Trading Analysis Report: META¶

Analysis date: 2026-10-21
Rating: Underweight
Generated: 2026-10-21 12:21:29
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: 707.95 Price As Of: 2026-10-21

The quote currency isn't stated in the tool output. I'm assuming USD because the exchange is NMS (Nasdaq).

META: Technical Report (as of 2026-10-21)¶

0. Data integrity notes¶

Last-bar mismatch: The get_stock_data CSV shows a 2026-10-21 close of 707.67 on volume of 11,002,526. The verified snapshot shows 707.95 on 11,004,932. Open, high and low match in both (697.74 / 713.22 / 692.38). I used the snapshot throughout. If the 10/21 bar is still forming, that would explain the gap, but the tools don't say so. Treat today's close and volume as provisional.
MACD rounding: The macd tool returns 10.4618 and the snapshot shows 10.47. That's a rounding-level difference and doesn't change the read.
Matching values: RSI, ATR, the 50 SMA and the 200 SMA agree with the snapshot.
Suspect CSV bar: The 2025-12-12 row has a high of 708.60 against an open of 647.60 and a close of 642.05. It looks like a bad print. It's outside the current indicator windows and doesn't affect this analysis.
Not pulled separately: The 10-EMA, Bollinger bands, MACD signal and MACD histogram come from the verified snapshot.

1. Indicator selection¶

I chose 8 indicators that cover different jobs and don't overlap: - Trend: close_50_sma, close_200_sma and supertrend (weekly, monthly and daily). - Momentum: macd and rsi. - Volatility: atr. - Volume: obv. - Stretch: z_score.

2. Big picture: a pullback inside a larger uptrend¶

Base: META made two lows near 520–525. The first was a 519.37 intraday low on 2026-03-27. The second was a 524.08 intraday low on 2026-07-30.
Breakout: On 2026-09-21 the stock gapped up from a 665.23 close to a 741.25 close, about +11.4%. Volume was 48.7M, against roughly 10–20M on normal days. My tools don't identify the catalyst.
Peak: The rally peaked at 777.59 (close) and 779.82 (intraday high) on 2026-09-24.
Pullback: The stock then made a lower high of 771.27 intraday on 10/13 (762.50 close). From there it fell to a 681.05 intraday low and a 694.25 close on 10/20. That is about −10.7% from the peak close.
Today: 10/21 closed at 707.95, a bounce of roughly +2% from 10/20.

3. Trend: bullish on every timeframe¶

Moving averages: Price is above both averages.
Price is about 7.0% above the 50 SMA (661.41).
Price is about 12.1% above the 200 SMA (631.78).
Both are rising. The 50 SMA has been climbing about 2–3 points per session over the past week.
Golden cross: The 50 SMA crossed above the 200 SMA on 2026-10-06. On 10/5 it was 627.61 against 627.82. On 10/6 it was 630.52 against 628.27. The cross is recent and driven by the September surge, so it's a lagging confirmation and not an early signal.
SuperTrend, in tier order:
Weekly (primary): UP, stop 569.73, close 24.26% above it.
Monthly: UP, stop 522.46, close 35.50% above it.
Daily: UP, stop 681.41, close 3.90% above it.
The timeframes don't conflict. The daily stop is only 26.54 points (about 1.1 ATR) below the close, so it's the first line to break in a deeper pullback. A daily close below 681.41 would flip the daily trend. The weekly and monthly trends would stay up.

4. Momentum: fading from an extreme, now neutral¶

MACD:
The line is 10.47, below its signal of 19.23. The histogram is −8.76.
The line peaked at 42.93 on 9/25 and has fallen almost every session since, with a small uptick on 10/13.
It's still above zero, so this is a momentum cooldown and not yet a bearish regime change.
A bullish re-cross needs the histogram to turn positive.
RSI: It is 50.21, down from 80.55 on 9/24 (overbought) and 77.87 on 9/21. It bottomed at 46.43 on 10/20 and has not touched 40 in the lookback window. That is a reset to neutral with no oversold signal, so RSI gives no edge for a mean-reversion entry.
10-EMA: Price is 10.02 points (about 1.4%) below the 10-EMA of 717.97. The short-term trend is slightly negative, and 10/21 did not reclaim the 10-EMA. Its high of 713.22 stayed below it.

5. Volatility: moderate, with gap risk the ATR can't capture¶

Bollinger Bands: The middle band is 730.06, the upper is 770.29 and the lower is 689.82. Price is below the midline and about 22% of the way up the band range, 18.13 points above the lower band. The midline is overhead resistance for now. The lower band sits near the 10/20 low area.
ATR: It is 24.31, about 3.4% of price. It ranged from 22.10 (10/9) to 28.01 (9/28) over the window, so volatility is normal and not expanding.
Gap risk: The CSV shows four large gap days, all with heavy volume:
2025-10-30: close 664.22 against 749.13 the day before, about −11%, on 88.4M shares.
2026-01-29: 736.41 against 667.01, about +10%, on 59.9M.
2026-04-30: 610.86 against 667.97, about −9%, on 52.8M.
2026-07-30: 538.60 against 585.15, about −8%, on 42.3M.
These fall in the last week of Jan, Apr, Jul and Oct. If the quarterly pattern holds, another event could be close. My tools don't confirm an earnings date, so verify it. Moves of 8–11% are roughly 2.3–3.2 ATR, so ATR-based stops would not protect against an overnight gap.

6. Volume: pullback volume is light, but OBV is weak¶

Pullback volume: On the 9/21–9/24 breakout, volume was 48.7M, 28.1M, 30.8M and 34.8M. The pullback sessions from 10/14 to 10/20 ran 16–21M. Today's 11.0M is the lowest in the 30-row snapshot window and may be incomplete. The pullback is happening on much lighter volume than the thrust, which is constructive.
OBV: It is 329.7M, near the low of the window. It peaked at 421.4M on 9/24. It fell from 405.7M on 10/13 to 318.7M on 10/20, a drop of about 87M over five straight down closes, then recovered 11.0M today.
Reading OBV: OBV is confirming the pullback and not diverging from price. The decline reflects consecutive down days at moderate volume and not a capitulation spike. The slope is negative and has only just ticked up. A rising OBV with price reclaiming 718–730 would be the confirming sign.

7. Stretch: not stretched¶

Weekly (primary): +1.19.
Monthly: +0.82.
Daily: −1.10.
No tier has |z| ≥ 2, so there is no statistical mean-reversion signal. The higher tier is still modestly above its 20-bar mean, and the daily negative reading just reflects the recent dip.

8. Levels from the tool data¶

Resistance (nearest first):
717.97 (10-EMA).
About 718–725: the closes of 718.67 on 10/9 and 725.18 on 9/30 sit in this zone.
730.06 (Bollinger midline).
762.50 (10/13 close).
770.29 (upper band) and 771.27 (10/13 high).
777.59 / 779.82 (the 9/24 close and high).
Support (nearest first):
689.82 (lower band) and 692.38 (today's low).
681.41 (daily SuperTrend stop) and 681.05 (10/20 low).
665.23–680.30: the 9/18 close and the 9/21 open, the base of the breakout gap.
661.41 (50 SMA, rising).
631.78 (200 SMA).
569.73 (weekly SuperTrend stop).

9. Scenarios¶

Constructive: A close back above the 10-EMA at 717.97, then above the 730.06 midline, would improve the picture. Histogram improvement and OBV turning up would add confirmation. The next test zone would be 762–780.
Cautionary: A close below 689.82 and then below 681.41 would break the daily SuperTrend and put 665–680 and the 50 SMA at 661.41 in play. MACD would likely cross toward zero. Weekly and monthly SuperTrend would remain up unless price fell much further.
Current posture: Higher-timeframe trend is bullish, short-term momentum is neutral to soft, and there are no oversold or overbought extremes. Timing signals are mixed: price is below the 10-EMA and the midline with a negative MACD histogram, but trend structure is intact.

10. Limits¶

The catalyst for the 9/21 gap and the next earnings date are not in my tools.
I did not run ADX, TD-9, MFI or KDJ. TD-9 and ADX would be reasonable follow-ups to check for exhaustion and trend strength.
All percentages above are arithmetic on the verified prices.

Summary table¶

Item	Reading (source)	Implication
Close / date	707.95 on 2026-10-21 (snapshot)	CSV shows 707.67 and a slightly different volume; the last bar may be provisional
50 SMA / 200 SMA	661.41 / 631.78, both rising	Price 7.0% and 12.1% above them; bullish structure
Golden cross	50 SMA crossed above 200 SMA on 2026-10-06	Lagging confirmation of the September surge
SuperTrend (W / M / D)	UP 569.73 / UP 522.46 / UP 681.41	All tiers up; daily stop is 3.90% (about 1.1 ATR) below price
10-EMA	717.97	Price 1.4% below it; short-term resistance
MACD / signal / hist	10.47 / 19.23 / −8.76	Momentum cooling from the 42.93 peak on 9/25; still above zero
RSI	50.21 (80.55 on 9/24, 46.43 on 10/20)	Neutral; no oversold or overbought signal
Bollinger (L / M / U)	689.82 / 730.06 / 770.29	Price below the midline, about 22% up the band range
ATR	24.31 (about 3.4% of price)	Normal volatility; gap risk is larger than ATR implies
OBV	329.7M, down from 421.4M on 9/24	Confirms the pullback; watch for an upturn
Z-score (W / M / D)	+1.19 / +0.82 / −1.10	No stretch on any tier
Key resistance	717.97, 730.06, 762.50, 779.82	Reclaiming 718–730 is the first improvement
Key support	689.82, 681.41, 665.23–680.30, 661.41, 631.78	A close below 681.41 flips the daily SuperTrend
Event risk	Four ≥8% gap days, all in the last week of Jan, Apr, Jul and Oct	Verify the earnings date; ATR stops don't cover gaps

Sentiment Analyst¶

Overall Sentiment: Mixed (Score: 5.0/10) Confidence: Low

Scope and data limits. Window: 2026-10-14 to 2026-10-21, for META (Meta Platforms, Inc., NMS). Reddit was disabled by config, so there is no Reddit read at all. The news feed has 19 headlines with no article bodies, so only titles can be assessed. StockTwits has 30 messages, of which only 9 carry a sentiment label (6 Bullish, 3 Bearish) and 21 are unlabeled. Q3 earnings are due after the close today (2026-10-21), so the outcome is unknown and every signal below is pre-event positioning. Confidence is low.

1. Source-by-source breakdown

News (Yahoo Finance feed, 19 headlines). The framing is event-driven and split between a legal shock and earnings anticipation. - Legal/regulatory: Reuters, AP, Axios and MT Newswires all report that Meta was ordered to pay New Mexico $25 billion in the Cambridge Analytica-era privacy case. The Reuters headline says a jury found the company misled consumers about data privacy. This is the one hard negative event in the window, and it landed on earnings day. It is covered by four outlets, which makes it the most heavily reported item. Appeals and a final payment timeline are not covered in the headlines provided. - Earnings framing: Yahoo Finance's live blog says Q3 'will test Wall Street's stomach for AI spending'. Barron's says 'If There's a Market Bubble to Pop, Meta Earnings Hold the Pin' and 'The Magnificent Seven Need to Tell a Story Along With an Earnings Beat'. IBD says 'Meta Stock Climbs With Earnings On Deck'. The tone is anticipatory and wary rather than outright negative. The core question is whether AI capex will be rewarded. - Positive/neutral: Barron's says Meta and Tesla are 'crushing their average earnings day returns' and credits SpaceX. TheStreet calls Zuckerberg's latest move his 'boldest AI move yet'. The market-wrap headlines (Dow, S&P 500 and Nasdaq gaining as tech rallies) show a constructive tape on 10-21. The body text of these items is not available, so I cannot say what drove the moves or what the 'AI move' is. - Peer context: 24/7 Wall St. notes Alphabet is down about $860B from its peak while Apple, which avoided the AI race, is near all-time highs. That is a cautionary backdrop for AI-capex-heavy names such as META.

StockTwits (30 messages). The labeled split is 6 Bullish (20%) vs 3 Bearish (10%), or 67/33 among labeled posts. Because 21 of 30 messages (70%) are unlabeled and only 9 are labeled, this is a thin sample and the ratio is only a weak lean. - Bullish: @FatmanEZ ('740 tonight'), @r3y ('beat.. 750-800'), @BigCheeze2014, @EquityEdgeTrader, @StockStinjo and @thefourwind (a SOXL post expecting +5% after META earnings, so it is semi-related). @EquityEdgeTrader gives levels: META +1.9% to 707.71, bouncing off a 692.38 low, still under the 20-day (730.05), with 713.22 as the next hurdle, then 730. A loss of 692 reopens 681. - Bearish: @2026Bear ('how much revenue was circled to meta this quarter… biggest bubble in history', a circular-revenue/AI-bubble thesis), @luffie, and @StrongArmor ('713 double top? red candles'). - Unlabeled but informative: @Apex_101 ('765 after earnings') is a bullish target. @Holdmybagz asks 'back to 600 after earnings?', which is a bearish worry. @Techdamentals frames META's data-center and chip spending as capex paid years before returns, pointing to a space company's $19.6B quarterly spend as the comparison category.

Reddit. Not available (disabled by config). No read is possible, and no inference should be drawn from its absence.

2. Cross-source divergences and alignments - Legal shock vs retail shrug: the news flow is heavy on the $25B penalty, but retail largely dismisses it. @Justwin100 writes 'Who cares about $25B when you can just mention Ai or Muse'. @JohnJohnson99 expects Meta to win at the Supreme Court. @StockStinjo, labeled bullish, says the state wanted roughly $260B, the judge cut it to $40B and then $25B, and Meta will appeal and drag it out. Those figures come from one unverified post and are not in the headlines. @musial12 suggests the timing was unhelpful for the stock. Retail is treating the penalty as a headline rather than a thesis-changer. That could prove complacent if the appeal outlook or other state actions are less benign than assumed. - Institutional caution vs retail optimism: Barron's and Yahoo frame earnings as a test of the AI-spending appetite and a possible bubble pin. Retail price chatter is mostly bullish ('740', '750-800', '765'), though bearish voices (bubble, double top, 600) are present. - Alignment: both sources agree that earnings are the dominant catalyst and that the AI-capex payoff is the key debate. Both also acknowledge that risk is two-sided. @GoodNewsBull suggests some portfolio managers are hedging via AAPL ahead of META's report to shield against the AI trade if META disappoints. @NeverLuckyNick argues QQQ would flush on a META sell-off.

3. Dominant narrative themes 1. Earnings-day binary: AI spending vs monetisation, with the market looking for a story as well as a beat. 2. The $25B New Mexico penalty, and whether it is noise or a precedent. 3. AI bubble/circular-revenue worry vs 'AI product' enthusiasm (Muse, the Zuckerberg AI move). One retail post praises Muse's usability on Apple Watch. 4. A tech tape lifted by the SpaceX/SPCX surge and positive earnings-day reactions, with sentiment tied to Tesla, QQQ and AI peers. 5. Short-term technicals: 692/681 support, 713 resistance (flagged as a possible double top), and the 730 20-day.

4. Catalysts and risks - Catalysts: Q3 results and guidance after the close today. Any capex and 2027 expense outlook, plus evidence of AI monetisation, would be the swing factor. A beat could trigger a short-term squeeze toward the 730 20-day and then retail targets of 740-765. Appeal progress on the New Mexico case would also matter. Tesla's report the same evening and a firm tech tape could add sympathy moves. - Risks: a capex or expense surprise that revives the AI bubble narrative. Legal and regulatory contagion, since a $25B state-level judgment could encourage similar actions (not evidenced in the data). Failure to clear 713 resistance, with a loss of 692 opening 681. Retail optimism ahead of the print may already be priced in, and 70% of messages carry no label, so the true retail skew is uncertain. Crowded positioning is possible but not verifiable from this data. Past sentiment is not predictive and this is not a price call.

5. Summary of key signals

Signal	Direction	Source	Evidence
$25B New Mexico penalty	Bearish (event)	News (Reuters, AP, Axios, MT Newswires)	Jury found Meta misled users on privacy; 43M+ violations per a StockTwits post; appeal expected per retail
Retail dismisses the penalty	Bullish/neutral	StockTwits	'Who cares about $25B', 'they'll go to the SC and win', 'appeal and drag this out'
Earnings as AI-spend test	Mixed/cautious	News (Yahoo, Barron's)	'Test Wall Street's stomach for AI spending'; 'bubble to pop'
Pre-earnings run-up and strong tape	Mildly bullish	News (IBD, Barron's, Yahoo)	'Meta Stock Climbs With Earnings On Deck'; tech rallies; earnings-day returns strong
Labeled retail ratio	Mildly bullish, low reliability	StockTwits	6 Bullish (20%) vs 3 Bearish (10%), 21 unlabeled of 30
Price targets after earnings	Bullish	StockTwits	740, 750-800, 765
Bearish retail view	Bearish	StockTwits	'biggest bubble in history', 'double top at 713', 'back to 600?'
Technical levels	Neutral/contested	StockTwits	707.71 (+1.9%) off 692.38 low; below 20-day 730.05; resistance 713.22; support 692/681
AI-capex skepticism, peer context	Mildly bearish	News (24/7 Wall St.), StockTwits	Alphabet down about $860B from peak, Apple near highs; capex-before-revenue concern
Reddit	No data	N/A	Disabled by config

Net read: Mixed. Retail leans mildly bullish and shrugs off the legal ruling, but the news flow pairs a large legal negative with a binary AI-capex earnings test. Sample sizes are small and Reddit is missing, so conviction is low.

News Analyst¶

META (Meta Platforms, Inc.): News and Macro Report, week of 2026-10-14 to 2026-10-21¶

0. Data coverage¶

The FRED macro tool failed on every call. I tried fed funds, 10-year Treasury, CPI, core PCE, unemployment, VIX and yield curve, plus the raw IDs FEDFUNDS and DGS10. Each returned DATA_UNAVAILABLE. I have no hard readings for rates, inflation, labor or volatility, and I haven't estimated any. Macro conclusions below rest on prediction markets and headlines only.
The news tools returned headlines without article text. Anything beyond a headline is marked as inference.
No META price feed was available. I infer the approximate price level only from prediction-market strike ladders.
Q3 results are not in the feed. Every META item is forward-looking, so I can't confirm whether results are out. Oct 21 is a Wednesday, and the coverage reads as if the report is imminent or due after the close. Verify the timing and the actual print before acting.

1. META company news¶

1. Q3 earnings are the dominant catalyst. - Yahoo Finance: "Meta Q3 earnings to test Wall Street's stomach for AI spending." - Barron's: "If There's a Market Bubble to Pop, Meta Earnings Hold the Pin." - Barron's: "The Magnificent Seven Need to Tell a Story Along With an Earnings Beat." - IBD: "Meta Stock Climbs With Earnings On Deck. These Are The Numbers To Watch." - Reading: META's stock reaction depends on whether investors accept the AI capex story, more than on the headline EPS beat. The stock appears to have run up into the print, per the IBD headline.

2. A $25 billion legal judgment in New Mexico is the main negative headline. - Reuters, AP, Axios and MT Newswires all report that Meta was ordered to pay New Mexico $25 billion. - Reuters says a jury found Meta misled consumers about data privacy. The AP headline says a judge ordered the payment for deceiving Facebook users. MT Newswires ties it to the Cambridge Analytica case. - The sources differ on whether the jury or the judge made the call, and MT Newswires hedges with "reportedly." - I have no information on appeal, accrual or payment timing. I also have no balance-sheet data to size it against cash. - Inference: expect questions on the earnings call about legal reserves, appeals and read-through to other state cases.

3. Zuckerberg's AI strategy. TheStreet runs "Mark Zuckerberg just made his boldest AI move yet." The headline has no detail. It adds to the AI-spending scrutiny.

4. Sector sentiment around AI spending is split. - 24/7 Wall St.: Alphabet has lost about $860B from its peak, while Apple, which "avoided the AI race," is near all-time highs. The market is questioning heavy AI capex, and META is one of the biggest spenders. - Offsetting that, tech rallied on Oct 21 (Dow, S&P 500 and Nasdaq all gained). A "blowout quarter" from SpaceX is lifting AI and chip names. - Barron's: "Meta and Tesla Are Crushing Their Average Earnings Day Returns. Thank SpaceX."

2. Broader market and macro (headline-level)¶

Equities: US indices rose on Oct 21. Futures were up ahead of the Meta, Tesla, IBM and Amazon reports. Earnings season is the main driver this week.
Global yields: Reuters reports France's 10-year yield hit 5%. That points to global long-end yield pressure, which hurts long-duration growth stocks like META. I have no US 10-year reading to compare against.
Commodities: The WSJ reports gold futures edging up while oil slips.
Tariffs (Polymarket): A US-Canada tariff-lowering deal by Oct 31 is at 5%, and by Dec 31 at 40%. Trade policy is not an immediate catalyst for META.

3. Prediction-market signals¶

Fed policy (the most important macro signal): - No Fed cuts in 2026: 97% (about $9.0M volume). - December 2026 meeting (about $1M volume per market): 25bp hike 78%, hold 18%, 25bp cut 1%, 50bp+ hike 2%, 50bp+ cut 0%. - The market is pricing a hike, not a cut. That is a hawkish tilt for long-duration equities. Volumes are moderate and the hike-versus-hold gap is large, so verify it against fed funds futures and actual inflation data. I couldn't do that because FRED was down.

Recession and AI bubble: - US recession by end of 2026: 8%. - AI bubble burst in 2026: 9% (up 3.1pp on the week). By June 30, 2027: 20%. - Reading: the market sees low near-term macro risk, but AI-bubble odds are rising.

META price-touch markets (thin, about $3K–18K each, so low reliability):

Strike (October)	Touch probability	1-week change
Low $680	62%	+17.0pp
Low $660	45%	+14.5pp
Low $640	30%	+12.5pp
Low $620	18%	+6.0pp
High $760	40%	−29.5pp
High $800	18%	−22.5pp
High $820	12%	−15.5pp
High $860	4%	−10.0pp
High $900	3%	n/a

Week of Oct 19 market: the high touching $750 is 40% (resolves Oct 23).
Over the week, downside-touch odds rose sharply and upside-touch odds fell sharply.
Inference only: spot looks to be in the low-to-mid $700s. That is bracketed by a 62% chance of touching $680 and a 40% chance of touching $760. Check the live quote. The two-sided ladder implies traders expect a large earnings-driven move, roughly ±8–10% over October.

META AI model benchmarks: The chance that a Meta model scores 55% or higher on Humanity's Last Exam in 2026 is 12% (down 9pp on the week). At 60% or higher it is 8%. Both are small, thin markets. They suggest low expectations for a near-term benchmark leadership claim.

4. Actionable takeaways for traders¶

Treat the earnings print as a binary event, and confirm timing first. The reaction will likely turn on capex guidance, 2027 expense commentary and how management ties AI spending to ad revenue. These are standard metrics, listed from general knowledge. The feed gives no consensus figures.
Account for legal risk. The $25B New Mexico award adds possible charges, reserve disclosures and headline risk. Watch for appeal news and any accrual.
Rate risk is skewed hawkish. The Fed market prices no cuts and a likely December hike, and French 10-year yields are at 5%. That is a headwind for high-multiple tech if long yields keep rising. Confirm with actual US yield data.
The AI-capex narrative cuts both ways. Sector sentiment is strong on Oct 21, helped by SpaceX and chips. Investors are also punishing perceived AI overspenders, as in Alphabet's drawdown, while rewarding Apple. A capex raise without clear revenue payback is the main downside scenario.
Prediction-market momentum is bearish into the print. Downside-touch odds are climbing and upside-touch odds are collapsing. This could reflect de-risking or a post-judgment drop. Because the markets are thin, don't weight them heavily.

5. Summary table¶

Category	Key point	Source/evidence	Implication for META	Confidence
Earnings	Q3 report imminent or due; focus on AI spending	Yahoo, Barron's, IBD, WSJ headlines	Binary event; capex and guidance drive the reaction	High that it's a catalyst; results not in feed
Legal	$25B New Mexico judgment over privacy/Cambridge Analytica	Reuters, AP, Axios, MT Newswires	Possible charge, appeal and read-through to other cases	High that it was reported; details and finality unverified
AI narrative	Alphabet down about $860B from peak; Apple near highs	24/7 Wall St.	Market skeptical of heavy AI capex	Medium
Market tone	Indices up Oct 21; SpaceX and chips leading	Yahoo, Stocktwits, Barron's	Supportive backdrop into the print	Medium
Fed	No 2026 cuts 97%; Dec hike 78%, hold 18%	Polymarket	Hawkish; pressures growth multiples	Medium (unverified against FRED)
Global yields	France 10-year at 5%	Reuters	Long-end yield pressure	Medium
Recession	US recession by end of 2026 at 8%	Polymarket	Low macro tail risk	Medium
AI bubble	Burst in 2026 at 9% (+3.1pp); by mid-2027 at 20%	Polymarket	Rising but still small tail risk	Low–Medium
META price ladder	Downside touches rising, upside touches falling; spot likely in the $700s	Polymarket (thin)	Bearish positioning into earnings; large expected move	Low (thin volume, inferred spot)
Macro data	Fed funds, 10-year, CPI, core PCE, unemployment, VIX and yield curve all unavailable	FRED tool errors	No hard macro confirmation	Data gap
Commodities	Gold up, oil lower	WSJ	Mild risk-hedging tone	Low

Handoff: The next analyst should get the live META price and the Q3 results, if released. They should also pull actual Treasury yield, inflation and VIX data to check the hawkish Fed pricing.

Fundamentals Analyst¶

META (Meta Platforms, Inc.) Fundamental Report, as of 2026-10-21¶

0. Scope and data caveats¶

Nothing new in the last week. The latest filed period is Q2 2026 (quarter ended 2026-06-30). The most recent insider transaction is dated 2026-10-12. There are no Q3 2026 financials in the data yet. Meta has historically reported Q3 in late October, so the print is probably close, but I could not confirm the date.
Derived figures. The quarterly feeds omit Q4, because Q4 sits inside the 10-K. I rebuilt Q4 2025 as the fiscal year minus the first three quarters. The cash-flow feeds are year-to-date, so I differenced them to get single quarters. Gross profit is "not tagged" in the income statement feed, so I computed it as revenue minus cost of revenue. Quarterly figures marked * are my derivations.
No quote in the tool output. The implied share price is about $707 to $708. Three metrics agree on it: P/E × EPS, P/B × book value, and market cap ÷ shares.

1. Snapshot and valuation (vendor data)¶

Metric	Value	Comment
Market cap	~$1.80T	About 2.55B shares implied
P/E (TTM) / Forward P/E	26.7x / 20.4x	TTM EPS of $26.54 is distorted in both directions (see §2)
PEG	0.94
P/B / Book value per share	6.9x / $102.52
Price / Sales (my calc)	~7.9x
Market cap / EBITDA (my calc)	~16.4x	Not EV-based, because debt and securities detail is unavailable
Dividend yield	0.3%
Beta	1.185
52-week range	$520.26 to $779.82	Implied price is about 72% of the way up the range, roughly 9% below the high
50-day / 200-day average	$659.49 / $632.42	Price is above both
ROE / ROA	29.8% / 14.6%
Debt/Equity	0.43x	Implies roughly $112B of debt, probably including leases (my inference)
Current ratio	2.23	I verified it against the balance sheet: 125,475 ÷ 56,379

Forward EPS check. The vendor's forward EPS is $34.65, but the tool doesn't say which period it covers. If it means FY2026, H1 EPS was $16.62 ($10.44 + $6.18), so H2 would need about $18, or roughly $9 a quarter. Q2 2026 was $6.18, and Q4 2025 was about $8.87*.

2. Income statement¶

Quarterly trend (USD millions)¶

Quarter	Revenue	YoY	Op. income	Op. margin	Net income	Diluted EPS
Q1'25	42,314	+16.1%	17,555	41.5%	16,644	6.43
Q2'25	47,516	+21.6%	20,441	43.0%	18,337	7.14
Q3'25	51,242	+26.2%	20,535	40.1%	2,709	1.05
Q4'25*	59,894	+23.8%	24,745	41.3%	22,768	~8.87
Q1'26	56,311	+33.1%	22,872	40.6%	26,773	10.44
Q2'26	60,801	+28.0%	18,775	30.9%	15,848	6.18

Annual¶

FY	Revenue	Op. income	Op. margin	Net income	EPS
2023	134,902	46,751	34.7%	39,098	14.87
2024	164,501	69,380	42.2%	62,360	23.86
2025	200,966	83,276	41.4%	60,458	23.49

The trailing-twelve-month figures I rebuilt match the vendor's:

TTM item	My figure	Vendor figure
Revenue	$228.2B	$228.2B
Net income	$68.1B	$68.1B
Gross margin	~81.7%	—
Operating income	$86.9B	—
Operating margin	~38.1%	34.8% (see §6)

Key observations¶

Top-line growth is strong. Revenue grew 28% in Q2 2026 to $60.8B, up 8% from Q1. Growth accelerated from +16% in Q1 2025 to +33% in Q1 2026, then eased slightly in Q2.
Profit is not following revenue. Q2 2026 operating income fell 8.2% YoY ($18.8B vs $20.4B) and 17.9% QoQ. Operating margin dropped to 30.9%, from 43.0% a year earlier and 40.6% in Q1. This is the biggest red flag in the data.
Cost of revenue rose 33% YoY ($11.3B), faster than revenue. Gross margin slipped to about 81.4% from about 82.1%.
Implied operating expenses (revenue minus cost of revenue minus operating income, my calculation) were about $30.7B. That is up about 65% YoY from $18.6B and up about 32% QoQ from $23.2B.
The tool doesn't break out opex. I can't tell whether the jump is structural (AI headcount, depreciation from the capex build) or one-off (legal or restructuring). The Q2 10-Q should answer this.
Q4 2025 was an earlier warning: revenue grew +23.8% but operating income grew only about +5.9%. The margin was 41.3%, against about 48.3% in Q4 2024.
Net income is noisy.
Q3 2025 net income of $2.7B against $20.5B of operating income implies a very large non-operating or tax charge.
Q1 2026 net income of $26.8B exceeds operating income of $22.9B, which implies a large benefit.
The tool doesn't identify either item. Both flow into the TTM EPS of $26.54, so the 26.7x trailing P/E is not a clean read. Operating income and the forward P/E are better guides.
FY2025 net income fell 3% ($60.5B vs $62.4B) even though revenue grew 22% and operating income grew 20%.

3. Cash flow and capex¶

Single quarters, derived from year-to-date data (USD millions):

Period	Op. cash flow	Capex	FCF (OCF − capex)	Capex % of revenue
Q1'25	24,026	12,941	11,085	30.6%
Q2'25	25,561	16,538	9,023	34.8%
Q3'25	29,999	18,829	11,170	36.7%
Q4'25*	36,214	21,383	14,831	35.7%
Q1'26	32,226	18,997	13,229	33.7%
Q2'26	31,862	30,116	1,746	49.5%

Annual:

FY	Op. cash flow	Capex	FCF	FCF margin
2023	71,113	27,045	44,068	32.7%
2024	91,328	37,256	54,072	32.9%
2025	115,800	69,691	46,109	22.9%

Capex is the main story. Capex grew 87% in FY2025. H1 2026 capex of $49.1B is 67% above H1 2025.
Q2 2026 capex was $30.1B, about 94% of that quarter's operating cash flow, so FCF was only about $1.7B.
Operating cash flow is healthy. Q2 OCF was up about 25% YoY.
TTM from the statements: OCF is about $130.3B, capex about $89.3B, and FCF about $41.0B, an 18% margin and about a 2.3% FCF yield on the $1.8T market cap.
Simple extrapolation, not guidance: if Q3 and Q4 each match Q2's ~$30B of capex, FY2026 capex would be about $109B.
Capital returns have shrunk.
Financing cash flow was +$9.4B in H1 2026, against −$35.5B in H1 2025.
In Q4 2025 it was roughly +$25B.
Together with rising liabilities (§4), this suggests debt-funded capex with less buyback activity. That is my inference, since the tool shows no debt or buyback lines.

4. Balance sheet (USD millions)¶

Date	Total assets	Cash & equiv.	Current assets	Current liabilities	Total liabilities	Equity	Liabilities / assets
2024-12-31	276,054	43,889	100,045	33,596	93,417	182,637	33.8%
2025-09-30	303,844	10,187	73,118	36,958	109,778	194,066	36.1%
2025-12-31	366,021	35,873	108,722	41,836	148,778	217,243	40.6%
2026-03-31	395,250	23,426	109,765	46,753	151,569	243,681	38.3%
2026-06-30	449,956	15,462	125,475	56,379	188,735	261,221	41.9%

Assets are up 48% since Sep 2025. They went from $304B to $450B, driven by AI infrastructure and related investment.
Liabilities rose by $79B (+72%) over three quarters. Q4 2025 added $39B and Q2 2026 added $37B.
Equity still grew by $67B, so book value is $102.52 per share.
Cash is thin relative to the balance sheet. Cash and equivalents are $15.5B, down from $35.9B at year-end and only about 12% of current assets. The remaining ~$110B of current assets is not itemized in the tool, so marketable securities and net cash can't be assessed here.
Liquidity is still adequate. The current ratio is 2.2x and debt/equity is 0.43x. The direction of travel is toward more leverage and less cash.

5. Insider transactions (150 rows, 2025-11-03 to 2026-10-12)¶

There were no open-market purchases. Every priced transaction is a sale or a $0 gift.
Blank-price rows are probably vesting events. Rows with shares but no price appear on the 13^th to 15^th of Feb, May, Aug and Nov. The tool doesn't label them, so I treat them as probable RSU vesting or withholding entries.
Nothing in the past week. The last filing is dated 2026-10-12.

Sales since 2026-07-27 (roughly $102M in total across about 10 insiders)¶

Insider	Role	Details	Approx. value
Christopher Cox	Officer	80,000 shares on 9/9, 9/15 and 9/21 at $650 to $725	$55.0M
Mark Zuckerberg	CEO	27,474 shares on 9/24 at ~$775 to $779, near the 52-week high	$21.4M
Javier Olivan	COO	Weekly sales of 1,575 shares (about $1.1M each) in Sep and Oct, plus 5,354 shares on 9/8	$10.8M
Susan Li	CFO	8/14 and 8/18 sales	$6.5M
Andrew Bosworth	CTO	7,848 shares on 8/18	$4.4M
Aaron Anderson	Officer	3,240 shares on 9/3	$2.0M
Curtis Mahoney	Officer	1,559 shares on 8/18	$0.8M
Directors (Andreessen, Kimmitt, Alford)	Directors	Small sales	~$0.8M

Zuckerberg gift: 591,690 shares on 2026-07-31, which I estimate at about $320M at that day's ~$544 price.
Susan Li's trailing 12 months came to about $106M. Her Feb 2026 sales alone were about $88M, at roughly $630 to $650.
Andrew Bosworth sold about $24M over the same 12 months.
Olivan's weekly sales are evenly sized and recur on a fixed schedule. That is consistent with a pre-arranged trading plan, though the tool doesn't confirm one. His share count per sale has risen from 517 to 1,555 and then 1,575, so his weekly sales are likely to continue.
Cox and Zuckerberg sold into the September rally. The stock went from about $550 in mid-August to about $779 on 9/24. Cox's sales are new in this dataset, with none before September.
Governance. Dina Powell appears as a Director in Nov 2025 and as President from Feb 2026.

6. Data-quality flags¶

Free cash flow definitions differ. The vendor shows $21.55B, but OCF minus capex from the statements gives about $41.0B. I can't explain the ~$19B gap, so verify the definition before using either.
Operating margin differs. The vendor shows 34.8%. I compute 38.1% TTM, 30.9% for Q2 2026 and 40.6% for Q1 2026. None matches the vendor figure.
H1 2026 cash doesn't reconcile. Operating plus investing plus financing cash flow gives about −$9.8B. Balance-sheet cash fell by $20.4B ($35.9B to $15.5B). The ~$10.6B gap could be restricted cash, FX or tagging, but the tool doesn't say. Earlier periods reconcile closely.
One insider price looks wrong. The CFO's 8/14 sale is shown at $689.85. Prints from 8/3 to 8/18 are around $545 to $600. Treat that figure as a possible artifact.

7. What the data supports¶

Supportive - Revenue is growing 28% on a $60.8B quarterly base, and operating cash flow is up about 25% YoY. - Forward P/E is about 20x and PEG is 0.94. - Returns are high (ROE ~30%), and liquidity ratios are comfortable. - Price is above the 50-day and 200-day averages.

Cautionary - Margins: operating margin compressed by 12 points YoY in Q2 2026. Implied opex rose about 65% YoY and cost of revenue grew faster than revenue. - Cash conversion: Q2 FCF was ~$1.7B on $30B of capex, and TTM capex absorbs ~69% of OCF. - Balance sheet: liabilities are up $79B in three quarters while cash is down to $15.5B. - Earnings quality: both the Q3 2025 and Q1 2026 net income figures are heavily affected by unexplained non-operating or tax items. - Insiders: there is no insider buying, and selling is concentrated among senior officers after the September rally.

Watch items 1. The Q3 2026 earnings date. Look at the opex trend, capex run-rate, FCF, debt issuance and the tax line. 2. The Q2 10-Q detail on debt, marketable securities, lease obligations and the opex drivers. These are the main gaps in this data. 3. Whether Cox's selling continues, whether Olivan's weekly sales persist, and any further Zuckerberg sales or gifts. 4. Whether H2 EPS can move toward the ~$9 a quarter implied by the vendor's forward EPS, if that figure is for FY2026.

This report covers only what the tools support. It is not a trade recommendation.

8. Summary table¶

Category	Key point	Evidence	Read
Valuation	Forward P/E 20.4x, TTM P/E 26.7x, PEG 0.94	Vendor data; TTM EPS distorted by one-offs	Neutral to positive
Price context	~$707 implied, 52-week range $520 to $780	Above 50-day ($659) and 200-day ($632) averages	Positive trend
Revenue	Q2'26 $60.8B, +28% YoY; Q1'26 +33%	Quarterly income statement	Positive
Profitability	Q2 op. margin 30.9% vs 43.0% a year earlier	Implied opex ~$30.7B, +65% YoY	Negative; key diligence item
Net income	Q3'25 $2.7B and Q1'26 $26.8B distorted	Large gaps vs operating income	Use operating income instead
Operating cash flow	Q2'26 $31.9B, +25% YoY; TTM ~$130B	Cash flow statement	Positive
Capex / FCF	Q2'26 capex $30.1B, FCF ~$1.7B; TTM FCF ~$41B (vendor $21.6B)	Capex 69% of TTM OCF	Negative near term
Balance sheet	Assets $450B; liabilities $189B (+$79B in 3 quarters)	Cash $15.5B; D/E 0.43x; current ratio 2.2	Leverage rising, liquidity OK
Capital returns	Financing flipped to +$9.4B in H1'26 from −$35.5B	Dividend yield 0.3%	Buybacks apparently de-emphasized (inference)
Insiders	~$102M of sales since 7/27; no buys in 12 months	Cox $55M, Zuckerberg $21M, Olivan $11M	Cautionary
Data issues	FCF and operating margin vendor mismatches, H1'26 cash gap, one suspect price	See §6	Verify against the 10-Q
Next catalyst	Q3 2026 earnings, likely late October (unconfirmed)	Historical cadence	Check the date

II. Research Team Decision¶

Bull Researcher¶

Bull Analyst: # Bull Case for META: Opening Argument (price $707.95, 2026-10-21)

Q3 earnings are reportedly due after the close tonight, so I'll say up front that I don't know the result. I don't need to. The case for owning META going in rests on growth, valuation, cash generation and trend, and the setup is better than the headlines suggest.

1. Growth at scale, and it's accelerating¶

Revenue grew +28% YoY in Q2'26 to $60.8B, after +33% in Q1'26. A year ago, in Q1'25, growth was +16%. Few companies with a $228B revenue base are speeding up.
Gross margin is still about 81%.
Operating cash flow was $31.9B in Q2 (+25% YoY), and about $130B over the trailing twelve months.

2. The valuation doesn't match the growth¶

Forward P/E is 20.4x and PEG is 0.94. That is a market-multiple price for a business growing revenue near 30%.
ROE is about 30%, the current ratio is 2.2x, and debt/equity is 0.43x.
The stock is about 9% below its 52-week high of $779.82 and well above its 50-day ($659) and 200-day ($632) averages.

3. The capex is the strategy, and revenue is the proof¶

The bear will say capex is $30B a quarter. I'd ask what it is buying. Capex rose 87% in FY25, and revenue growth went from +16% to +33% over the following five quarters. That is the pattern of spending that earns a return. Trailing OCF of about $130B covers about $89B of capex and leaves roughly $41B of FCF. The vendor's own FCF figure is lower ($21.6B), but it is positive on either definition. Meta is funding this build from operations, not from desperation.

4. The technical picture is a healthy reset¶

All three SuperTrend timeframes are UP (weekly, monthly and daily). The golden cross came on 10/6.
The 7/30 low of $524 was followed by a ~49% rally to the 9/24 peak.
The pullback of about 10.7% brought RSI from 80.55 to ~50, so overbought conditions are cleared without becoming oversold.
The breakout gap (base $665) has not been filled. The 10/20 low was $681.
Breakout volume was 28–49M shares a day, while pullback volume has been 16–21M. Sellers have shown little conviction.
No z-score is stretched (weekly +1.19, daily −1.10).

5. Positioning into the print¶

Prediction markets show downside-touch odds rising sharply and upside-touch odds collapsing (the $760 high fell 29.5 points on the week). Bearishness into earnings is the kind of setup that can fuel a relief move on decent numbers. Barron's also notes META has been "crushing" its average earnings-day returns. I'd weight the thin prediction markets lightly, but the direction is clear.

Pre-empting the bear¶

"Q2 operating margin fell to 30.9% from 43%." This is the bear's best card, and I won't hide from it. Implied opex rose about 65% YoY. But Q1 was 40.6%, so one quarter is not yet a trend. Even at 30.9%, Q2 operating income was $18.8B. The data can't tell us whether the step-up is structural (AI talent, depreciation) or one-off. That is the first question for management tonight. A 20x forward multiple already prices in some margin skepticism.

"Q2 FCF was only $1.7B and liabilities are up $79B." Q2 was the peak of an investment cycle. Equity also rose $67B over the same period, and OCF is growing 25%. Liquidity is fine and leverage is moderate at 0.43x D/E. Cash is thin at $15.5B, but it is supported by about $11B a month of operating cash flow.

"The $25B New Mexico judgment." That is about $9.80 a share, or 1.4% of the price. It is a one-time item against about $130B of annual OCF, and the headline says Meta was ordered to pay, with appeals widely expected. The stock has already fallen about 10.7% from its peak. Retail seems complacent about the legal risk, and I'd be more careful than they are. But this is a bounded, non-recurring liability. The real risk would be contagion to other states, and nothing in the data shows that yet.

"Insiders are selling." Total sales since late July are about $102M against a $1.8T market cap, or roughly 0.006%. Olivan's evenly sized weekly sales look like a pre-arranged plan. Zuckerberg's 9/24 sale of $21M is small beside his ~$320M share gift in July. This is routine compensation-driven selling, not a signal.

"The Fed may hike, and the AI bubble is a risk." Prediction markets put a December hike at 78%, but that market is thin and unverified against actual data. The same sources give 8% odds of a US recession and 9% odds of an AI-bubble burst in 2026. Advertising revenue is growing 28% in that environment. A bubble scare is the real risk, but it needs a catalyst, and tonight's print is where Meta can show AI spending turning into revenue.

"Earnings gaps have been brutal." Three of the last four ≥8% gap days were down (−11%, −9%, −8%). But the stock was bought both times. The April and July drops set up the $520–525 double bottom, and the stock then rose about 49% in under two months. The market has been rewarding Meta for fundamentals.

What would change my mind¶

Capex guidance rises sharply without a matching revenue-growth outlook.
Management can't explain the opex step-up, or says it's permanent and still growing.
A daily close below the $681 SuperTrend stop that continues toward the 50 SMA at $661.
Evidence of other state actions following New Mexico.

Bottom line¶

The company is growing 28–33% at $60B a quarter, trades at about 20x forward earnings, generates more than $130B a year in operating cash flow, and sits in an intact uptrend after a low-volume pullback. The risks are real: margins, capex intensity and legal exposure. But they are priced and bounded, and the growth that justifies the spending is showing up in the numbers.

Bear, tell me which of these you think breaks the thesis: the margin step-down, the capex burden or the legal overhang. I think I have an answer for each. Bull Analyst: # Bull Rebuttal for META (price $707.95, 2026-10-21)

You've sharpened this thesis, and several of your hits land. I'll concede those first and then show where I think your numbers overreach.

What I'm conceding¶

"Accelerating" was the wrong word. Growth slowed from 33.1% to 28.0%. The two-year stack is steady at 49.2 → 49.6, so "sustaining" is accurate and "accelerating" is not.
"Q2 was the peak" was an assertion. Nothing in our data shows capex peaking. My case doesn't need a peak. It needs the spending to earn a return.
My earnings sequence was misdated. The 519 low (3/27) came before the April print. The accurate pattern is a ~29% run into April, then −8.5%.
The positioning argument was weak. There are 9 labeled posts, the prediction markets are thin, and the tape is up today. I'm dropping it.
The golden cross is lagging, and a $681 stop won't protect through a gap. Agreed. That is a sizing problem, which I come back to.
20x isn't conservative. It embeds a partial margin recovery (see below).

Where I'm holding the line¶

1. Your 28.6x freezes revenue¶

$24.70 is Q2 EPS × 4. That holds revenue at $243B for a year, at Q2's margin and tax and other-income conversion. You agree the business is growing 28%, so a no-growth year isn't a base case. My arithmetic, assuming next-twelve-month revenue of ~$275B (only ~20% growth) and 84–90% net-to-operating-income conversion:

NTM operating margin	Forward P/E at $708
31% (Q2 repeated)	~24–25x
35%	~21–22x
38%	~19–21x

The vendor's $34.65 implies a mid-to-high-30s margin, which is a partial recovery and not a return to 43%. I accept that. But if Q2 is the permanent base, you're paying ~25x for ~20–28% growth, not 28.6x. Normalized trailing EPS is also ~$29–31 once the Q3'25 charge (Q3'25 net income was $2.7B on $20.5B of operating income) offsets the Q1'26 benefit. The trailing multiple is therefore ~23–25x, not 26.7x.

2. The margin is a step, not a slope, and "same fact" is unproven¶

Margins: Sequentially they ran 41.3% → 40.6% → 30.9%. Your "three straight quarters" is −7, −1 and −12 points YoY, which isn't a steady slide. Q1'26 was nearly flat YoY on 33% growth.
Capex intensity: For five quarters it ran 31–37% of revenue while margins held at 40–43%. Only in Q2, with capex at 49.5%, did both break. You're right that depreciation lags. But a 10-point drop in one quarter is a lot for a lagging cost to produce.
Where the costs are: Of the ~$15B YoY rise in total costs, ~$12.1B sits outside cost of revenue, and cost of revenue (+$2.8B) grew in line with revenue. The tools don't show where depreciation lands, so some could be in R&D. But "capex equals margin loss" is then an inference, not a finding.
Cash margin: Q2 OCF margin was 52.4% against 53.8% a year earlier. Q1 was 57.2% against 56.8%. OCF rose $6.3B YoY while net income fell $2.5B. So the GAAP squeeze is at least partly non-cash recognition of the build, while cash earnings power is intact. If that is depreciation, it persists. It then has to be outgrown by revenue, which is all a 20–25x multiple requires. Costs only need to grow slower than revenue, not fall.

3. The H1 table is really one quarter¶

Your 13 cents on the incremental dollar is Q1 and Q2 averaged. Q1'26 incremental margin was 38% ($5.3B of operating income on $14.0B of added revenue), in line with the 41.5% base. Q2 was negative (−$1.7B on +$13.3B). The table restates "Q2 was bad," which is the open question for tonight's call. FY25 shows the same thing. Operating income grew 20% on 87% more capex, and the net income decline came from the unexplained Q3'25 charge.

4. "Not funded from operations" is wrong in the data¶

OCF exceeded capex in all six quarters shown, including Q2 (1.06x). H1 coverage was 1.3x. The H1 financing inflow of $9.4B is ~19% of H1 capex. That is a modest top-up, and the swing from −$35.5B mostly reflects less capital returned, a choice. I'll grant that Q2's $1.7B of FCF is thin and that the TTM FCF yield is only ~2.3%. This is a growth stock whose return rests on earnings power, not a cash-yield story.

5. The legal item has the wrong denominator¶

You measured $25B against cash alone ($15.5B) and against FCF. Against the rest of the balance sheet it is: - ~10% of equity ($261B) - ~20% of current assets ($125B) - ~19% of TTM OCF - 1.4% of market cap

I also can't attribute the timing, but the stock is $54.55 below its 10/13 close of $762.50, about $139B of market cap, or ~5.5x the headline number. If the verdict is part of that drop, it's well priced. If it isn't, the drop is about earnings. Contagion is the real tail. I agree our feed can't rule it out, and that is on my kill list.

6. The print: the stats are noise, and the payoff is symmetric¶

The sample: The four gaps average −4.4% with a standard deviation of ~10 points. With n=4 the standard error is ~5 points, so the mean is statistically indistinguishable from zero. Three of four down is not a pattern at n=4.
The payoffs: From $708, the three down gaps cost $56–$80 a share and the one up gap adds $74. That is roughly symmetric, and a break above $780 would be new highs, not a cap.
The pullback: The 10/20 low retraced ~38.6% of the 7/30–9/24 rally, and the breakout gap base ($665) is intact. I agree short-term momentum is soft: price is below the 10-EMA, MACD is negative and OBV is weak. That is why this isn't a technical call.
Sizing: You're right that stops fail through gaps, so the risk control is position size. At a 5% position, an 11% gap costs about 0.55% of the portfolio.

The crux¶

We now agree on the facts. Revenue is growing 25–30%, OCF is growing, and Q2 operating margin was 30.9% for reasons we can't yet identify. We disagree on whether Q2 is a new base or a step. At 31% forever and 20% growth, the stock trades at ~25x. At the mid-30s margin the Street's EPS implies, it is ~21x. That is a fair price for the growth rate in the second case and a full one in the first, so the downside is a multiple compression, not a broken thesis.

What I need tonight¶

Opex: A credible breakdown of Q2's ~$7.5B sequential step-up, and Q3 operating margin of at least ~35%.
Revenue: Growth of ~23% or better, which holds the two-year stack at ~49.6 (≈$63B).
Capex: A guide that keeps FCF positive and OCF margin above ~50%.

What would change my mind¶

Q3 margin near 31% again with no one-off explanation, or guidance that opex keeps growing faster than revenue.
Revenue growth below ~23%.
A capex raise alongside falling OCF margin.
Copycat state actions.
A post-print daily close below $681.

Bottom line¶

I haven't seen the print, and neither have you. Your strongest argument is that the multiple embeds a margin recovery. I accept that, but the price already reflects a 10–15% drawdown, a 24–25x multiple even at Q2's margin, and positive OCF coverage of capex every quarter. I'd own META into the print at a size where an 11% gap is a nuisance, and add on a hold of $681–$690 or a reclaim of $718–$730 after the numbers. Waiting for the print costs you the +$74 case in exchange for avoiding the −$56 to −$80 case. That is a sizing preference, not a verdict on the business. Bull Analyst: # Bull Closing for META (price $707.95, 2026-10-21)

We now agree on most of the facts. Revenue is growing 25–30%, operating cash flow is up year over year, and tonight turns on one question: is Q2's 30.9% operating margin a step or the new base? Below is what I concede, where I still disagree, and what I'd do.

What I'm conceding¶

Your table is right. Each margin point is worth about $0.94 of EPS, so four points is roughly $80 a share at a constant multiple. A margin miss is a real outcome, not a tail.
I misdescribed cost of revenue. It grew +33% against +28% for revenue, not "in line," and gross margin slipped about 0.7 points (82.1% to 81.4%). Roughly 11.4 of the 12.1 points of margin lost still sit below gross profit, but I overstated it.
Incremental margins are weak. Over the last three quarters they blend to about 13% (12%, 38% and −13%), against 38% in FY25. Q1'26 is my only counterexample.
"Cash earnings power intact" was too strong. OCF is up 25% YoY but flat sequentially ($32.2B to $31.9B), with OCF margin falling from 57.2% to 52.4%. Q2 FCF margin was 2.9% against 19.0% a year ago.
The rally is unexplained. The 10–15% drawdown I cited was peak to trough (−12.7% intraday). Today it is −9%. The 7/30 close of $538.60 looks like the reaction to the Q2 print, which I'm inferring from timing. The stock is +31% since then with no new filing in our data. On annualized Q2 EPS ($24.72), that is about 22x in July against about 29x today. I can't explain the 9/21 gap, and the tail is therefore wider than the −8% to −11% gap history. This is why I size for a −15% gap.

Where I'm holding¶

1. Your table used my most conservative revenue input¶

$275B of NTM revenue is 20% YoY growth in each quarter. You say revenue isn't fading, so test the growth you believe in. The columns below use the same 87% conversion and 2.55B shares as your table:

NTM revenue	31% margin	35% margin	37% margin
$275B (+20%)	24.3x	21.6x	20.4x
$285B (+25%)	23.5x	20.8x	19.7x
$292B (+28%)	22.9x	20.3x	19.2x

The Street's EPS needs about 35%, not 36–38%. The vendor's $34.65 implies about $101.6B of operating income. That is 36.9% of $275B but about 35.6% of $285B and 34.8% of $292B. My 35% pass mark is the Street's number on a growth-consistent revenue base. This assumes the EPS is NTM, which the data doesn't confirm.
Margin is a risk, not a certainty of loss. If Q2's margin repeats and growth holds, you pay 23–24x for 25–28% growth, which is not cheap but not a bubble multiple.

2. Tonight is one question, not a conjunction¶

Of my "need tonight" items, three are floors Q2 already cleared: - Revenue growth of 23% or better (Q2 was 28%). - OCF margin above 50% (Q2 was 52.4%). - Positive FCF (Q2 was $1.7B).

You're right that FCF-positive is a low bar, and I set it low on purpose. The only open item is a margin of about 35% or better, with an opex bridge. On that we agree.

3. "Every candidate cause is a problem" is too strong¶

Depreciation persists, and I accept that the ramp is real. Your $3B-a-quarter figure assumes all $67B of added non-current assets is depreciable. Construction in progress, land, investments and leases may not be, or may start depreciating later. You flagged it as illustrative.
AI compensation persists, but it is investment with an identifiable payback test.
A one-off, such as a legal accrual (which can sit in G&A), is not a problem. It is the bull case.

I'll retract "step, not slope" and say "step now, slope unknown." Tonight's margin and the forward expense guide settle it.

4. Your legal dilemma is false¶

If the stock's drop is the verdict, the market has priced about 5.5x the headline, with a margin of safety.
If it isn't, the direct hit is 1.4% of market cap.

In both branches the verdict itself is bounded. Only precedent is unbounded, and it needs copycat states and further adverse rulings. We have no evidence of either, and I've kept it on my kill list. A Q3 accrual is plausible, and you're right that penalties are generally not deductible. I agree we should grade Q3 ex-charge, so the first question for management should be whether the margin is shown ex-charge.

5. On "zero edge," you're partly right¶

I'm not claiming a statistical edge from four gaps. My claim is narrower. The expected gap is about zero, not negative, and the thesis is a 12-month one at 23x or less for a business growing in the high 20s. A starter position is regret management for the case where the margin recovers and the stock gaps +10%. Your plan and mine both act on the print, and they differ only in whether to hold a starter into it. At a 2.5% starter, a −15% gap costs about 0.4% of the portfolio.

I can't fully answer the Alphabet and Apple contrast from headlines. The market sorting spenders by payback is the right frame, and it is why 28% revenue growth against a 31% margin is tonight's test.

The plan¶

Now: A half-size starter (about 2.5%, target 5%) that I can hold through a −15% gap.
Add if Q3 margin is about 35% or better with an opex bridge, revenue growth is 23% or better, and capex guidance keeps FCF positive.
Hold the starter if margin is 31–35% with a credible bridge and the multiple compresses. Do not add.

What would change my mind¶

Q3 margin near 31% again, with no one-off explanation.
Guidance that opex keeps growing faster than revenue.
A capex raise with falling OCF margin.
Revenue growth below 23%.
Copycat state actions.
A post-print daily close below $681.

Bottom line¶

META is growing at $60B a quarter, generates about $130B a year of OCF, and trades at 23–24x NTM EPS even if Q2's margin repeats. At the margin the Street needs, it is closer to 20x. The unexplained rally is a real risk, and it is why I would hold a starter now and add only on evidence. I'd rather be wrong by half a percent of the portfolio than miss the stock if the margin comes back. Bull Analyst: # Bull Final Rebuttal for META (price $707.95, 2026-10-21)

Your arithmetic mostly holds, so I'll concede first and then say what is left.

What I'm conceding¶

My 35% pass mark was too low for the revenue base we've both landed on. Holding the two-year stack at ~49.6 gives ~$278B NTM, and the vendor's $34.65 then needs ~36-37% at 87% conversion. I'm moving my pass mark to 36%.
"The one-off is the bull case" has no candidate in our data. The New Mexico verdict landed after Q2 closed, so it can't explain Q2's opex. A one-off is something management has to prove, not a branch I can count on.
It's two questions. One is Q3's margin and the other is the slope (the 2027 expense and capex guide). A 36% print with a re-compressing guide fails.
My add rule didn't discriminate. Q2's $1.7B already cleared "FCF positive," and 23% growth is the stack-hold base case, which is in line and not a beat.
Your July comparison is your best point. The stock fell 8% on this same 28%-growth, 30.9%-margin quarter at ~22x annualized Q2 EPS ($538.60 / $24.72). It is now ~29x on the same base, up 31% with no new filing in our data. Expectations are higher, and I can't explain why.
A match to the Street is fair value by construction. A +10% gap needs a beat or a re-rating, and a −11% gap needs only a Q2 repeat. The dollar payoffs are symmetric, but the conditions aren't.

Where I'm still standing¶

1. The gap between 35% and 36-37% is about one conversion-rate wobble¶

Using $278B, 2.55B shares and the two conversion rates that bracket the debate:

NTM op. margin	EPS @87%	P/E	EPS @91%	P/E
31%	$29.40	24.1x	$30.75	23.0x
35%	$33.20	21.3x	$34.72	20.4x
37%	$35.09	20.2x	$36.71	19.3x

35% hits the Street's $34.65 at ~91% net-to-operating conversion.
Clean periods ran 89.7% (Q2'25), 92.0% (Q4'25) and 89.9% (FY24). Q2'26 was 84.4%.
A 3-point swing in conversion moves EPS by ~$1.15, which equals ~1.2 margin points.
Q2'26's 84.4% may reflect less interest income on a smaller cash pile and a bigger interest bill. That is my inference, and if it holds, 87% is generous and your 4% miss is fair.

Either way, the pass-mark dispute is worth 1-2 margin points, or $20-40 a share. It doesn't change the structure.

I'll also state the valuation plainly. If Q2's 31% repeats, META trades at ~24x for ~22% growth. That is fair, not cheap, and the 0.94 PEG exists only in the recovery case. My claim is "fair if no recovery, ~20-21x if recovery."

2. The slope is the real 12-month bet¶

At a constant multiple, the return equals EPS growth. The stock earns that if expense growth stays at or below revenue growth. That is the slope question, and it is where this thesis lives or dies. It is also where I've said I'd be wrong.

3. On timing, I'll say it's close¶

I don't claim an edge from the gap. At 2.5% the stakes are ~0.4% of the portfolio on a −15% gap and ~0.25% on a missed +10%.
Waiting is reasonable, and I'm not calling it a mistake.
But your buy list is stricter than the Street needs. It asks for ≥37% ex-charge, a non-compressing 2027 guide and FCF above ~$10B a quarter.
The FCF bar needs capex of roughly $22-27B a quarter (depending on OCF) against Q2's $30.1B. Nothing in our data points to that, in a debate that is about the size of the build.
You called my list a conjunction. Yours is a longer one, and it works more like a veto than a trigger.

Tighter rules (replacing the ones you showed don't discriminate)¶

Add only if all of these hold: - Q3 operating margin is ≥36% ex any legal accrual, with the accrual itemized. If management doesn't itemize it, the quarter is ungradeable and I don't add. - There is an opex bridge that separates persistent costs (depreciation, AI compensation) from one-offs. - The 2027 guide has total expense growth at or below revenue growth. That is the slope test. - Revenue is ≥~$64B (~25% growth). $63.2B holds the stack but doesn't earn an add.

Hold the starter, don't add if margin is 31-36% with a credible bridge and a neutral guide.

Exit the starter on any of these: - Margin of ~31% with no bridge. - Guidance that opex again outgrows revenue. - Revenue below ~$62.5B (<22%). - A capex raise with falling OCF margin. - Copycat state actions. - A post-print daily close below $681.

Bottom line¶

The debate has converged on most of it: - Revenue is growing 22-28%. - OCF is up year over year. - Q2's margin was a step, and the slope is unknown. - The stock is ~20-21x NTM if the Street's margin arrives and ~24x if it doesn't. - The unexplained 31% rally means I can't say what's priced in, so I size for a −15% gap.

I'd hold a 2.5% starter into tonight, pre-sized for the gap, and add only on the evidence above. If you'd rather wait, I won't argue you're wrong. My claim is narrower. The exposure is small, the thesis is a 12-month slope bet, and the first real evidence arrives after tonight's close. Bull Analyst: # Bull Closing Rebuttal for META (price $707.95, 2026-10-21)

Your cost-dollar table is the sharpest margin analysis in this debate, so I'll start by crediting it. I'll then show where the conclusion goes further than the table does. The Q3 result is still unknown to both of us.

What I'm conceding¶

The arithmetic ties out. Q2 costs were $42.0B, up $8.6B (+26%) from Q1. On Q3 revenue of $63.2–65.5B, a Q2-style 31% needs costs up 4–8%. A 36% print needs costs roughly flat to down 4%.
The recent cost record is unfavorable. Costs outgrew revenue in Q4'25, Q1'26 and Q2'26, and the gap widened. I called the slope "unknown," but for the last three quarters it is known, and it points the wrong way.
The starter pays in one bucket. My exit rules are all post-print, so the starter's only job is to own the gap. I have no demonstrated edge on that and won't claim one.
A small correction to your numbers: at 84.4% conversion, 36% on $278B gives about $33.1 of EPS, not $32.9. That is still about 4% under the Street.

Where I still disagree¶

1. 36% is an NTM figure, not a Q3 requirement¶

I derived 36% to reconcile an undated vendor EPS with NTM revenue. You applied it to a single quarter. This is an illustration, not a forecast. It uses 2025–26 seasonal patterns (Q4 +15% sequentially, Q1 −6%, Q2 +8%), which are conservative against recent history:

Quarter	Revenue	Costs	QoQ cost	Margin
Q3'26	$63.2B	$43.6B	+3.8%	31.0%
Q4'26	$72.7B	$44.9B	+3%	38.2%
Q1'27	$68.3B	$43.6B	−3%	36.2%
Q2'27	$73.7B	$45.3B	+4%	38.5%
NTM	$277.9B	$177.4B		36.2%

Q3 doesn't have to be the proof. A 31% Q3 can sit inside a Street-consistent NTM if cost growth slows hard afterward.
It is still a big ask. Costs would have to go from +26% QoQ to about +3–4% a quarter. That is a trajectory test, not a Q3 dollar-decline test, and the guide is where it shows up.
The pass mark is itself uncertain. The vendor doesn't date the $34.65. On roughly $300B of FY27 revenue it implies about a 33% margin. If it is FY26, H2 needs about 37–39%, which favors you. Neither of us knows which, so 36% is not a bright line.

2. Three quarters isn't the whole record¶

Our annual data shows costs moving a lot relative to revenue: - FY24: costs +7.9% on revenue +21.9%. Margin went from 34.7% to 42.2%. - FY25: costs +23.7% on revenue +22.2%. - Q2'26: costs +55% on revenue +28%.

Within two years, relative cost growth has swung by more than 40 points, so costs are partly a management choice. I accept that FY24 had no $30B-a-quarter depreciation ramp, and I'm not claiming it repeats. I agree the burden of proof is on the company, and tonight is the first chance for it to carry that burden.

3. Your buckets sort on the wrong variable¶

You said earlier that the guide matters more than Q3's margin. Your §4 then sorts outcomes by Q3 margin. The gap history doesn't map to margin level:

Gap	Prior-quarter margin	Revenue growth	Reaction
10/30/25	40.1%	+26%	−11%
1/29/26	41.3% (−7 pts YoY)	+24%	+10%
4/30/26	40.6%	+33%	−9%
7/30/26	30.9%	+28%	−8%

I'm inferring from timing that these gaps are the earnings reactions, and the causes aren't in our data. It's n=4, so these are anecdotes, not a statistic. They do show that the market has paid up for a margin-down quarter, and that it has sold quarters with healthy margins. So "a Q2 repeat is −11% in plain arithmetic" assumes a constant multiple, and the multiple moves with the guide.

What I'm doing¶

Trim the starter to 2%, from 2.5%. Your cost table raised my odds of the bad bucket, and the unexplained +31% rally means I can't say what's priced in. A −15% gap costs about 0.3% of the portfolio.
Keep the add rules. I add only if Q3 margin is at least 36% ex-charge with the accrual itemized, there is an opex bridge, the 2027 guide has expense growth at or below revenue growth, and revenue is at least $64B. The guide is the gate. A 36% print with a re-compressing guide is not an add.
Keep the exits. I exit on a ~31% margin with no bridge, guidance that opex outgrows revenue, revenue below ~$62.5B, a capex raise with falling OCF margin, copycat state actions, or a close below $681.

Bottom line¶

We agree on the facts: revenue growth of 23–28% at $60B+ a quarter, OCF up year over year, Q2's margin as a step, and a slope that tonight's guide will test. We disagree on whether to own the gap. The expected gap is about zero, and the stakes are 0.25–0.4% of the portfolio either way.

I'd keep a small starter because the 12-month question is whether costs grow slower than revenue, and the company has shown it can do that. You'd wait and pay up for confirmation. If the numbers are good, that costs you the gap, and if they're bad, it saves you the gap. Neither choice is a mistake. The decision that matters is what each of us does after the print.

Bear Researcher¶

Bear Analyst: # Bear Case for META: Response (price $707.95, 2026-10-21)

I'll concede the real strengths: revenue is growing 28%, operating cash flow is up 25%, and the trend is up. I'm not arguing the business is broken. I'm arguing that at $708, before a binary print, you are paying for a margin recovery the latest numbers contradict, and that your risk controls can't work against the moves this stock makes on earnings.

You asked which of margin, capex or legal breaks the thesis. They aren't three separate risks. The margin step-down and the capex burden are the same fact, the cost of the build, and the legal judgment lands on the thinnest part of the balance sheet. Your thesis also contradicts itself. A 20x forward multiple needs costs to normalize, while the growth story needs $30B-a-quarter capex to continue, and that capex drives depreciation and opex up.

1. The 20x forward multiple assumes the recovery you haven't seen¶

The $34.65 forward EPS is undated. If it's FY26, H1 EPS was $16.62, so H2 needs about $18, or roughly $9 a quarter. Q2 was $6.18, so that is a 46% jump from the latest print.
H1 also includes Q1's $10.44, where net income ($26.8B) exceeded operating income ($22.9B) because of an unexplained benefit. The clean run-rate is lower than it looks.
Annualizing Q2's $6.18 gives about $24.70, or ~28.6x, not 20x.
You said 20x "already prices in some margin skepticism." It's the reverse. The estimate behind that multiple assumes margins recover.

2. "Accelerating" is wrong¶

Growth went from +33.1% in Q1'26 to +28.0% in Q2'26. That is a deceleration.
The two-year stack is flat: Q1 was 16.1 + 33.1 = 49.2, and Q2 was 21.6 + 28.0 = 49.6. Q1's headline was flattered by an easy comp.
Comps get harder. Q3'25 grew 26.2%, so merely holding 28% means lifting the two-year stack to about 54.

3. Revenue is proof of revenue, not of return¶

H1 YoY	H1'25	H1'26	Change
Revenue	$89.8B	$117.1B	+30%
Operating income	$38.0B	$41.6B	+10%
Capex	$29.5B	$49.1B	+67%
FCF (OCF − capex)	$20.1B	$15.0B	−26%

Only about 13 cents of each incremental revenue dollar reached operating income, against a ~41% average margin.
This isn't new. In Q4'25, revenue grew 23.8% and operating income grew 5.9%. In FY25, net income fell 3% and FCF fell $8B on 87% more capex.
Revenue per dollar of period-end assets has fallen from 0.60 (FY24) to 0.55 (FY25) to 0.51 (TTM).
Margin compression is now three straight quarters YoY: −7 points in Q4'25, −1 in Q1'26, and −12 in Q2'26. You call one quarter "not a trend." It's the third.
Capex is a cash outlay now but an expense over many years. Even flat capex pushes depreciation higher for years.

4. "Q2 was the peak" is an assertion¶

Q2 capex jumped ~58% QoQ to $30.1B, which is 49.5% of revenue and 94% of OCF. Nothing in our data says that's a peak. Q3 and Q4 at Q2's pace imply about $109B for FY26, up 57% on FY25.
Your $41B TTM FCF blends cheaper quarters. Q2 annualizes to ~$7B, or a 0.4% yield. The vendor's $21.6B is 1.2%, and even your figure is only 2.3%.
The "$11B a month of OCF" is already spoken for. Capex ran about $10B a month in Q2.
Cash is $15.5B, about half of one quarter's capex.
This isn't "funded from operations." Financing flipped from −$35.5B in H1'25 to +$9.4B in H1'26, and liabilities are up $79B (+72%) in three quarters. The $67B equity rise isn't liquidity, and part of it is that unexplained Q1 benefit.

5. The legal item is measured against the wrong denominator¶

You sized $25B against $130B of OCF, but $89B of that goes to capex. Against what's left, $25B is 61% of your TTM FCF, 116% of the vendor's, 14x Q2's FCF, and 1.6x cash on hand.
"Bounded" holds only for this case. A finding that Meta misled users on privacy is a template for other state AGs. You say nothing shows contagion, but our feed is headlines only, so that is an absence of evidence.
We don't know whether a jury or a judge made the call, since the sources disagree. We don't know the reserve or appeal-security requirements either. The "$260B → $40B → $25B" story comes from one unverified StockTwits post.
You concede retail is complacent, and "Who cares about $25B" is exactly that. Complacency is a risk, not a comfort.

5. The "healthy reset" read doesn't hold up¶

The 49% rally ran from the 7/30 low to the 9/24 peak and has no identified catalyst. The tools can't explain the +11.4% gap on 9/21, and there is no new filing since Q2. You say the market is "rewarding fundamentals," but the last fundamentals print sold off 8%.
Light pullback volume just before a binary event shows people waiting, not that sellers lack conviction. Today's 11.0M is the lowest in the window, so buyers haven't shown up either.
OBV fell ~87M over five straight down closes and sits near the window low. Today's high of 713.22 failed to reclaim the 10-EMA (717.97). Price is below the Bollinger midline, and the MACD histogram is −8.76.
Weekly and monthly SuperTrend being "up" is trivial after a 49% rally. The golden cross is, in the technical report's own words, a lagging artifact of one gap day.

6. The earnings history doesn't say what you say¶

The four ≥8% gaps were −11.3%, +10.4%, −8.5%, −8.0%, an average of −4.4%, with three of four down. Barron's "crushing average earnings-day returns" is a headline with no body text in our data.
Your "bought both times" sequence is off. The 519 low was 3/27, a month before the April print. After that print the stock fell from $611 to a $524 low in July, about 14%.
Your exit doesn't work. A $681 daily-close stop can't execute through a gap. A repeat of −8% to −11% from $708 lands around $628–$651, through $681, the $665 gap base and the $661 50 SMA, and down to the 200 SMA at $632.
A repeat of the best case, +10.4%, gets you to about $782, which is the $780 high where Zuckerberg and Cox just sold. That payoff shape is poor.

7. The positioning argument cuts both ways¶

You cite bearish prediction markets as contrarian fuel after admitting they're thin. Meanwhile retail is labeled 6–3 bullish with 740–765 targets, the tape rallied into the print on the SpaceX blowout, and IBD says "Meta Stock Climbs With Earnings On Deck." That isn't washed out.

The AI-bubble-burst odds are 9% for 2026, up 3.1 points on the week, but 20% by mid-2027. Alphabet is down ~$860B from its peak while Apple, which skipped the AI race, is near highs. The market is already sorting AI spenders by payback.

I'm not leaning on the Fed numbers because FRED was down and the markets are thin. But 97% "no cuts" and a priced hike aren't friendly to a long-duration multiple.

8. Insiders are a minor point¶

Your $102M-versus-$1.8T denominator is meaningless. What matters is behavior: no open-market buys in 12 months, Cox's $55M of sales starting in September (none before), and Zuckerberg selling $21.4M at roughly $775–779, essentially the top. A gift isn't conviction. This isn't decisive, but it isn't "nothing."

What would change my mind¶

Q3 operating margin back near 40% with a credible one-off explanation for Q2's opex step-up.
Flat-to-lower capex guidance alongside revenue growth of at least 28%.
FCF back above about $10B a quarter.
A clear reserve and appeal path on New Mexico, with no copycat state actions.

Bottom line¶

I'm not betting that revenue fades. I'm saying that at $708, before the print, the price pays for the best-case margin story. The recent gap distribution is negative, cash is thin, and the legal item is large relative to FCF. If the print shows margins rebounding, I'd rather pay up after seeing it than before. I wouldn't buy ahead of tonight. Bear Analyst: # Bear Rebuttal for META (price $707.95, 2026-10-21)

You gave real ground and I'll give some back. Then I'll show where your answers still leave the print as a coin flip with a price attached.

What I'm conceding¶

28.6x froze revenue. Your NTM framing is the right one, and your table's arithmetic checks out under your assumptions.
"Not funded from operations" overstated it. OCF covered capex in all six quarters. The financing swing is also partly a choice to return less capital. What survives is that coverage is razor-thin (1.06x in Q2) and management is prioritizing capex over returns.
n=4 can't establish a negative mean. I'm dropping "three of four down" as a pattern.
Q1'26 incremental margin was 38%. My H1 table blended a good quarter with a bad one.

1. Your table is my argument¶

I'll use your inputs: $275B NTM revenue, 87% net-to-operating conversion (the midpoint of your range) and 2.55B shares.

NTM op. margin	EPS	P/E at $708
28%	~$26.3	~27.0x
31% (Q2 repeated)	~$29.1	~24.3x
35%	~$32.8	~21.6x
37%	~$34.7	~20.4x

Each point of margin is worth about $0.94 of EPS. Going from 31% to 35% is about $3.75 of EPS, or roughly $80 a share at 21.6x. The margin question alone is worth the whole −$56 to −$80 you call tolerable.
The vendor's $34.65 needs about 36–38% on your revenue. Your own pass mark of 35% is slightly short of what 20.4x assumes.
A "no recovery" print is a −11% outcome in plain arithmetic. If the market is paying 21.6x for a 35% world and Q2's 31% repeats at that same multiple, the stock is about $628. That matches the −11.3% gap, so it isn't a tail.
A 28% margin gives ~27x even on your growth assumption. That is the risk if the Q2 step is the first of several.

2. "Step, not slope" doesn't survive the balance sheet¶

You're right that depreciation can't explain Q2's whole $7.5B sequential opex jump. I can't see D&A either, and talent costs or accruals could be in there. But look at the asset side.

Assets keep arriving. Total assets rose $84B from Dec '25 to Jun '26, and non-current assets rose $67B. Most of that has not yet been through a full quarter of expense.
Illustrative math, not a finding. $67B over a ~5.5-year life is about $3B a quarter at full run-rate. That is roughly 5 margin points on a $63B quarter, and it is already bought. It comes before any H2 capex.
Every candidate cause for the step is a problem. If Q2's opex is depreciation, it ratchets up. If it is AI compensation, it persists. If it is a one-off, we need the bridge to know.
"Costs only need to grow slower than revenue" is true, but Q2 opex grew ~65% against 28% revenue growth. The incremental margins of the last three quarters were about 12%, 38% and −13%. Q4'25 is derived from reported growth rates. Two of three were far below the ~41% average.
"Cash earnings power intact" ignores the capex line. OCF adds back depreciation on assets you've already paid for and must keep expanding. FCF margin was 2.9% in Q2 against 19.0% a year earlier. Q2 current liabilities also rose $9.6B, so some OCF may be working-capital timing. The data can't separate that.

3. The build is bigger than the capex line¶

H1'26 cash capex was $49B, but non-current assets rose $67B, even before any depreciation. Candidates include non-cash lease additions, investments, and capex accrued but unpaid. The tools can't say which. Cash also fell $20.4B in H1 to $15.5B, and the report leaves about $10.6B of that unreconciled. That is a reason to wait for the 10-Q, not a conclusion. But "1.06x coverage" measures one definition of the build.

4. Legal: your two branches cut against you¶

I accept your denominators. I never argued solvency. Two points remain.

The 5.5x squeeze. You note the stock is $139B below its 10/13 close, about 5.5x the headline. If that drop is the verdict, the market is pricing precedent, and "1.4% of market cap" is the wrong frame. If it isn't, the verdict is not in the price.
Your Q3 test may be contaminated. The verdict landed after quarter-end and before the 10-Q. Whether Q3 books an accrual depends on whether the loss is probable and estimable, and on Meta's view of its appeal. That is an inference I can't verify. If a charge is booked, up to $9.80 a share against Q2 EPS of $6.18 would hit GAAP. Government penalties are generally not tax-deductible, so verify before assuming a shield. Your "Q3 margin ≥35%" test then depends on whether management presents it ex-charge. Ask that question before reading the margin.
The "$260B → $40B → $25B" path is still one unverified StockTwits post.

5. On the print, you've argued yourself to zero edge¶

You say n=4 is noise for the mean but good enough to call the payoff "roughly symmetric." You can't have it both ways.

If the expected gap is zero, owning ahead earns nothing for the variance. Sizing caps the damage at 0.55% of the portfolio. It doesn't create an edge.
Your plan concedes this. You'd add on a hold of $681–$690 or a reclaim of $718–$730 after the numbers. Post-print entry is acceptable to you.
Your "what I need tonight" list is a conjunction. It needs an opex bridge, ≥35% margin, ≥23% growth, a capex guide that keeps FCF positive with OCF margin above 50%, and no copycat suits. Failing any one is on your kill list. The "FCF positive" bar is also one that Q2's $1.7B (0.4% yield) already clears.
Information is worth more than the $74. The answer to the margin question moves the stock by about $80.

The drawdown anchor. You cite a 10–15% drawdown. The stock is about 9% below the 9/24 high, and that peak came off a +11.4% gap on 9/21 that neither of us can explain. There has been no new filing since Q2. The stock is still 6.4% above its pre-gap close of $665 and 35% above the July low. You also didn't answer the Alphabet (−$860B) and Apple contrast, which says the market is already sorting AI spenders by payback.

Bottom line¶

I'm not arguing revenue fades or that Meta is broken. The stock is priced for a margin recovery nobody has seen, with a discretionary pre-print exposure that has no demonstrated edge and a legal item that can distort the number you're grading.

What would change my mind: - Q3 operating margin of about 37% or better, on a basis that is clear about any legal charge, with an itemized Q2 opex bridge. - Capex guidance with FCF back above about $10B a quarter. - A clear New Mexico reserve and appeal path, with no copycat state actions.

If that is what the call says, I'll pay up after the print and accept missing the $74 case. If it isn't, waiting saved the −$56 to −$80. At $708, I wouldn't own it into tonight. Bear Analyst: # Bear Closing for META (price $707.95, 2026-10-21)

You've given up most of the factual ground, so I'll be brief. I'll hand back what's fair, then show where the table, the "one question" framing and the plan still don't hold.

What I'm conceding¶

The starter is small. A 2.5% position through a −15% gap costs about 0.4% of the portfolio. I'm not saying it endangers anything, only that it earns nothing.
The OCF margin floor will probably clear. Q3'25 OCF margin was 58.5% against 53.8% in Q2'25, so seasonality favors Q3'26. I'm dropping that line.
The verdict itself is bounded. I never argued solvency.
"Step now, slope unknown" is the right frame. I'll use it below.
I'm not claiming depreciation explains Q2. I can't see D&A either. My understanding is that Meta spreads infrastructure costs across cost of revenue and the opex lines, so "below gross profit" doesn't rule depreciation out. That is general knowledge, not our tools, so check the 10-Q.

1. Your 35% pass mark rests on growth you retracted¶

Your $285B and $292B rows need 25–28% NTM growth. You've conceded that "accelerating" was wrong, so test those rows against your own two-year-stack framework. Holding 28% in every quarter requires:

Quarter	Prior-year growth	Two-year stack at 28%
Q3'26	26.2%	54.2
Q4'26	23.8%	51.8
Q1'27	33.1%	61.1
Q2'27	28.0%	56.0

That averages about 56, against 49.2 and 49.6 actually printed. It's a re-acceleration. Holding the stack at ~49.6 gives about 22–23% growth, or $275–281B. The Street's implied $101.6B of operating income is then 36–37%, not 35%. At 35% on ~$278B, EPS is about $33.20, roughly 4% under $34.65. Your "pass" is a small miss, assuming the vendor's EPS is NTM, which the data doesn't confirm.

2. At the pass mark, the price already assumes you pass¶

The 20.4x forward P/E is built on Street EPS. If Q3 delivers the margin the Street needs, the stock is at fair value by construction, around 21x on your table. For a +10% gap (~$779), you need one of two things: - EPS about 10% above the Street at today's multiple, which is roughly a 40% margin, back to Q1'26. - A multiple that expands on guidance.

A −11% gap needs only a repeat of Q2. The dollar payoffs are symmetric, but the conditions to reach them aren't. You're paid for beating consensus, not for the margin "coming back."

3. "One question" is really two, and a floor isn't a cleared bar¶

The slope is the guide, not Q3's margin. You say the step is now and the slope is unknown. A 36% Q3 margin with a 2027 guide pointing to renewed compression fails, because the multiple is built on forward numbers. Q3 margin alone can't settle it.
Your revenue floor of 23% sits on the base case. Holding the stack gives 23.4%, and extrapolating the last deceleration (−5.1 points) gives 22.9%. Clearing it in Q2 at 28% says little about Q3.
Your add rule doesn't discriminate. "Capex guidance keeps FCF positive" was met by Q2's $1.7B. If capex stays near $30B a quarter, FCF is only about $2B a quarter (~$8B a year, a 0.4% yield), and that passes.
You can't grade the margin cleanly. The New Mexico verdict was reported this week, after Q2 closed, so it can't be the Q2 one-off. Your third branch, the "one-off is the bull case," has no candidate in our data. A Q3 accrual is another matter. $25B on a ~$63B quarter is about 40 margin points if it lands in opex (that depends on probable-and-estimable and the appeal posture, which I can't verify). Management will choose what "ex-charge" means. You can't use Q3 margin as a rule until you know that.

4. The unexplained rally raises the bar¶

You've conceded that the 9/21 gap has no catalyst in our data, that the tail is wider than the −8% to −11% history, and that the stock is +31% since 7/30 with no new filing. In July the market sold a 28%-growth, 30.9%-margin quarter by 8%, at about 22x annualized Q2 EPS. Today it's about 29x on the same base.

When you can't explain a move, you can't say what's priced in. Either the market knows something we don't, or the move can reverse. Either way, expectations for tonight are higher than the ones that produced July's drop, and you're paying that higher price before the information arrives. A gap fill alone is −6% to $665, and an 8–11% gap lands at $630–651, through the $661 50 SMA and into the 200 SMA at $632.

5. Your plan has no edge¶

You've conceded that the expected gap is about zero and called the starter "regret management." Compare the stakes: - At a 2.5% size, missing a +10% gap costs about 0.25% of the portfolio. - A −15% gap costs about 0.4%.

Both are tiny, so the decision turns on information. Your post-print actions and mine are identical. Waiting gives up the +$74 case, which needs a beat. It avoids the −$56 to −$80 case, which needs only a repeat. With a 12-month thesis, a post-print entry 10% higher is cheap compared with the cost of being wrong on the slope.

What would change my mind¶

Q3 margin of about 37% or more, ex-charge, with an itemized Q2 opex bridge.
A 2027 expense and capex guide that doesn't extend margin compression.
Revenue growth comfortably above 23%.
FCF back above about $10B a quarter.
A clear New Mexico reserve and appeal path, with no copycat suits.

Bottom line¶

I'm not betting that revenue fades. I'm saying that at $708 the stock already prices the margin recovery the Street needs, that the good case is a match to consensus, and that the bad case needs nothing more than Q2 repeating. You concede there's no edge and no explanation for the run-up. If the print delivers, I'll pay up after it. If it doesn't, I'll have avoided the gap. I wouldn't hold META into tonight. Bear Analyst: # Bear Final Rebuttal for META (price $707.95, 2026-10-21)

We now agree on most of the structure. Below I concede what's fair, then add one piece of arithmetic that I think settles the pass mark.

What I'm conceding¶

The 35% vs 36-37% dispute is worth 1-2 margin points. I won't fight over it. We both use ≥36%.
Revenue isn't where I should attack. I derived Q2→Q3 sequential growth from the reported YoY rates. It was +3.9% in 2024 and +7.8% in 2025, which puts Q3 at about $63.2B to $65.5B (+23% to +28% YoY). My stack-hold base case sits at the low end of that range.
"FCF above $10B a quarter" was a bad bar. It would force me to guess capex. I'm replacing it below.
Your revised rules are better. They discriminate, and the "ungradeable if not itemized" clause is the right call.
The portfolio stakes are small. About 0.25-0.4% either way. We're arguing about process, not survival.

1. The pass mark is a cost-dollar problem¶

Total costs here are revenue minus operating income. Q2 costs were $42.0B, up $8.6B (+26%) from Q1's $33.4B. Here is what each margin outcome requires in Q3 (revenue bookends derived as above):

Q3 revenue	Margin	Implied total costs	vs Q2 costs
$63.2B	31% (Q2 repeats)	$43.6B	+3.8%
$63.2B	36% (your pass)	$40.4B	−3.8%
$65.5B	31%	$45.2B	+7.6%
$65.5B	36%	$41.9B	−0.2%

"No recovery" needs ordinary cost growth. Costs rise 4-8% sequentially, which is slow next to the last three quarters.
Your pass mark needs costs flat or down in dollars, right after a $8.6B step-up, with $450B of assets still being put to work.
Persistent costs can't fall that way. Depreciation and AI compensation, the candidates we've named, ratchet up. Only a real Q2 one-off of roughly $0.1-1.6B or more does it. You've agreed we have no candidate for one in the data.
A one-off may exist, and the 10-Q or the call may show it. But at 20.4x on the Street's EPS, the price already assumes the recovery. It doesn't wait for proof.

2. The slope has a record, and it points the wrong way¶

You call the slope unknown. We have three quarters of it:

Quarter	Revenue YoY	Total cost YoY
Q4'25*	+23.8%	~+40%
Q1'26	+33.1%	+35%
Q2'26	+28.0%	+55%

*Q4'25 and its prior-year comparison are backed out from reported growth rates and the ~48.3% prior-year margin, so treat the figure as approximate.

Costs outgrew revenue in every one. Q1 was close, and Q4 and Q2 were not.
Your 12-month bet needs a reversal of the only observed trend. Your add rule asks for a 2027 guide with expense growth at or below revenue growth. The data shows no quarter like that, and I don't see a reason for the guide to differ.
That doesn't make the bet wrong. It means the burden of proof is on the company, and it hasn't been met.

3. The conversion argument cuts both ways¶

Your 91% conversion case is fair. Clean periods ran 89.7%, 92.0% and 89.9%.

Those periods had far more cash. Cash was $43.9B at the end of 2024 and is $15.5B now, with liabilities up $79B in three quarters.
You offered the likely cause yourself: less interest income and a bigger interest bill. If that's right, reverting to 91% is the less likely path, and Q2's 84.4% is closer to the current run-rate.
At 87% and a 36% margin on $278B, EPS is about $34.0, roughly 2% under the Street, so that part is close. At 84.4% the same margin gives about $32.9, a 5% miss. This is an inference, and the 10-Q would settle it.

4. The starter's role in your own rules¶

All of your exit triggers are post-print. None stands between the position and the gap. Walking your own buckets:

Margin ≥36% ex-charge, bridge, benign guide, revenue ≥$64B: you add. The stock already prices the Street's margin, so this is fair value by construction. The starter collects a +10% move only if the print beats. Per §1, that needs costs flat or falling in dollars.
Margin 31-36%, credible bridge: you hold without adding. The starter rides the multiple compression that we agree is the −11% case.
Anything failing your exit list: you sell after the gap, locking in the loss.

The starter earns its keep in one bucket, and that bucket requires the unusual outcome. The other two buckets cost it money. Meanwhile your add rule and mine are nearly identical, so the choice is just whether to own the stock for the gap. Waiting for the evidence costs the beat-and-re-rate case. For a 12-month slope thesis, I'd take that trade.

Revised bear buy list (replacing the FCF bar)¶

Q3 operating margin ≥36% ex-charge, with an itemized Q2 opex bridge.
A 2027 guide where total expense growth and capex growth are at or below revenue growth.
Revenue ≥~$64B.
A clear New Mexico reserve and appeal path, with no copycat state actions.

That list now matches your add conditions closely, which is why I consider the debate converged on substance.

What would change my mind on owning it tonight¶

Only information I can't get before the close: - A credible Q2 one-off, such as a charge in the 10-Q I haven't seen. That would make the cost arithmetic in §1 achievable. - Evidence that the 9/21 gap and the +31% rally since July reflect upward estimate revisions. I'd want to know what the Street expects before judging whether the bar has risen further.

Bottom line¶

Revenue isn't my objection. Growth of 23-28% at $60B+ a quarter is real. At $708 the price already assumes the margin recovery, and the recovery requires Q3 costs to fall in dollars after a 26% sequential jump, after three straight quarters of costs outgrowing revenue. A repeat of Q2 needs only normal cost growth. The payoffs are symmetric in dollars, but the conditions that produce them aren't.

You said waiting is reasonable and the exposure is small. I agree on both. The first real evidence on the slope arrives after tonight's close, so I would buy it then. I wouldn't hold META into the print. Bear Analyst: # Bear Final Response for META (price $707.95, 2026-10-21)

Your corrections are fair, so I'll take them first. Then I'll show where your NTM table depends on one assumption that last year's data contradicts.

What I'm conceding¶

$33.1, not $32.9. 36% on $278B at 84.4% conversion is $33.13, about 4.4% under the Street. That's your number.
I applied an NTM figure to one quarter. A 31% Q3 can sit inside a Street-consistent year if costs flatten afterward. "Costs must fall in dollars in Q3" was too strict. Q3 alone isn't the test.
Your gap table undercuts my buckets. I'm assuming the four gaps are the prints. Reactions were −11% on a 40.1% margin, +10% on a quarter with margin down 7 points YoY, −9% on 40.6%, and −8% on 30.9%. The 30.9% quarter was sold the least, so margin level doesn't explain the moves. I'm dropping "a Q2 repeat is −11% in plain arithmetic." That assumed a constant multiple, and the multiple moves with the guide.
Costs are partly a choice. FY24 costs rose 7.9% on revenue up 21.9%. That precedent is real.

1. Your table reaches 36% by letting Q4 costs barely move¶

Your Q4'26 has revenue up 15% sequentially and costs up 3%. In Q4'25, revenue rose 16.9% (to $59.9B) and costs rose 14.5%, from $30.7B to $35.1B. That is nearly one-for-one. Holding your revenue path and your Q1 (−3%) and Q2 (+4%) cost steps off whatever Q4 base results, and using 87% conversion and 2.55B shares:

Q4'26 cost growth	NTM margin	NTM EPS	P/E at $708
+3% (your table)	36.2%	$34.3	20.6x
+10%	32.9%	$31.2	22.7x
+14.5% (Q4'25 pattern)	30.8%	$29.2	24.2x

Your row reproduces the Street's $34.65. That means the Street's number needs Q4 costs to ignore the biggest revenue quarter of the year.
A Q4'25-style cost pattern leaves you at Q2's margin. That is roughly 24x, which you've called "fair, not cheap."
Caveats: Q4'25 is derived (fiscal year minus nine months), and it's the only Q4 cost seasonality in our data. Some of that rise may not repeat, and some Q4 costs are revenue-linked by construction. This is a stress test, not a forecast. It does show that "costs grow slower than revenue" means something specific in your table, and last year's Q4 didn't do it.

2. The undated EPS doesn't rescue the calibration¶

You built the table so that $34.65 is calendar NTM. That is the assumption in dispute. - If $34.65 is FY26, H2 needs $18.03 ($34.65 − $16.62). Your table's H2 operating income of $47.4B gives about $16.2, roughly 10% short. - If it's FY27, the 20.4x is a multiple on earnings that haven't started.

Either way, "20x" isn't a clean NTM multiple. I can't tell you which reading is right, and neither can you.

3. What the starter is for¶

Your gap table implies the reaction turns on guidance, and neither of us can see it. You've said you have no demonstrated edge on the gap. The stakes are about 0.2–0.3% of the portfolio, so this is a process disagreement, not a risk one. - My process: don't carry exposure to an outcome I can't read when the information arrives within hours. If the guide is good, I pay up and accept missing a +10% day. - Your process: hold a small starter and add on evidence. - Ambiguity: a New Mexico accrual could land in Q3, and under your own rule an unitemized quarter is "ungradeable, don't add." In that case the starter has carried the gap and earned nothing.

What would change my mind (post-print)¶

Q4 implied costs. If management gives a full-year expense range, back out Q4. Up about 5% or less from Q3 actual is your table's path. Up 10% or more is the Q4'25 pattern. (The expense-guide format is general knowledge, not in our feed.)
Q3 margin ex-charge of about 35% or better, with an itemized Q2 opex bridge.
A 2027 guide where total expense growth and capex growth are at or below revenue growth.
Revenue of about $64B or more, and a clear New Mexico reserve and appeal path.

Bottom line¶

We now agree on the facts: 23–28% growth at $60B+ a quarter, OCF up YoY, Q2's margin a step, and a slope that tonight's guide tests. What remains is that the Street's EPS needs costs to decouple from seasonality, which they didn't last Q4. Without that decoupling the stock is at 24x with higher spending ahead. Your 2% starter is sized sensibly for that risk, and I don't call it reckless. But the stock already pays for the best case, so I'd rather buy it after the guide than hold it into the print.

Research Manager¶

Recommendation: Underweight

Rationale: This was a high-quality debate, and by the end both sides agreed on most of the facts. The disagreement narrowed to one question: do you carry exposure into tonight's print, and how much? On the business, the bull's case is real. Revenue grew 28% to $60.8B in Q2, after 33% in Q1, on a two-year stack holding near 49.6. Gross margin is about 81%. Trailing operating cash flow is about $130B, and OCF covered capex in all six quarters shown. The $25B New Mexico judgment is about 10% of equity and 1.4% of market cap, so it is bounded on its own. The bull also correctly showed that n=4 earnings gaps say nothing about the expected reaction. The gap table (−11% on a 40.1% margin, +10% on a margin-down quarter, −8% on 30.9%) shows the reaction turns on guidance, not margin level. The bear also had to retract his 28.6x 'frozen revenue' multiple and the 'not funded from operations' claim. The bear nonetheless won the arguments that decide sizing at $707.95 right now. First, the price already assumes the recovery. The 20.4x forward multiple rests on Street EPS of $34.65 (undated), which needs roughly a 36% NTM operating margin, against Q2's 30.9% (down from 43%). The bull conceded the pass mark and that a match to consensus is fair value by construction. A +10% move needs a beat or a re-rating. Second, the cost slope has a record and it points the wrong way. Costs outgrew revenue in Q4'25 (~40% vs 23.8%), Q1'26 (35% vs 33%) and Q2'26 (55% vs 28%). Q2 costs rose $8.6B sequentially. Q2 FCF was $1.7B, a 2.9% FCF margin, on $30.1B of capex, against 19% a year ago. The bull's NTM table reaches 36% only by letting Q4 costs rise 3% on 15% revenue growth. Q4'25 costs rose about 14.5% on 16.9% revenue growth, and repeating that pattern puts NTM at roughly 30.8% margin and about 24x. Third, the 31% rally since the 7/30 close (from about 22x to about 29x annualized Q2 EPS) has no identified catalyst, including the unexplained +11.4% gap on 9/21. Neither side can say what is priced in. Fourth, the bull conceded there is no edge on the gap. The bull also concedes that stops cannot execute through it, and the decisive information arrives within hours. Secondary bear points add weight: cash is thin at $15.5B, liabilities are up $79B in three quarters, there have been no insider open-market buys, the bubble odds are rising, and the market is sorting AI spenders by payback. The 10-Q may also carry a legal accrual that makes Q3 hard to grade. None of this says the business is broken, and the bull is right that at a 31% margin META is roughly fair (about 24x for about 22-25% growth) rather than a bubble. That is why this is not a Sell. But the bull's own final position is a 2% starter against a 5% target, which is itself an underweight. The bear's argument that the stock already prices the best case, with the cost trajectory unproven, carried the debate. So the call is cautious: hold below standard weight into the print, then rebuild only on evidence about the cost slope.

Strategic Actions: (1) Before today's close, reduce exposure. Earnings are reported after the bell, so any trim must happen in regular-session trading today. Holders at or above a standard allocation should trim to about 40% of standard weight (for example, 2% on a 5% standard weight). Holders already below that level can sit tight. If you are flat, the preference is to wait rather than open new exposure before the print. Do not exceed a 40%-of-standard starter. (2) Size for a −15% gap, which takes the stock to about $600. Do not rely on stops: the $681 SuperTrend and the $665 gap base cannot protect against an overnight move. At 40% of standard, a −15% gap costs about 0.3% of the portfolio. (3) Post-print rebuild to standard weight, in tranches over 1–2 weeks rather than chasing the open. Move to about 70% of standard if all of the following hold. Q3 operating margin is at least ~36% excluding any New Mexico accrual, with the accrual itemized. If it is not itemized, treat the quarter as ungradeable and do not add. There is a credible opex bridge separating persistent costs (depreciation, AI compensation) from one-offs. Revenue is at least ~$64B. Then move to 100% of standard if the 2027 guide shows total expense growth and capex growth at or below revenue growth. Also back out implied Q4 costs from any full-year expense range: up about 5% or less from Q3 actual supports the recovery path. Paying up toward the $760–780 area is acceptable if these pass. If the stock gaps down but these criteria pass, add into $630–665 (gap base, 50 SMA at $661, 200 SMA at $632). (4) Hold at the reduced weight, without adding, if margin is 31–36% with a credible bridge and a neutral guide, or if implied Q4 cost growth is 5–10%. (5) Exit to zero on any of the following: margin near 31% with no bridge; guidance that opex keeps outgrowing revenue; implied Q4 cost growth of 10% or more (the Q4'25 pattern was ~14.5%); revenue below ~$62.5B; a capex raise alongside falling OCF margin; copycat state AG actions after New Mexico; or a post-print daily close below $681 that extends toward $661. (6) Open items to verify, since they are missing from our data: the period behind the $34.65 forward EPS (FY26, FY27 or NTM); the D&A breakdown and its placement across cost lines in the 10-Q; net-to-operating conversion (Q2 was 84.4% versus about 90% in clean periods); the ~$10.6B unreconciled H1 cash decline; the catalyst for the 9/21 gap and any upward estimate revisions; and New Mexico verdict details (judge or jury, reserve, appeal bond). The trader and portfolio manager should apply these weights relative to the actual holdings.

III. Trading Team Plan¶

Trader¶

Action: Sell

Reasoning: The research plan's Underweight maps to a Sell, which here means trimming exposure ahead of the print and not liquidating. At $707.95 the 20.4x forward multiple already needs roughly a 36% NTM operating margin, against Q2's 30.9%. Costs have outgrown revenue in three straight quarters, and Q2 FCF was only $1.7B on $30.1B of capex. The 31% rally since 7/30 has no identified catalyst. Technically the higher-timeframe trend is intact (price above the rising 50 and 200 SMAs, SuperTrend up on all tiers). But price sits below the 10-EMA (717.97) and the Bollinger midline (730.06) with a negative MACD histogram, and OBV is weak. The daily SuperTrend stop at 681.41 is only about 1.1 ATR away, and four prior earnings gaps of 8-11% (2.3-3.2 ATR) mean stops cannot protect through the print, so the cut has to be made before the close. The business is not broken and the stock is roughly fair at a 31% margin, which is why this is a trim and not a full exit.

Entry Price: 707.95

Stop Loss: 681.41

Position Sizing: Before today's close, trim to about 40% of standard weight (e.g., 2% on a 5% standard). Holders already below that level can sit tight, and flat accounts should wait and not open new exposure before the print. Size for a -15% gap to about $600, which costs about 0.3% of the portfolio at 40% of standard; do not rely on stops. Post-print, rebuild in tranches over 1-2 weeks only if the plan's criteria pass: about 70% of standard if Q3 operating margin is at least ~36% ex-accrual with revenue of at least ~$64B, and 100% if the 2027 guide shows expense and capex growth at or below revenue growth. Add into $630-665 on a gap down only if those criteria pass. Exit to zero on margin near 31% with no bridge, implied Q4 cost growth of 10% or more, revenue below ~$62.5B, or a post-print daily close below 681.41 extending toward 661.

FINAL TRANSACTION PROPOSAL: SELL

IV. Risk Management Team Decision¶

Aggressive Analyst¶

Aggressive Analyst: I'll open, since nobody else has spoken yet. The trade is a Sell, meaning trim META to about 40% of standard weight before the close, and I think it's the aggressive call, even though "sell" sounds timid. Aggression means taking the asymmetric bet. Here the asymmetry sits in having cash and conviction after the print, not in holding full size through a coin flip.

The stock is up about 31% since the 7/30 close of 538.60. It gapped 11% on 9/21 on 48.7M shares with no catalyst my tools can identify, and it has already given back about 10.7% from the 9/24 peak close. At 707.95 the forward multiple is 20.4x, and the trader's math says that needs roughly a 36% operating margin over the next twelve months. Q2 printed 30.9%, down from 43.0% a year earlier. The vendor's forward EPS of $34.65, if it covers FY2026, means H2 has to run about $9 a quarter against Q2's $6.18, a 46% jump. Q2 capex was $30.1B, about 94% of operating cash flow, leaving $1.7B of free cash flow. Liabilities are up $79B in three quarters while cash sits at $15.5B. The market is paying a full price for a recovery nobody has seen yet.

Now the conservative case, which I expect to be "exit completely." I'd resist that, because it throws away the bullish branch. The weekly and monthly SuperTrends are up, with the weekly stop 24% below price. The golden cross happened on 10/6, revenue is growing 28% on a $60.8B base, and operating cash flow is up about 25%. Nobody knows yet whether the Q2 opex jump was structural or a one-off. Q1 ran 40.6%, so the 36% bar isn't fantasy. If management bridges the margin and shows 2027 cost growth at or below revenue growth, this stock can retest 760 to 780. The prediction markets still put about 40% on a touch of 760 and 18% on 800, thin as they are. Going to zero means watching that move with nothing on. The 40% stub keeps me in the game and costs very little.

The neutral camp will say the trend is intact, the stock is roughly fair, so hold full size and let the 681.41 stop do the work. That stop doesn't work here. It's 26.5 points, about 1.1 ATR, below price, while the last four earnings gaps were 8% to 11% (2.3 to 3.2 ATR). Three of those four were down. A -11% gap lands near 630 and a -15% gap near 600, so the stop is irrelevant, and you take the fill at the open. Trend indicators are also lagging. The golden cross came off a single gap day, and a weekly stop 24% away says nothing about tomorrow morning. Short-term evidence is leaning the wrong way too: price is below the 10-EMA at 717.97 and the Bollinger midline at 730.06, the MACD histogram is -8.76, and OBV is 329.7M against a 421.4M peak. The prediction markets moved the same way in a week. Odds of touching $680 rose 17 points, and odds of touching $760 fell 29.5 points. The light-volume pullback isn't comfort either. Volume always thins before a binary event, and today's 11.0M may be an incomplete bar.

The neutral analyst may also say "wait and see." Doing nothing is a decision to carry 100% of the gap risk, and the fundamentals and tape give no clear edge on which way it breaks. I don't want to be paid for that risk at this price.

On the numbers, a -15% gap costs about 0.3% of the portfolio at 40% of standard weight, versus 0.75% at full weight on a 5% standard. That saves about 0.45% in the bad case. If the stock gaps up 10% instead, the trim forgoes about 0.3% and I rebuild at higher prices. I'm paying a small, known premium for the right to act decisively after the print.

The aggressive part comes next. The add zone of 630 to 665 sits on top of the 50 SMA at 661.41 and the 200 SMA at 631.78, so if the stock gaps down and the criteria pass, I buy hard into support at the moment of maximum fear. Otherwise I rebuild to 70% on a margin of at least ~36% ex-accrual with revenue of at least ~$64B, and to 100% if the 2027 guide shows expense and capex growth at or below revenue growth. I exit to zero on a margin near 31% with no bridge, implied Q4 cost growth of 10% or more, revenue below ~$62.5B, or a daily close below 681.41 extending toward 661. That is conviction in both directions. The accrual language matters because of the New Mexico $25B judgment. If it gets booked, it would swamp GAAP operating income of roughly $19B to $23B a quarter, so I judge the margin ex-accrual. The reports disagree on jury versus judge, and appeal timing is unknown.

Retail looks complacent about the judgment ("who cares about $25B") and is anchoring on 740 to 800 post-print targets. Only 9 of 30 StockTwits messages were labeled, so I wouldn't lean on that either way. Cox sold $55M and Zuckerberg $21.4M near 775 to 779, and nobody has bought on the open market in twelve months. Much of the insider selling is probably routine, so I weight it lightly, but it isn't a vote of confidence. Macro adds a headwind for long-duration names: the prediction markets price no 2026 cuts at 97% and a December hike at 78%, and France's 10-year hit 5%. FRED was down, so I'd verify those against actual yield data.

Two checks before acting. Confirm the earnings timing, since the sentiment feed says after the close today but the other tools couldn't confirm it, and the last price bar may be provisional. If the print is already out, the post-print rules govern, not the pre-print trim. Also, I don't know the caller's portfolio. If you're already at or below 40% of standard weight, sit tight, and if you're flat, wait for the print.

My recommendation is to trim to about 40% of standard weight before the close, then rebuild or buy the dip on the post-print criteria.

FINAL TRANSACTION PROPOSAL: SELL Aggressive Analyst: Let me give ground where you've earned it, because it makes the part I'm holding firm on easier to see.

Neutral, you're right that I over-built the tape case. Price under the 10-EMA, a negative MACD histogram and soft OBV are rounding errors against an event where the last four gaps ran 8 to 11 percent, and the technical report itself calls the posture mixed. You're also right that the prediction-market shift is mostly mechanical. Spot slid from the low 760s to 708, so the touch odds just followed it. The ladder is close to symmetric, about 30 percent to touch 640 and about 31 percent to touch 776. I'm dropping all of that. The trim rests on two things only: a stop that can't protect me through a gap, and a 20.4x forward multiple that needs a margin the company hasn't printed since Q1.

I'll also concede my $64B revenue bar was loose. It's about 25 percent growth off $51.2B, and the trend is 28 percent, so roughly $65.6B. I'll move the rebuild bar to about $65B. Your three fixes to the post-print rules are good:

The 32 to 36 percent middle needs a rule. Hold the trimmed size unless management gives a credible bridge, and then add only a first tranche.
The 681.41 close-to-zero rule is stale after any real gap, so reset it off the new structure.
A gap into 630 to 665 gets one third of the planned rebuild on day one, with the rest after the gap-day low holds.

Your history on that last one is persuasive. The October 2025 gap to 664 and the April gap to 611 both went on to lower lows, and only July's gap was the low. So "buy hard" was the wrong phrase. Decisive means pre-written, not all at once.

Now where I push back.

Conservative, you want 20 percent or flat because the stub has no edge. Neutral's arithmetic shows what that is worth. Twenty points of standard weight is one percent of the portfolio on a five percent standard, so a 15 percent gap on it is 0.15 percent. You're calling a tenth of a percent the difference between boring and not. That's a loss-budget choice, as you said yourself. I'll take 35 as the number, and 25 to 30 for anyone with correlated mega-cap exposure elsewhere. But "no edge, therefore zero" proves too much. The upside tail is just as real as the downside tail, and Neutral's bracket leaves a real slice of outcomes at 36 percent or better. Being fully flat into a symmetric print and then chasing a gap-up isn't safer, it just makes the decision at a worse price.

Your -24 percent round trip is a three-week retrace, not one night, and the worst print-day move in the data is -11.3 percent. The exit rules exist for the retrace. Size tonight for it and you pay for it twice.

Your cost-base math is the best work in this debate, and it's why I won't go bigger. But look at what it says. Costs went +4.4B, then -1.7B, then +8.6B. That last step is out of pattern. Flat costs of $42.0B on $65.6B of revenue gives about 36 percent. So the bar sits between a one-off scenario near 40 percent and a no-one-off scenario near 33 percent. That's a genuine toss-up, which is the case for a trim and against an exit.

You say the top tier, expense and capex growth at or below revenue growth, asks for a reversal of the whole trend. Look at the dollars. If Q3 and Q4 each look like Q2's $30B, FY2026 capex is about $109B. That's a simple extrapolation, not guidance. Capex growth in the mid-20s on that base is still $25B or more of additional capex in 2027. That's a deceleration from 60-plus percent, not a cut. It's hard, which is why it's the 100 percent tier and not the 70 percent one.

I'd also drop "roughly fair at a 31 percent margin" from the trader's wording. You're right that it sits badly against the 36 percent math. At 31 percent the stock is priced for a recovery nobody has seen. That justifies the trim, but not an exit, because the recovery is plausible.

On the $25B, I'll move part of the way. Neutral showed it's under a fifth of trailing operating cash flow and about 1.4 percent of market cap, so it's a headline and leverage-direction risk, not a funding scare. I'd still judge the operating margin ex-accrual, because a booked charge would swamp the read. But I'll add a separate check on timing, any debt issuance tied to it, and any change to the capex plan.

One more fact cuts across everyone's rebuild rules. If those four gap days are the earnings reactions, they don't line up with the margin. Q3'25 printed 40.1 percent and gapped -11. Q1'26 printed 40.6 and gapped -9. Q4'25 printed 41.3 and gapped +10. Q2'26 printed 30.9 and gapped -8. The reaction has been about capex and forward costs, not the trailing margin. So a margin pass alone only earns the first rebuild tier, and the 2027 guide decides the rest.

On execution, confirm the print timing first. The sentiment feed says after the close today, nothing else confirms it, and the last bar looks provisional. If results are already out, none of the pre-print trim applies. I don't know the caller's book, so the guidance is relative. Trim to about 35 percent of standard weight, go to 40 if there's a large embedded tax gain, and sit tight if already at or below that. If flat, wait for the print.

The aggressive move here isn't carrying the whole gap. It's being the one with cash and written rules when the print lands.

FINAL TRANSACTION PROPOSAL: SELL Aggressive Analyst: Let me start with what I'm taking from both of you, because it's more than last round.

Neutral, the Q4 cost-growth exit is the best catch of the round, and it was my rule, so I'll say plainly that it's broken. Last year total costs went from about 30.7 billion in Q3 to 35.1 billion in Q4, up 14.5 percent. Revenue rose 16.9 percent, the margin went to 41.3, and the stock gapped up 10. A flat 10 percent cap would have sold me out of the best reaction in the table. I'm replacing it with yours: exit if the guide midpoints imply a Q4 margin below Q3's actual with no named one-off. I'd run it ex-accrual, since the full-year expense guide gets contaminated if the New Mexico charge is booked.

I'm also taking your point on costs. Flat costs give exactly 36 percent, but flat isn't the no-news case. Variable cost grows with revenue and depreciation keeps climbing, so I called it a toss-up and "less than a coin flip" is the better description. That strengthens the trim rather than weakening it.

Conservative, your best point is that tonight's size is probably the size we carry afterward. Neutral's conditional hold answers it. Hold at 30 to 35 if the guide implies a Q4 margin at or above Q3's, or if management names a one-off in Q2 costs. Step down to 20 or 25 if the guide implies lower with no explanation. The smaller carried size then shows up in the branch where the cost question resolves against us, and nobody pays for it in the branch where it resolves for us. I'd make that step-down in the next regular session after the call, not into after-hours quotes.

I do push back on "the tape trades the headline, and you're reading ex-accrual." You cited the -11.3 percent day, with 20.5 billion of operating income and 2.7 billion of net income. Now look at last April's -9 percent day. Net income was 26.8 billion against 22.9 billion of operating income, a 3.9 billion benefit. So one big down day came with an 17.8 billion hit below the line, and another came with a benefit below the line. Two data points aren't a model, but they don't show the tape keying off the net income line in either direction. That isn't comfort. It's why I'm sizing by loss budget and not by a prediction of which way it breaks.

On the $25 billion, I accept your arithmetic. If Q3 and Q4 capex match Q2, free cash flow for the year is about 19 billion, so the judgment is roughly a year of FCF on that path. That's a reason to ask three things on the call: whether it's accrued, whether there's debt behind it, and whether the capex plan moves. It's also a reason the stub is 30 to 35 and not 40-plus. It isn't a reason to be at 20 before we know. I'd keep two separate tests. Read the margin ex-accrual, and treat the charge as a real liability when you look at the cash and capex plan.

On correlation, you say the carve-out is the default because everyone owns index funds. META is a small slice of an index fund. What stacks risk is owning other AI-capex names directly, and that's a book-level check. If it fails, the fix may be cutting somewhere else instead of cutting META harder. On tax, Neutral's roughly 9 basis points is about what the extra selling saves on the worst print in the data, so it's close to a wash and the caller's situation decides.

Neutral, you're right that "cash and written rules" doesn't separate 20 from 30. Either size leaves far more dry powder than a first tranche needs. The written rules are what give the optionality, not the stub. The stub keeps the bullish branch from being a regret event, and it costs somewhere between 6 and 9 basis points more than 20 on the worst print on record. The ladder is symmetric, so it earns about the same on a gap up. I'll meet you at 30, with 35 on a clean book, and 20 for a tax-deferred account or a tight loss budget. Your formula is the right way to hand it over: take the loss you'd shrug off on a -15 percent gap and divide by 15 percent.

On the look-through, agreed, and I'd add that the biggest leg of the rally, the 9/21 gap, had no catalyst I can find, and it wasn't a better margin. The market paid for the recovery before seeing it. That's the case against exiting. It's also the case for trimming, because this print has to deliver what the price already assumes.

One cleanup on the rebuild, because the 100 percent rung got lost in your package. Take the first third of the rebuild only if margin is at least 36 ex-accrual, revenue is about 65 billion or better, and the guide doesn't imply a falling Q4 margin. Reach 70 percent in thirds over one to two weeks if the call and the 10-Q hold up and the stock holds the gap-day low. Go to 100 only on a 2027 guide with expense and capex growth at or below revenue growth. Revenue between 62.5 and 65 billion, or margin between 32 and 36, means hold. Exit on margin near 31 with no bridge, revenue under 62.5, or an implied Q4 margin below Q3's with no named one-off. The 681.41 close rule is retired, and I reset the stop off the post-print structure.

Finally, a practical gap nobody has covered. All of this assumes there's still regular session left. Today's bar looks provisional, which suggests we're mid-session. If the bell has rung and results aren't out, "no after-hours orders" means carrying the entire gap. I agree with that rule after the print, when spreads are widest and information is thinnest. Before the print, I'd rather work a limit order than carry full size through a gap the whole panel says to avoid. I have no spread data, so that's a judgment call. Also confirm the print timing first, because only the sentiment feed says after the close today. If results are already out, none of the pre-print trim applies and the post-print rules run.

So for META, I'd trim to about 30 to 35 percent of standard weight, scaled by the caller's loss budget. If you're already at or below that, sit tight, and if you're flat, wait for the print. The aggressive part isn't carrying a gap you can't price. It's showing up after the print with cash, a written plan, and the discipline not to buy the first red candle, which history says was early two times out of three.

FINAL TRANSACTION PROPOSAL: SELL Aggressive Analyst: Conservative, Neutral, I'll take the corrections first, because several of them are right.

Conservative, you're right about the limit order. A trim's failure mode should be benign, and a resting limit that doesn't fill is the one outcome this whole exercise exists to prevent. Enter a marketable limit now, at or just below the bid, and check the fill before the bell. Today's tape helps. META bounced about 2 percent off yesterday's 694 close, ran into 713.22 (the level retail is already calling a double top), and is still under the 10-EMA at 717.97. Selling into a strong tech tape beats selling the same shares after the print. That's execution, not thesis, but it's free.

You're also right that the conditional hold only protects the leg after the gap, so tonight's number really is the number. The sentiment report puts the $25B on earnings day, so I withdraw anything that sounded like "priced in." At about $9.80 a share against Q2 EPS of $6.18, a booked charge could make the headline quarter ugly. Whether it gets booked is an accounting call on whether a loss is probable and estimable with an appeal live. That's my understanding, and I can't verify it here, so it's the first question for the call.

I'll take Neutral's bands as written, including the rule that if I can't compute the ex-accrual margin from the release, I'm in the hold band and add nothing until the filing quantifies it. A "named one-off" has to carry a dollar figure. I'm retiring the 10 percent cost-growth exit and the 681.41 close rule. In the hold band, a falling implied Q4 margin with no quantified one-off steps me down to 20 or 25 percent of standard. Below 32 percent with no quantified bridge, or revenue under $62.5B, I'm out.

Now where I hold.

Conservative, you adopted the loss-budget formula and then put a ceiling on top that overrides it. Your own threshold is 0.22 percent on a minus 15 percent gap. That's a 1.47 percent position, or 29 percent of a 5 percent standard. Then you cap at 25 and make 30 conditional on a large embedded gain. A tax gain is a cost of selling, not a measure of risk tolerance. Either the loss budget sets the size or it doesn't.

What we actually disagree on is one input, the gap you underwrite. Divide by 15 percent and a 0.25 percent budget gives a third of standard. Divide by 25 percent and it gives a fifth. I'd underwrite 15, because it already sits above the worst print in the record (minus 11.3 percent), and the worst margin print in the table drew only an 8 percent gap. That was Q2's 30.9 percent against 43.0 a year earlier. The thin ladder also puts the same 18 percent on a touch of 620 and a touch of 800, so it doesn't see a fatter left tail. Your 2022 memory is fair context, but revenue was stalling then and is growing 28 percent now. I'd rather not litigate it. Hand the caller both denominators, and if they want to survive a 2022-style night, they divide by 25. The gap between 30 and 20 is about 12 basis points at that tail, about 6 on the worst print in the data, and roughly the same size in the other direction on a gap up. That's a matter of taste, not evidence.

Your "when I can't verify, default to smaller" rule proves too much. I can't verify the print or the upside tail either. If unverifiability alone cuts size, the rule has no stopping point short of zero, and you stopped at twenty.

On the book check, you and Neutral are right that Amazon reports this week and Alphabet is down about $860B from its peak. The AI-capex read-across is live, and the tools can't show these gaps were stock-specific. That's a one-minute question for the caller. The answer is 30 percent of standard if they hold no other direct AI-capex names, and 25 if they do or can't say.

Your first-tranche gate list has grown to margin, revenue, implied Q4 margin, dollar-quantified one-offs, capex run rate and judgment funding, plus waiting a session and a hold of the gap-day low. A list that long never opens. Neutral's numeric version is the right one: Q3 capex not above roughly Q2's $30B, no implied Q4 step-up, and any debt tied to the judgment sized in the filing. The first tranche is only about 13 points of standard, so the heavy cash tests belong at the 70 and 100 percent rungs. The 10-Q usually lands within a day or two of the release (from memory), so waiting for it before the first tranche costs little.

On the collar, I have no options data. A collar also caps the upside, which is the thing the stub exists to keep. Pick the highest-basis lots first, check the one-year mark and wash-sale rules, and only then ask whether tax justifies carrying more.

Here is the plan. Confirm the print timing first, since only the sentiment feed says after the close today and the last bar looks provisional. If results are out, none of the pre-print trim applies and the post-print rules run. If the print is still ahead, trim to about 30 percent of standard through a marketable limit in the regular session, 25 if the book has other direct AI-capex exposure, and 20 if you divide by 25 instead of 15. Take no after-hours orders in either direction. Anyone already at or below their number sits tight, and anyone flat waits.

After the print, the first third of the rebuild needs 36 percent or better ex-accrual, revenue around $65B, no falling implied Q4 margin, and the capex gate. It goes in at the next regular session once the gap-day level holds, whether the gap is up or down. Rebuild to 70 percent in thirds over one to two weeks, and go to 100 only on a 2027 guide with expense and capex growth at or below revenue growth. Reset the stop off the new structure.

That is the aggressive position. You shouldn't carry a gap you can't price or hold a stop that can't act. Show up after the print with cash and a written plan, and be bold on the branch where the numbers actually come in.

FINAL TRANSACTION PROPOSAL: SELL Aggressive Analyst: Between you two, the argument is nearly settled, so I'll start with what I'm taking.

Neutral, I oversold "asymmetric." Going from full weight to 30 percent moves 3.5 percent of the portfolio on a 5 percent standard. That saves about 0.4 percent if the stock repeats the worst night in the window and costs about 0.35 percent if it repeats the best. The ladder is symmetric, and the trailing margin hasn't predicted any of the four reactions. So this is a variance cut, not an option. What I'd still claim is about where the risk gets spent. Tonight's exposure is a coin flip I can't price. After the print, the exposure is to what the guide says about 2027 costs, which is what the 20.4x multiple depends on. I'd rather spend my variance there, though that's a judgment, not a measurement.

Conservative, target weights are the right way to write the post-print table, and Neutral's ceiling-and-pace fix stops it from churning anyone. The cost is that someone who missed the trim sells to the ceiling at the next open, after the gap. That's the best argument for entering the marketable limit now and checking the fill before the bell.

I'm also taking your three tiers. "Can't say" and "does hold" shouldn't be the same number, so I'd use 30 for a clean book, 25 for an unknown one, and 20 for a known overlap or a tight budget, with Neutral's definition of overlap. One correction to my own shorthand: tax-deferred isn't a tier. The loss-budget formula sets the number, and taxes only decide what it costs to get there.

I'm fine with the other fixes too: - If the release is muddied by the accrual, hold the smaller of tonight's size and 25. - Wait for Amazon once and no further. - Allow a few billion of tolerance for finance-lease principal in the capex comparison. - Use Neutral's slope on the low end. A margin of 30 to 32 with no dollar bridge goes to about 15, and below 30 or revenue under $62.5B goes to zero. - Fire the implied Q4 margin trigger only when it sits a point or two under Q3's actual.

Now I'll push on two things.

First, the capex line in your 70 percent test isn't as heavy as advertised. H1 capex of $49.1B plus two more quarters at Q2's $30.1B is $109.3B. So "FY2026 pace no higher than $109B" means "Q3 and Q4 no higher than Q2," which is the same bar as the first-tranche gate. You said passing that only means no worse than the worst free-cash-flow quarter in six, then put it on the rung where you wanted something stronger.

The real new content in your 70 test is the other two items: the $25B quantified in dollars with its funding and any bond requirement, and the Q2 cost step explained in dollars. I'd keep both. If you want the cash line to bite, write it in free cash flow, with Q3 FCF above Q2's $1.7B alongside the capex pace line. That's a deliberately low bar, and the threshold is my judgment, not derived. Operating cash flow is growing about 25 percent a year and was $30.0B in Q3 last year. Straight-lined, that is around $37B, so a $30B capex quarter leaves several billion of free cash flow. It passes if the trend holds and fails if cash flow stalls, which is the thing you're actually worried about. Demanding a return to the $9B to $15B range would just be a way of never buying.

Second, the real money in this plan isn't where we've been arguing. We've spent five rounds on five points of standard weight, about four basis points on a 15 percent night. The 70 points between 30 and 100 are worth about 35 basis points for every 10 percent the stock moves. Count the gates on that climb: - margin ex-accrual - revenue - implied Q4 margin - dollar-quantified one-offs - the capex line - free cash flow - judgment funding - the 10-Q - Amazon - the gap-day hold - a 2027 guide for the last rung

Each is reasonable. Together they can leave someone at 30 percent while the stock walks to 780, with the weekly SuperTrend 24 percent below price and the monthly 35 percent below. This stock rallied 31 percent from the 7/30 close without a better margin print, and its biggest leg was a gap on 48.7M shares that no tool can explain. A rule set that demands proof at every rung will always be late, and I accept that as the price of discipline. But the stub is only tolerable if lateness is cheap.

So keep the first tranche to four numbers readable from the release that night: margin ex-accrual of at least 36 percent, revenue around $65B or better, no implied Q4 margin drop of a point or two, and Q3 capex no higher than about $30B. Put everything else on the 70 rung. Any gate still open at the 10-Q should be decided as a pass or a fail at the 10-Q, not rolled to the next catalyst. Neutral's "neither waived nor permanent" is the right wording, and I'd just add a date to it.

Here is the package.

Confirm the print timing first. Only the sentiment feed says after the close today. The last bar looks provisional, and its 11.0M volume is the lowest in the window, which fits a session still in progress. If results are already out, skip the trim and run the table from the next regular session.

If the print is still ahead, enter a marketable limit now, check the fill before the bell, and place no after-hours orders in either direction. The target is the loss you'd shrug off on a minus 15 percent gap, divided by 15 percent. That works out to 30, 25 or 20 percent of standard weight by tier, and dividing by 25 instead lands near 20. Sell the highest-basis lots first and check the one-year mark and wash-sale rules. Anyone already at or below their number sits tight, and anyone flat waits.

After the print, the ceilings are: - Zero below 30 percent margin or on revenue under $62.5B. - About 15 for a 30 to 32 percent margin with no dollar bridge. - 30 in the hold band, or 25 if the implied Q4 margin falls with no dollar-quantified one-off. - 70 on a pass. - 100 only on a pass plus a 2027 guide with expense and capex growth at or below revenue growth.

The pace is a third at a time from the next regular session once the gap-day level holds. Reset the stop off the new structure.

I'm not going to pretend 30 percent is a bold position. The bold part is being willing to climb fast and far once the numbers are on the table, and we should write the rules so that happens.

FINAL TRANSACTION PROPOSAL: SELL

Conservative Analyst¶

Conservative Analyst: I'll start with what we agree on, because it narrows the argument. The 681.41 stop is useless through a print when the last four gaps ran 8 to 11 percent and the stop is only about 1.1 ATR away. Holding full size is also a decision to carry the whole gap, not a neutral default. I'm not defending hold-and-let-the-stop-work, and I'm not claiming META is broken, since revenue is up 28 percent and operating cash flow is up about 25. Where I part ways is the 40 percent stub and the claim that it costs very little.

You said yourself that the fundamentals and the tape give no clear edge on which way this breaks, and that you don't want to be paid for gap risk at this price. That logic doesn't stop applying at 40 percent. The right to act decisively after the print comes from having cash and written rules, and you get that by selling, not by keeping shares. The stub is exposure with no edge. Since you won't add until your criteria are checked, it only earns anything on the gap itself, which you've said isn't where your edge is. "Keeps me in the game" is a regret argument, not a risk argument.

I'd also push back on minus 15 percent as the stress case. META closed at 538.60 on July 30, and the rally began with a gap day no tool can explain. A full round trip is about minus 24 percent, which is roughly 0.48 percent of the portfolio at 2 percent weight, not 0.3. I'd size so the tail is boring. About 1 percent on a 5 percent standard, which is 20 percent of standard weight, puts a minus 15 percent gap at 0.15 percent and a full round trip near 0.24. That budget is my choice, not something the data dictates, so the caller should scale it to their own tolerance.

On the bullish branch, you say the 36 percent bar isn't fantasy because Q1 ran 40.6. But look at the cost base. Q1 total costs, meaning revenue minus operating income, were about 33.4 billion. Q2 was about 42.0, up roughly 8.6 billion, or 26 percent, in a single quarter. At 64 billion of revenue, a 36 percent margin needs total costs near 41 billion. So costs have to fall about a billion right after that jump, while 30 billion a quarter of capex feeds depreciation, which is my inference since the tools don't break it out. If costs simply hold flat at 42, the margin is about 34.4 percent, a miss on your own test. Maybe Q2 had a one-off and it snaps back, but you said nobody knows, and not knowing is a reason for smaller size. The trader's note also calls the stock roughly fair at a 31 percent margin, which sits badly next to its own math that 20.4x needs 36. On the data we have, it's priced for a recovery nobody has seen. And your top rebuild tier needs 2027 expense and capex growth at or below revenue growth. Capex rose 87 percent in FY2025 against 22 percent revenue growth, and H1 2026 capex is up 67 percent, so that tier asks for a reversal of the entire trend.

The "buy hard into 630 to 665" idea worries me most. Those levels are the 200 and 50 day averages, which are lagging. They're also where an average-sized earnings gap of 8 to 11 percent lands from here, roughly 630 to 651, so they aren't a floor. A gap into that zone that also passes your margin and revenue test is a narrow slice, mostly a capex scare on a good quarter. If it comes with margin near 31, your rules say exit. Even if it passes, buying hard on the day of maximum fear means sizing into the widest spreads with the least information. I'd rebuild in tranches over a week or two, as the trader's plan says.

On the accrual, excluding the New Mexico judgment is a fair way to read operating trends, but it can't be how you judge risk. Twenty-five billion is only about 1.4 percent of market cap, so retail isn't crazy to shrug. But it's about 14 times Q2 free cash flow, more than the vendor's 21.6 billion FCF figure, and about 1.6 times the 15.5 billion of cash and equivalents. Liabilities are already up 79 billion in three quarters and capex absorbs about 69 percent of trailing operating cash flow. I can't see the securities breakdown or appeal requirements, so I'm not claiming a funding problem. My point is that your margin rule ignores real money competing with the capex program, and the reports don't even agree on whether it was a jury or a judge. That argues for less size, not a rule that lets the stock off.

On the smaller points, you cite 40 percent odds of touching 760 and 18 percent for 800, but those markets trade $3,000 to $18,000 each. The same ladder gives 30 percent for 640 and 18 percent for 620, so the tails are close to symmetric. That is more evidence of no edge, not a reason to stay in. Insider selling is probably routine, as you say, but there's also no buying, and Cox's sales only started in September. The rate backdrop is unverified because FRED was down, so I wouldn't build on it either way.

My recommendation is to confirm the print timing first. The last bar looks provisional, and if results are already out, none of this pre-print trim applies. If the print is still ahead, I'd cut to about 20 percent of standard weight, or go flat if the position is already large for the caller's book or one stock on one night can't cost them 0.3 percent. I don't know their holdings, cost basis or tax position. If realizing gains is the obstacle, a defined-risk hedge is an alternative, but I have no options data to price it and I'd expect it to be expensive into earnings. After the print, rebuild in tranches only on your criteria, with the exit rules kept mechanical.

FINAL TRANSACTION PROPOSAL: SELL Conservative Analyst: I'll start with what I'm giving up, because some of your corrections are right. Neutral, the minus 24 percent round trip is a three-week retrace, not a number to size tonight against, so I'm dropping it. You're also right that comparing the $25 billion to Q2 free cash flow flattered my case, since that quarter is depressed by discretionary capex. Aggressive, I accept that the tape evidence is thin and that a margin pass only earns a first tranche. That narrows our disagreement to 35 percent versus 20, but I still think the gap matters.

First, think about what tonight's number actually is. Neutral's fix for the middle zone is to hold the trimmed size if the margin lands between 32 and 36 percent, and Neutral says that's where the print most likely lands. Aggressive adopted that rule. So 35 percent isn't a one-night stub. In the most likely outcome, it's the size you carry into next quarter while the cost question is still open. The argument that being flat means chasing a gap-up at a worse price also fails to separate 35 from 20. The rebuild runs in thirds over one to two weeks either way, so the plan already accepts not chasing. The difference between us is about 0.75 percent of the portfolio in position, or roughly a tenth of a percent on a 15 percent gap. If the expected value is a wash, as you both say, a conservative mandate takes the lower variance.

Second, the cost math. I'll grant that at $65.6 billion of revenue and flat costs the margin is 36.0 percent. That's a knife edge. A revenue miss of $1.6 billion, about 2.5 percent, takes you to roughly 34 percent on the same costs. And flat isn't the central case I'd pick. Q2 capex jumped nearly 60 percent in one quarter, from $19.0 billion to $30.1 billion, and depreciation follows capex with a lag, so I'd expect costs to keep rising from here. That's inference, since the tools don't break out depreciation. Costs were already up 55 percent year over year in Q2 against revenue up 28. A $4 billion one-off is plausible, but nothing in the data shows one. I wouldn't carry the larger size on an assumption.

Third, Aggressive's own table of gaps. Q3'25 printed a 40.1 percent margin and gapped minus 11, and Q1'26 printed 40.6 and gapped minus 9. You drew the right lesson, that the reaction isn't about the trailing margin. But that also means the line we're judging doesn't predict the move, so a margin pass is no reason to hold more tonight. The largest drop in the record, minus 11.3 percent on 88 million shares, came in a quarter with $20.5 billion of operating income and only $2.7 billion of net income. I can't tell you what that charge was or that it drove the move. But it's the closest precedent for a $25 billion judgment landing in the headline numbers. You plan to read the margin ex-accrual, while the tape trades the headline. For the first few hours you'd be holding 35 percent through a print your own rules say to adjust.

Fourth, the $25 billion. I'm not claiming a funding problem. Neutral is right that there's about $110 billion of other current assets I can't see, and payment timing is unknown. But take Aggressive's straight-line extrapolation. If Q3 and Q4 look like Q2, FY2026 capex is about $109 billion. If operating cash flow also holds near $32 billion a quarter, free cash flow comes out around $18 to $19 billion. The judgment would be larger than a full year of free cash flow on that path. It competes with the 2027 capex plan for the same dollars, while liabilities are up $79 billion in three quarters and cash is $15.5 billion. That doesn't justify an exit, but it argues against starting the middle-zone hold from the bigger number.

Fifth, your own carve-outs. You both say to go to 25 or 30 percent if the caller holds correlated mega-cap tech elsewhere. Nearly every book does, if only through index funds, and Tesla reports the same evening. The stock's beta is 1.185. So your exception is closer to the default. I'd start at 20 percent and treat 25 as the ceiling. A large embedded tax gain doesn't change that. Tax is a known, finite cost, while a gap is an unknown one, and I'd rather pay the tax than let it set the risk. A defined-risk hedge is the alternative, though I have no options data and would expect it to be expensive into earnings.

On execution, I'd do the trim during the regular session rather than leave it to after-hours spreads. Confirm the print timing first, because only the sentiment feed says after the close today, and if results are out, none of this applies. After the print, if the headline is muddied by the accrual, I'd wait for the call and the 10-Q detail before acting on a subjective call like whether there's a credible bridge. A smaller stub makes that waiting cheap. Anyone already at or below 20 percent should sit tight, and anyone flat should wait.

We agree on the sell, on the stop being useless through a gap, on verifying the timing, and on tranche rebuilds. Our remaining difference is whether to carry 35 percent of standard weight into a print whose reaction function we can't predict, when the most likely outcome leaves us holding that size afterward. I'd carry 20.

FINAL TRANSACTION PROPOSAL: SELL Conservative Analyst: I'll start with where I'm moving, since you both landed some hits. Neutral, the tax point is fair, and I overreached when I said I'd rather pay the tax than let it set the risk. For a caller with a large embedded taxable gain, 30 percent of standard weight is defensible. Aggressive, my index-fund point was weak, and Neutral is right that beta describes ordinary days, so I'm dropping both. I'm also adopting Neutral's loss-budget formula as the handoff. We're now arguing over six to nine basis points on the worst print in the data, so I'll spend my time on where the plan still carries risk nobody has priced.

First, the conditional hold. Neutral's fix is to hold 30 if the guide implies a Q4 margin at or above Q3's and step down to 20 or 25 if not. Aggressive took that as the answer to my point that tonight's size is the carried size. It only half answers it. The step-down happens in the next regular session, after the gap. In the branch where the cost question resolves against us, we eat the gap at the bigger size and then sell lower. The rule protects the second leg, not the first. It also leans on judgment calls made the morning after a gap. A "named one-off" is something management can always supply, so I'd want it quantified in the 10-Q. The implied Q4 margin needs a full-year expense guide that Aggressive himself says gets contaminated if the New Mexico charge is booked. And an exit at "near 31" with a hold band starting at 32 leaves a hole, which is where discretion creeps back in.

Second, the rebuild gates skip half the thesis. The trim rests on a 36 percent margin, but also on Q2 capex at 49.5 percent of revenue, $1.7B of free cash flow, liabilities up $79B and cash at $15.5B. The first-tranche gates are margin, revenue and the implied Q4 margin, and none of them touch cash. With capex running this far ahead of depreciation, the income statement understates cash intensity, so a margin pass can coexist with a heavier 2027 depreciation load. Aggressive's own table shows the tape has traded capex and forward costs, not the trailing margin. A first tranche gated on the trailing margin uses the variable you say doesn't drive the reaction. I'd want the capex run rate and the funding for the judgment on the table before the first tranche, not only before the 100 percent rung.

Third, the $25B. Neutral, you said the headlines have been out for days and the stock fell about 9 percent, and in the same paragraph said the tools don't date the ruling. Those can't both support "priced in." The one source that speaks to timing, the sentiment report, says it landed on earnings day. The slide from 762 to 694 has an easier explanation: the stock was giving back a September gap. Aggressive, I agree two below-the-line data points aren't a model. But I can't tell from the tools whether the Q3'25 item was cash, and a judgment is cash eventually. My understanding, which I can't verify here, is that a penalty paid to a government is generally not deductible. In that case the after-tax cost is close to the full $25B, about $9.80 a share on 2.55B shares, against Q2 EPS of $6.18. If it's booked in Q3, which Neutral and I both think is plausible for a case pending at quarter-end, the headline GAAP quarter could show a loss. I'm not claiming the tape will care. I'm saying nothing in the table is a close precedent, and with no precedent I'd rather be smaller.

Fourth, "stock-specific." Neutral, you said the four gaps look stock-specific as far as the tools show. The tools only carry META's own price series, so they can't show that either way. The world report has Alphabet down about $860B from its peak and Amazon reporting this week, so the AI-capex read-across is live. Aggressive says to cut elsewhere before cutting META harder. That's a fair fix, but we don't know there's an elsewhere to cut. A "standard weight" framework suggests a multi-name book, and hyperscalers and chips are among the likeliest holdings. That's judgment, not data. When I can't verify, I default to the smaller number.

Fifth, the tail we're sizing to. A minus 11.3 percent gap is the worst of four observations in the tool window, and the trader's minus 15 percent is only a modest haircut to it. From memory, not these tools, META gapped about minus 25 percent twice in 2022 on earnings, in a margin-and-capex scare. The regime differs, so this is context and not a forecast. But at a 25 percent gap, 20 percent of a 5 percent standard costs 25 basis points and 35 percent costs 44. That 19-basis-point difference is more than double the number we've been trading. The thin prediction ladder also gives about 18 percent for a touch of 620 this month.

Last, Aggressive, working a limit order has a failure mode. If it doesn't fill before the close, we carry the whole gap, which is exactly what the trim exists to prevent. A trim's failure mode should be benign, so use a marketable limit, enter it now, and check the fill before the bell.

Here is where I land. Trim to about 20 percent of standard weight, with 25 as the ceiling, in the regular session and with no after-hours orders. The caller can go up to about 30 only if all three hold: a loss budget of at least 0.22 percent of the portfolio on a minus 15 percent gap, no direct AI-capex holdings elsewhere, and a large embedded taxable gain they'd rather not realize. For that last case, a wide defined-risk collar is worth pricing against carrying more naked. I have no options data, and constructive-sale rules matter. Anyone already at or below target should sit tight, and anyone flat should wait.

Confirm the print timing first, since only the sentiment feed says after the close today. If results are already out, none of the pre-print trim applies. Afterward, take a first third of the rebuild only if margin ex-accrual is at least 36 percent, revenue is about $65B or better, the implied Q4 margin doesn't fall below Q3's, the one-offs are quantified in the 10-Q, and the capex run rate and judgment funding are disclosed. Do it no sooner than the next regular session, once the stock holds the gap-day low. The 100 percent rung stays tied to a 2027 guide with expense and capex growth at or below revenue growth, and the exits stay as Aggressive rewrote them.

FINAL TRANSACTION PROPOSAL: SELL Conservative Analyst: Let me start with what I'm giving up, since you both landed real hits.

Aggressive, you're right that I adopted the loss-budget formula and then put a ceiling on it that a tax gain could override. A tax gain is a cost of selling, not a measure of risk tolerance. I'm dropping that condition and the collar with it. I have no options data, and a collar caps the upside the stub exists to keep. Pick the highest-basis lots, check the one-year mark and wash-sale rules, and let the budget set the size.

Neutral, you're also right that standard weight is meant to carry a normal print's tail, and that cutting 70 percent from standard is already the response to this quarter's excess risk. I'll hand the caller both denominators, 15 percent for the worst print in the window and 25 percent for a 2022-style night, and stop arguing over which is right. Aggressive, "when I can't verify, default to smaller" isn't a principle with a stopping point. I'll use it as a tie-breaker, and I've used it here to move up, not to go to zero.

That leaves a smaller gap than we've been acting out. Both of you said 30 requires a caller who can say they hold no other direct AI-capex names. So the caller who hasn't told us anything gets 25 by your own rules, and I'll take 25 as the default. That is 1.25 percent on a 5 percent standard, about 19 basis points on a minus 15 percent gap. Against my old 20, the difference is about 4 basis points, so I won't spend another round on it.

I still object to one thing in the middle of your rule. Aggressive says 25 if they hold other AI-capex names or can't say, and Neutral said the same. Those shouldn't be the same number. "Can't say" is a blend of people who do and people who don't. The caller who knows they own other direct AI-capex names is the bad half of that blend. With Amazon reporting this week and Alphabet down about $860B from its peak, that overlap is a correlated gap, not diversification. So I'd put a clean book at 30, an unknown book at 25, and a known overlap or a tight budget at 20.

The bigger holes are after the print.

First, the plan has no branch for a trim that doesn't get done. Neutral says that if the close is missed, don't sell after-hours and run the post-print rules from the next session. But those rules are written as add, hold or step-down from a trimmed base. Someone still at 100 percent has no instruction. I'd write the post-print table as target weights, not actions: exit means zero, the hold band means the stated hold size, and a pass means the first tranche. Then it works for the person who got the trim done, the person who missed it, and the person who was already below. The trim's failure mode should be benign, and right now this one isn't.

Second, you both criticized my gate for always passing, but the 70 percent rung has the same flaw. "If the call and the 10-Q hold up" is as soft as "disclosed." Both of you said the heavy cash tests belong at 70 and 100, and neither plan writes one for 70. Here's a numeric version: - FY2026 capex pace no higher than the roughly $109B straight-line (H1 of $49.1B plus two quarters at Q2's $30.1B). - The $25B either accrued or quantified in dollars, with how it's funded (cash or new debt) and whether an appeal needs a bond. - The 10-Q showing the Q2 cost step explained in dollars.

I also want to be clear about what the first-tranche capex gate means. Q3 capex no higher than $30B is calibrated to the worst free-cash-flow quarter of the last six. Q2's $1.7B compares with a prior range of $9B to $14.8B. Passing it means no worse than the worst quarter, not better. That is fine for a 13-point tranche, but it shows why 70 needs a real test.

Third, ambiguity should resolve small. If the release is muddied by the accrual and you can't compute the ex-accrual margin, hold the smaller of tonight's size and 25. The 10-Q usually lands within a day or two, as Aggressive said from memory, and re-adding costs only a bit of price.

Fourth, timing. Amazon is due this week, though I can't see the date in the data. If it lands inside the first-tranche window, I'd wait for it, because the first tranche is small and the AI-capex read-across is live. Neutral's rule against buying the first green candle, and Aggressive's against buying the first red one, both support this.

On execution, I agree with the marketable limit, entered now. The bounce toward 713 is not a reason to work the order for a better price. This is a time decision, not a price decision. Confirm the print timing first, since only the sentiment feed says after the close today and the last bar looks provisional. If results are already out, none of the pre-print trim applies.

So my package is to trim to about 25 percent of standard weight for an unknown book. Use 30 if the caller confirms a clean book and a budget that supports it, and 20 for a known AI-capex overlap, a tight budget or a tax-deferred account. Anyone already at or below their number sits tight, and anyone flat waits. After the print, write the rules as target weights, and require a numeric cash test before the 70 percent rung.

FINAL TRANSACTION PROPOSAL: SELL Conservative Analyst: Most of what's left is detail, so I'll start with what I'm giving up.

Neutral, you're right that "the stub has no edge" is equally true of the full position. My case for the smaller number never rested on edge. It rests on this quarter's excess risk, meaning a 20.4x multiple that needs a margin nobody has seen since Q1 and a stop that can't act, and on not knowing the caller's book. The three tiers settle that. I'll take 30 for a clean book, 25 for an unknown one, and 20 for a known overlap or a tight budget. Aggressive, I also accept that tax isn't a tier, and that the capex pace line isn't the heavy part of the 70 rung.

I still have four things to fix.

First, your replacement cash test is nearly empty. Requiring Q3 free cash flow above Q2's $1.7B passes if everything stays exactly as bad as Q2. Operating cash flow flat near $32B against $30B of capex gives about $1.9B. Your $37B comes from applying 25 percent year-over-year growth to Q3 last year, but that growth was about 34 percent in Q1 and about 25 in Q2. The extrapolation assumes the slowdown stops. I'm not asking for a return to $9B to $15B, which would be a way of never buying. I'd ask for Q3 free cash flow of about $5B or better, which is roughly $35B of operating cash flow against $30B of capex. That's 17 percent year-over-year growth, so it tolerates further slowing. The threshold is my judgment, not derived, and it goes on the 70 rung.

Second, the pace line isn't redundant if you keep your four-number first tranche. That tranche tests only reported Q3 capex, and the full-year capex range in the release tests the forward quarter. Q3 could print under $30B while the range implies a Q4 well above it. So I'd move the pace line, with a few billion of tolerance for finance-lease principal, into the first tranche. It's readable from the release that night and costs nothing in lateness.

Third, the plan conflates two things. Neutral's package says the target is the loss you'd shrug off divided by 15 percent, and then says the defaults are 30, 25 or 20 percent of standard weight. Those only agree at a 5 percent standard, which was just the trader's example. At a 10 percent standard, 30 percent of standard is 3 percent of the portfolio, and a minus 15 percent gap costs 45 basis points, nearly double the 25 basis point budget we've been using. At a 2 percent standard the tier undershoots the formula. I'd tell the caller to take the lower of the formula's number and the tier percentage of their standard weight. The formula is the loss constraint and the tier is the book-overlap constraint, so you need both.

Fourth, there's a timing inconsistency. If the print is after the close today, the next regular session is the gap day itself. A rule that says to start the first tranche at the next regular session once the gap-day level holds can't be satisfied on that day, because you don't know the low until the close. So the earliest first tranche is the session after the gap day, and later if you wait once for Amazon. Selling to the ceiling at the gap-day open is accepted, but buying waits a day. That is de-risk fast, re-risk slow, and it should be written down so nobody reads "next regular session" as "buy the open."

Aggressive, I don't think this makes the plan too slow. You said lateness has to be cheap. The first tranche is about 13 points of standard, or 0.65 percent of the portfolio on a 5 percent standard, so a 5 percent move while you wait costs about 3 basis points. The 70 points between 30 and 100 are worth about 35 basis points per 10 percent move. Being at 30 while the stock runs 10 percent costs that much, and being at 30 when it falls 10 percent saves it. That's the variance cut working as designed, not a flaw in the gates. Your date-stamp is the right way to bound it. Decide each open gate as pass or fail at the first session after the 10-Q is filed. If the filing hasn't appeared within about a week, decide on what's in hand and don't let it drift. I'd also drop "climb fast and far." Your own table shows the trailing margin hasn't predicted any of the four reactions, so the pace should stay at thirds over one to two weeks.

On execution, a trim that goes unfilled is the failure case the whole exercise exists to avoid. Enter the marketable limit now. If it hasn't filled by mid-afternoon, cross the spread. This is a time decision, not a price decision, and the bounce toward 713 isn't a reason to hold out. Place nothing after hours. Confirm the print timing first, since only the sentiment feed says after the close today and the last bar looks provisional. If results are already out, skip the trim and run the post-print table from the next regular session.

My package is to trim before the close to the lower of the loss-budget number and 30, 25 or 20 percent of standard weight by tier. Anyone already below that sits tight, and anyone flat waits. After the print, use the ceilings and the pace as written, with the first tranche gated on your four numbers plus the capex pace line, and the 70 rung adding the $5B free cash flow bar, the $25B quantified in dollars with its funding, and the Q2 cost step explained in dollars.

FINAL TRANSACTION PROPOSAL: SELL

Neutral Analyst¶

Neutral Analyst: Both of you already agree on more than you're admitting. You both want to sell into the print, you both say the 681.41 stop does nothing against a gap, and you both say to confirm the earnings timing first. I agree with all three. The real fight is 40 percent versus 20 percent, plus what happens after the print. The first part matters much less than either of you is making it sound.

Take the difference between you. Twenty points of standard weight is 1 percent of the portfolio on a 5 percent standard, which is just the trader's example. A 15 percent gap on that slice costs 0.15 percent, and a 10 percent gap up makes 0.10. The prediction ladder is close to symmetric. It gives about 30 percent to touch 640, and interpolating between 760 and 800 gives about 31 percent to touch 776, the same distance up. So Aggressive, the stub costs little, but Conservative is right that it earns little too. And Conservative, "exposure with no edge" is worth about a tenth of a percent either way. The real exposure is in the post-print playbook, and I think each of you has holes there.

Aggressive, first. You stack the price below the 10-EMA, the MACD histogram and OBV into a bearish case. The technical report calls the posture mixed, with RSI at 50 and no stretch on any tier. Into an event where the typical move is 8 to 11 percent, those are rounding errors. Your prediction-market point is also mostly mechanical. Spot slid from the low 760s on 10/13 to 708, so the 680 touch got likelier and the 760 touch less likely, and that says little about the print. Conservative read the ladder correctly.

The 630 to 665 zone isn't support either, as Conservative said. It's the 200 and 50 SMA, which is where an ordinary 8 to 11 percent gap lands. The data suggests caution here. After the October 2025 gap to 664, the stock later made its 519 low, about 22 percent lower. After the April gap to 611 it later made 524, about 14 percent lower. Only the July gap was the low. So buying on the fear day was early two times out of three, and you'd be buying into wide spreads.

Your rebuild bars are also loose. A $64B revenue floor is about 25 percent growth, below the 28 percent trend, which implies roughly $65.6B. Your 681.41 close rule is stale after any real gap, and it's tight even without one. The stock traded to 681.05 yesterday and closed at 694. A close only two percent below that would send you to zero.

Conservative, your cost-base math is the most useful thing in this debate, but it's built on $64B. At $65.6B of revenue, flat costs of $42.0B give about 36 percent. So the question isn't whether costs must fall, it's where revenue lands and how much of the Q2 step repeats. That step was $8.6B in one quarter. The prior two quarters moved +$4.4B and −$1.7B. I can't prove a one-off because the tools have no depreciation or opex breakdown, but that step looks larger than a steady depreciation ramp would produce. Two stylized cases make the point. If $4B is one-off and costs otherwise grow $1.5B, the margin is about 40 percent. If none is one-off and costs rise another $2B, it's about 33 percent. The 36 percent bar sits in the middle of those, so it's neither fantasy nor a full reversal. That's a real chunk of probability on the bullish branch, and it argues against going nearly flat.

On the $25B, comparing it to Q2 free cash flow flatters your case, because that FCF is depressed by discretionary capex. Against roughly $130B of trailing operating cash flow it's under a fifth. The $15.5B of cash ignores about $110B of other current assets the tool doesn't itemize. I'd treat it as headline and leverage-direction risk, not a funding scare. A verdict on a case pending at quarter-end might get booked in Q3, which is my inference and not in the data. If so, the GAAP print could look awful while the operating story is fine. That argues for small first tranches after the print, not a deeper cut now.

Your minus 24 percent round trip is also a three-week retrace, not one night. The largest print-day move in the data is minus 11.3 percent. Sizing tonight's exposure to a multi-week retrace double-counts what the post-print exit rules are for.

My version is to trim to about 35 percent of standard weight before the close. That's a judgment call, not a derivation. On a 5 percent standard it's a 1.75 percent position. The worst print in the data costs about 0.2 percent of the portfolio, and a 15 percent gap about 0.26. Go to 25 or 30 percent if the caller has a lower loss budget or holds lots of correlated AI and mega-cap tech elsewhere. Go to 40 if there's a large embedded tax gain. Someone already below 35 percent should sit tight, and someone flat should wait.

After the print, the trader's rules need three fixes. First, they leave the middle undefined. Rebuild is at 36 percent or better and exit is near 31 percent with no bridge, but my bracket says 32 to 36 percent is where the print most likely lands. In that zone, hold the trimmed size and let the 2027 commentary and next quarter decide. Second, set the revenue bar at trend, around $65B, not $64B. Rebuild toward 70 percent in thirds over one to two weeks, and go to 100 percent only on a 2027 guide with expense and capex growth at or below revenue growth. Third, on a gap into 630 to 665 that passes the criteria, buy at most a third of the planned rebuild on day one and add after the stock holds the gap-day low. Reset the stop off the new price structure after the print. The exits for a 31 percent margin with no bridge, 10 percent or more implied Q4 cost growth, or revenue below $62.5B stay as written.

Before acting, confirm the print timing, since today's bar looks provisional. If results are already out, none of the pre-print trim applies. I wouldn't lean on the rate backdrop either, because FRED was down and the Fed pricing is unverified.

FINAL TRANSACTION PROPOSAL: SELL Neutral Analyst: Here's how I see where we are. You both sell, you both say the 681.41 stop does nothing through a gap, you both want the print timing confirmed first, and you both rebuild in thirds. What's left is 35 versus 20 percent of standard weight. On a 5 percent standard that's 0.75 percent of the portfolio, which is about 8 or 9 basis points on the worst print in the data, minus 11.3 percent. I'd rather spend the time on things that move more than that: the middle zone, how the exit rules are written, and tax.

First, where I'm moving. Conservative, you're right that the middle outcome is the likeliest, so tonight's size is also the size we carry afterward. I was also too generous on costs. Each $1B of total cost is about 1.5 points of margin on $65.6B of revenue, and flat costs of $42.0B give exactly 36 percent. But flat isn't the no-news case. Cost of revenue runs about 18.6 percent of revenue, so another $5B of revenue adds roughly $0.9B on its own. At $30B a quarter of capex, depreciation should keep climbing by something like a billion a quarter if servers are written off over five or six years. That's my assumption, since the tools don't break it out. So no news lands near $43 to 44B of costs, or 33 to 34.5 percent. Getting to 36 needs a billion or two of the $8.6B Q2 step to be non-recurring. That isn't fantasy, because depreciation from the Q2 capex jump can't explain most of that step. But it's less than a coin flip, so Aggressive, "genuine toss-up" is generous. I'm moving my number from 35 to 30, which is 1.5 percent on a 5 percent standard and about 0.22 percent of the portfolio on a 15 percent gap.

I'm not going to 20, though, and Conservative's remaining arguments don't get me there.

On tax, selling 15 more points of standard weight means selling 0.75 percent of the portfolio. If half the sale value is gain and the rate is near 24 percent, that's about 9 basis points of tax. It's partly a timing cost, since rebuying resets your basis, so I won't oversell it. But it's paid for sure, and it roughly equals what the extra selling saves on the worst print we've seen. In a tax-deferred account it's zero, and 20 is fine there.

Beta of 1.185 describes how META moves with the market on ordinary days. It says little about its own earnings night, and as far as the tools show, the four big gaps look like stock-specific reactions. Tesla reporting the same evening doesn't change that. The overlap that matters is AI-capex sympathy, which is a book-level check on total AI exposure. That's why my lower range is for heavy holders, not a ceiling for everyone.

On the Q3'25 precedent, operating income that quarter was a clean $20.5B at a 40.1 percent margin, and the hit sat below the line. The stock still fell 11 percent, which supports Aggressive's point that the tape trades capex and forward costs. It tells us nothing about how the tape treats a charge inside operating income. And "we can't predict the reaction" doesn't favor 20 over 35, because unpredictability is symmetric. If I can't call a gap down, I can't rule out a gap up.

On the $25B, you've conceded it isn't a funding problem. The headlines have been out for days, and the stock fell about 9 percent from 762 on 10/13 to 694 on 10/20. The tools don't date the ruling, so I can't say how much is priced in. What's still open is whether and how much gets accrued, which is a management and auditor call I can't see. That's a real reason to shade from 35 to 30, and the call should clear it up.

You also keep invoking a conservative mandate, but we don't know the caller's mandate. I'd hand them a formula instead: take the loss you'd shrug off on a minus 15 percent gap, divide by 15 percent, and that's your position as a share of the portfolio. A loss budget of 0.15 percent gives 1 percent, or 20 percent of a 5 percent standard. A budget of 0.3 percent gives 2 percent, or 40.

Aggressive, you have three problems. First, "cash and written rules" doesn't separate 20 from 30. In the middle zone, which we both think is likeliest, the rule is hold, so no cash gets deployed. In the tails you rebuild in thirds over one to two weeks, and either size leaves far more cash than a first tranche needs.

Second, nobody has caught that the exit clause for implied Q4 cost growth of 10 percent or more is miscalibrated. Last year total costs went from $30.7B in Q3 to $35.1B in Q4, up 14.5 percent. Revenue rose 16.9 percent, the margin rose to 41.3 percent, and the stock gapped up 10 percent. A flat 10 percent cap would have fired an exit in a quarter the market liked. If it means year over year, it's the opposite problem, since costs are up 55 percent year over year now. I'd write it against revenue. If the guide midpoints imply a Q4 margin below Q3's actual with no named one-off, that's the exit trigger. It matches the 100 percent tier you already like. Q4'25 is derived as full year minus nine months, and I'm assuming management gives its usual revenue range and full-year expense range.

Third, the stock bottomed the day the 30.9 percent margin printed, gapping down 8 percent, and then rallied 31 percent without a better margin print. That's your argument against exiting, since the market already looked through the trough once. It's also the argument for trimming, because that look-through is now in the price, and this print has to deliver the recovery it assumed.

Conservative, you're right that the middle-zone hold is where tonight's number really bites. I'd make that hold conditional instead of shrinking tonight's size. Hold at 30 if the guide implies Q4 margin at or above Q3's, or if there's a named one-off in the Q2 costs. Step down to 20 to 25 if the guide implies lower with no explanation. You get the smaller carried size exactly where the cost question resolves against us, and Aggressive keeps the larger size where it resolves for us.

I agree with you on execution too. Trim in the regular session. Place no orders in after-hours, in either direction. Make the first rebuild tranche no sooner than the next regular session after the call, ideally after the 10-Q detail on the opex drivers.

My package is to trim to about 30 percent of standard weight. Use 20 for a tight loss budget or a tax-deferred account, and 35 to 40 for a large embedded taxable gain. If you're already at or below your target, sit tight, and if you're flat, wait for the print. If results are already out, none of this applies. Afterward, take a first tranche of a third of the rebuild only if margin is at least 36 percent, revenue is about $65B or better, and the guide doesn't imply a falling Q4 margin. Go to the full 70 percent only on a 2027 guide with expense and capex growth at or below revenue growth. Reset the stop off the new structure.

FINAL TRANSACTION PROPOSAL: SELL Neutral Analyst: We're arguing over a few basis points now, so I'll say where you each moved me and where I still disagree.

Conservative, you're right on three things. I can't call the $25 billion priced in. I said the headlines had been out for days and also that the tools don't date the ruling, and the sentiment report places it on earnings day. I drop that. The conditional hold also protects the leg after the gap, not the gap itself, and "named one-off" has to mean a dollar figure in the release or 10-Q, not a sentence on the call. And the tools only carry META's own price series, so they can't show the gaps were stock-specific. I also agree on the marketable limit. A trim's failure mode shouldn't be not getting filled, so Aggressive, enter it now and check the fill before the bell.

I think 30 versus 20 is a simpler disagreement than it looks. It's the same loss budget with a different gap. If you'd shrug off 0.25 percent on a 15 percent gap, that's a 1.67 percent position, a third of a 5 percent standard. Underwrite 25 percent instead and the same budget gives 1 percent, or 20 percent of standard. So the question is whether tonight's size should cover a 2022-style tail. The worst print in the window is minus 11.3 percent, and the trader's 15 percent is already a haircut above that. The prediction ladder's lowest strike is 620, about minus 12 percent at 18 percent odds, and nothing there speaks to minus 25.

Your 2022 memory is fair context, and I'd flag it as memory. The cost side of that scare, margin and capex, rhymes with today. The revenue side doesn't, because growth was stalling then and is 28 percent now. More important, standard weight is supposed to already carry a normal print's tail. We're trimming for excess risk this quarter, meaning a 20.4x multiple that needs a 36 percent margin, a 31 percent rally with no catalyst, and a stop that can't act. Cutting 70 percent from standard does that. Underwriting a minus 25 tail at every print would be a rule about standard weight, not a view on this stock. The whole gap between 30 and 20 is about 6 basis points on the worst print in the data and 12 on a 2022-style night, with similar size on the upside.

Your default-to-smaller instinct is still right when the caller can't vouch for the rest of the book. Amazon reports this week and Alphabet is down about $860B from its peak. So my number is 30 for someone who can say they hold no other direct AI-capex names, and 25 for someone who can't. We can't verify their book, but they can in a minute.

On the $25 billion, I can't verify deductibility either. But even at the full $9.80 a share it's about 1.4 percent of the price, smaller than any gap in the window. The risks are the headline EPS, competition with capex for cash, and whether it's accrued. Q3'25's below-the-line hit was, as I recall and haven't verified, a non-cash tax item, so I agree there's no precedent for a legal accrual. The answer to no precedent is a quantification rule, not a deeper trim.

Your rebuild gate has the flaw you pinned on us. "Capex run rate and judgment funding disclosed" always passes, because every company discloses. Give it numbers. I'd use Q3 capex no higher than Q2's $30 billion, no step-up in implied Q4, and any debt raised against the judgment sized in the filing. Also scale the gates to the add. The first tranche is a third of a 40-point rebuild, about 13 points of standard, or 0.65 percent of the portfolio on a 5 percent standard. The heavy cash tests belong at the 70 and 100 rungs.

Aggressive, you have two things to fix. "Don't buy the first red candle" has a twin, which is don't buy the first green one. Your own table shows 40.1 percent printed with minus 11, 40.6 with minus 9, 41.3 with plus 10, and 30.9 with minus 8. The trailing margin hasn't driven the reaction. If the stock gaps up toward 780 on a margin pass, the first tranche still waits for the next regular session and for the gap-day level to hold. Second, the stub's value as regret insurance is real, but it belongs to the caller's temperament, not the tape. If the reason for holding more is tax, sell the highest-basis lots first before deciding you need 35 or 40. That cuts the realized gain on the same sale and shrinks the tax argument we've been trading. Check the one-year mark, and the wash-sale rule if any lot is at a loss, since we rebuild within weeks.

Conservative's hole in the bands is real, so here they are without gaps. At 36 percent or better ex-accrual, with revenue around $65 billion or better and a guide that doesn't imply a falling Q4 margin, plus the capex gate, take the first third. Do it in the next regular session, once the gap-day level holds. From 32 to 36 percent, or revenue between 62.5 and 65, hold tonight's size. Step down to 20 or 25 if the guide implies a lower Q4 margin and no one-off is quantified in dollars. Below 32 with no quantified bridge, or revenue under 62.5, exit. If you can't compute the ex-accrual margin from what's released, treat it as the hold band and add nothing until the filing quantifies it. The 70 percent level comes in thirds over one to two weeks, and 100 percent only on a 2027 guide with expense and capex growth at or below revenue growth. Reset the stop off the new structure.

My package is to trim to about 30 percent of standard weight, which is 1.5 percent on a 5 percent standard. The formula is the loss you'd shrug off on a minus 15 percent gap divided by 15 percent. Use 25 if you can't vouch for your other AI-capex exposure, and 20 for a tight budget or a tax-deferred account. Go to 35 or 40 only with a large embedded taxable gain, after picking lots. Anyone already at or below their number should sit tight, and anyone flat should wait.

Confirm the print timing first. Only the sentiment feed says after the close today, and the last bar looks provisional. If results are out, the pre-print trim doesn't apply. If the close is missed, I wouldn't sell into after-hours on either side of the release, and the post-print rules run from the next regular session. Rates are unverified because FRED was down, so I wouldn't lean on them.

FINAL TRANSACTION PROPOSAL: SELL Neutral Analyst: We're down to about five points of standard weight, which is 0.25 percent of the portfolio on a 5 percent standard, or under 4 basis points on a 15 percent gap. So I'll start with what the trade is, then fix the places where each of you is still stretching.

Going from full weight to 30 percent takes 3.5 percent of the portfolio off the table. On the worst print in the window, minus 11.3 percent, that saves about 0.4 percent. On the best one, plus 10, it gives up about 0.35. That is a variance cut, not a directional bet.

Aggressive, that undercuts your framing. You call this the asymmetric trade because of the cash you'll hold afterward, but the ladder is symmetric and the reaction hasn't tracked the trailing margin in any of the four prints. In the pass branch you rebuild at higher prices, 13 points at a time. In the fail branch, your own history says buying the first red candle was early two times out of three. The plan's value is that nobody has to make a decision on a gap. It doesn't buy you an option.

Conservative, the same fact cuts against you. "The stub has no edge" is equally true of the full position. Neither size has an edge, so the only question is how much variance to carry. The loss-budget formula answers that, which also answers Aggressive's complaint that "default to smaller" has no stopping point. I'll take your three-way split, with 30 for a clean book, 25 for an unknown one, and 20 for a tight budget or known overlap. I'd define overlap as more than one other hyperscaler or AI chip name at full weight. That threshold is my judgment, not data.

Conservative, target weights are the best addition this round, but they need one fix. If "pass" means "first tranche," someone who missed the trim and sits at 100 sells down to about 43 after a good print, then buys back toward 70 days later. That is churn with a tax bill. The fix is to separate the ceiling from the pace.

The ceiling is what you're willing to hold given what the print showed: - Zero on the exit branch. - 30 in the hold band, or 25 if the implied Q4 margin falls with no dollar-quantified one-off. - 70 on a pass. - 100 on a pass plus a 2027 guide with expense and capex growth at or below revenue growth.

The pace is a third at a time, starting the next regular session once the gap-day level holds. If you're above the ceiling, sell down to it. If you're below, climb at pace. That works for the person who trimmed, the person who missed the close, and the person who was already low.

I'll also accept your ambiguity rule, hold the smaller of tonight's size and 25. It only bites a clean-book caller, so it's a five-point question. I'd keep your numeric 70 percent test with two repairs: - The $109B pace comes from the cash-flow capex line. From memory, the company's own capex guide includes finance-lease principal, so compare like with like or allow a few billion of tolerance. - Wait for Amazon once and no further. Several hyperscalers report in this stretch, and a rule that waits for the next print will always find one.

Gates also have to be satisfiable. A failed gate gets re-tested at the 10-Q, neither waived nor permanent. The weekly and monthly SuperTrends sit 24 and 35 percent below price, and I don't want a pile of reasonable checks to leave someone at 30 percent of a trending stock indefinitely.

Aggressive, I owe you a correction on a rule I wrote. Below 32 percent with no bridge goes to zero, while 32.1 holds 30 points of position. That is a cliff on a tenth of a point of margin. Also, the one time margin printed 30.9, the stock made its low that day and rallied 31 percent. A repeat of Q2 is worse than the first, because the rally assumed recovery, but it isn't a reason to dump the rest into a gap. I'd use a slope. A margin of 30 to 32 with no dollar bridge goes to about 15. Below 30, or revenue under $62.5B, goes to zero. The stub is regret insurance on the way up, and a small piece of it is regret insurance on the way down too.

The implied Q4 margin test is also a knife edge. Each $1B of cost is about 1.5 points of margin on $65B of revenue, so a full-year expense range two billion wide leaves a point and a half of slack either way. The guide's width isn't in our data. I'd trigger only when the implied Q4 margin is a point or two below Q3's actual, and treat anything inside that as the hold band.

My package is this. Confirm the print timing first, since only the sentiment feed says after the close today and the last bar looks provisional. If results are out, skip the trim and run the post-print table from the next regular session.

If the print is still ahead, enter a marketable limit now, in the regular session, and check the fill. Place no after-hours orders in either direction. The target is the loss you'd shrug off on a minus 15 percent gap divided by 15 percent: - The defaults are 30, 25 or 20 as above. - Divide by 25 instead if you want to survive a 2022-style night, which lands near 20. - Sell the highest-basis lots first, and check the one-year mark and the wash-sale rules. - Anyone already at or below their number sits tight, and anyone flat waits.

After the print, use the ceilings and pace above, and reset the stop off the new structure.

FINAL TRANSACTION PROPOSAL: SELL Neutral Analyst: We're down to disagreements worth a few basis points, so I'll take the live items quickly and then raise two things that matter more.

Aggressive, your cash-flow bar is the weak spot. Requiring Q3 free cash flow above Q2's $1.7B means a quarter exactly as bad as the worst of the last six passes. Your own straight line, 25% growth on last year's $30.0B of operating cash flow, gives about $37.5B, which leaves roughly $7B against a $30B capex quarter. Conservative's $5B sits below your central estimate and above "no change," so you can't object without disowning your forecast. It does need OCF of about $35B, roughly 17% year over year and 10% above Q2, after growth slowed from about 34% in Q1 to about 25% in Q2. That's reachable without being a gift. I'd also define it as statement OCF minus statement capex, since the vendor's $21.6B and the roughly $41B we derive for TTM disagree and nobody has explained why.

When you compressed the first tranche to four numbers, you also dropped half of the capex gate. It was never just Q3 capex. It was Q3 capex plus no step-up in implied Q4, which comes from the full-year capex range in the same release. Conservative is right to put it back, and it costs no lateness. "Climb fast and far" doesn't fit a plan built on thirds either. The speed should come from dated gates, not from skipping steps. And the 31% rally with no better margin print is an argument against exiting, but it's also evidence that the trailing margin predicts little. That's why the first tranche stays small.

Conservative, I accept the lower-of rule, with the loss formula as the cap and the tier as the target. I also accept the timing point. If the print is after the close, the next session is the gap day, so the earliest buy is the session after it. Two objections, though.

First, "cross the spread by mid-afternoon" describes what a marketable limit already does in a $1.8T stock during the regular session. If it isn't filling within minutes, it isn't marketable, so reprice it. Spread worries belong to after-hours.

Second, your 70-rung gates have to be satisfiable. "The $25B quantified in dollars with how it's funded and whether an appeal needs a bond" may simply not appear in the 10-Q. If management accrues nothing and says it will appeal, that is an answer, just not a dollar figure. A gate that can't pass is a refusal to rebuild dressed up as a test. Define it by what must be stated: accrual status and, if unaccrued, whether payment is stayed pending appeal. Likewise "Q2 cost step explained" should mean within a couple of billion, not to the penny. If a gate is still open at the date-stamp, freeze at the weight reached and re-test at the next print. Don't sell on it. The bands decide when you sell, and the gates decide when you add. Otherwise someone who missed the trim gets whipsawed by a gate that only ever meant "don't add yet."

Neither of you has flagged that every revenue bar is our own extrapolation. $65B is about +6.9% sequentially, against +7.8% in last year's Q3 and +8.0% in Q2, so it's fair rather than a stretch. But year-over-year growth already slid from 33% to 28%, and the tools show neither the company's Q3 revenue guide nor consensus, which is what the tape reacts to. If the guide's low end is above $62.5B, use that as the zero line.

The bigger point is that someone who doesn't get the trim done carries about 70 more points of standard weight than this plan wants. On a 5% standard that's 3.5% of the portfolio, or about 40 basis points on the worst print in the data. That's roughly ten times the under-4 basis points separating 25 from 30. So the most valuable line in this plan is "enter it now and check the fill," not the tier.

Here is the package. Confirm the print timing first, since only the sentiment feed says after the close today and the last bar looks provisional. If results are out, skip the trim and run the bands from the next regular session.

If the print is still ahead, enter a marketable limit now in the regular session, confirm the fill, and place nothing after hours. The target is the lower of two numbers. One is your loss budget on a 15% gap divided by 15%. The other is 30% of standard weight for a clean book, 25% for an unknown one, or 20% for a known overlap or tight budget. I'd define overlap as more than one other hyperscaler or AI-chip name at full weight. Anyone already below the target sits tight, and anyone flat waits.

After the print, the selling bands work like this: - Go to zero on a margin ex-accrual below 30%, or revenue below $62.5B. - Go to about 15% on a margin of 30 to 32% with no dollar bridge. - Hold 30% on a margin of 32 to 36% or revenue of $62.5B to $65B. If you can't compute the ex-accrual margin, hold the smaller of tonight's size and 25%. - Step to 25% if the implied Q4 margin is more than a point or two below Q3's with no dollar-quantified one-off. - The ceiling is 70% on a pass, and 100% only with a qualifying 2027 guide, which from memory usually comes on the Q3 call.

For buying, the first third needs four things: - Margin ex-accrual of at least 36%. - Revenue of about $65B or better. - No implied Q4 margin drop beyond a point or two. - Q3 capex at or below about $30B, with the full-year range not implying a materially higher Q4. Allow a few billion for finance-lease principal.

That tranche goes in no sooner than the session after the gap day, once the gap-day low holds, and with one wait for Amazon and no more. The second and third thirds need free cash flow of at least $5B, the judgment disclosed as above, and the Q2 cost step explained. Decide those at the first session after the 10-Q or day seven, whichever comes first. Retire 681.41. The gap-day low is a pause line for adds, not an exit.

That keeps the sell, drops the three weakest pieces of the original plan, and makes every gate answerable on a known date.

FINAL TRANSACTION PROPOSAL: SELL

V. Portfolio Manager Decision¶

Portfolio Manager¶

Rating: Underweight

Executive Summary: Underweight META. If tonight's Q3 print is still ahead (only the sentiment feed confirms after-the-close timing, so verify it first), trim during today's regular session with a marketable limit and confirm the fill; place no after-hours orders. Target the lower of two numbers: (a) the loss you would accept on a -15% gap divided by 15%, or (b) 30% of standard weight for a clean book, 25% if you don't know your other AI-capex exposure, or 20% if you hold other hyperscaler or AI-chip names or have a tight budget. Holders already below that sit tight and flat accounts wait. Do not rely on the 681.41 stop through the gap. After the print, apply the margin, revenue and capex bands below, rebuilding in thirds toward 70% and then 100% of standard over 1-2 weeks only if they pass; the horizon is 1-3 months with a $658 target.

Investment Thesis: Why reduce. At $707.95 the 20.4x forward P/E on $34.65 Street EPS needs roughly a 36% NTM operating margin. Q2 printed 30.9%, down from 43.0% a year ago. Costs outgrew revenue in Q4'25, Q1'26 and Q2'26, and Q2 costs rose $8.6B sequentially to about $42.0B. Q2 FCF was $1.7B on $30.1B of capex (49.5% of revenue), against $9-15B in prior quarters. The stock is up about 31% from the 7/30 close of 538.60, with its largest leg an unexplained +11.4% gap on 9/21. The Neutral analyst's cost math was the most decisive work in the debate. Each $1B of cost is about 1.5 margin points on ~$65.6B of trend revenue, so flat costs give exactly 36%. Variable cost of revenue (about 18.6%) plus depreciation from $30B-a-quarter capex puts the no-news case near $43-44B of costs, or 33-34.5% margin. Clearing 36% therefore needs part of the Q2 step to be non-recurring, which is plausible but less likely than not. The price already assumes the recovery the print has to deliver. On top of that, the daily SuperTrend stop at 681.41 is about 1.1 ATR away, while the last four earnings gaps were 8-11% (2.3-3.2 ATR), so stops cannot protect through the print. The $25B New Mexico judgment is only about 1.4% of market cap. But at roughly $9.80 a share against Q2 EPS of $6.18, a Q3 accrual could produce an ugly headline quarter. Deductibility is unverified. Why not Sell. Revenue grew 28% to $60.8B and TTM operating cash flow is about $130B, covering capex in all six quarters shown. The weekly and monthly SuperTrends are up, sitting 24% and 35% below price, and the golden cross came on 10/6. The last time margin printed 30.9%, the stock bottomed that day and rallied 31%. The thin prediction ladder is roughly symmetric: about 30% odds to touch 640 versus about 31% to touch 776. The tape's reaction has tracked guidance and capex rather than trailing margin: 40.1% margin gapped -11%, 41.3% gapped +10%, and 30.9% gapped -8%. Neither side has an edge on direction, so this is a variance cut, not a short. Going from full weight to 30% on a 5% standard saves about 0.4% of the portfolio on a repeat of the -11.3% gap and forgoes about 0.35% on a +10% gap. A -15% gap costs about 0.22%, 0.19% or 0.15% at the 30%, 25% or 20% tiers. Callers who want to survive a 2022-style -25% night should divide their loss budget by 25%, which lands near 20%. Sell the highest-basis lots first and check holding periods and wash-sale rules. I deepened the Research Manager's 40% trim to 20-30% because the Conservative analyst showed tonight's size is also the size carried through the most likely 32-36% middle outcome. Rating label: the trader's 'Sell' explicitly meant a trim, not an exit, which is Underweight on this scale. Post-print target weights (sell down at the gap-day open if above them, buy at the pace below). Go to 0% if margin ex-accrual is below 30% or revenue is below $62.5B; use the guided low end instead if it is higher. Go to about 15% if margin is 30-32% with no dollar bridge. Hold 30% if margin is 32-36% or revenue is $62.5-65B. Step to 25% if the implied Q4 margin is more than 1-2 points below Q3's actual with no dollar-quantified one-off. If the ex-accrual margin can't be computed, hold the smaller of tonight's size and 25%. The ceiling is 70% on a pass and 100% only on a 2027 guide with expense and capex growth at or below revenue growth. The first third of the rebuild requires four things: margin ex-accrual of at least 36%; revenue of about $65B or more; no implied Q4 margin drop beyond 1-2 points; and Q3 capex of about $30B or less, with the full-year range not implying a materially higher Q4 (allow a few billion for finance-lease principal). It goes in no sooner than the session after the gap day, once the gap-day low holds, with one wait for Amazon. The second and third thirds require: Q3 FCF of about $5B or more (statement OCF minus capex); the judgment's accrual status disclosed, or a stay pending appeal; and the Q2 cost step explained to within a couple of billion. Decide these at the first session after the 10-Q or day 7, whichever comes first. A gate that stays open freezes adds; it does not trigger sells. I am retiring the 681.41 close rule and the flat 10% Q4 cost-growth exit. The latter would have fired on Q4'25, when costs rose about 14.5% and the stock gapped +10%. The stop resets off the post-print structure. What changes the call: a 36%+ margin plus a disciplined 2027 cost and capex guide moves this toward Overweight or Buy; a margin near 31% with no bridge, or revenue below $62.5B, moves it to Sell. Price target ($658) uses a valuation method: $34.65 forward EPS × (33.5% central-case margin ÷ 36% implied margin) = about $32.24, × the current 20.4x forward multiple = about $658. It sits just below the 50 SMA at 661.41 and the 665.23 gap base. Bull case: 36%+ margin and a retest of $760-780. Bear case: about 30.8% margin, EPS about $29.6, and about $604, close to the -15% gap scenario. Data gaps lower confidence: the period behind the $34.65 is undated; the earnings date is unconfirmed; today's bar may be provisional; and the tools lack the D&A and opex breakdown, the Q3 revenue guide or consensus, New Mexico verdict details, the 9/21 catalyst, the ~$10.6B H1 cash gap, options data and the caller's holdings.

Current Price: 707.95

Price Target: 658.0

Confidence: Medium

Time Horizon: 1-3 months