTM (20% margin), and FCF conversion is above 100% of net income. - The balance sheet is very strong: current ratio 2.6, D/E 0.06, and cash about equal to debt. - The forward P/E of 35x and PEG of 0.53 look reasonable if the forward EPS of $15.76 is delivered.

Cautionary evidence - At about 22x sales, 142x GAAP earnings and about 100x+ FCF, the price needs a roughly 4x earnings step-up. There's little room for a revenue or margin disappointment. - Beta of 2.45 means large swings in both directions. The 52-week range is 3.5x. - Q2 FY26 EPS may include about $0.27 of non-operating or tax items. - FCF fell sequentially as capex doubled and working capital grew. - Insiders sold about $335M this year with no purchases. - The Q2 FY26 statements are about 3.5 months old. The Q3 FY26 report, which the tools don't date, is the next major information event.

Items to monitor and verify 1. Q3 FY26 results. Check for revenue of about $12.5B or more (continuing the +12% sequential pace), gross margin of at least 53–54%, and operating margin at or above 17%. 2. Earnings quality. Check the tax and non-operating items in the Q2 FY26 10-Q. 3. Working capital. Check inventory and receivables growth against revenue. 4. Capex. Check whether the $808M quarterly rate persists, and what the roughly $4B of non-capex investing outflows were. 5. Further insider sales. Watch Norrod's monthly pattern (the next one would be around mid-November) and whether Su's sales resume at the larger May–June size. 6. Price levels. Reference points are the 50-day average at about $539.5, recent insider execution at $535–543 and $498–516, the 200-day average at about $397, and the 52-week high at $658.52.

Not covered by these tools: segment mix, guidance, consensus revenue, customer concentration, export-control exposure and competitive dynamics. These are material to the thesis and need another source.

8. Summary table¶

Category	Key point	Data	Implication
Valuation	Priced for steep earnings growth	Market cap $907B; P/E 141.8x; forward P/E 35.3x; P/S about 22x; P/FCF about 103–108x	Little margin for error
Forward EPS	About 4x TTM GAAP EPS	$15.76 vs $3.92	Implies about $26B net income vs $6.4B TTM; basis likely non-GAAP
Growth	Accelerating	Q2 FY26 revenue $11.5B, +50.1% YoY, +12.5% QoQ	Strong momentum
Margins	Expanding	GM 53.8%, op. margin 17.25%, opex 36.5% of revenue	Operating leverage evident
Earnings quality	Net income above operating income	Q2 FY26: $2,297M vs $1,990M; about $0.27 EPS unexplained	Verify in the 10-Q
Cash flow	Strong but softened in Q2	TTM FCF about $8.4B; Q2 FCF $1.56B (-39% QoQ)	Capex doubling and working-capital build
Capex	Stepping up	Q2 FY26 $808M (7% of revenue), about 2.9x YoY	Watch the investment ramp
Balance sheet	Very strong	Current ratio 2.61; D/E 6.4%; equity $67.2B; cash $5.1B	Low financial risk
Book value	Acquisition-inflated	Assets jumped $12B to $67B in Q1 FY22	P/B and ROE of 10.2% are less meaningful
Volatility	Very high	Beta 2.45; 52-week range $188–$659	Size positions accordingly
Past-week insider	Norrod sale	17,262 shares at about $535–543, $9.3M (10/15)	Routine monthly cadence
2026 insider selling	Heavy and broad	About $335M across 7 insiders; Su $212M	Mostly formulaic; CEO share count tapering
Insider buying	None recently	Last buy May 2025 (about $1M)	No insider-conviction support
Data flags	Provider figures differ from SEC-derived	TTM gross profit $23.0B vs $22.0B; FCF $8.84B vs $8.40B	Prefer the SEC-derived numbers
Next catalyst	Q3 FY26 report	Date not in tool data	Verify the date; it is the key near-term event

II. Research Team Decision¶

Bull Researcher¶

Bull Analyst: # Bull Case for AMD at $555.60

I'll start with what the numbers say, then address the objections I expect from you.

1. Growth is accelerating, not just continuing¶

Revenue growth stepped up. YoY growth ran about 32-36% through FY25, then hit +50.1% in Q2 FY26 ($11.54B). That's a larger business growing faster, not a maturing one. H1 FY26 is up 44%, and the FY20-25 revenue CAGR is about 29%.
Operating leverage is visible. Gross margin is 53.8%, up from about 50%. Opex grew 32% against 50% revenue growth, so opex fell from 41.5% to 36.5% of revenue year over year. Incremental operating margin in Q2 was about 40%. Operating income was $1.99B, up about 35% sequentially.
Cash generation has inflected. TTM FCF is about $8.4B (20% margin). FY25 FCF was 2.8x FY24, and H1 FY26 FCF is up 68% YoY.
The balance sheet funds growth without leverage. The current ratio is 2.61 and debt/equity is 6.4%. Capex can double without straining anything.

2. Valuation: 142x trailing is a rear-view number¶

Quarterly GAAP EPS has gone $0.54 → $0.75 → ~$0.92 → $0.84 → $1.38. A trailing P/E blends in quarters that no longer describe the business. On a forward basis AMD trades at 35x with a PEG of 0.53.

The crux is that forward EPS of $15.76 is about 4x TTM, and the tool doesn't say whether it's GAAP or non-GAAP. The market is paying for a trajectory, and Nov 3 is the next test of it. But a stock growing revenue 50% with 40% incremental margins is the kind that can grow into its multiple.

3. Demand, flows and macro¶

AI demand: Anthropic reportedly has over $600B of compute locked in and wants more, Supermicro raised its revenue forecast, and the data center segment is described as "finally delivering profits." I'll flag that the Anthropic vendor split is unconfirmed, so I'm using it as a demand signal, not a contract win.
Sell-side: Susquehanna raised its target to $620 from $550, about 12% above today's price.
Flows: $332.1M of net ETF buying on Oct 19, with SOXX adding AMD. That's real dip demand, though I can't tell how much is mechanical.
Macro: US recession odds are only 8%. The demand backdrop isn't what's being questioned.

4. The pullback looks like a pullback¶

The weekly and monthly SuperTrends are both UP, with stops at 453.83 and 336.14. Per the framework, those outrank the daily DOWN signal.
The drop is -14.4% from the peak close after a ~240% run from the March low. Price is still 39% above the 200 SMA and reclaimed the rising 50 SMA (541.16) today.
RSI reset from 72 to a low of 37.77 without reaching oversold, so the froth is gone without a breakdown.
OBV shows no divergence, so there's no sign of aggressive distribution.
The daily TD-9 buy setup is complete, MACD's rate of decline is slowing, and the daily z-score is -1.16 against +0.52 weekly.

Anticipating the bear¶

"22x sales, 100x+ FCF. It's priced for perfection." Fair, and the stock is volatile for that reason: beta is 2.45. But it has already lost about 14% going into the print, so some of that perfection has been repriced. Revenue growth is accelerating, not decelerating, which is the opposite of what a peak-multiple setup usually looks like.

"Q2 net income ($2.30B) exceeds operating income ($1.99B), so EPS quality is suspect." It's worth checking in the 10-Q. But even if you haircut EPS to ~$1.11, that's still above Q1's $0.84 and Q3 FY25's $0.75. The cleaner metric, operating income, is up about 35% QoQ and about 57% over Q3 FY25. The growth story doesn't depend on that line.

"FCF fell 39% sequentially." Capex doubled to $808M (7% of revenue) and working capital grew while revenue rose 12.5% sequentially. That's what building ahead of demand looks like. I'll concede the tools don't split receivables from inventory, so that's a 10-Q item. Even with the dip, FCF is 20% of TTM revenue.

"Insiders sold $335M and nobody is buying." The sales are regular and formulaic: Norrod sold 17,261-17,262 shares in each of the last three months, with recurring exercises at the same $84.85 strike. The CEO's share volume fell from 250k in May-June to 110k in August-September. Selling after a roughly 3x move isn't a shock, and none of it is a lump-sum exit. I agree there's no insider buying, but that's a lack of a positive signal, not a negative one.

"A December hike is 78% priced." This is the bear's best macro point. It's a thin-ish prediction market I couldn't verify against FRED, but I won't wave it away. My response is that a hike in an environment with 8% recession odds signals strong growth. The earnings are compounding faster than the discount rate is rising.

"The bounce is on the lowest volume in 30 sessions, and the daily trend is down." Agreed on both. The bar may be partial, and a daily flip needs a close above 617.04. I'm not calling a bottom. I'm saying the higher-timeframe trend is intact and the risk/reward from the 50 SMA is reasonable.

"Retail is 100% bullish." I'm not leaning on that. Zero bearish tags on 30 messages is noise and, if anything, a mild contrarian caution. My case rests on the financials.

What would change my mind¶

A close back below 526.83 / 523.78 with OBV making new lows and the MACD histogram widening.
A Q3 print with sequential growth well below the ~12% run rate (roughly $13B would match it) or gross margin sliding below ~53%.
Confirmation in the 10-Q that the Q2 EPS beat was mostly non-operating, combined with a build in inventory.

My questions for the bear¶

If revenue growth is accelerating at this scale, why is a 35x forward multiple "perfection" and not a reasonable price for compounding?
What specifically in the Q2 operating line, the part you can't blame on tax or investment items, is the red flag?
Weekly and monthly trends are up, the 50 SMA held, and a 13% drawdown has already happened. What is the bear case from here, and not from $649? Bull Analyst: # Bull Rebuttal: AMD at $555.60

You landed some real hits, so I'll start with what I'm giving up, then show where I think the bear case overreaches. (mine) marks my own arithmetic on the tool data.

What I concede¶

"Accelerating" was the wrong word. YoY growth is the product of the last four sequential moves (+20.3%, +11.1%, -0.2%, +12.5%). The honest claim is sustained growth: about 10.7% per quarter on average, but only about 6% per quarter over the last two (mine).
The 40% incremental margin was a Q1-to-Q2 number. Rebased, it is about 19% over two quarters (from Q4 FY25) and about 31% over three (from Q3 FY25) (mine). Half over half, operating margin is flat at 15.5% in H2 FY25 and 15.9% in H1 FY26 (mine). Leverage is unproven, not disproven.
My 53% gross margin tripwire was sloppy. Q1 was 52.8%. I've reset it below.
I'm retiring OBV and TD-9 as support. OBV tracking price is neutral, and the report calls the daily TD-9 a short-term signal only. I also can't dispute 140x operating income. It's arithmetic.

Where I'm holding¶

1. The "mechanical" decel assumes Q3 repeats Q2's +12.5%. History says Q3 has been the stronger sequential quarter. Backing prior-year quarters out of the YoY column (mine):

	Q1→Q2	Q2→Q3
FY24	+6.6%	+16.9%
FY25	+3.3%	+20.3%

Q1 FY26 was also better than last year's Q1 (-0.2% vs -2.9%). Two years isn't a law, and Q3 FY25 was partly a rebound from a weak Q2. Still, a +17-20% Q3 means $13.5-13.9B and a YoY of 46-50%. So your $13.9B "prove me wrong" bar is not an outlier. It's last year's seasonal pattern. A decel that anyone can compute from last year's published $9.25B is also not news. The print will turn on sequential growth and margin versus expectations, and neither of us has consensus.

2. Cash flow compares Q2 to the best of four quarters. - OCF/net income fell from 2.1x to 1.0x, but TTM OCF of $10.1B is still about 1.57x TTM net income of $6.4B (mine). - Q2 OCF of $2.37B is above Q3 FY25's $2.16B. - The $3.4B non-cash current asset build is a fair 10-Q question. But my third tripwire was conjunctive: non-operating EPS confirmed and an inventory build. Neither half is confirmed, so it hasn't tripped. - The balance-sheet table shows no short-term investments, so "cash fell to $5.1B" may understate liquidity. Securities purchases would still be AMD's money. Unitemized investing outflows are not the same as leakage.

3. Insiders, in dollars. - The trailing 90 days were $115M of the $335M year to date. That's about a third of the dollars in about a third of the year, so on pace and not accelerating. - The last 30 days were only Norrod's $9.3M, against roughly $34M a month on average (mine). - His Oct 15 sale came after quarter-end and before the print, a window when discretionary trades are typically restricted. That fits a pre-set plan. It's an inference, since there's no 10b5-1 flag. - I still agree there have been no buys in 17 months.

4. The evidence gap cuts both ways. You're right that the AI demand headlines are weak and the vendor split is unconfirmed. But the tools also show nothing on capex digestion. The hard evidence we have is AMD's own P&L, with revenue up 50% and gross margin at 53.8%. A capex pause would show up in both. Meta's capex commentary is reportedly due this week (timing unverified), which is a cheap early test before Nov 3.

5. Targets and flows. The Susquehanna raise ($550 to $620) was reported this week, with the stock about 14% off its peak. That is raising into weakness. Nothing in the data says when the $550 was set. I'm not leaning on it, and I'm not leaning on the ETF flows either, since I can't separate mechanical buying.

Risk/reward¶

You compared regime stops (-18%, -28%) with first resistances (+11%, +17%). Like for like, the tactical levels are 541.16 (-2.6%) and 526.83 (-5.2%) on the downside and 617.04 (+11%) and 649.42 (+17%) on the upside.

I accept your ATR point, though: 526.83 is only 1.1 ATR away, which is noise. A 2-ATR stop (about $503, -9.6%) gives roughly 1.2:1 to 1.8:1 (mine). That's decent, not compelling. The chart doesn't carry this trade. The print does.

Valuation: what I'll defend¶

Your three-year illustration is arithmetically right. Two clarifications: - Each 5 turns of exit multiple moves that $35B outcome by about $175B, roughly 19% of today's cap (mine). "Roughly nothing" is one draw from a wide distribution, and the exit multiple is the unknowable part. - I'm dropping PEG and the $15.76 forward EPS from my case, since neither of us knows the horizon or basis.

What remains is a bet that 40%+ growth persists while operating margin holds 17% or better. At 140x operating income, that is a high-beta compounding bet, not a cheap stock. Size it for beta 2.45.

Revised tripwires¶

Item	Thesis intact	Thesis broken
Q3 revenue	≥ $13.0B (+12.5% seq)	< $12.5B (+8% seq)
Gross margin	≥ 53.8%	< 52.8% (Q1 level)
Operating margin	≥ 17%	< 15.5% (back to the half-year average)
10-Q	Non-operating items explained, asset build mostly receivables	Non-operating EPS confirmed and inventory build
Price	Closes hold above 541.16	Two closes below 523.78 (closing basis, to avoid the noise you flagged)

Questions for the bear¶

What does waiting cost? Every confirmation on your list (a close above 617.04 on volume, a $13.9B print) is only visible at a higher price. At 617 you'd be paying about 11% more for the confirmation. What's your entry if the print is in line, and are you really buying above your own flip level?
What's an in-line print worth? If Q3 is $13.0B, gross margin 53.8% and operating margin 17%, is AMD a buy, a hold or a short at $555? Your bottom line sounds like a timing and sizing call, not a thesis call.

Bottom line¶

I agree the chart doesn't justify aggression and the valuation leaves no cushion. The bear case is strongest on the cash flow items, which the 10-Q can settle. It is weakest on the "mechanical decel," because Q3 has been the strongest sequential quarter in both of the last two years.

My stance is a partial position now, sized for a ±5% daily range, with the rest added only after the print confirms. That keeps exposure to a Q3 near the seasonal pattern while respecting that a 140x-operating-income stock can gap either way on Nov 3. Bull Analyst: # Bull Closing Rebuttal: AMD at $555.60

(mine) marks my arithmetic on the tool data.

We agree on more than the transcript suggests. The stock is fully valued, the chart doesn't carry the trade, and the print decides it. Neither of us is calling a collapse. What's left is narrow: is a small, sized position through a binary print defensible? I'll concede what you earned, then defend what's left.

What I concede¶

The tripwire was miscalibrated. If seasonality implies $13.5-13.9B, then $13.0B is a miss against the pattern, not "intact." I'm replacing it with tiers:

Q3 revenue	YoY (mine)	Read
≥ $13.5B	≥ 46%	Beat vs. seasonal pattern
$13.0-13.5B	41-46%	In line: hold, don't add
$12.5-13.0B	35-41%	Soft: cut
< $12.5B	< 35%	Thesis broken

Q2 cash conversion is a real yellow flag. At 20.5% OCF/revenue, Q2 is the weakest of the four quarters. Q4 FY25 grew a similar +11% sequentially and converted at 25.3%, so "growth consumes cash" doesn't fully explain it.
A stop is not a loss cap. The 10-15 open was -5.4% through the prior close, and the peak-to-trough drop was -18.6% in eight sessions. Size has to cap the loss, not the stop.
The 10-Q tripwire shouldn't be conjunctive. Either finding is a yellow flag, and both together is red.
Your valuation table ties. I re-ran every row. I'm also withdrawing "waiting costs 11%," since you showed it's symmetric.

Where I'm still holding¶

1. The $3.4B build is a gross number. The same balance sheet shows current liabilities up $1.6B, so net non-cash working capital rose about $1.8B (mine). Some current liabilities could be non-operating, so that's a ceiling on the netting, not proof. The two explanations also aren't exclusive. Receivables or inventory would hit OCF, and securities purchases would hit investing and explain the cash decline. Part of each fits the tables. The signal is real, but it's smaller than $3.4B suggests.

2. The questions you're deferring to a later 10-Q are answerable today. The latest statements (quarter ended Jun 27) are SEC-derived, so the Q2 10-Q is already filed. It holds the answers to three questions: - What explains the $0.27 EPS gap? - Is the asset build receivables, inventory or securities? - What were the roughly $4.2B of non-capex investing outflows?

The Q3 10-Q arrives after the print, so waiting for it doesn't resolve these. The tools are the limitation, not the world. Reading the Q2 10-Q is the highest-value step before Nov 3, and I'm making my starter conditional on it.

3. The valuation table omits two things. I accept the framework, but: - Retained earnings. Cumulative net income is about $44-53B over three years and $115-142B over five (mine). That adds roughly 1.5-2.5 points a year to your returns. - Opex scaling. Q2 opex grew 32% against 50% revenue growth, a ratio of 0.64. If opex grows at 0.64 × 40% ≈ 25.6% a year with gross margin held at 53.8%, year-3 operating margin is about 27% (mine). That is close to your 29% hurdle, not the 20-25% rows. But this is exactly the leverage I conceded is unproven. H2 FY25 and H1 FY26 operating margins are flat.

Together, these move the three-year base case at a 30x exit from a loss to about break-even (mine). This is not a cheap stock. At an in-line print it's about 103x annualized operating income, and at a strong print about 86x. Returns above break-even need growth above 40%, margins near 27% or higher, or a multiple above 30x. That is the ceiling of my valuation argument.

4. Rates, burden of proof and digestion. - Rates. The Dec 9 FOMC comes after the print, and the hike is already 78% priced. A hold (18%) is the upside surprise and a 50bp+ hike (2%) is the downside one. I'm not claiming earnings outrun the discount rate. Rate sensitivity is a sizing input, which is part of why beta is 2.45. - Burden of proof. My evidence is the P&L: +50% in Q2, +44% in H1, gross margin 53.8%. I'm retiring "AI share gain" as a claim, since no tool supports it. - Digestion shows late. Agreed. Meta's capex commentary is the earliest read. The news summary and a retail post both say it's imminent, but the date is unverified.

Position sizing¶

Replace the stop with a loss budget. Position size = loss budget ÷ assumed gap. Plan on a 20% adverse move, which is a bit worse than the realized -18.6%. A 0.5% portfolio loss budget then implies a position of about 2.5% (illustrative). Staging it means a starter either now or after Meta's commentary, and a second tranche only after the print. I don't have the options-implied move, so I can't price a collar or put spread.

Final stance¶

Read the Q2 10-Q first. If it shows non-operating gains and an inventory build, there's no starter. I'd agree with you then.
If it's clean, take a small starter sized by the loss budget. My edge is not informational. It's a modest event-risk premium on a stock with a primary uptrend and 44% H1 growth.
Act on the Q3 print by tier. Add only at ≥$13.5B with gross margin ≥54% and operating margin ≥19%. Hold at in-line, and cut at soft.

Questions for the bear¶

If the 10-Q is clean, would you still refuse a 2.5% starter? If so, your objection is to the price and not the evidence.
Is there any pre-print price at which you'd initiate, such as the 523.78 lower band? If the answer is "only after confirmation, at any price," that's a legitimate evidence-first rule. But then call it that, not a valuation objection, because you've said you'd pay about 86x after a good print. Bull Analyst: # Bull Final Rebuttal: AMD at $555.60

(mine) marks my arithmetic on the tool data.

We've converged on a lot: the stock isn't cheap, the chart doesn't carry the trade, the Q2 10-Q is the gate, and sizing should come from a loss budget. Below are four places where I think your closing doesn't hold, then my final stance.

What I concede¶

The 27% margin was the favorable window. I'll use 22-25%. The FY25 full-year window (ratio 0.71, about 25%) is less noisy than one quarter's YoY, but the range is fair.
I have no informational edge pre-print. On your math the nine-session beta premium is about 0.4%, and I can't dispute it.
Meta is a second binary event. I'd use it as information, though (see section 4).

1. The 35x exit isn't "generous"¶

Today's 35.3x is price over forward EPS. You applied 35x to year-3 trailing earnings. The like-for-like exit is a forward multiple on year-4 earnings. If growth is still 30% in year 4 (my assumption), your 35x trailing equals about 27x forward. Changing only the multiple convention in your table:

Exit multiple	Trailing equivalent	Mid (22%)	Best (27%)
25x forward	32.5x	-3.5%/yr	+2.7%/yr
~27x forward (your table)	35x	-1.3%	+5.1%
35x forward (today's)	45.5x	+7.1%	+14.2%

If $15.76 is non-GAAP, the GAAP forward multiple is higher, which cuts against me. The point is that the sign of the return depends on whether the market still pays today's regime multiple. Only the top-right cell clears a double-digit hurdle for a beta-2.45 stock. This is a bet on persistence, not a cheap stock, and I'm not calling it one.

2. Q2 cash flow is Q1 giving back¶

You said Q1's flat revenue flattered working capital. Then average the two quarters, the same half-over-half device we both used on operating margin:

	H2 FY25	H1 FY26
OCF / revenue (mine)	24.4%	24.4%
Capex / revenue (mine)	2.5%	5.5%
FCF / revenue (mine)	21.9%	18.9%

Flat operating margin counted against my leverage claim, so flat cash margin counts against your deterioration claim. The entire 3-point FCF-margin decline is capex, which is investment and not weaker conversion.

This also answers "the two explanations can't both be benign." H1 ex-cash current assets rose about $5.0B against about $4.2B of non-capex investing outflows. That fits securities purchases (inference; the 10-Q settles it). The OCF "puzzle" is then just Q1/Q2 timing. Part of the OCF/NI drop may also come from the $0.27 EPS gap, worth about $0.45B. If that is non-cash, OCF/NI is about 1.28x, not 1.0x.

3. Calibrate the gate¶

EPS gap. You value the stock on operating income (140x), so a non-recurring gain doesn't change your thesis. It resets run-rate EPS to about $1.11. NI already exceeds operating income, so this item will very likely trip. A gate that almost certainly trips isn't informative, so I'd call it yellow.
Inventory build. Red, agreed.
Acquisitions. Non-current assets rose only about $3.0B in H1, including $1.2B of capex. Xilinx left a $12B→$67B signature, and nothing like that is here. That bounds acquired revenue without eliminating it, and at any normal revenue multiple it's a small slice of the $3.85B YoY increase. The 10-Q confirms.

4. Meta, symmetry and your $485 rule¶

You showed that waiting for the print is symmetric (11% up is as likely as 11% down). The same logic applies to paying the gap after Meta. Meta's capex commentary also tests the risk you named, AI-capex digestion, and it lands days before the print. So I'm changing my stance. The starter comes after the 10-Q and Meta, not "now."

On your pre-print entry of about $485 or lower, your own 50%-growth, 25%-margin case gives about 10%/yr from $555. If a strong print gaps the stock to $620, the same fundamentals give about 6%/yr (mine). So evidence that raises the price also lowers your return. That makes your rule "wait for price," not "wait for evidence," and the two aren't the same. I accept the first as a rule. The second needs a stated hurdle rate.

Final stance¶

Gate the starter on a Q2 10-Q with no inventory build and no material acquisition, plus Meta capex commentary that isn't digestion. If either fails, no starter and I agree with you.
Size by loss budget. A 0.5% portfolio loss at a 20% gap gives about 2.5%. Size caps the loss, not a stop.
Add after the print only on a strong result. That means Q3 revenue of at least $13.5B, gross margin of at least 54%, operating margin of at least 19%, and OCF of at least $2.5B (borrowing your confirmations). Hold at in-line ($13.0-13.5B) and cut at soft (under $13.0B).
Thesis broken by Q3 revenue under $12.5B, gross margin under 52.8%, or an inventory-led build.

Where we still differ¶

This is now narrow. You require an edge or a ~$485 price before taking risk. I'm willing to take a small, capped position when the primary trend is up, 44% H1 growth is reported, and the table's outcomes run from about -3% to +14% depending on whether the multiple regime persists. I can't show positive expected value pre-print, and I won't claim it. I'm claiming that a 0.5%-of-portfolio loss budget is a reasonable price for staying exposed to the upper half of that range. Bull Analyst: # Bull Closing: AMD at $555.60

(mine) marks my arithmetic. Assumptions are stated where I make them.

Your closing is the strongest version of the bear case, and much of it is right. Below is what I'm giving up, the three places I still disagree, and where I land.

What I concede¶

The add rule. Your 50%-growth, 25%-margin case returns about 10%/yr from $555 and about 6% from $620. Adding after a strong print, at the post-gap price, is a momentum add. I'm withdrawing the print tiers as a standalone add trigger. Adds now require re-running the table at the post-print price and clearing the same bar I'd apply to a starter.
No event premium. Nine sessions of beta premium is about 0.4%, or about 1 bp on a 2.5% position. I'm not defending a nine-session trade. Anything I hold is multi-year exposure, sized by loss budget.
Meta. It screens one risk (capex digestion). It says nothing about AMD's share, and the tools have no customer-concentration data. I'm dropping it as a gate.
The evidence boundary. The tools show no segment mix, guidance, consensus or concentration. My case is the consolidated P&L: +50% in Q2 and +44% in H1, 53.8% gross margin, 17.25% Q2 operating margin (15.7% TTM), 20% TTM FCF margin and 6.4% debt/equity.

1. $485 isn't derived from the hurdle you cited¶

You set CAPM at 14-16% and called $485 "lenient" against it. Holding your exit values fixed, and using the ~27x-forward row your $485 comes from (mine):

Best-case cell. The +5.1%/yr from $907B implies a year-3 value of about $1.05T. A 14% return needs an entry cap near $711B, about $435 a share.
Mid-case cell. The -1.3%/yr implies about $872B. A 14% return needs about $589B, about $360 a share.
At $485, your table gives +10.0% and +3.3%, about 6.7% averaged.

So $485 doesn't clear CAPM either. It's a margin-of-safety heuristic (10% in the best case, non-negative in the mid case), and a reasonable one. My 2.5% starter is also a heuristic. We have different tolerances, not a derived price against "no edge."

2. A standing bid fills when something went wrong¶

You showed that evidence which raises the price also lowers the return. The mirror holds too. A bid 12.7% below market fills when price falls. The likely causes before Nov 3 are a Meta capex scare, a rates repricing or a digestion headline. Those are news about the table's inputs (growth, exit multiple), not noise around fixed inputs. At $485 the 40% growth and 22-27% margin may no longer be the right inputs. Your rule needs the condition you put on my add: re-run the table at fill. My level carries the same condition.

3. "Cash-like" is a statement about weights¶

Your equal-weight average of about +4%/yr is correct arithmetic. But:

The grid fixes growth at 40%, below H1's 44% and Q2's 50%. Ten points of growth is worth about 7-8 points a year, against 10-11 points for the 25x-to-35x forward swing (mine, interpolated from your cells). That cuts both ways, since 30% growth would hurt badly. The grid omits a dimension that matters as much as the multiple.
Equal weights centered on 40% aren't neutral. I have no base rate for sustaining 40%+ growth on a $41B base, and the tools don't give you one either. "Cash-like" follows from your weights.
The $15.76 basis cuts both ways. I'm not rebuilding the case on it. But if it's non-GAAP, as the fundamentals report suggests, today's GAAP forward multiple is above 35x. Applying 35x to GAAP-like earnings then asks the market for less than today's multiple, not the same.
Your 4% cash rate is an assumption (FRED was down). If a December hike is coming, cash likely yields more, which cuts against me.

I'm not claiming positive expected value. I'm claiming the grid can't settle it. A loss-budgeted position is how I'd carry a view I can't resolve, and the reported fundamentals tilt my priors toward the right tail more than yours.

Where I land¶

Starter on a level, not at market. I'd use about $525 (lower Bollinger 523.78, Oct 16 low 526.83), 5.5% below today and inside your $485-525 range. If it doesn't fill, I don't chase, and I accept that cost as you accept yours.
Conditions at fill. Read the Q2 10-Q first. An inventory-led build is red, the EPS gap is yellow, and what the ~$4.2B of non-capex outflows funded is a capital-allocation question. Also require that the drop isn't a growth warning.
Size by loss budget. A 0.5% portfolio loss at a 20% gap gives a 2.5% maximum. Size caps the loss, not a stop.
After the print. The tiers stand: ≥$13.5B beats, $13.0-13.5B holds, $12.5-13.0B is soft and means cut. Adds follow the re-run rule above.
Thesis broken. Q3 revenue under $12.5B, gross margin under 52.8%, or an inventory-led build.

What's left between us¶

We agree the business is real, the stock isn't cheap, the weekly and monthly trends are up, and size should cap the loss. The remaining gap is $525 versus $485, and whether to carry 2.5% at $525, where your table shows +7.2% in the best case and +0.6% in the mid case. I say yes because I weight the right-tail cells more heavily, given reported growth of 44-50% and 20% FCF margins. You say no. The tools can't settle that difference in priors, but the Q2 10-Q and the Nov 3 print will give us much more to settle it with.

Bear Researcher¶

Bear Analyst: # Bear Case for AMD at $555.60

The business is real. Revenue is growing fast and the balance sheet is clean. My argument is that at a $907B market cap, 13 days before a print, the price already contains the good news. Most of your "acceleration" and "operating leverage" evidence also looks different once you pick a different starting quarter. Figures marked (mine) are my arithmetic from the fundamentals tables.

1. The +50% is mostly arithmetic¶

YoY growth is the product of the last four sequential changes: +20.3% (Q3 FY25), +11.1%, -0.2%, +12.5%. That is lumpy, not accelerating. The +20% quarter drops out of the comparison next print, so Q3 FY26 YoY falls to about 40% even if AMD repeats Q2's +12.5% sequential growth (mine).

Your own tripwire proves the point. $13B is +40.6% YoY. Holding 50% needs $13.87B, another +20% sequentially. You have set the "no change" line at a print that would show a roughly 10-point deceleration in the headline.

2. Operating leverage is a recovery from a weak Q1¶

You measured from Q2 FY25, which was a trough (39.8% gross margin, a -$134M operating loss). Use Q4 FY25 instead:

	Q4 FY25 (derived)	Q1 FY26	Q2 FY26
Gross margin	54.3%	52.8%	53.8%
Operating margin	17.1%	14.4%	17.25%
Opex % revenue	37.2%	38.4%	36.5%

Two-quarter view: revenue is up 12.3% and operating income is up 13.6%. The incremental operating margin is about 19%, not 40%. The 40% and the "+35% QoQ" are measured off Q1, when operating income had fallen 16%.
Gross margin is below Q4's level. Your kill criterion was "GM below ~53%", but Q1 FY26 was 52.8%. That line was already crossed two quarters ago without the thesis noticing, so it isn't informative.
There is nothing to blame on tax or investment items. TTM operating income is about $6.5B, so $907B is roughly 140x operating income (mine).

On the net income gap, "still above Q1" misses the point. The 142x P/E is built on GAAP EPS. Net income exceeded operating income in Q2 FY25 (by about $1.0B), in FY25 (by about $0.6B) and again in Q2 FY26. If that recurs, operations are not where the earnings come from. If it's a one-off, run-rate EPS is about $1.11, which annualizes to $4.44, or 125x. Even the full $1.38 annualizes to about 101x.

3. The 35x forward multiple depends on one number¶

$15.76 × ~1.66B shares is about $26B of net income, against $6.4B TTM. At Q2's 19.9% net margin, that takes roughly $130B of revenue. At a generous 25% net margin it takes about $105B. TTM revenue is $41B. The tool doesn't say whether $15.76 is GAAP.

Here is an illustration. Assume 50% growth for three years, which gives about $139B of revenue. Apply a 25% net margin, well above the 15.6% TTM margin, for about $35B of net income. At 25x that is $871B, below today's $907B. Three years of best-case growth would earn roughly nothing if the multiple compresses to 25x. That is what "priced for perfection" means.

A 14% pullback doesn't dent this when the stock is still up about 190% from its 52-week low and 39% above its 200 SMA.

4. In Q2, profit rose and cash from operations fell¶

Q2 changes: operating income rose 35%, net income rose 66%, and operating cash flow fell 20% ($2,955M to $2,366M). OCF/net income dropped from 2.1x to 1.0x.
Capex doesn't explain this. It affects FCF, not OCF. Non-cash current assets rose about $3.4B in a quarter when revenue rose $1.3B. You concede the tools can't split inventory from receivables. If it's inventory, that is your third tripwire, already tripped.
FCF isn't reaching shareholders. H1 FCF was $4.1B, and about $4.2B went to non-capex investing outflows nobody has itemized (FY25 was similar, about $4.5B). Financing was -$365M and the share count is flat. The FCF yield is 0.9% and cash fell to $5.1B.

5. The macro and demand evidence is weaker than presented¶

8% recession odds measure the wrong risk. AMD's demand is a few hyperscaler capex budgets, not GDP. The risk is AI-capex digestion, and one retail poster flagged Meta concentration cutting both ways.
The rate argument assumes the earnings arrive. With no cuts at 97% and a December hike at 78% (unverified against FRED), you say earnings are compounding faster than the discount rate. Most of the value sits in $26B of earnings that don't exist yet. That is the long-duration exposure.
The AI demand evidence is headlines. The Anthropic $600B has an unconfirmed vendor split, which you concede. Supermicro's raise is a general AI-server read-through with no AMD attribution. "Finally delivering profits" is a title. The tools contain no segment mix, guidance, consensus, share data or customer concentration, so the claim that AMD is winning AI accelerators isn't evidenced. Several headlines frame NVDA as the cheaper, better name.
Flows and targets are small. $332M of ETF buying is 0.04% of market cap. Susquehanna's $620 is +11.6%, about 2.3 daily ATRs on a 4.8% ATR stock. Its old $550 target sat below the stock when it printed 615–649, so the analysts are following price.

6. The tape doesn't confirm a bottom¶

Volume is light. The bounce is on the lowest volume in 30 sessions (10.1M, possibly a partial bar), against 28.8M on the 10-15 gap down. The cause of that gap is unknown.
The 50 SMA reclaim is thin. Three closes were below it, and one is above it by 2.7%, about half an ATR.
Short-term momentum is still bearish. Price is below the 10 EMA, the MACD histogram is -13.11, and the MACD line fell again on a +3.9% day.
OBV "no divergence" confirms the selling. A bullish signal would be OBV holding up while price fell. Instead OBV fell about 112M with price.
The oscillators give weak support. The report calls the daily TD-9 a short-term bounce signal. Weekly z is +0.52 and monthly z is +1.50, so only the daily looks cheap.
The weekly supertrend is lagging. It is a 3×ATR trailing stop, so it flips at 453.83, 18.3% below price. That is about the size of the drop that just happened (-18.6% peak close to trough).
Risk/reward is skewed. The downside is -18% to the weekly stop and -28% to the 200 SMA. The upside is +11% to the daily flip level (617.04) and +17% to the prior high, through resistance at 566.57, 599.13 and 617.04. Your 526.83 invalidation level is only about 1.1 ATR away, inside one normal day's range.

7. Insiders and crowding¶

Insiders sold about $335M this year, $212M of it Su's. There have been no buys in 17 months, and the last buyer paid $107–$114. "Formulaic" is plausible, but the data has no 10b5-1 flag. The CEO's share count tapered, but each share is worth more, so I'd want dollars. On the crowd, 14 bullish and 0 bearish tags, squeeze talk, call-heavy positioning and dismissal of the Griffin bubble comment all point to a crowded long going into the print. The sample is small and noisy, so I treat this as supporting evidence only.

Your three questions¶

Why is 35x perfection? It only holds if EPS goes up about 4x. The YoY growth rate is set to fall mechanically next quarter, and three years of 50% growth would earn roughly nothing at 25x.
What's the red flag in the operating line? It isn't that the line is bad. Over two quarters it shows no leverage (19% incremental margin, flat operating margin, gross margin below Q4), and operating cash flow is falling.
What is the bear case from $555? It's a hard comparison on Nov 3, a priced-in hike, an unconfirmed bounce on thin volume, and an 18% distance to your "primary trend" stop against 11% to the trend flip.

What would prove me wrong¶

Q3 revenue of about $13.9B or more, with gross margin of 54% or higher and operating margin above 19%.
A 10-Q that explains the net income gap with items that recur, and shows the $3.4B asset build was mostly receivables or securities rather than inventory.
OCF and FCF recovering to $2.5B or more.
A daily close above 617.04 on rising volume.

Bottom line: I'm not calling a collapse, and the weekly and monthly trends are up. But with the information we have, it's better to wait for the print than to pay 140x operating income for it. If you do own it, size for a ±5% daily range. Bear Analyst: # Bear Rebuttal: AMD at $555.60

You conceded the points that mattered most: "accelerating," the 40% incremental margin, OBV and TD-9, and 140x operating income. The dispute is now narrower, and I think your remaining position has three internal inconsistencies and a valuation problem. (mine) marks my arithmetic on the tool data.

What I concede¶

Seasonality. I checked your table against the YoY column. Q2→Q3 was +16.9% in FY24 and +20.3% in FY25. I withdraw "mechanical" decel, because that assumed Q3 repeats Q2's +12.5%.
Cash conversion. H1 OCF is 1.45x net income and TTM is 1.57x. The $3.4B asset build could include securities, so I can't call it leakage.
Insiders. The 90-day dollar pace is in line with the year, and an Oct 15 sale before the print fits a pre-set plan. I'm dropping this argument.
Risk/reward levels. Comparing tactical levels to tactical levels is the fair framing.

1. Your seasonality argument undercuts your own tripwire¶

Your "intact" line is below your own expectation. You say the seasonal pattern implies $13.5-13.9B. Your table calls anything at or above $13.0B "thesis intact," and that is 4-6% below the pattern you're using to justify holding into the print (mine). If the pattern is the expectation, $13.0B is a miss. A miss at ~22x sales is what a stock with this setup gets punished for.
The "everyone can compute it" argument cuts both ways. You said a decel anyone can compute from last year's $9.25B isn't news. The same is true of seasonal Q3 strength. If the pattern is public, it's in the price, and a pre-print position has no stated edge. You also said the chart doesn't carry the trade and you have no consensus.
The sample is thinner than it looks. You conceded Q3 FY25 was partly a rebound from a trough quarter. That leaves FY24 as the one clean observation.

2. Cash flow: the trend is the problem, and the tripwire is built not to trip¶

Your third tripwire needs two things at once. It trips only if non-operating EPS is confirmed and inventory built. A conjunctive tripwire on two unknowns is designed to stay untripped.
"Q2 is above Q3 FY25" ignores that revenue is 25% larger.

	Q3 FY25	Q4 FY25	Q1 FY26	Q2 FY26
Operating margin	13.7%	17.1%	14.4%	17.25%
OCF / revenue (mine)	23.3%	25.3%	28.8%	20.5%

Versus Q3 FY25, operating margin is up 3.5 points while cash margin is down 2.8 points. Q2 is the weakest of the four quarters, not just weaker than the best one. - Q2 is the better guide to your bull case. Q1 revenue was flat (-0.2%), which probably flattered working capital. Your thesis needs 12%+ sequential growth, and growth consumes cash. The 0.9% FCF yield is the price of that growth. - The two explanations can't both be benign. If the $3.4B was mostly securities, it can't also explain why OCF fell. Then the shortfall most naturally means part of the $2.3B net income wasn't cash. That is my inference, and the 10-Q settles it.

3. The 1.2-1.8:1 only works if stops execute¶

A 2-ATR stop at ~$503 assumes you can exit at ~$503. The 10-15 session opened at 540.00 against a prior close of 570.82 (-5.4%) on 28.8M shares, in a session the tools don't tie to earnings. An earnings gap in a 4.8% ATR stock can open through $503. The loss isn't capped at 9.6% (mine). I don't have the options-implied move, but the ratio assumes a smooth path that a binary print doesn't offer.

"Partial position now, rest after confirmation" also means the half you hold before the print is, by construction, the half exposed to the unconfirmed outcome.

4. Your revised bet doesn't clear the bar¶

You dropped PEG and the $15.76 forward EPS, so nothing in your case says "cheap" anymore. What remains is 40%+ growth with a 17%+ operating margin. Here is what that bet is worth, starting from $41.3B TTM revenue, with net income at 93% of operating income (the Q3 FY25-Q1 FY26 average) and a 30x exit multiple (mine):

Horizon	Revenue	Op. margin	Net income	Implied cap	Annual return vs $907B
3 years	$113B	20%	$21B	$632B	-11%
3 years	$113B	25%	$26B	$790B	-4.5%
5 years	$222B	20%	$41B	$1.24T	+6.4%
5 years	$222B	25%	$52B	$1.55T	+11.3%

A margin hurdle. Breaking even over three years at 30x needs about a 29% operating margin. The best year in the record is 22.2% (FY21).
Even a five-year run is middling. Five years of 40% growth, which makes AMD a $222B-revenue company, returns 6-11% a year from here with margins above anything in the record.
The multiple is the swing factor. You said the exit multiple is the unknowable part, and that's true. But the table needs the multiple above 30x to work, and it only reaches ~40x if growth is still very high at that point.

5. Your two questions¶

What does waiting cost? About nine sessions remain to the print (Oct 22 through Nov 3). At a 4.8% ATR, an 11% move in either direction is roughly one standard deviation of ordinary noise (rough, mine). So "waiting costs 11%" is as likely to be "waiting saves 11%."

On whether I'd buy above my own flip level: yes, if the evidence arrives. The 617.04 is the daily supertrend's level, not mine. If Q3 shows ~$13.9B, gross margin of 54%+, operating margin above 19% and OCF back to $2.5B or more, my operating objections are retired and I'd pay up. That print annualizes to ~$10.6B of operating income, or about 86x (mine). That's a valuation hold, not a table-pounding buy. Paying before the evidence means paying a lower price for the same unknown, but it also means holding the downside the stop doesn't cover.

What is an in-line print worth? By your own pattern, in-line is $13.5B or more, not $13.0B. But take your number: $13.0B, 53.8% gross margin and 17% operating margin annualize to $52B of revenue (17x sales) and ~$8.8B of operating income, or about 103x (mine). That is a hold if you own it and not an initiation at $555. I'm not recommending a short. Weekly and monthly trends are up, beta is 2.45, and a crowded 14:0 retail tape plus call-heavy positioning is squeeze fuel into a print. My argument is against paying this price, not for betting on a collapse.

What you didn't answer¶

Rates and duration. You dropped "earnings compound faster than the discount rate." With a December hike at 78% (unverified against FRED), the value is in earnings that don't exist yet.
Burden of proof. You said the evidence gap cuts both ways. At 140x operating income, it doesn't. The price asserts continued AI share gain, so the owner has to evidence it. The tools offer headlines, a vendor split you conceded is unconfirmed, and no segment, guidance or consensus data.
Capex digestion shows up late. You said it would appear in AMD's P&L. By then the multiple has already moved.

Bottom line¶

Your stance is a partial position now, sized for ±5%. That means pre-print exposure with no edge you can name, a stop that doesn't hold across a gap, and a valuation that returns low-teens at best even if you're right for five years. My stance is to not initiate before Nov 3, trim into 599-617 if holding, and revisit after the 10-Q shows what the $3.4B asset build and the gap between net and operating income really were.

If you want to keep your tripwires, make the 10-Q one disjunctive: either non-operating EPS confirmed or an inventory build trips it. Otherwise it can't fail. Bear Analyst: # Bear Closing: AMD at $555.60

(mine) marks my arithmetic on the tool data. Stated assumptions follow each calculation.

We agree that the stock is fully valued, the chart doesn't carry the trade, and the print decides it. What's left is whether a small pre-print starter has positive expected value. I'll concede what you earned, then explain why I think it doesn't.

What I concede¶

The working-capital build is about $1.8B net, not $3.4B. Current liabilities rose $1.6B, so the net build is about $1.8B (mine). If those liabilities are payables or accruals, they already sit inside Q2 operating cash flow. That shrinks the question without answering it, because $1.8B is still about 78% of Q2 net income.
The Q2 10-Q is already filed. My "revisit after the 10-Q" was loose. Neither of us can read it through these tools, so both our positions are conditional on it. I accept it as the gate.
Your sizing arithmetic is sound. A 0.5% loss budget over a 20% gap gives a 2.5% position, and that is a better tool than a stop.
Retained earnings count. They add about 2 points a year, and I include them below.
Your revenue tiers are fine.

1. The 27% margin is the best of four windows¶

Your 27% year-3 margin comes from Q2's opex/revenue growth ratio of 0.64. Here is the same method (40% growth, 53.8% gross margin, TTM opex of $15.5B) on other windows (mine):

Opex growth ÷ revenue growth	Window	Year-3 op. margin
0.64	Q2 FY26 vs Q2 FY25	~27%
0.71	FY25 vs FY24	~25%
0.81	Q3 FY25 → Q2 FY26, the base we both used	~22%
0.91	Q1 FY26 vs Q1 FY25	~19%

You picked the most favorable window. Operating margin was 15.5% in H2 FY25 and 15.9% in H1 FY26, which you've already conceded is flat. I'll use 22% as the mid-case.

2. What each entry price buys¶

Assumptions: 40% growth for 3 years ($113B revenue), net income at 93% of operating income, retained earnings added at 1x, no dilution, and a 35x exit. I'm being more generous than the 30x we used before, since 35x is today's forward P/E.

Entry	Best case (27% margin)	Mid case (22% margin)
$555.60	+5.1%/yr	-1.3%/yr
$523.78 (lower band)	+7.2%/yr	+0.6%/yr
~$485	+10.0%/yr	+3.3%/yr

For a beta-2.45 stock, even your best case from today returns about 5% a year. A 10% best-case return needs roughly $485.

Evidence can change this. If the print supports 50% growth for three years at a 25% operating margin, the same table gives about 10%/yr from $555 (mine). A good print can earn the price. A pre-print starter can't.

3. The "event-risk premium" is too small to see¶

The premium is small. With a 5% equity premium (my assumption) and beta 2.45, nine sessions of beta premium is about 0.4%. The distribution around the print is plausibly ±10% or wider. I don't have the implied move, and neither do you.
Sizing caps the loss but doesn't create a gain. You said your edge isn't informational, so a 2.5% starter risks 0.5% of the portfolio for an expected return near zero.
Meta isn't a cheap test. Retail posts today point to Meta reporting after the close (unverified). "Starter now" means holding through that event, and "after Meta" means paying whatever gap it produces. It is a second binary event.
Rates matter through the exit multiple. A 35x multiple in a hiking cycle, with the December hike 78% priced (unverified against FRED), is the assumption in the table most at risk.

4. Your questions¶

"If the 10-Q is clean, would you still refuse a 2.5% starter?" Yes, and you've named it correctly. A clean 10-Q retires my earnings-quality objection. What remains is price and the lack of an edge. That is a valuation objection, and it is linked to the evidence bar, because the 140x price is what makes the bar so high.

Your point about 86x is fair. I said I'd "pay up" after a strong print, but that multiple was computed at today's market cap, and a stock that rallies on the print carries a higher multiple. What I meant is that a strong print raises my growth input and I re-run the table. It isn't an automatic buy.

"Is there a pre-print price at which you'd initiate?" About $485 or lower. $523.78 only improves the best case to about 7%. The cost of this rule is that the market may never offer that price. If the print is strong and the stock gaps to $620, I miss about 11%. That is the price of the discipline, and you showed it is symmetric.

What should the 10-Q gate be? Make it disjunctive. Any one of these is a red flag: - The $0.27 EPS gap comes from non-recurring items. - The asset build is inventory rather than receivables. - The roughly $4.2B of non-capex investing outflows is acquisitions or strategic investments rather than securities. If acquisitions, the +50% growth may include acquired revenue.

Final stance¶

I'm not short, and I'm not predicting a collapse. The weekly and monthly trends are up, and the 50 SMA held.

Pre-print, I don't initiate above roughly $485-525. If you hold, trim into 599-617. If you take the starter anyway, your 2.5% sizing rule is right, but treat it as a bet with no demonstrated edge.
Read the Q2 10-Q before Nov 3, and apply the disjunctive gate above.
Act on the print by your tiers, with my confirmations added. A strong print is $13.9B or more, gross margin of 54% or more, operating margin above 19%, and OCF back to $2.5B or more. That earns a re-run of the table at higher growth. An in-line print is a hold with no add. A soft print is likely to gap through any stop, which is why I'd cap the loss with position size.

The bull case is a high-beta compounding bet on 40%+ growth with 22-27% margins. I think the evidence needed to justify $555 comes from the print and the 10-Q, so I'd wait for them. Bear Analyst: # Bear Final: AMD at $555.60

(mine) marks my arithmetic on the tool data. Assumptions are stated where I make them.

The gap has narrowed. You now say you have no pre-print edge, the stock isn't cheap, and you can't show positive expected value. You also moved the starter to after the Q2 10-Q and Meta. I'll concede what you earned, then argue that those admissions already decide the question.

What I concede¶

Cash conversion is flat half over half. H2 FY25 OCF/revenue is 4,759/19,516 = 24.4%, and H1 FY26 is 5,321/21,789 = 24.4% (mine). I rejected single-quarter noise on margins and then leaned on it for cash. "Q2 is the weakest quarter" is withdrawn as a trend claim, and the FCF-margin decline is capex (2.5% → 5.5%).
The securities reading fits the tables. Ex-cash current assets rose about $5.0B in H1 against about $4.2B of non-capex investing outflows. If that was securities, the cash is still AMD's, and "FCF isn't reaching shareholders" was too strong. The OCF question is unresolved, not deteriorating.
Acquired revenue is not a growth-quality gate. Even a multi-billion deal is a small slice of the $3.85B YoY revenue increase. Your "no Xilinx signature" bound is loose, because net non-current assets are after roughly $1.5B per half of D&A (mine, provider EBITDA minus operating income). But that doesn't change the growth rate. What the $4.2B bought is a capital-allocation question for the 10-Q.
The EPS gap is yellow. I value on operating income, so it doesn't move my thesis.
The exit-multiple convention. 35x on year-3 trailing earnings isn't like-for-like with a forward P/E, so "generous" was the wrong word.
My entry rule is a price rule. It comes from a hurdle rate, which I state below.

1. Your table, probability-weighted, is a cash-like return¶

Return/yr (your table)	25x fwd	~27x fwd	35x fwd
Mid (22% margin)	-3.5%	-1.3%	+7.1%
Best (27% margin)	+2.7%	+5.1%	+14.2%

Equal weight across the six cells gives about +4%/yr. Putting half the weight on the 35x row gives about +6% (mine).
That is roughly cash. Assuming cash pays about 4% (my assumption, since FRED was down), the excess return is near zero for a beta-2.45 stock.
CAPM sets the hurdle higher. With a 5% equity premium (assumed), a Blume-adjusted beta of about 2.0 requires about 14%, and the raw 2.45 requires about 16%. Only the top-right cell reaches that, and you conceded 27% was the favorable margin window. At a 25% margin with full persistence, I get about 11-12% (interpolated).
The one row that works rests on the number you retired. You dropped the $15.76 forward EPS two rounds ago because neither of us knows its basis or horizon. The "35x forward" row is built on it. That row also needs the market to keep paying today's multiple after growth has slowed to about 30%, with a December hike 78% priced (unverified against FRED).

2. You wait for Meta but not for the print¶

You moved the starter from "now" to "after Meta" because paying the gap is symmetric. That logic applies to Nov 3 too. Waiting is symmetric, you concede no edge, and nine sessions of beta premium is about 0.4%.

A loss budget is a ceiling, not a reason. On a 2.5% position, 0.4% is about 1 basis point of portfolio (mine), against a 0.5% loss budget. A position small enough not to hurt can't meaningfully help.
Meta is a weak gate. Higher capex shows the pool is growing, not AMD's share of it. The tools have no customer concentration data, and the headlines lean toward NVDA. Good Meta news will probably lift AMD, so you pay the gap. Bad news, you skip. The Nov 3 risk, the larger event, stays either way.

3. "Evidence raises the price and lowers the return" cuts against your add rule¶

Your arithmetic holds. The 50%-growth, 25%-margin case returns about 10%/yr from $555 and about 6% from $620 (about $1.01T cap) (mine). That is a competent market. A print strong enough to earn your add is mostly paid for in the gap. I said a strong print earns a re-run of the table, not an automatic buy, and I wouldn't buy at $620 for 6%/yr either.

But your stance adds after a strong print, which by your number is a ~6%/yr base case for a beta-2.45 stock. That is a momentum add, not a valuation add, and it needs a trend-based exit rather than a thesis.

My hurdle is at least 10%/yr in the best case and non-negative in the mid case. That gives about $485 now (+10.0% and +3.3%), and it is lenient next to the CAPM figure above. The cost is that the market may never offer it, and a strong print could gap the stock to $620 first. I accept that cost.

Where this leaves us¶

No new money before Nov 3 above about $485-525. If you hold, trim into 599-617.
Read the Q2 10-Q. An inventory-led build is red. The EPS gap is yellow. What the $4.2B funded is a question to answer.
Act on the print. Strong means ≥$13.9B revenue, ≥54% gross margin, ≥19% operating margin and OCF ≥$2.5B, which earns a re-run of the table. In-line means hold. Soft means likely gapping through any stop, which is why size must cap the loss.
The tools can't see segment mix, guidance, consensus or customer concentration. There is no Q2 data-center baseline, and the Polymarket Q3 median ($8.1-8.3B) has only $200-500 of volume behind it. The thesis rests on the consolidated P&L, and the price asks for much more than the P&L shows.

I'm not short, and I'm not predicting a collapse. The weekly and monthly trends are up, and the stock can run further. But at $555, ahead of two binary events, the probability-weighted return in your own table is close to cash for a beta-2.45 stock, and you've said you have no edge. Until the 10-Q and the print improve either the return or the evidence, my position size for new money is zero. If you take the 2.5% anyway, treat it as a no-edge bet and let size, not a stop, cap the loss. Bear Analyst: # Bear Final Word: AMD at $555.60

(mine) marks my arithmetic. Assumptions are stated where I make them.

Your closing is a disciplined position: a limit bid, a 2.5% cap, a 10-Q gate, no chasing. The gap between us is now one price level and one variable. I'll concede what you earned, then explain why I still wouldn't take the other side of it before Nov 3.

What I concede¶

$485 isn't a CAPM price. Your arithmetic holds. Clearing 14% from the ~27x-forward cells needs about $435 (best case) or $360 (mid case). $485 is a margin-of-safety heuristic, and calling it "lenient" against CAPM was loose.
A standing bid fills on bad news. Any fill needs the same table re-run you put on my add rule.
"Cash-like" was my weighting. The six equal-weighted cells average about 4%/yr (mine). That's a judgment, not a theorem, and the grid left out growth.
The GAAP/non-GAAP basis cuts both ways. A higher cash rate would also cut against me. Neither is resolved.
My rule is mostly "wait." By the rough estimate below, a $485 bid fills about 1 time in 8 before the print. Call it an evidence-first rule with a dislocation bid attached.

1. The dispute is one variable: 40% or 50% growth¶

Entering at $525 instead of $555.60 adds about 2 points a year over three years (mine). Equal-weighted, the grid average moves from about 4% to about 6%. Add your 7-8 points for 50% growth and it's roughly 12-14% (interpolated, so rough). That is close to the ~14% Blume-adjusted CAPM figure, assuming 4% cash and a 5% equity premium.

So your starter is effectively a bet on 50% growth for three years, which means $139B of revenue, 3.4x TTM (mine). Growth readings depend on the window, as the margin readings did: - ~26%: the last two quarters annualized (-0.2%, then +12.5%). Q1 is seasonally soft, so I don't treat this as the base case. - 44%: H1 YoY. - 50%: Q2 YoY, which equals the compounded last four sequential changes.

The grid's 40% sits mid-range, and 50% is the top, as 27% was for margin. Reported 44-50% growth is also the most-watched number on this stock, so it's already in a price of 22x sales and can't justify tilting your weights above the market's. Neither of us has a base rate for sustaining it on a $41B base.

2. The input that decides it arrives on Nov 3¶

No tool has guidance, consensus or segment mix. The print supplies forward guidance and the first test of a 46-50% Q3 YoY. You accepted that waiting nine sessions costs about 0.4% of beta premium. That is a cheap price for resolving the one variable that moves the return by 7-8 points a year.

3. A "level, not market" bid still holds the print about half the time¶

Rough assumptions: driftless path, daily σ ≈ ATR/1.6 ≈ 3.0%, nine sessions so σ ≈ 9%, and barrier-touch odds. - $525 (-5.5%): touch probability about 53%. - $485 (-12.7%): touch probability about 13%.

These are crude, since fat tails and mean reversion matter. But half the time your starter fills pre-print with a 5.5% cushion, about 0.6σ of the nine-day range, against a print distribution of ±10% or wider. Neither of us has the implied move. The loss budget caps the damage, but it doesn't supply a reason to take the risk.

Your fill condition, "the drop isn't a growth warning," also needs a diagnosis. After a full debate, no report explains the 10-15 gap (-5.4% open, 28.8M shares) or the -18.6% peak-to-trough drop. If the diagnosis comes from news you'll read at fill, then it is the diagnosis doing the work, not the price.

4. The loss budget is a ceiling¶

A 2.5% position at beta 2.45 carries about 6.1% of beta-adjusted exposure (mine). That scarce risk budget needs a reason to expect a return above the hurdle. At 40% growth the $525 average is about 6%, and at 50% it approaches the hurdle. You said a loss-budgeted position is how you'd carry a view you can't resolve. When resolution is nine sessions away at a ~0.4% cost, I'd wait for it rather than carry the view.

Where we land¶

Agreed: the business is real, the stock isn't cheap, the weekly and monthly trends are up, the Q2 10-Q is the gate, size caps the loss and not a stop, and adds must clear the table at the post-print price.

Not agreed: whether a pre-print starter at $525 is worth the risk budget.

My stance: 1. No new money at $555, and no pre-print starter at $525. A $485-or-lower bid is a dislocation bid, with the table re-run at fill. 2. If holding, trim into 599-617 and size for a ±5% daily range. 3. After the print, I buy if revenue is ≥$13.9B, gross margin ≥54%, operating margin ≥19%, OCF ≥$2.5B, and guidance supports about 50% growth. The test is that the table at the post-gap price gives ≥10%/yr in the best case and a non-negative mid case. If a gap prices that out, I pass. 4. I'm not short. If you fill at $525 and the print confirms 50%, you'll have been right, and I'll have paid the cost I named: missing a move I wouldn't chase.

Bottom line: at $555 you're paying for the top of the growth range three years out, ahead of a binary event that will tell us whether it's the top or the middle. Waiting is cheap and a pre-print position isn't, so I'd wait.

Research Manager¶

Recommendation: Underweight

Rationale: The debate was productive, and by the end the two sides agreed on most of the facts. The weight of evidence went to the bear on price, not on the business. The bull opened with five pillars: accelerating growth (+50% YoY in Q2 to $11.54B), a 40% incremental margin, a 35x forward P/E with a 0.53 PEG, $8.4B of TTM FCF on a clean balance sheet, and up weekly and monthly SuperTrends with a reclaim of the 50 SMA at 541.16. Almost every pillar got rebased. 'Accelerating' became 'sustained' once the bear showed YoY is just lumpy sequential moves compounded (+20.3%, +11.1%, -0.2%, +12.5%). The 40% incremental margin was a Q1-to-Q2 artifact. Measured from Q4 FY25 it is about 19%, and half-over-half operating margin is flat (15.5% vs 15.9%). PEG and the $15.76 forward EPS were dropped because nobody knows whether that figure is GAAP. OBV and TD-9 were retired. The bull also conceded several more points: ~140x TTM operating income, no informational edge before Nov 3, that a stop is not a loss cap on a 4.8%-ATR, beta-2.45 stock, and that adding after a strong print is a momentum add, not a valuation add. The bear gave ground too, and that matters for why this is not a Sell. Q3 has been the strongest sequential quarter in both prior years (+16.9%, +20.3%), so the 'mechanical decel' was withdrawn. Cash conversion is flat half over half at 24.4% OCF/revenue, and the FCF-margin dip is capex (2.5% to 5.5% of revenue). Insider selling is on pace and looks plan-driven. The $3.4B working-capital build nets to about $1.8B and may be partly securities. Nobody argued for a short, and the primary uptrend is intact. What decided it was the expected-return math, which the bull ultimately accepted. At ~$907B, 22x sales and ~140x operating income, the bear's three-year grid averages roughly 4-6% a year from $555. That grid assumes 40% growth, 22-27% operating margins, a 25-35x forward exit and retained earnings added back. A 4-6% return is close to cash for a beta-2.45 name and well below a 14-16% CAPM hurdle. Only the cells assuming 50% growth for three years (about $139B of revenue, 3.4x TTM) plus today's multiple regime clear that bar. That regime also has to survive a hiking cycle, with a December hike about 78% priced, unverified. The bull's final case is explicitly a prior-weighted right-tail bet that he says he cannot show has positive expected value, and he himself would not buy at $555. The variable that decides the return, 40% versus 50% growth, gets its first real test with guidance on Nov 3, nine sessions away, and waiting costs only about 0.4% of beta premium. The tape does not argue for paying up either. The bounce came on the lowest volume in 30 sessions. The MACD histogram is still negative and price is below the 10 EMA. The daily SuperTrend needs a close above 617.04 to flip, while the weekly regime stop sits about 18% below. Underweight fits better than Hold. Both sides agree there is no reason to carry full exposure into a binary print at a near-cash expected return, and the reported growth and higher-timeframe trend argue against exiting. Trim to a loss-budgeted core and let the 10-Q and the print earn back the weight. Several things are missing from the evidence: segment mix, guidance, consensus, the options-implied move, customer concentration, Q2 10-Q line items (the $0.27 EPS gap, the composition of the asset build, the ~$4.2B of non-capex investing outflows) and verified rate data. Nothing in the evidence explains the Oct 15 gap down on 28.8M shares.

Strategic Actions: 1) Target pre-print exposure of about 50% of a standard allocation, with a hard cap: a 20% adverse earnings gap must cost no more than 0.5% of the portfolio, which means a portfolio weight of 2.5% or less. Use whichever limit is lower. Size for a ±5% daily range and do not rely on stops to cap gap risk. 2) For holders above target, trim in stages. Take about one-third of the reduction now, around $555-567 (first resistance 566.57). Work the rest with limit sells at $599-617 (599.13, 617.04). If those levels have not filled by the close of the last session before the Nov 3 report, finish the trim at market, because the goal is reduced event exposure. 3) No new money at $555 and no pre-print initiation at $525. The only pre-print buy is a dislocation bid at $485 or lower, sized to at most 25% of a standard allocation and never above the 50% target. It requires three conditions at fill: re-run the valuation grid, confirm the drop is not a growth or capex-digestion warning, and clear the 10-Q check in step 4. 4) Before Nov 3, read the filed Q2 10-Q and treat each finding separately. An inventory-led build in current assets is red: cut to 25% of standard. Net income above operating income driven by non-recurring items is yellow: use run-rate EPS of about $1.11 for valuation. Identify what the ~$4.2B of non-capex investing outflows funded. If acquisitions, adjust the growth figures for acquired revenue. 5) If Meta's capex commentary signals AI-capex digestion, cut to 25% of standard. 6) Technical risk rules, on a closing basis. Two daily closes below 523.78: reduce to 25% of standard. A weekly close below the 453.83 weekly SuperTrend: exit the remainder. A daily close above 617.04 on rising volume: stop trimming and hold the residual, without chasing. 7) After the Nov 3 print. Strong means revenue of $13.9B or more, gross margin of 54% or more, operating margin of 19% or more, OCF of $2.5B or more, and guidance consistent with about 50% growth. On a strong print, re-run the grid at the post-gap price. If the best case is at least 10% a year and the mid case is non-negative, rebuild toward 100% of standard in tranches over 2-3 weeks. If the gap has priced that out, stay underweight. Revenue of $13.0-13.9B with gross margin at least 53.8% and operating margin at least 17%: hold the underweight, no adds. Revenue of $12.5-13.0B: cut to 25% of standard. Thesis broken: revenue under $12.5B, gross margin under 52.8%, operating margin under 15.5%, or a confirmed inventory-led build. Exit if any of these occur. 8) Gaps to close if possible before acting on adds: the options-implied move for Nov 3 (to calibrate the 20% gap assumption), consensus Q3 revenue and margins, and data-center segment figures.

III. Trading Team Plan¶

Trader¶

Action: Sell

Reasoning: The research plan's Underweight maps to a Sell, meaning a trim to a loss-budgeted core rather than a full exit. At about $907B market value, 22x sales and roughly 140x TTM operating income, the three-year return grid averages only 4-6% a year from $555. That is close to cash for a beta-2.45 stock and well below a 14-16% hurdle, and the bull conceded he would not buy at this price. The tape does not support paying up: the bounce came on the lowest volume in 30 sessions, the MACD histogram is -13.11, price is below the 10 EMA (566.57), and the daily SuperTrend only flips on a close above 617.04. It is a trim rather than an exit because the weekly and monthly SuperTrends are up (weekly stop 453.83), the 50 SMA (541.16) was just reclaimed, and Q3 has been seasonally strong. Guidance on Nov 3 is the first real test of 40% versus 50% growth, so exposure should be reduced into that binary event. Missing from the evidence: the options-implied move, consensus, segment mix, verified rate data, and any explanation for the Oct 15 gap down.

Entry Price: 566.57

Stop Loss: 617.04

Position Sizing: Portfolio context is not provided, so apply this to your own position. Target pre-print exposure of about 50% of a standard allocation, capped so a 20% adverse earnings gap costs no more than 0.5% of the portfolio (a weight of 2.5% or less); use whichever limit is lower. For holders above target, take about one-third of the reduction now around $555-567, then work the rest with limit sells at $599-617. If those have not filled by the close before the Nov 3 report, finish the trim at market. Do not add new money at $555. The only pre-print buy is a dislocation bid at $485 or lower, sized at no more than 25% of a standard allocation. Cut to 25% of standard on two daily closes below 523.78 or if Meta's commentary signals AI-capex digestion, and exit the remainder on a weekly close below 453.83. A daily close above 617.04 on rising volume ends the trimming (the stop level is a thesis check, not a loss cap, given the 4.8% ATR and gap risk). Size for a ±5% daily range.

FINAL TRANSACTION PROPOSAL: SELL

IV. Risk Management Team Decision¶

Aggressive Analyst¶

Aggressive Analyst: I'll open, since neither of you has spoken yet, and I'll take on the arguments I expect to hear.

The first thing to settle is what "aggressive" means on a Sell. At about $555 AMD has a $907B market cap, trades at roughly 22x sales and about 140x trailing operating income, and our return grid averages only 4-6% a year over three years. For a beta-2.45 stock, that is cash-like compensation for equity-plus risk. Holding full size is the comfortable, passive choice. The bold move is to recognize that the capital is badly priced for the risk and to act before the Nov 3 print. Even the bull conceded he wouldn't buy here, which tells me the case for adding is hollow.

I expect the conservative analyst to say, "If it's that expensive, exit completely." I disagree, and not out of timidity. A full exit throws away real convexity. Revenue growth accelerated to +50.1% in Q2 FY26, margins are expanding, and the weekly and monthly SuperTrends are both up, with the weekly stop at 453.83. The stock reclaimed its 50 SMA today and Q3 has been seasonally strong. Selling everything means betting that every one of those things fails at once. A loss-budgeted core keeps me in a 240% uptrend while capping the damage. That is how you stay in the game for the next leg and still harvest the valuation.

The conservative case also misses the other side of the plan. The trade isn't only about selling. I'm placing a dislocation bid at $485 or lower, sized at up to 25% of a standard allocation. If a bad print or a Meta capex scare gaps this stock down 12-15%, I want to be a buyer into forced selling, not a spectator who sold everything and has no framework to re-enter.

The neutral analyst will probably say it's balanced: strong fundamentals, a stretched price, so hold and see. That sounds sensible, but it isn't neutral. Holding full size into a binary event, with a 4.8% ATR, an unknown implied move, and a stock that already dropped 13% in a week, is a decision to take maximum variance for a low expected return. The neutral view also treats the bullish evidence as equal weight to the bearish evidence, when it isn't. Today's bounce looks like a relief rally. MACD histogram is -13.11, price sits below the 10 EMA at 566.57, and the daily SuperTrend only flips on a close above 617.04, about 10% away. I'll concede the volume point is softer than it looks, since the 10.1M-share bar may be unfinished. I don't need it, because the trend and momentum readings carry the argument without it.

Take the bulls' best card, the 35x forward P/E and 0.53 PEG. Forward EPS of $15.76 is about four times trailing GAAP EPS and nearly three times the Q2 GAAP run-rate. That implies about $26B of net income against $6.4B over the last twelve months. It's probably non-GAAP, so it isn't like-for-like, but either way the price assumes a giant earnings step that hasn't been reported yet. Meanwhile, Q2 net income exceeded operating income by about $0.27 a share of unexplained benefit, FCF fell 39% sequentially as capex doubled, and insiders sold about $335M this year with no buys in 17 months. None of that kills the story, but it means the margin for error is thin.

Sentiment backs the trim too. StockTwits shows 14 bullish tags and zero bearish, and posters are dismissing the "AI bubble" talk and predicting short squeezes. When the crowd has no bears, I get paid for being the one who sizes down. Susquehanna's $620 target is only about 12% above the current price, which isn't much reward for this risk.

I'll acknowledge the gaps. We lack the options-implied move, consensus, segment mix, verified rate data, and any explanation for the Oct 15 gap down. That's exactly why the plan is a trim to about half of standard size, capped so a 20% earnings gap costs no more than 0.5% of the portfolio, rather than a naked bearish call. It also has hard triggers: cut to 25% on two closes below 523.78 or a Meta capex-digestion signal, exit on a weekly close below 453.83, and stop trimming on a close above 617.04 on rising volume.

I'd rather be the analyst who reduces exposure at 140x operating income with a defined plan to buy dislocation than the one who rides a coin flip at full size and calls it patience.

FINAL TRANSACTION PROPOSAL: SELL Aggressive Analyst: I'll give back what you two have fairly taken first, because the rest of my case is stronger without the weak parts.

The conservative is right that selling one-third of the reduction now leaves you at about 83% of standard until the stock rallies 8 to 11%. I can't call this a relief bounce and also need it to run to 599 or 617. Both of you are right that the $485 bid collides with my own rule to cut at 524, and that it would fill on a gap before anyone knows whether Meta or AMD's own guide caused it. I'm also dropping "Q3 is seasonally strong" and the 50 SMA reclaim. Nothing in our reports supports the first, and the second is half an ATR on a possibly unfinished bar. The case doesn't need either. It rests on price against a 4-6% return grid, a binary event in about two weeks, and an implied move nobody has looked up.

So the mechanics change. Sell about half the reduction in the next session or two, taking 100 to roughly 75. Put the next tranche on limits around the 10 EMA at 566 to 580, not 599 to 617. Finish with a market order keyed to the confirmed report time, because if AMD reports after the close on Nov 3, "the close before" is that same day's session. Check the Meta date today as well. Replace the $485 bid with a post-print buy ladder triggered by what guidance says, not by price.

Where I won't move is the conservative's 25%-or-flat. Our own grid says AMD returns 4-6% a year. That is inadequate pay for the variance, but it isn't negative. Going flat is a call that expected return is below zero, and nobody has shown that. Meanwhile the weekly and monthly trends are up, and the biggest gaps in the window were upward. The reward-risk comparison also used the wrong marker. Susquehanna's $620 is one firm's number, and the stock closed at 649 two weeks ago. Against the real high, the upside is +17% to +18.5%, and the weekly stop is 18% down. That is roughly symmetric. Only the 200 SMA is 28% away, and that is where a trend break ends, not a marker for the next 13 days.

On the 20% gap as a central case, the 18.6% slide was a grind with one roughly 5% gap-down, not a series of gaps. Twenty percent is about 4.2 ATR, so I treat it as a stress test, which is how the plan uses it. The conservative is right that it runs through 453.83. That is why the cap is computed on the gap and why the plan calls the weekly stop a thesis check, not a loss limit.

The conservative's growth arithmetic is correct. At Q2's 12.5% sequential pace, revenue is about $13.0B, only +40% against $9.25B, and holding +50% needs about 20% sequential growth. I'll use that against the neutral. His "company growing 50%" leans on a Q2 FY25 base that was the softest in the series, with a 39.8% gross margin and an operating loss. That is a weak reason to hold more. The reason to keep a core is that margins are expanding, incremental operating margin was about 40%, and a 40% headline on 12.5% sequential growth isn't a miss in dollars. We don't have consensus, so we can't say how the market will read it. That is the argument for halving, not for exiting.

On the neutral's 40% versus my 50%, the difference is 0.1% of the portfolio on a 20% move either way. That isn't something the evidence can settle. The 0.5% cap, the 0.4% and the 0.25% are all mandate choices, and I won't pretend the data picks 40 over 50. If your risk budget says 40, run 40 with these mechanics. What we can actually argue from the data is the shape: front-load, sell near the 10 EMA, back it with a date-based order, and rebuild after the print.

I'll take the neutral's put-spread idea further, and it turns the conservative's best point into a tool. He's right that a share with a stop is not an option payoff. So if the implied move is about 10% or less and puts are fairly priced, hedge the last tranche with a put spread instead of selling it. That is real convexity, and for a holder sitting on a 240% gain it avoids a taxable sale. If the implied move is 14-15%, premium is too rich and you sell the shares. We need to pull that number before the final tranche.

My case doesn't depend on rates, the StockTwits tags or insider selling. I agree none of those should carry weight when FRED is down, the sample is thin, and the selling looks formulaic. A flat position also carries a cost. It means buying back after the print, probably higher if the print is good, with no core to benefit from the upside gaps this stock has already shown.

FINAL TRANSACTION PROPOSAL: SELL Aggressive Analyst: Let me start with what I'm taking, because the conservative and the neutral are right about the order of operations, and fixing it makes my own case stronger.

The conservative is right that "check the Meta date today" and "sell over the next session or two" don't fit together. The retail posts about a Meta after-hours reaction run from about noon to 1:20 Eastern, and one poster quotes 555.55. That suggests our close is a midday print and today's bar is unfinished. If so, the 10.1M volume is mostly a clock artifact, and I'm dropping it from the thesis entirely. I'm also dropping "relief rally," because I can't read a bar that's still printing. The tape evidence I'd keep comes from completed bars: the unexplained 10-15 gap down on 28.8M shares, and 18.6% lost in eight sessions with no earnings.

So the lookups go first, and they take minutes: Meta's report time, AMD's confirmed date and time, the implied move, and consensus. Today is a Wednesday, which is a likely Meta night, but nobody has verified it. If Meta reports tonight, I'd be at about 60 of 100 before the close. If it doesn't, I'd get to 70 over a session or two.

On 50 versus 60, I side with the neutral. The conservative's four-times test treats a limit order as a bet that loses 13% against a 3% gain. But a gap up fills the limit at the open, well above 573, so the order only underperforms if the stock falls. That makes it a directional call, and nobody has argued for negative drift. What's left is a variance budget. Ten points of standard through a 5-13% Meta reaction is roughly 0.03-0.07% of the book, assuming a 5% standard weight. If you want 50, take 50. I wouldn't spend more time on that gap.

I'm also taking the conservative's rebuild point. A 40-versus-50 growth gate fails on arithmetic, because 12.5% sequential growth gives about $13.0B, only +40% against a base that included the one 20% sequential quarter in the series. Write the gate in dollars before the print. My proposal is revenue near $13B, gross margin at or above about 53%, operating margin at or above about 17%, and a Q4 guide measured against consensus once someone pulls it. Those thresholds anchor on Q2's 53.8% and 17.25%, so they are my proposals, not data on what the market expects. I'd add a price condition too, since the same good print is a different buy from a stock that has already gapped up than from one that is 12% lower.

The put spread comes in at half credit, as the neutral said. A 525/450 spread covers about two-thirds of a 20% gap and under half of a 30% drop. The tax-gain argument depends on a holder we know nothing about, so I'm withdrawing it.

Here is where I'm staying put. The conservative's 25-33% floor comes from a 30% stack: a 20% gap plus a week of follow-through, treated as something you can't act on. The gap is the part you can't act on. After that you can, so the neutral's post-print rule is the right answer. I'd make it two-sided. If AMD's guide lands below the bar we wrote beforehand, cut to a quarter within two sessions with no open-ended judging. If the numbers clear the bar and the stock still gaps down on a sector or Meta-driven move, hold the core and buy in steps. That is what my $485 bid was trying to do, minus the blind buying.

I also don't want the stress number to be a fight between 20, 25 and 30. Let the implied move set it. A reasonable rule is stress at about twice the implied move, with a floor of 20%, and the weight is 0.5% divided by that. If the market prices 10-12%, you get 2-2.5%, which is the trader's number or the neutral's. If it prices 15%, the stress is 30%, the weight is about 1.7%, and the conservative is right. That is my heuristic, not a law, but it lets data decide.

On "thin edge on huge variance means a small core," I only half agree. Our own grid says 4-6%, which the trader calls close to cash. That doesn't prove the edge is zero, and it's an average across scenarios, not a point forecast. And "opportunity cost versus lost capital" is an asymmetry only if your mandate says so. Fifteen points of standard move the book by the same dollars up or down. A stability-first mandate is legitimate, and nobody has given us the portfolio context. If the mandate is strict capital preservation, go flat. We agree on that. Otherwise the lower-of rule already builds in the 0.5% cap, and nobody should have to rebuild a core after a good print.

The conservative also made the best argument for convexity. A beta-2.45 share is linear, and a stop doesn't turn it into an option. So for the upside you don't want to hand back, if the implied move is moderate, a small call spread paid for out of the same 0.5% budget is a cleaner way to stay in than 15 extra points of shares. It's fairly priced, not free, and the premium counts as capital at risk. If the implied move is 14-15%, skip it and just sell shares.

So the plan is this. Today, do the lookups. Sell to about 60 before the close if Meta is tonight, or to 70 over a session or two if it isn't. Put roughly 15 on limits at 566-580. Finish at market on the confirmed AMD report day. The final target is the lower of half of standard or the budget weight from the stress rule. There is no new money at 555, no $485 bid, no 599-617 tier, and no seasonality or volume arguments. Anyone already at or below target does nothing. After the print, follow the two-sided rule and the dollar gate.

We all say sell, and the remaining disagreement is worth a few hundredths of a percent of the book. The reason to act now is that the reward is cash-like, the event is binary, and we have time to get the order of operations right.

FINAL TRANSACTION PROPOSAL: SELL Aggressive Analyst: Before I argue anything, here is what I'm taking from you, because you've both earned it.

The conservative is right that tranches should be written as fractions of the distance to target. Sixty today plus fifteen on limits lands at forty-five, which either sells through a fifty target on a bounce none of us can read or leaves a forty target short. I should have absorbed that when the neutral raised it. Size the limit tier to the gap between sixty and the target, cancel it on arrival, and let the dated market order finish.

I'm also taking the attribution trap. My "hold if it was sector-driven" carve-out is the story people tell themselves to justify holding, and the 28.8 million share gap on the fifteenth is a live example that nobody can explain. It goes. The neutral is right that a gate anchored to Q2's 53.8% and 17.25% is necessary but not sufficient, because consensus decides the rest. One nuance is that forward EPS of $15.76 is probably non-GAAP and covers a forward year, so "three times Q2" overstates how much of a step-up the next print must show. But the gate still can't carry the decision alone.

The call spread comes out of the baseline. The neutral is right that selling fifteen points of shares and buying back a fraction of that delta in calls is a smaller sale with more moving parts, bought into pre-earnings implied volatility that we haven't priced. The put spread gets less than half credit until someone has the chain, since it pays nothing between 555 and 525, and that zone contains a gap the size of the one we just saw. Both are optional extras for someone with a chain in hand, and the baseline is shares.

On sequencing, I'm fine with sixty today unconditionally. My Meta-dependent split was only there because I hadn't done the lookups. Put Microsoft, Alphabet, Amazon and the October FOMC on the list, but all four dates come from memory, not our reports. I agree with the neutral that "more events means sell sooner" has no stopping point, because the target and the budget are what stop it. Don't size off the 78% December hike print either. FRED was down and nobody has verified it.

Now the places where I'd still push.

The post-print rule is the live argument. I side with the neutral over the conservative on the first step. An in-line guide is the base case, and cutting to the floor on "at or below consensus" double-counts the valuation worry you already acted on. "In line" needs a band, though. I'd say a guide midpoint within roughly 1-2% of consensus, which is my rule of thumb and not data. But the neutral's second step has a flaw he hasn't noticed. The 50 SMA is under 3% below price and rising. Any gap of 7% or more from here opens below it, so "closes below the gap-day low or the 50 SMA" fires almost automatically. The rule quietly becomes "gap down 7%, go to the floor." If you want that, say it. I'd use the gap-day low as the second test and drop the 50 SMA from it. Cut half the distance to the floor at the open, and cut the rest only if the stock closes under that day's low. That is still attribution-free. It just refuses to sell the low of a session the stock hasn't yet confirmed, and it fails fast when the stock does confirm.

On Meta, I agree the trader's trigger needs numbers. A cut or flattened capex range fires it and nothing else does. We don't have AMD's customer mix, though, so don't lean on Meta as a proxy for AMD's demand. If Meta holds its range and AMD drops 7% or more the next day, the same price rule applies.

On keeping a core, the conservative is right that the big gaps up came off 348 and 421, not 555. I'm not asserting that a gap up is the base case. I'm keeping a core for the reason the neutral gave. We can see accelerating growth, roughly 40% incremental operating margin, a clean balance sheet and weekly and monthly trends still up. We can't see the $0.27 of unexplained EPS, the Q2 capex doubling, or what moves the multiple next. When we can't tell which way it breaks, a budgeted core costs a few hundredths of a percent of the book. Going to the floor and rebuilding slowly after a good print isn't free either, but it is a price worth paying to cap the other tail.

The remaining unknowns are the implied move, consensus, the confirmed AMD date and time, Meta's capex range and the other dates. They should set the target and the hedge decision, but they shouldn't hold up the first sixty. That sale is justified by price against a 4-6% return grid alone, so do the sale and the lookups in parallel.

This is the bold version of a trim. Sell forty points of standard at 555 into variance the grid says you aren't paid for, keep a core sized so a stress gap stays inside the budget, and write the rebuild and cut rules before the print, while nobody is emotionally attached to the answer.

FINAL TRANSACTION PROPOSAL: SELL Aggressive Analyst: I'm taking most of that, because the last round made the plan better.

The conservative is right that a rule calibrated on a gap that went from a 5.4% open to 7.4% under the old close by the 16^th can't be written to look only at the open. The neutral is right that a close can't print below its own day's low, so my gap-day-low test only works as a later close. I'm fine with the 50 SMA clause being gone, with look-through exposure going into the definition of standard once, and with Meta's trigger written as a number against its last capex range. I also accept pulling the earnings-implied move the same way twice. One practical addition on the re-pull: if the second pull raises the target, nobody buys back the day before the print. You cancel whatever limits are still resting. That gives the conservative his ratchet without a buyback.

Here is where I still push, limited to things that change an order.

First, the 5% and 7% post-print triggers. The conservative justified 5% as half the expected swing at a 10% implied move, but we don't have the implied move. If it comes back at 14%, 5% is a third of the swing and fires on an ordinary outcome. We also calibrated these on a gap with no earnings in it. So write them as half and seven-tenths of the earnings-implied move, with 5 and 7 as floors and as defaults if the lookup fails. At a 10% implied move it's the identical rule. It uses the same logic as the stress rule we already adopted, one input instead of hand-picked numbers.

Second, the guide-versus-consensus trigger. Both of you kept "guide below consensus, go to the floor in the first session." But the conservative's own argument in his last message was that for a stock up 240% the bar that matters is the buy-side's, and the tape is the only way to read it. Consensus is a sell-side number. If the guide is a hair under and the stock is up 6%, that tree sells to the floor into strength on a story, which is the attribution trap with a different label. I proposed a guide-below-bar cut myself earlier, and I'm retiring it for this reason. Use consensus as a veto on rebuilding, not a trigger to cut, and let price decide the cuts. That deletes a branch, which is what the neutral asked for. The plan already defaults to price alone when consensus can't be sourced, which suggests price was carrying the weight anyway.

Third, the pre-print levels. The conservative is right that dropping the 523.78 cut silently isn't a decision, and I'll take the neutral's half-step version. But 523.78 is a Bollinger band. It recomputes every session and will almost certainly have moved by the time it matters. Pin the rule to a printed price instead, the 10-16 intraday low of 526.83. A close under it is a close below the low of the entire sell-off. The same standard applies to the trader's clause that ends trimming on a close above 617.04. Nobody kept it, so I'm dropping it on purpose. The budget is sized for a gap, and a rally into the print makes a gap more expensive in dollars. The return grid also gets worse at 617, not better, so a rally shouldn't stop the trim.

Fourth, the clock on the rebuild. The neutral is right that "rebuild slow" without a clock becomes never. The cuts are same-session, and the justification is follow-through: 5.4% down at the open to 7.4% by the close, and 18.6% in eight sessions. The same tape shows follow-through the other way, the 43% from the July 29 close to the September 21 close, which is the conservative's own number. If follow-through is why we cut at the open, it's also why the first rebuild step shouldn't wait more than two closes above the gap-day low. Rebuild is capped at the valuation cap, so the worst case is returning to half of standard, a size we already called acceptable. I'd do the second step a week after the print, and cancel both if the stock closes under the gap-day low. That's a clock on the existing branch, not a new one.

None of this is the bold part. The bold part is selling about forty points of standard today into a reward our own grid calls cash-like. We keep a core because we can't see which way it breaks, and we write every rule before the print so neither fear nor FOMO writes it after the gap.

To keep the plan from waiting on a lookup, here are the defaults. If there's no implied move, use a 25% stress, which is about a 2% weight at a 5% standard. If there's no consensus, price decides alone. If AMD's report time isn't confirmed by Oct 30, finish the sale by the Nov 2 close. Anyone already at or below target does nothing, and a strict capital-preservation mandate goes flat. Nobody adds money at 555.

FINAL TRANSACTION PROPOSAL: SELL

Conservative Analyst¶

Conservative Analyst: I agree with the aggressive analyst on the diagnosis. By his own grid the stock returns 4-6% a year with a beta of 2.45, and that doesn't pay for the risk. My problem is that the plan he built doesn't do what he says it does.

Start with the mechanics. For someone at a standard allocation today, the plan sells about a third of the reduction now. That is roughly 17 points out of 100, so you're still holding about 83% of standard. The rest sits in limit orders at 599 to 617, which only fill if the stock rallies another 8 to 11%. So the "trim to half" gets you smaller only if the stock goes up. If Meta spooks the group first, those limits never fill and you take the gap at five-sixths size. Retail is already talking about an after-hours Meta reaction, though the date isn't verified. He criticizes full size into a binary event, then leaves most of the trim dependent on a rally he calls a relief bounce. If it's a relief bounce, sell into it. The 10 EMA at 566 is the nearest resistance, and 599 is a long way off.

The triggers have the same flaw, because they all fire after the damage. Two closes below 523.78 means you've already taken the first gap and part of the second. That level is only 5.7% away, about 1.2 ATR. The loss budget also treats a 20% gap as the ceiling, and a 20% gap lands near 444, already through his 453.83 exit. This stock fell 18.6% in under two weeks with no earnings print, and the 10-15 gap came on the heaviest volume since September 21 with no explanation. The 200 SMA is 28% below. I'd treat 20% as a central bad case, not a tail.

On the $485 bid, that isn't much of a dislocation. At 485 the stock is still around 124x trailing GAAP earnings and 19x sales. If 140x is badly priced at 555, it's only 12% less badly priced there. If the gap comes from guidance, the whole 35x forward premise is what breaks, so the cheapness is illusory. The bid also sits between his own rules, which say cut to 25% below 524 and exit under 454, and a standing order to buy before you know why it gapped tends to buy the first leg of a repricing. Decide after the print, with guidance in hand.

On convexity, a beta-2.45 share is linear and gappy. A stop doesn't give you an option payoff when the stock jumps over it. If you want upside through the print with a known cost, that's a defined-risk structure, and we have no implied move to price one. That's another reason not to hold shares as a stand-in.

He also uses today's bar inconsistently. The 50 SMA reclaim is 14 points, about half an ATR, on the lowest-volume session in 30 that he concedes may be unfinished. He cites it as a reason to keep a core, then calls the same bounce a relief rally. And nothing in our four reports supports "Q3 is seasonally strong."

The fundamentals add a setup problem. If Q3 keeps Q2's 12.5% sequential pace, revenue is about $13.0B, only around +40% against the $9.25B base. Holding +50% needs roughly 20% sequential growth. So the headline growth rate likely steps down on arithmetic alone, at 22x sales. We also saw what a bad quarter looks like inside the last six: gross margin fell from 50.2% to 39.8% and operating income went negative. The cause is unlabeled, but it happened. Now capex has doubled and non-cash current assets grew about $3.4B in a quarter, with no inventory split.

I'll pre-empt the neutral view too. The balance sheet is strong, with a current ratio of 2.6 and cash about equal to debt. That speaks to solvency, not to what a 22x-sales equity does when the multiple compresses. Prediction markets, which are deep and unverified against FRED, price a December hike at 78% and no 2026 cuts at 97%. That's a headwind for this kind of stock, and the plan never mentions it.

I won't lean on the zero bearish StockTwits tags, since half those posts only mention AMD in passing. I also won't lean on insider selling, which looks mostly formulaic. The case stands on price, event risk and what we can't see.

My adjustment is that Sell stays, but I'd go deeper and sooner:

Cut the pre-print target to about 25% of standard, and halve the loss cap to 0.25%, which is a weight of roughly 1.25%.
Do the bulk of it now, at market, into this bounce, not at 599 to 617.
Drop the standing $485 bid and revisit after the print.
If the mandate is capital preservation, go flat and re-enter on evidence.

The cost is giving up some of a post-earnings pop. But the sell-side's best marker, $620, is only about 12% above, while the downside references at 454 and 399 are 18% and 28% away.

FINAL TRANSACTION PROPOSAL: SELL Conservative Analyst: I'll start with what I'll give back, because you've both earned it. Calling a 20% gap a central bad case was an overreach, since the only gap down in the window was about 5%. Setting Susquehanna's $620 against the 200 SMA was apples to oranges. Against the 649 close, the upside is about +17% and the weekly stop is about -18%, and I'll call that roughly symmetric. Limits at 566 to 580 are also a fair ask, because 580 is under one ATR away, nothing like 599 to 617. And I never said expected return is negative. My case is about size, and I'll come back to that.

The three of us now agree on direction and land within about ten points of standard on the destination. You both say 40 versus 50 is worth about a tenth of a percent of the book, and that figure leans on a 5% standard weight nobody has given us. So I won't spend your time on 25 versus 40. Where I think you're still exposed is the order of operations.

Both plans keep 70 to 75% of standard on through the next session or two, and the one event that could matter in that window is Meta. Today is a Wednesday, which is when Meta has usually reported, though that's my recollection and not in our reports. Retail posts timestamped around midday today already talk about Meta after hours, and the news report says Meta earnings were due this week. I can't verify the date, but "check it today" and "sell in the next session or two" don't fit together. If it's tonight, the next session comes after the reaction. Today's bar also looks unfinished, so there may still be time to act in this session. I'd make tranche one bigger and do it today, getting to roughly the trader's own 50% now rather than treating 50 as the pre-print destination.

Waiting for the limit costs little to skip. Selling near 555 instead of about 573 gives up roughly 3% on those shares. The aggressive's own Meta scenario, a 12-15% gap, costs four to five times that on the same shares. So waiting only pays if the chance of reaching your limit is more than four times the chance of that gap. The aggressive himself calls this a relief rally under the 10 EMA, with the MACD histogram at -13 and an unexplained high-volume gap down still on the tape. Use the limits for the second half of the reduction, not the first.

You both also make the last tranche depend on an implied move nobody has pulled, while scheduling the first two without it. That number is a lookup, and it sets the stress test and the hedge-versus-sell choice. Pull it today.

On the stress test, you've both said the weekly stop is a thesis check and the real protection is the size cap on a 20% gap. Then the gap assumption is the only parameter doing any work. The event doesn't end at the opening print. This stock lost 18.6% in eight sessions with no earnings at all, and it rose 43% from the July 29 close to the September 21 close. A gap plus a week of that follow-through is a 30% drawdown, which is about 390, just under the 200 SMA. Re-run the trader's formula at 30%. Half a percent divided by 0.30 is a weight of about 1.7%. If that's below half of standard, it's your number. On the neutral's assumed 5% standard it's about a third, so I'd now set the pre-print floor at 25 to 33%.

I'm for the put spread as a tool, since I raised defined risk in the first place. But premium on a beta-2.45 name two weeks before earnings is probably rich. Retail chatter already includes 525 put hedges, which hints at demand but proves nothing about price. A spread protects between its strikes, so a gap through the short leg leaves exactly the tail the cap was meant to cover. Implied volatility usually collapses after the print, so a moderate drop can leave the hedge worth less than you paid. The taxable-sale argument also assumes a particular holder, and we know nothing about the portfolio. Count a hedged tranche at what it actually covers, not as sold.

On the rebuild, the neutral wants growth holding 45 to 50% before buying back. The aggressive has now agreed that at Q2's pace the headline lands near 40%, and holding 50 needs about 20% sequential growth. The only 20% sequential quarter in our series was Q3 FY25, the rebound off the trough everyone now calls a soft base. So that gate likely fails on a perfectly decent print. Then either you never rebuild, or you quietly loosen the gate after the fact. Write it before the print, in dollars and margins: gross margin around 53-54%, operating margin of 17% or better, and a Q4 guide whose bar you set in advance.

The neutral also leans on a 20% FCF margin. That's trailing twelve months. The latest quarter was 13.5%, down 39% sequentially, with capex doubled, non-cash current assets up about $3.4B, and about 27 cents of Q2 EPS unexplained. None of that kills the story, but it's the freshest data and it's softer than the trailing number.

On size versus sign, positive expectancy is necessary to hold anything, but how much you hold scales with edge relative to variance. A 4-6% grid on a 2.45 beta stock with a 4.8% daily range is a thin edge on a huge variance. That supports a small core, and shares with a stop aren't convex, as the aggressive has agreed. Buying back higher after a good print costs opportunity, while a gap through every level costs capital. With our targets about a tenth of a percent of the book apart, that asymmetry should break the tie for a stability-first mandate. I'd only go flat if the mandate is strict capital preservation.

So sell. Get to about half of standard today, then work down to a quarter to a third into the print using limits at 566 to 580 and a date-based market order. Treat any hedge as partial cover, and write the rebuild gate now.

FINAL TRANSACTION PROPOSAL: SELL Conservative Analyst: I'll start with what you've earned.

The neutral is right that my four-times test was wrong. A limit that fills on a gap up fills at the open, not at 573, so it's a directional bet. Under a zero-drift view it has the same expected value as selling now, with more variance. I also won't fight sixty versus fifty through Meta. At a 5% standard weight, ten points of standard through a 5-13% move is a few hundredths of a percent of the book, and the 5% is an assumption anyway. And the aggressive's stress rule, twice the implied move with a 20% floor, beats my fixed 30%. It lets the options market set the number instead of the three of us, and it shrinks the weight exactly when insurance gets expensive. I'll take it.

What's left is mostly what the plan doesn't say.

First, the lookup list is too narrow. Everyone has Meta on it, but Meta is the first capex read-through, not the only one. My recollection, which is not in our reports, is that Microsoft, Alphabet and Amazon usually report in the last week of October, and that the October FOMC falls the week before AMD's print. With a December hike priced at 78%, a hawkish October statement is a real event for a beta-2.45 stock. None of that is verified, so put the whole calendar on the list. It also changes the sequencing. If there are several gap risks before November 3, then "sixty if Meta is tonight, seventy over a session or two if not" just walks the extra ten points into the next event. Go to sixty today either way. The bar looks unfinished, so you can do it at today's prices.

Second, write the tranches as fractions of the distance to target, not fixed numbers. Sixty today plus fifteen on limits lands at forty-five. If the lower-of rule says fifty, you've sold through target on a bounce you both say you can't read. If it says forty, you're five short and need the market order anyway. Size the limit tier to the gap between sixty and target, and cancel it once you get there.

Third, the post-print rule, where I'd push hardest. The aggressive says if the numbers clear the bar and the stock gaps down on a sector or Meta-driven move, hold the core and buy in steps. How would you know it was sector-driven? We have a live example: a 28.8 million share gap down on the fifteenth that none of our four reports can explain. Attribution after a gap is something everyone does with a position, and it always lands on the answer that justifies holding. The bar also isn't tied to consensus yet, because nobody has pulled it. If a company clears a dollar bar we wrote blind and still gets sold, the market is telling you the bar was too low.

So I'd write it the other way. A guide at or below consensus takes you to the floor in the first session, not within two. The neutral's 10% trigger also leaves a hole at 5-10%, which is the size of the gap we just saw. That 5.4% gap extended to about 7% below the prior close within two sessions, on its way to 18.6% off the high. Cut fast, rebuild slow, and rebuild only after the stock holds a level for a couple of sessions, not into the gap.

Fourth, the hedges. The put spread covering two-thirds of a 20% gap is gross of premium. It also pays nothing between 555 and 525, a 5.5% dead zone that contains the 10-15 gap. Half credit is generous, so I'd give it less until we see the chain.

The call spread worries me more. We're saying the stock is badly priced for the risk, then proposing to pay premium two weeks before the print, probably into elevated implied volatility that collapses afterward, to keep upside on that same stock. The aggressive does say to skip it if the implied move is high, which is fair. But if it's in the plan at all, count its delta as shares and leave it out until it's priced.

One more thing on keeping a core for the upside gaps. The big gaps up we keep citing came with closes near 348 and 421, not 555, off a far lower base. They show the stock can gap up. They don't show it will from here, at 140 times trailing operating income. A flat position isn't my ask unless the mandate is strict capital preservation, but a core for upside gaps isn't free either. It costs real variance for an edge our own grid calls close to cash.

So here is where I land: - Go to about sixty today, whatever Meta's date turns out to be. - Finish by date, with tranches sized to the target, using the lower of half of standard or 0.5% divided by twice the implied move (20% floor). - Write the post-print rules now, with consensus in hand: fast down, slow up, and no buying on attribution. - Use hedges only once priced, at partial credit, with their delta counted.

FINAL TRANSACTION PROPOSAL: SELL Conservative Analyst: Let me start with what I'm giving you, because this plan is now much closer to what I'd want than where we started.

The neutral is right that an in-line guide shouldn't trigger a second cut on its own. I wrote that rule too mechanically. We will already have sold forty points for the valuation worry, and cutting again on the base case counts the same worry twice. A guide below consensus still takes you to the floor in the first session, and I'd keep that. For an in-line guide, though, price has to decide. The aggressive is also right about the 50 SMA clause. It sits under 3% below the stock and is rising, so any 7% gap opens through it and the clause adds nothing. Drop it. I'll also agree that the call and put spreads stay out of the baseline, and that nobody should spend more time on 40 versus 50.

The post-print price rule is still wrong, though, and that's where I'd push hardest. The neutral calibrated the 7% on how the gap on the fifteenth extended, but the rule measures it at the open. Run it against the one episode we have. The stock opened 5.4% under the prior close, which is below 7%, so no action. By the close on the sixteenth it was 7.4% under the old close. A rule that only looks at the open never looks again, so it would have stayed silent on the only gap in our data until the damage was done. Measure where you calibrated. Cut half the distance to the floor if the stock opens more than about 5% down. That's roughly one ATR, and at a 10% implied move it's half the expected swing. Go the rest of the way if either of the first two closes is more than 7% under the pre-print close. That's still attribution-free, and it would have caught the fifteenth.

The aggressive's version says to cut the rest only if the stock "closes under that day's low." A close can't print below its own day's low, so I assume you mean a later session. That means holding the second half through at least one more full session after the gap, and the rule fires only if the decline continues. I'd move the gap-day low to the rebuild side. Don't add back while the stock trades under it.

Second, nobody's summary keeps the trader's pre-print trigger, the cut to a quarter on two closes below 523.78. Either keep it or drop it on purpose. I'd keep it, as one close instead of two. That level is 5.7% away, and a close under it is a lower low than the 10-16 intraday low of 526.83, with the 50 SMA lost along the way. On a name with a 4.8% ATR, the extra confirming session is expensive. That's my judgment, not data, and the neutral may fairly say it double-counts. But dropping it silently isn't a decision.

Third, the 0.5% budget is computed on AMD alone. A $907B company is a meaningful weight in broad-index and semiconductor funds, and the same Meta night or hawkish FOMC hits the rest of the AI complex too. Portfolio context isn't provided, so the caller should count total exposure, including funds and correlated holdings, against the budget, not just the direct line. Otherwise "standard" understates the position and the target comes out too high.

Fourth, the target depends on the implied move, and the implied move changes. Set it from today's number, then re-pull on the last day before the print and use whichever gives the lower weight. A rising implied move is information about risk. A falling one isn't a reason to add back in front of the print.

Fifth, the whole post-print tree leans on consensus, and nobody has pulled it. Put it on the lookup list with a default: if you can't source it by the print, the price rules decide alone. For a stock up 240%, the bar that matters is the buy-side's, which we can't see, and the tape is the only way to read it. That's another reason the price tests need to be calibrated properly.

On the core, I accept the neutral's reasoning: we keep it because we can't tell which way it breaks, not because we expect a gap up. My only point is that the core is sized by the stress rule, and anything larger needs an argument the evidence hasn't given us. Anyone who needs strict capital preservation should go flat and rebuild on evidence.

You can call this the bold version or the moderate version. It's the same plan, and the version I'd sign off on has the sixty sold today and the lookups run in parallel. It has a target set from the implied move and checked again on the last day, and a date-based market order that finishes the job. After the print, I'd cut on closes against the pre-print level, not on opens or stories.

FINAL TRANSACTION PROPOSAL: SELL Conservative Analyst: I'll start with what I'm taking, because most of this round is right.

Scaling the post-print triggers to the implied move is better than my fixed 5 and 7, and I can say why. When the loss cap is the binding one, the weight is 0.5% divided by twice the implied move. A trigger at half the implied move then costs about 0.125% of the book, and one at seven-tenths costs about 0.175%, whatever the implied move turns out to be. The dollars at the trigger stay put. The cost is that at a 14% implied move the triggers sit at 7% and about 10%. The 5.4% open and 7.4% close we calibrated on would then have fired nothing. That's acceptable only because the position is already smaller by then, and only if the implied move is the earnings figure, pulled the same way twice, as the neutral said.

I'm also taking consensus as a veto on rebuilding in place of a cut trigger. A sell-side number shouldn't sell us into strength when the buy-side bar is what moves the stock. Dropping the 617.04 clause on purpose is right too, since a rally makes the gap more expensive and the grid worse. So is pinning the pre-print rule to the printed 526.83 instead of a Bollinger band that recomputes. And no buyback on the re-pull is the ratchet I wanted.

Four things still change orders.

First, we keep describing the target in points of standard, but the budget is a weight. If AMD rallies 10% into the print, a share count that was 2.5% of the book at 555 is about 2.75%, and a 20% gap on that costs 0.55%, over budget. So mark the target to the last-day price and sell down to it. Size the 566-580 limits in weight at those prices, too. The ratchet runs one way: if the stock falls, you don't buy back to restore the share count.

Second, the second post-print step shouldn't depend on the first having fired. As written, the gap-day-low test only exists if the open trigger tripped. A stock that opens down 4%, below the open trigger, then closes at minus 6%, minus 7%, minus 9% would slip through. This stock lost 18.6% in eight sessions mostly by grinding, so make the close test free-standing. Cut the rest if any of the first three closes is more than seven-tenths of the implied move under the pre-print close, with the 7% floor. That's a clarification, not a new branch.

Third, the lookup list is missing two items that cost almost nothing. One is the reason for the Oct 15 gap. We've calibrated the whole tape-based tree on a 28.8M-share gap that none of our four reports explains, and a news archive search for Oct 14-15 may say whether it was AMD-specific or sector-wide. The other is the Q2 10-Q. The fundamentals report flagged the roughly $0.27 of EPS above the operating-income run-rate, the working-capital build and the $4.2B of non-capex investing outflows. The "four-fold step-up" argument, and what a good print has to look like, both rest on that base, and the filing has been public for months.

Fourth, the rebuild needs one adjustment. The clock is fair, but the first step lands two closes after the tape told us to sell, so make the steps thirds of the room to the cap, not halves. After a gap up, "half of standard" was a conclusion about $555, so rerun the grid at the new price before using it as a cap. Otherwise the clock has us buying back at a worse price on a rule written for a better one. The 18.6% slide with no print also says the stress after the event isn't "ordinary." At roughly 18-19% it gives a weight of about 2.7%, so the room to add after a clean print is real but smaller than it sounds.

One small point on dates. Oct 30 is too late a deadline for a lookup that takes minutes, so do it today. Let the fallback be the close before the earliest date any source shows. Today that's Nov 2, off a single headline.

So the plan stands as converged. Sell to about sixty today and run the lookups in parallel, including the two I added. Set the target as a weight from the lower of half of standard or 0.5% over twice the earnings-implied move, floored at a 20% stress. Put limits at 566-580 sized in weight, and finish at market before the print. Cut on closes against the pre-print level, scaled to the implied move. Rebuild in thirds, with consensus as a veto, and add no new money at 555. A strict capital-preservation mandate still goes flat.

FINAL TRANSACTION PROPOSAL: SELL

Neutral Analyst¶

Neutral Analyst: I agree with where both of you landed on direction. At 22x sales, with a 4 to 6 percent grid return and a beta of 2.45, nobody should sit at full size into November 3, and I'm not going to argue for a hold. Where you both go wrong is the mechanics, and how much weight you put on a few shaky data points.

I'll take the aggressive plan first. The conservative analyst is right that it sells only about 17 points of a 50-point reduction up front. The rest hangs on limit orders 8 to 11 percent above the market, even though the aggressive analyst himself calls this move a relief rally. You can't believe the bounce is weak and also need it to run to 599 or 617 for your sizing to work. The plan does include a market-order backstop on the close before the print, which the conservative analyst skipped over. But that does nothing for Meta, and the retail chatter suggests Meta may report very soon. That date is unverified, so check it today.

The $485 bid is worse. It contradicts the plan's own rule to cut to 25 percent below 523.78, and a 12 to 15 percent gap fills it at the open, before anyone knows whether Meta or AMD's own guidance caused the move.

I'd also drop "Q3 has been seasonally strong." Nothing in our four reports supports it, and the only Q3 evidence is last year's quarter, measured off a trough. Leaning on 14 bullish tags and zero bears is thin too. Half of those posts only mention AMD in passing, and the same feed has people discussing 525 put hedges.

The conservative analyst overcorrects. Calling a 20 percent gap a central bad case isn't supported by what we have. The 18.6 percent slide was a grind over about two weeks, and the only gap down in the window was about 5 percent at the open on 10-15. The biggest gaps in the data are upward, in late April and early May. Twenty percent is a good stress test, and that is how the trader's plan uses it, but it isn't the median bad outcome.

His point that the triggers fire after the damage is true but beside the point. The plan's protection is the size cap, which is computed on a 20 percent gap straight through every level. The triggers are thesis checks, not the loss limit.

His reward-to-risk comparison is also rigged. He sets one firm's $620 target against the 200 SMA and the weekly stop as if they were the same kind of number. A target is a point estimate, and supports are worst-case markers. The stock closed at 649 two weeks ago, above that target. Susquehanna is catching up to price, not capping it.

Selling most of the position at market today means selling on a possibly partial bar, the lowest volume in 30 sessions. It also comes right after a completed daily TD-9 buy setup, with the daily z-score at -1.16. That is a poor place to be a forced seller. The 10 EMA at 566 is a sensible place to sell into, and it's 2 percent away, not 8.

Cutting to a quarter of standard also gives up most of your participation in a company growing 50 percent, with expanding margins, a 20 percent FCF margin, and weekly and monthly trends still up. If the print is good, you'd be sitting at 1.25 percent weight. That's a real cost, not a free hedge.

On rates, a December hike priced at 78 percent is a headwind, but it's a slow variable that is unverified against FRED and already in a price that's up 240 percent. It explains the low grid return. It isn't a reason to dump three-quarters of the position this week.

Here is the middle path. Target about 40 percent of standard before the print. The trader's own 2.5 percent cap implies a standard weight of 5 percent. So 40 percent is a 2 percent weight, and a 20 percent adverse gap costs 0.4 percent of the portfolio. That budget sits between your two and still keeps 40 percent of the upside.

Get there in three pieces: - Sell about half the reduction right away, taking 100 down to roughly 70 over the next session or two. Do it before Meta reports if that is imminent. - Put another 15 or so on limits between 566 and 580, around the 10 EMA, instead of 599 to 617. - Sell whatever remains at market on the close before the Nov 3 report, as the trader already planned. The target is then reached by date, not by price.

Drop the 485 bid before the print, but keep the idea behind it. After the print, with guidance in hand, rebuild toward standard in two or three steps. That means growth holding around 45 to 50 percent, the weekly stop at 453.83 intact, and the stock either pulling back to the 50 SMA or closing above 617.04 on rising volume. That gives the aggressive analyst his framework for buying dislocation without buying a gap blind.

Nobody has checked the options-implied move, and it should be checked before the last tranche. If it's around 10 percent or less and puts are reasonably priced, a put spread on part of the position is a cleaner way to cut gap risk than selling shares, especially for anyone with large embedded gains. If it's 14 to 15 percent, insurance is expensive and selling shares is the better route.

All of this is relative to your own position. If you're already at or below about 40 percent of standard, there's nothing to sell. Just don't add at 555.

Keep the 453.83 weekly exit as a thesis check, not as protection. A 20 percent gap would blow through it, and if that happens you should judge the guidance before dumping the remainder on the first print.

This plan takes the conservative analyst's points on timing and the bid, and the aggressive analyst's points on keeping a core and having a re-entry plan. It doesn't depend on a rally, a seasonal pattern, or a 20 percent gap being the norm.

FINAL TRANSACTION PROPOSAL: SELL Neutral Analyst: Before I argue, I want to say how close the three of us now are. We all say sell. We've all dropped the 599 to 617 limits, the $485 bid, the seasonality line and the 50 SMA reclaim. The destination is somewhere between a quarter and half of standard. What's left is sequencing, one shaky calculation on each side, and a few things I got wrong myself.

I'll start with my corrections. The conservative is right about Meta. Today is a Wednesday, the news report says Meta was due this week, and the retail posts about an after-hours Meta reaction are timestamped between about noon and 1:20 Eastern. Those same timestamps, plus one poster quoting 555.55, suggest our close is a midday snapshot. That's my inference, not something the tools say. If so, today's session still has hours left, and the 10.1 million share volume is almost certainly an unfinished bar. The trader's rationale still cites that volume as a bearish tell, so it should come out of the thesis, not just be discounted. And "check the Meta date today" doesn't fit with "sell over the next session or two." Look it up first. If Meta reports tonight, I'd take the first tranche to about 60 of 100 before the close, not 70.

I'd stop at 60 because the conservative's four-times arithmetic doesn't hold up. He says waiting for a limit near 573 only pays if reaching it is four times likelier than a 12 to 15 percent gap down. But a gap up fills that limit at the open, around 620, not 573. The limit loses only if the stock falls and wins if it rises, so his test is a directional call, and nobody has argued that expected return is negative. The real question is how many shares to carry through Meta, and that's a budget question. A Meta reaction like the 5 percent gap on the 15^th, or a bit more, on ten extra points of standard at a 5 percent weight is about 0.04 percent of the book. That isn't worth a rewrite.

His 30 percent stress test has a similar problem. It stacks a 20 percent gap on a week of follow-through and treats the week as something you can't act on. The gap is the part you can't act on. After the print you can, so write the rule now: if the stock gaps down more than about 10 percent on guidance, cut the rest to a quarter within two sessions, with no open-ended judging. That keeps his protection without sizing for the stacked path.

On sizing, both of you are arguing the wrong unit. The trader's rule is the lower of half of standard or a weight where a 20 percent gap costs 0.5 percent of the book. My 5 percent standard was an assumption, as the conservative said, and 40 percent only falls out if that's true. At an 8 percent standard the cap binds near 31 percent. At a 3 percent standard half binds. So the 25, 40 and 50 you're trading are each right for some book. I'd keep the trader's rule, move the stress from 20 to 25 percent as a midpoint between the two of you, and target the lower of half of standard or a 2 percent weight. If your standard is 5 percent, that's my 40.

On the put spread, the conservative scores some points and loses others. Implied volatility falling after the print is just the price of insurance, which is why the implied move has to be pulled before anyone buys one. But he's right that a spread covers part of the loss, and that my taxable-gain argument assumed a holder we know nothing about. Take a spread from about 525 to 450. Against a 20 percent gap to 444 it pays 75 on a 111-point loss, about two-thirds. Against a 30 percent drop it covers under half. So count a hedged tranche at half credit, and only use one if the implied move is on the cheap side.

I also concede the rebuild gate. At Q2's 12.5 percent sequential pace, Q3 revenue is about $13.0B, which is only +40 percent year over year, because Q3 FY25 was the one 20 percent sequential quarter in our series. The headline slows on arithmetic alone. That also means the trader's "40 versus 50 percent growth" framing should be rewritten. The gate should be dollars and margins: revenue near $13B, gross margin around 53 to 54 percent, operating margin near 17 percent, and a Q4 guide against consensus, which nobody has yet. I'd add a price condition, since a good print into a stock that already gapped up isn't the same buy as one into a stock that's down 12 percent. The conservative is also right that I leaned on the 20 percent trailing FCF margin when the latest quarter was 13.5 percent. I'd stop citing it, but I wouldn't move the target either way on it. Without the 10-Q we can't tell whether it's a capex ramp or something worse.

The aggressive should hear that "the data can't pick 40 over 50" is true, but it argues for the lower-of rule, not for staying at 75 while a likely Meta night is unchecked. He also keeps the 599 to 617 tier in his plan, and the trader's text still carries the $485 bid, the seasonality claim and the light-volume bounce. Cut all of those.

So my plan has a lookup phase and an execution phase. Today, pull the Meta report time, AMD's confirmed report date and time (Nov 3 comes only from a Motley Fool headline), the options-implied move, and consensus. Then sell to about 60 before the close if Meta is tonight, or to about 70 over a session or two if not. Put roughly 15 on limits at 566 to 580, and finish at market before the print, keyed to the confirmed time. The final target is the lower of half of standard or a 2 percent weight. No new money at 555, no $485 bid, and anyone already at or below target does nothing.

FINAL TRANSACTION PROPOSAL: SELL Neutral Analyst: We're now arguing over a few hundredths of a percent of the book on sizing, so I'll spend my time on the post-print rules and on a few places where each of you is still too tidy.

First, what I'm taking. The conservative is right that tranches should be written as fractions of the distance to target. Sixty today plus fifteen on limits lands at forty-five, which either overshoots a fifty target on a bounce none of us can read or leaves you five short if the target is forty. So size the limit tier to the gap between sixty and the target, cancel it once you arrive, and let the dated market order finish the job. I'll also take sixty today unconditionally. My "sixty if Meta is tonight, seventy if not" resolves within the hour anyway, because the lookups take minutes and the bar is still printing. I don't buy the conservative's reason, though. "There may be more events, so sell sooner" has no stopping point. The target and the budget are what stop it. Put Microsoft, Alphabet, Amazon and the FOMC on the lookup list, but remember they come from memory, not from our reports.

The aggressive restated his sell-to-60-or-70 plan without absorbing the sizing point. His two-sided post-print rule has a bigger problem. "If the numbers clear the bar and the stock gaps on a sector or Meta-driven move, hold and buy in steps" invites the attribution trap. We have a live example in the 28.8 million share gap on the fifteenth that nobody can explain. His bar is also anchored to Q2: 53.8% gross margin and 17.25% operating margin. Forward EPS of $15.76 is nearly three times the Q2 GAAP run-rate, so the market's bar sits well above Q2. A company can clear a Q2-anchored gate and still get sold. The dollar gate should be necessary for rebuilding but not sufficient. Consensus has to decide the rest.

The conservative's "cut fast, rebuild slow" is right in spirit and too mechanical as written. He says a guide at or below consensus takes you to the floor in the first session. An in-line guide is the base case, and we'll already have trimmed once for the same valuation worry. Cutting again on an in-line print double-counts it, in a stock whose weekly and monthly trends are both up. I'd use the one attribution-free signal we have, which is price.

A guide below consensus takes you to the floor in the first session.
If the guide is in line or better but the stock opens down more than about 7%, cut half the distance to the floor immediately. Cut the rest if it then closes below the gap-day low or the 50 SMA. The 7% comes from how the 5.4% gap on the fifteenth extended, so it's a rule of thumb, not a measurement. That also closes the hole he found at 5 to 10%.
If it holds, stand pat.
If it gaps up, don't chase. Rebuild in steps only after it holds a level for a couple of sessions.

That keeps his protection without making us guess why the stock moved.

The same trap is in the trader's own Meta trigger. "Cut to 25% if Meta's commentary signals AI-capex digestion" will be read tonight in whatever direction suits each person's position. Write it in numbers against Meta's last stated capex range, which someone has to pull. A cut or a flattened guide fires the trigger and anything else doesn't.

On hedges, the conservative is right that the put spread is gross of premium and pays nothing in the 555 to 525 dead zone, so I'd give it less than half credit until we see the chain. I'd go further than he does on the call spread and drop it from the baseline. Sell fifteen points of shares, then buy calls carrying a fraction of that delta, and you've made a smaller sale with more moving parts and no implied move to price it against. Both spreads are optional, for someone with a specific reason and a chain in hand, with delta counted as shares. The baseline is shares.

The conservative is also right that the old gaps up came off closes near 348 and 421, not 555. But that isn't a reason to hold less. The reason for a core is what we can see: growth accelerating, about 40% incremental operating margin, a strong balance sheet, and weekly and monthly trends still up. Against that sit a price that needs a roughly four-fold earnings step-up and a few things we can't see, like the $0.27 of unexplained EPS and a Q2 FCF drop with capex doubled. We keep a core because we can't tell which way it breaks, not because we expect a gap up. The 0.5% cap bounds what that costs.

Here's the plan.

This hour, pull Meta's report time and prior capex range, AMD's confirmed date and time, the implied move, consensus, and the other hyperscaler and FOMC dates.
Sell to about sixty today.
Set the target at the lower of half of standard or 0.5% divided by twice the implied move, with a 20% floor. At a 5% standard weight that's fifty at a 10% implied move and about thirty-three at 15%.
Put limits at 566 to 580 sized to the gap to target, and cancel them on arrival.
Sell whatever remains at market on the confirmed report day.
Add no new money at 555, no $485 bid and no 599 to 617 tier. Anyone already at or below target does nothing, and a strict capital-preservation mandate goes flat.

The moderate version isn't a compromise between our numbers. It's the one plan that doesn't need a rally, a gap, or a story about why the stock moved to be right.

FINAL TRANSACTION PROPOSAL: SELL Neutral Analyst: Let me start with what I'm taking. The conservative's catch on my open-only rule is right. I calibrated the 7% on how the fifteenth extended, from a 5.4% open to about 7.4% under the old close by the sixteenth, and then wrote the test at the open, where that same episode would have read 5.4% and stayed silent. Measure where you calibrated. I'll also take his point that the aggressive's "closes under that day's low" can't work as written, because a close can't print below its own session's low. It has to be a later close under the gap-day low. And the aggressive is right that the 50 SMA clause was dead weight.

My first push goes to both of you. The 5% and 7% thresholds, the 1-2% in-line band, and one versus two closes under 523.78 are all rules of thumb calibrated on a single gap nobody can explain. I'm fine using them, because being wrong in either direction costs a few hundredths of a percent of the book, assuming that 5% standard weight, which is itself an assumption. But we shouldn't present them as measurements, and we should stop adding branches. Plenty of outcomes, like an in-line guide with the stock inside plus or minus 5%, trigger nothing under any version of this tree. The plan's value sits in the pre-print size and the tails, so spend the care there.

On the pre-print trigger, the conservative is right that dropping the 523.78 cut silently isn't a decision. I'd keep it as a half-step. One close under 523.78 takes you half the distance to the floor, and a second close finishes it, which is the same shape as the post-print rule. I won't make it one-shot because we're selling on price against return. At 524 the stock is about 6% less stretched, so if it gets there without news, the case for a full cut is weaker, not stronger. The tape can inform the decision, but it shouldn't be the only voice.

On the implied move, I like twice the implied move with a 20% floor, and I like the conservative's re-pull on the last day. But there's a measurement trap. A straddle bought today on an expiry just after November 3 prices roughly two weeks of ordinary trading plus the print, and on a name with a 4.8% ATR the ordinary part isn't small. The raw number overstates the earnings move today and decays mechanically toward the event move. So "take whichever gives the lower weight" would always pick today's inflated reading. Use the earnings-expected-move figure, or strip out the base volatility, and pull the same measure both times. Only act on the re-pull if it moves the target by more than a few points of standard, or you're trading noise.

Here's what I think the group has skipped, and it matters for the rebuild. The target is the lower of two caps that do different jobs. The 0.5% budget over the stress gap is an event cap, and it expires with the print. After guidance you're sizing for ordinary but still large beta-2.45 volatility, not a 20-30% gap. Half of standard is a valuation cap, and a clean print doesn't automatically touch it. So the aggressive is half right that the core survives a good print, but the event cap does loosen, so there is room to add. I was also too loose earlier when I said to rebuild toward standard. Rebuild in two steps, each after the stock holds above the gap-day low, up to the valuation cap. Go past it only if someone reruns the return grid on post-print numbers and it clears the hurdle. The conservative's "rebuild slow" needs a clock, or it quietly becomes never. The dollar gate is necessary, not sufficient, and price completes it.

So the post-print rules, written now, are these. A guide below consensus takes you to the floor, the trader's 25% of standard, in the first session. If consensus can't be sourced, price decides alone. Otherwise, an open more than 5% down takes half the distance to the floor. The rest goes if either of the first two closes is more than 7% under the pre-print close, or if a later close prints under the gap-day low. If it's in line or better and holding, do nothing. If it gaps up, don't chase.

Two small items. The conservative's point about look-through exposure is right, but it belongs in the definition of standard, applied once. Count funds and correlated holdings before you compute the target, not as a separate haircut. And Meta's trigger needs a number against its prior capex range. A cut or a flattened range fires it, and a flat or higher range doesn't.

The plan is the one we've converged on. Sell to about sixty today whatever Meta does, and run the lookups in parallel: Meta's time and prior capex range, AMD's confirmed date and time, the earnings-implied move, consensus, and the other hyperscaler and FOMC dates. Those dates are from memory, not our reports. Set the target from the two caps. Put limits at 566 to 580, sized to the gap to target and cancelled on arrival. Finish at market on the confirmed report day. Anyone already at or below target does nothing, and nobody adds at 555. A strict capital-preservation mandate goes flat.

FINAL TRANSACTION PROPOSAL: SELL Neutral Analyst: Most of this round is right, so I'll take that first and then spend my time on three things that still change an order.

Scaling the post-print triggers to the earnings-implied move beats any hand-picked 5 and 7. The conservative's arithmetic shows why: the cost at each trigger stays at roughly 0.125 to 0.175 percent of the book whatever the implied move turns out to be. I'll also own that the guide-below-consensus cut was in my last plan, and the aggressive is right to retire it. Consensus is a sell-side number and the stock moves on the buy-side bar, so my rule would have sold us into strength on a story. Consensus becomes a veto on rebuilding. The free-standing close test fixes a real dependency in my rule. Marking the target to the last-day price fixes the confusion between share count and weight. The 10-Q and the reason for the 15^th belong on the lookup list, since both bear on the four-fold earnings step-up we've all been leaning on.

My first push is on the rebuild. The aggressive's re-entry line, two closes above the gap-day low, mirrors his cut line, one close under the gap-day low. A stock chopping around that level sells you out and buys you back at about the same price, paying the spread both ways on a 4.8% ATR. A tape rule needs a band between exit and re-entry. The simplest fix adds no branch: never buy back below one ATR above the price of your last cut, about 5 percent at today's volatility. The clock still runs, so slow doesn't become never. It just can't run backwards through your own sale.

Second, we haven't said how much there is to rebuild. Take the 5 percent standard weight we keep borrowing. At an implied move of 10 percent or less, the loss cap and half of standard are both 2.5 percent. The pre-print target is then already the ceiling, and a clean print gives you nothing to add. Rebuild room appears only if the implied move comes back above about 10 percent, and then it's the gap between the event cap and half of standard. A larger standard weight makes the room appear sooner and a smaller one removes it. This also settles the conservative's 18 to 19 percent post-event stress. It gives about 2.7 percent, which is roughly where half of standard already sits, so we agree on the ceiling. Rerunning the grid at the new price should work in both directions. A 12 percent gap down improves the grid, but I wouldn't lift the cap above half of standard unless it clears the 14 to 16 percent hurdle.

Third is Meta. I wrote a contradictory line last round: "a cut or a flattened range fires it, and a flat or higher range doesn't." Flattened and flat mean different things to different readers, which is exactly the freedom we meant to remove. Write it as the midpoint of the full-year capex range coming down. That fires the trigger and nothing else does. Then hold it to the standard we just applied to consensus. It's a read-through, and AMD's open is the market's reading of it. If the range is lowered and AMD opens down more than a couple of percent (a rule of thumb), take the half step, the same shape as the 526.83 close. If the range is lowered and AMD shrugs it off, do nothing, because cutting into strength on a headline is the attribution trap again. The original cut to 25 percent on a Meta signal alone should go.

A note on proportion. At a 5 percent standard, selling to sixty today moves about two points of book weight into cash. The whole post-print tree we've spent three rounds on moves at most about a point and a quarter, and the follow-through it guards against costs roughly a tenth of a percent. It's written well enough now, so stop adding to it. The number everything hangs on is the 4 to 6 percent grid, and all any of us has seen is its average. It's a three-year, scenario-weighted figure being used to time a thirteen-day event. That's a good reason to treat half of standard as a cap rather than defend it to the decimal, and a good reason for the implied move and the 10-Q to matter more than another trigger tweak.

Here's the plan as I'd sign it. Sell to about sixty today and run the lookups in parallel. Set the target as the lower of half of standard or 0.5 percent over twice the earnings-implied move, with a 20 percent floor, marked to the last-day price and ratcheting one way. Put limits at 566 to 580, sized in weight to the gap and cancelled on arrival. Finish at market before the confirmed report, with the Nov 2 close as the fallback and the date confirmed today, not on Oct 30. Before the print, a close under 526.83 takes the half step. After it, price decides. An open down by half the implied move (5 percent floor) takes half the distance to the floor. Any of the first three closes down seven-tenths of the implied move (7 percent floor) takes the rest. Consensus only vetoes rebuilding. Rebuild in thirds on a clock, never within an ATR of your last sale, and only up to the rerun cap. No new money at 555. Anyone already at or below target does nothing, and a strict capital-preservation mandate goes flat.

FINAL TRANSACTION PROPOSAL: SELL

V. Portfolio Manager Decision¶

Portfolio Manager¶

Rating: Underweight

Executive Summary: Underweight AMD: cut exposure to a loss-budgeted core and keep it, rather than exiting. Today, sell to about 60% of a standard allocation whatever Meta's report date turns out to be. Then work down to the final weight, which is the lower of (a) 50% of standard or (b) 0.5% of the portfolio divided by twice the earnings-implied move (minimum 20% stress; use 25% if the implied move cannot be sourced). Use limits at $566-580 sized to the remaining gap, and finish at market by the close before the confirmed report (fallback: the Nov 2 close). No new money at $555, no $485 bid, no $599-617 limit tier. Before the print, a close below $526.83 takes half the distance to a 25%-of-standard floor and a second close finishes it; a weekly close below $453.83 exits the remainder; after Nov 3, price-based rules decide cuts and rebuilds.

Investment Thesis: DECISION AND MAPPING. The trader's 'Sell' is explicitly a trim to a core, not an exit. All three risk analysts converged on direction and differ only by a few hundredths of a percent of the book on size, so this is Underweight on the 5-tier scale. Sell (exit or avoid) is reserved for a strict capital-preservation mandate, and nobody showed negative expected return.

WHAT DECIDED IT: PRICE AGAINST RETURN, NOT THE BUSINESS. Valuation at $555.60: market cap $907.2B, 141.8x TTM GAAP P/E, about 22x sales, about 95x EV/EBITDA, 103-108x FCF. The $15.76 forward EPS (35.3x) implies about $26B of net income against $6.4B TTM, and its basis is unspecified. The Research Manager's three-year grid averages about 4-6% a year from $555. That is close to cash for a beta-2.45 stock and far below a 14-16% hurdle, and the bull conceded he would not buy here.

Why the headline growth rate is likely to slow: - Q2 FY26 revenue grew 50.1% YoY. - Repeating Q2's 12.5% sequential pace gives about $13.0B in Q3, only about +40% against the $9.25B Q3 FY25 base. - That base contained the only ~20% sequential quarter in the series.

Earnings-quality flags: - Net income ($2,297M) exceeded operating income ($1,990M), leaving about $0.27 of EPS unexplained against a run-rate of about $1.11. - Q2 FCF fell 39% quarter over quarter to $1.56B as capex doubled to $808M. - Non-cash current assets built by about $3.4B, with no inventory split. - About $4.2B of H1 investing outflows went to non-capex uses that are not itemized.

The tape: - MACD -1.94 with a -13.11 histogram. - Price below the 10 EMA (566.57); daily SuperTrend down with a 617.04 flip level. - An unexplained 10-15 gap down on 28.8M shares. - An 18.6% drop in eight sessions with no earnings.

I am discarding today's 10.1M-share volume as evidence, because the bar is likely partial. I am also discarding the seasonality claim and the 50 SMA reclaim (about half an ATR).

WHY NOT SELL OUTRIGHT. The weekly SuperTrend is up (stop 453.83, 22% below) and the monthly is up (336.14). The 50 SMA is rising (541.16). Margins are expanding (GM 53.8%, operating margin 17.25%). The balance sheet is strong (current ratio 2.61, D/E 6.4%). The grid return is thin but positive. We keep a core because we cannot tell which way this breaks, and the size cap bounds what that costs.

SIZING MECHANICS (from the converged debate): - Count look-through and correlated AI/semiconductor fund exposure inside 'standard'. - At a 5% standard weight, the final weight is 2.5% at a 10% implied move, 2.0% at 12.5%, and about 1.7% at 15%. - Pull the earnings-specific implied move the same way today and again on the last pre-print day. Use whichever gives the lower weight, mark it to the last-day price, and ratchet one way only (no buybacks). - The trader's clause to stop trimming on a close above 617.04 is dropped on purpose, because a rally makes the gap more costly and the grid worse. - Meta trigger: if Meta lowers the midpoint of its full-year capex range AND AMD opens down more than about 2%, take the half step. Otherwise do nothing.

POST-PRINT RULES: - Open down more than half the implied move (5% floor): cut half the distance to the floor. - Any of the first three closes more than 0.7x the implied move (7% floor) below the pre-print close, or a later close below the gap-day low: go to the floor. - Thesis broken (revenue under $12.5B, GM under 52.8%, operating margin under 15.5%, or an inventory-led build): exit. - Rebuild only if revenue is about $13.0B or more, GM about 53% or more and operating margin about 17% or more. Consensus can veto a rebuild but is not a cut trigger. - Rebuild in thirds, starting after two closes above the gap-day low, and never within about 1 ATR (about 5%) above the price of the last cut. - Cap the rebuild at 50% of standard unless the grid, rerun at the new price, clears the 14-16% hurdle.

WHAT WOULD CHANGE THE CALL. Upgrade to Hold or Overweight: a rerun grid at 10% a year or better after a print with revenue of $13.9B or more, GM of 54% or more, operating margin of 19% or more and OCF of $2.5B or more. Downgrade to Sell: the thesis-break levels above, a red-flag 10-Q, or a weekly close below 453.83.

MISSING EVIDENCE (look up today): - Earnings-implied move and consensus. - Confirmed AMD report date and time (Nov 3 comes from one headline). - Meta's report time and prior capex range. - The Q2 10-Q line items. - The cause of the Oct 15 gap. - Hyperscaler and FOMC dates (from memory only). - Verification of the 78% December-hike pricing.

PRICE TARGET (method 1, valuation objective). Apply the midpoint of the Research Manager's 4-6% a year return grid over 12 months: 555.60 x 1.05 = 583.38, about $583. The 4-6% range gives $577.82-588.94. Cross-checks: the target sits between the 10 EMA (566.57) and the Bollinger middle band (599.13), and below the single sell-side $620 cited in the debate. The 14-16% hurdle would require $633-644, and the shortfall against that is the core of the Underweight call.

Current Price: 555.6

Price Target: 583.0

Confidence: Medium

Time Horizon: 12 months