Office loans are going bad faster than in 2008. The 10-year has been above 4% nearly every day this year. Banks are lending into the AI data center boom. And big-bank earnings are days away. Here's what history says: BAC, JPM and WFC have beaten the S&P 500 from the October print to the January print in 14 of the last 16 years. This year, the backdrop is different: - Office CMBS delinquency: 12.0%, above the 2008 crisis peak - $875B of commercial mortgages mature this year, roughly half on bank balance sheets - 10-year: above 5%, higher than at any October print in our data - Data centers: WFC and BAC both flagged risk in long-dated AI infrastructure lending Our read: the pattern still holds, at about half strength. +3% to +5% vs. the S&P 500, not the usual +7% to +8%. Our 4,000+ subscribers got it first yesterday. Full report in the reply. $JPM $BAC $WFC $C $GS $MS alphatica.io/research/octobe…

For the last year you've seen our research and daily reads, the published studies, and the results, including the ones that missed. That's a fraction of what we run. Over the next quarter, our full research goes to email subscribers first. We're getting closer to launch. Join the list at alphatica.io $SPY

The S&P 500 only looks cheap if margins do something they've never done. Consensus: 28.7% EBITDA margins by 2028. 20-year record: 20.8%. And the cash is going the other way. Free cash flow as a share of EBITDA dropped from 48% to 41% in two quarters. Lowest since 2015. That's the capex bill. It doesn't disappear. It comes back as depreciation, right into the margins the street is counting on. 2027 P/E on consensus: 18.3x. 2027 P/E if margins just hold: 22.4x. The market isn't pricing growth. It's pricing a margin miracle. What breaks first? $SPY

MU reports tonight after the close. The biggest print of the new season. 13 straight beats. The last three: +21%, +31%, +20%. Up 550% in 52 weeks. And after guiding to a $50B quarter in June, larger than every full year Micron reported through FY2025, the stock has gone from $1,049 to $1,065. Three months, a record guide, and the stock went nowhere. Tonight we also make a call on the split. Street EPS: $31.82 Street Revenue: $51.5B Alphatica Estimates: EPS: $34.00 Revenue: $54.0B OUR CALL: BEAT OUR SPLIT CALL: YES, 10-for-1 Micron guided this quarter at $50.0B plus or minus $1.0B in revenue and $31.00 plus or minus $1.00 in EPS, with gross margin near 86%. The street is at $51.5B. That is above the top of management's own range. EPS at $31.82 sits near the ceiling. The street has already priced in a beat. We are sizing above it at $34.00 and $54.0B, about 10% above the guide midpoint on EPS. The last three quarters beat the street by 20% or more. We are not assuming a fourth. We sized MU at $22.50 in June. It printed $25.11. We were closer than the street and still too low. This cycle has run ahead of every estimate. The difference tonight: the street is already above the guide. WHAT MANAGEMENT SAID MID-QUARTER: At KeyBanc on August 10, Chief Business Officer Sumit Sadana said aggregate customer demand signals have increased even further since earnings. Calendar 2027 is now expected to be even tighter than 2026. Still no line of sight on when supply meets demand. U.S. investment plan raised from $200B to $250B. 16 strategic customer agreements, $22B in commitments, $18B of it cash. Take-or-pay terms, meaning customers pay whether they take delivery or not, running through 2030. More signed since. The detail most will miss: most volume under those agreements carries a price band. The floor sits at a margin well above any prior peak. The ceiling on the first 16 is set at calendar Q2 pricing. If memory prices keep climbing, part of Micron's volume is capped. Watch whether gross margin clears the 86% guide or stalls near it. OUR FOUR SIGNALS: Earnings Quality: 3/6 NEUTRAL. Operating trends, earnings quality, and capital position score BULLISH. Stock compensation at 2.6% of revenue is the lowest in our semiconductor database, and cash from operations ran 2x net income. Profitability scores NEUTRAL on memory's cycle history. Disclosure Signal: NEUTRAL, down from BULLISH. Last year's annual report produced the strongest tone improvement in our entire universe. We held BULLISH in June waiting for the quarterly filings to confirm it. Four straight quarterly filings later, tone is still stuck in negative territory. Prediction Model: BEAT. 13 straight beats. September reports: 4 of 4 beats since 2022, averaging +6.1%. Estimates rose 29.6% over 90 days, then locked for 60. Two downward revisions in the last 30 days, the first downward activity on MU in months. One analyst raised EPS 13% two days ago. The street is still grinding higher, but the unanimity has cracked. Earnings Risk Signal: +/-5.0% NEUTRAL. Calls and puts priced about even at +1.66. No lean. Options are pricing a $53 move. Micron's average post-earnings move over three years is 9.3%. 8 of its last 11 prints moved more than 5%. The options are pricing roughly half the move this stock usually delivers. THE SPLIT QUESTION: $1,065 a share. $1.23 trillion market cap. One options contract controls about $106,500 of stock. Since 2024, our split database shows 22 Russell splits of 4-for-1 or larger. 10-for-1 is the most common ratio, 7 of them. The four semiconductor names in that group, NVDA, AVGO, LRCX, and SMCI, all announced alongside earnings. LRCX and SMCI announced with fiscal year-end results. Tonight is Micron's fiscal year-end print. The share count allows it. 3.0B shares authorized, 1.13B outstanding. That covers a 2-for-1 as is. Anything larger needs more authorized shares, the same step NVDA and AVGO cleared within weeks without a shareholder vote. The case against: Micron has not split once in our data going back to 2014, memory companies manage for the cycle, and SNDK sits at $1,745 without announcing one. Our call: YES. 10-for-1. It is the lowest-conviction call on this card, and we are putting a timestamp on it anyway. THE SETUP: The guide is the trade. FQ1 consensus is $34.95 EPS and $56.9B. If Micron prints $34 tonight, that FQ1 number is already stale. $1,065 is the price. $50B was the guide. $51.5B is the street. $31.82 is the bar. 13 straight is the streak. Options are pricing half a normal Micron move. Three yellow circles, a green, and a split call on the table. Tonight the memory supercycle prints its fiscal year. Not Investment Advice $MU $SPY $QQQ

🚨 SPY WEEKLY OPEX: THE POSITIVE-GAMMA REGIME IS BACK The Fed hike landed. Monthly OPEX cleared. The structure survived. Two weeks ago: Gamma: −1.08M Delta: +8.17M Today: Gamma: +229.8K Delta: +37.59M G+D+ has been restored. This is the same regime that powered the May rally and August recovery. Dealers are buying dips. Ranges are compressing. The mechanical bid is at its strongest since early August. Friday erased 85.2% of pre-OPEX gamma, leaving the chain thin but still positive. The window is narrow: Positive gamma through October 1. October 2 OPEX begins the reversal. October 16 carries −222.4K gamma and 2.51M open interest. The 22-week pattern is clear: Positive gamma builds between monthly expirations. OPEX strips it. The structure resets. Then it rebuilds. We are inside the positive window now. Quarter-end rebalancing arrives Tuesday. After that, the structure supports the grind higher until October OPEX strips it again. Right on schedule for a possible push to new all-time highs in SPY. $SPY $SPX $QQQ

Shit we have a live trade. The USD/JPY will print a death cross within two sessions. It's already locked in. Everyone will trade it as risk-off. Nine prior crosses: the S&P returned +1.34% over 20 days, against +1.02% for a random window. The only scary one was March 2025, and that was tariffs. The cross is noise. The setup before it isn't. The trade: every time the gap between the USD/JPY 50-day and 200-day moving averages narrowed to less than 1.0, buy the Nasdaq 100 (QQQ), short the Russell 2000 (IWM), and hold for 20 trading days. Here's what that spread returned each time: Oct 2015 +3.46pp Jun 2017 +0.72pp Jan 2018 +1.95pp Jan 2019 +0.01pp Mar 2020 +10.20pp Jan 2023 +4.47pp Feb 2024 +1.85pp Sep 2024 +2.36pp Mar 2025 +2.85pp Nine for nine. Median +2.36pp. Worst +0.01pp. Six times the baseline spread. It paid more when the S&P fell (+3.93pp) than when it rose (+2.43pp). A stronger yen raises funding costs. Levered small caps pay it. Cash-rich mega-cap tech doesn't, whether the tape is up or down. Long QQQ, short IWM. We will post the 20-day result either way. $QQQ $IWM $SPY

🚨🚨THE PATTERN IS UNDENIABLE We tracked six major events since late July. Each one followed the same cycle: panic, recovery, sell. Here is what the institutions did every time. 1. July FOMC: $729 low → $779 recovery 2. August OPEX: $769 low → $775 recovery 3. Pre-NVDA: $762 low → $774 recovery 4. Jackson Hole: $769 low → $772 recovery 5. Iran shock: $761 low → $774 recovery 6. September FOMC: $752 low → $763 recovery (developing) The panics produce 3-5% recoveries in 3-6 sessions. Every time. The buying shows up in the dark pool before the event, holds through the panic, and sells into the recovery. Build before. Hold during. Sell after. Six events. Same playbook. But look at the recovery highs. $779. $775. $774. $772. $774. $763. The ceiling is compressing. Each recovery fails to reach the prior high. The panics are consistent. The recoveries are shrinking. The macro explains why. July had three dissents. August killed forward guidance. September added a rate hike at 3.75-4.00% with another expected. The 10-year near 5%. Each event added a structural headwind that didn't exist before. The institutions recover from each panic into a lower ceiling because the cost of capital rose between events. This is not a crash. The lows are holding. $757-$762 has been the floor for events 3 through 6. The falling wedge breakout at $751.72 held by $2.33 on the worst day. The support is intact. This is compression. The range is narrowing from the top. The recoveries are getting shorter. The institutions are distributing at progressively lower levels. Controlled. Methodical. Event by event. The opportunity is on both sides of the book. The longs work during the panic phase. The shorts work at the recovery peak. The dark pool flow tells you when each phase starts. We are watching whether this is the market finding its footing at a new rate regime or the early stage of something larger. The Five Conditions scorecard is still 0 out of 5 (we are not at a market top). The overnight repairs keep firing. The correlations are green. The structure has not broken. But the lower highs are real. And the pattern is undeniable. $SPY $QQQ $IWM

SPY is 2.08% from its all-time closing high. 25 sessions since the $777.88 close on August 13 (intraday high: $779.37). FinTwit continues to call the top. We tested what actually happens from here. Every day since 2004 where SPY sat within 2.5% of its ATH. 2,000 days. 39 pullback episodes lasting 20+ days. Here's the base rate. When SPY is within 2.5% of ATH, probability of a new all-time high: Within 20 trading days: 73% Within 40 trading days: 87% Within 60 trading days: 91% Within one year: 98% Today's specific analog (within 2.5%, ≥20 days since ATH): Within 20 days: 64% Within 40 days: 83% Within 60 days: 87% Roughly 2-in-3 by mid-October. 85% by mid-November. The base rate says SPY makes a new high. The race table is the real finding. When within 2.5% of ATH: 🚨SPY hits a new ATH before falling 5%: 83% 🚨SPY hits a new ATH before falling 10%: 95% Once you drop below -2.5% to the -2.5% to -5% band, "ATH before -5%" collapses from 83% to 43%. A 40 percentage point cliff. The 2.5% zone is a genuine regime boundary. Inside it, gravity pulls higher. Outside it, the race becomes a coin flip. The honest caveat: every major top had a "touch the band, then fail" phase. October 2007. January 2018. September 2018. February 2020. January 2022. Five of 39 episodes, 13% came within 2.5% and then broke down. That's the tail risk. Today could be one of those. But you don't trade the 13% as the base case. 39 pullback episodes. Median resolution: 15 trading days to a new ATH. 34 of 39 resolved within two months. The market has done 97.5% of the work to get back. The last 2.5% resolves the vast majority of the time. $SPY $QQQ $IWM

HOW TO READ ALPHATICA A lot of you followed us through the SPX 7,462 call. We published that target in October 2025 and then we reemphasized the target in March when SPX was in the low 6,600s. It hit in May. Along the way, we flagged a 7/7 rising wedge, a head and shoulders forming, accelerators firing, and gamma cliffs. Some of you asked: how can you be bullish AND flag bearish patterns at the same time? Here is the answer. And if you only read one post from us this year, make it this one. WE OPERATE ON TWO TIMEFRAMES. ALWAYS. The long-term thesis tells you where the market is going over months. SPX 7,462 was a year-end target. We think SPX 8,600 is reachable within 12-14 months. We are working on an updated year-end target now. The short-term structure tells you how it gets there. Not in a straight line. Never in a straight line. The path from 6,600 to 7,462 included a -7.72% drawdown, a war, a ceasefire, a gamma cliff that drained 37% of suppression overnight, four 25+ point composite swings in a single week, and two confirmed bearish patterns. Both were right at the same time. The target was right. The rising wedge was right. The gamma cliff was right. They operated on different timeframes. WHAT THE SHORT-TERM SIGNALS MEAN: When we flag a rising wedge, a head and shoulders, or a gamma cliff, we're not saying "the market is crashing." We're saying "the next 10-20 trading days have elevated risk of a 3-6% pullback." A 5% correction inside a 15% annual move is normal. It's not a contradiction. It's the path. The institutions we track don't sell their entire book when a wedge forms. They add hedges, adjust timing, and buy the dip when the pattern resolves. That's what $700 put OI at 451K tells you. The same institutions building bullish flow records are simultaneously hedging 6% below. Long AND protected. They believe in the direction. They respect the volatility. WHAT THE LONG-TERM TARGETS MEAN: When we publish a year-end target, we're saying "the structural, corporate earnings, macro, and flow data supports this level over this timeframe." It doesn't mean every week is green. It doesn't mean every pattern is bullish. It means the weight of evidence points to that destination. 7,462 took 10 weeks and included the worst single-week flow reversal since the ceasefire. The path was violent. The destination was reached. HOW TO USE BOTH: If you are a swing trader: the short-term signals are for you. The rising wedge, the accelerators, the gamma magnets, the composite score. These tell you about the next 1-20 trading days. Use them for timing. Respect the levels. If you are a position trader or investor: the long-term targets are for you. The short-term signals are noise in your timeframe. A 5% dip that our data flags is your buying opportunity, not your exit. If you are both: welcome to what the institutions do. They hold the long-term thesis and trade around the short-term structure. That's what "long and protected" means. That's what 451K puts at $700 alongside record call flow looks like. WHAT WE WILL NEVER DO: We will never tell you to buy or sell. We provide data and a framework. What you do with it depends on YOUR timeframe, YOUR risk tolerance, and YOUR strategy. A post that flags a bearish pattern is not a sell signal. A post that flags bullish flow is not a buy signal. Both are data. The decision is yours. THE BOTTOM LINE: We can publish SPX 8,600 as a 12-month target and flag a bearish pattern in the same week. Both can be right. The target is the destination. The pattern is the turbulence along the way. An analogy or how to think about it is the plane lands where it's going. It just doesn't fly in a straight line. Our job is to map both the flight plan and the turbulence. Your job is to know which one applies to your seat. $SPY $QQQ $VIX

We are heading to all time highs soon. $SPY

$SPY $740 should be the floor.

HERE WE GO. FIRST ONE AS WE MARCH CLOSER TO TUESDAY: Dark pool alert. After hours. 5:48 PM. 17:48:46 GOOGL BUY $500,000,093 NASDAQ Carte $500M $GOOGL buy. NASDAQ Carteret. Today's session: zero dark pool buyers. Every name sold. The heaviest dark pool sell of sixteen weeks. $GOOGL was sold -$452M in the dark pool yesterday. Both tapes agreed. After the close: $500M buy. One print. The dark pool that sold everything during the session is buying $GOOGL after hours. The after-hours print is the positioning. The session print was the exit. Both documented. Watching the tape. $GOOGL $SPY $QQQ

Five major tech names report this week. We ran the event study first. 399 earnings events back to 2001. MEDIAN 10-DAY ABNORMAL MOVE AFTER EACH REPORTS: TSLA +1.37% TXN +0.40% INTC +0.07% GOOGL -1.64% NOW -1.88% Three of those five are noise. Their confidence intervals straddle zero. Two hold up: GOOGL wins only 37.3% of the time across 83 events. NOW wins 32.7%. Pooled across all 399 events, the average 10-day move is +0.06%. The p-value is 0.86. There is no post-earnings drift in this group. HOW FAR THEY ACTUALLY MOVE: Share of quarters where the 10-day abnormal move exceeds 5%: TSLA 45.0% GOOGL 39.8% NOW 34.6% INTC 24.5% TXN 20.6% TSLA clears 5% after nearly half its prints. TXN after one in five. Same event. Same week. More than twice the frequency. The mean says nothing happens. The tails say otherwise. THE SKEW PROBLEM: The typical quarter is negative. A small number of very large winners drag the average back to flat. TXN gained 19.4% in a single session in April. That one print offsets a lot of small losses. Post 2010, the median 20-day abnormal return is -0.68% and statistically significant. Following a low-surprise quarter: -1.38%. Anyone quoting an average on these names is describing something that doesn't happen to most quarters. THREE THINGS WE TESTED AND KILLED: Pre-earnings momentum predicts nothing. Eighteen tests. Zero significant. Guidance revision predicts nothing. TXN's April guidance moved consensus 21.6% in one session. Correlation with the next 20 days: 0.07. Seasonality predicts nothing. Twenty-four tests. Zero survived. All five names are going into this week down. On this evidence, that tells you nothing about what comes next. THE BOTTOM LINE: The announcement gap is the event. The drift after it is not. The information is priced by T+1. Beat rates going in: TXN 80.2%, $INTC 75.5%, NOW 75.0%, GOOGL 68.7%, TSLA 46.7%. GOOGL has beaten thirteen quarters in a row. TXN, $GOOGL, $TSLA, and NOW report Wednesday after the close. INTC reports Thursday after the close. We will have full earnings coverage on each name this week.

*****We want to drive the follower count up for the Alphatica YouTube channel. It will help supplement costs for the platform at launch***** Please repost. We will have coverage on the following topics on our youtube channel. It will be a gold mind of information. Macro / Fed / Rates Analyst Semiconductors / AI Hardware Analyst Equity Strategy / Sectors Analyst Short Research / Forensic Analyst Earnings / Company Deep Dive Analyst Youtube: @Alphatica $SPY $QQQ $IWM

We will have coverage on the following topics on our youtube channel. It will be a gold mind of information. Macro / Fed / Rates Analyst Semiconductors / AI Hardware Analyst Equity Strategy / Sectors Analyst Short Research / Forensic Analyst Earnings / Company Deep Dive Analyst Youtube: @Alphatica $SPY $QQQ $IWM

We want to give our channel a quick update as we are seeing unusual patterns in the Dark Pool today. There was a recent $1B+ of simultaneous selling across QQQ, IWM, TLT, and BRK.B. All sells. Zero buys. We think there are two forces hitting the same session. The equity selling (QQQ, IWM, BRK.B) is calendar-driven. Most likely US pension funds all trimming equities to target weights. Today is day one of the window. This flow ends June 30. The bond selling (TLT, the largest print at $342.7M) is structural. Warsh eliminated forward guidance last Wednesday. The dot plot that anchored long-duration rate expectations since 2012 is gone. TLT prices the 20-year rate path. Without dots, that path is unknowable. Selling duration is the rational response to permanently higher rate uncertainty. In a normal rebalancing, institutions sell stocks and buy bonds. Today they're selling both. The equity selling we believe has an expiration date. The bond selling may not. $TLT $SPY $QQQ

Kevin Warsh just made his first policy decision. Rates held at 3.50%-3.75%. The hawkish surprise: 9 of 18 FOMC participants now see a rate hike this year. March projections called for cuts. That's a full reversal. We studied every new Fed Chair's first decision day over the last 20 years. Here's what the data said and what just happened. The pattern: Day 1 always sells off. Bernanke (2006): SPY -0.62%, QQQ -0.41% Yellen (2014): SPY -0.53%, QQQ -0.54% Powell (2018): SPY -0.19%, QQQ -0.44% Six data points. Six selloffs. The market prices uncertainty under every new Chair. Then nothing happens. 20-day forward returns after the first decision are indistinguishable from any other FOMC day. No statistical significance at any horizon. We pooled the first three meetings under each new Chair (n=9) and compared to all 182 FOMC days. Smallest p-value: 0.25. There is no "new Chair effect" beyond Day 1. The macro shift is real though. The dots moved from projecting cuts to projecting hikes. PCE forecast jumped from 2.7% to 3.6%. GDP trimmed from 2.4% to 2.2%. The Fed is acknowledging what our inflation research has been saying: America has an inflation problem. Two hot prints (CPI and PPI) in the last month. The data confirmed it. Now the Fed confirmed it. Three data points in 20 years. Not enough for a p-value. Enough for a pattern. If history holds, the market processes the uncertainty today and moves on within a week. The 20-day, 30-day, and 60-day clock starts now. We'll publish the results. $SPY $QQQ $IWM