Market Pulse · Daily Scan
Daily Market Pulse
Aug 28, 2026 · sentiment/flow window ≈ 48h
Pre-marketnot publishedPost-market
TLDR;
Two sessions that answered the year's two biggest questions in opposite directions, and the market could not hold both. On Aug 27 Nvidia's FQ2 print - revenue $96.22B, more than doubled year over year, with projected fiscal-2028 revenue growth of 70% against a 44% consensus - was supposed to be the session's story, and it was not. The larger event was that AI-exposed SOFTWARE re-rated violently: Salesforce +22.5% on an FQ2 beat and a Claudeforce collaboration with Anthropic, Okta +28%, CrowdStrike +20.5% on its best day ever, Palo Alto +12.8%, Synopsys +13%. The narrative that flipped is the important one: for eighteen months enterprise software traded as the VICTIM of AI, and in one session the market decided the agent era creates billable surface area rather than destroying it - Okta's management put a number on it, estimating roughly 90 machine identities per human worker requiring security. Then on Aug 28 Kevin Warsh used his first Jackson Hole keynote to say that better summer PCE and CPI readings do not indicate underlying trends have meaningfully improved, September HIKE odds jumped to 57%, and the tape gave back what it could: Nvidia -4.45%, the Russell 2000 -1.37%, gold down roughly 3%. The window's cleanest through-line is about DISTANCE, not direction. Nvidia was paid 8.7% for an out-year number. Marvell beat, RAISED FY27/FY28 growth guidance to 45%/50% from 40%/45%, disclosed a Google custom-chip arrangement worth up to $120B through fiscal 2033 - and fell 10.3%, because the money is too far away. Rubrik beat by 400% on EPS, raised full-year revenue guidance by more than $400M, missed gross margin by 70 basis points, and fell 8%. Autodesk beat and guided light, and fell. PayPal, which had done nothing at all, fell 12.5% because a takeover premium evaporated. In a 57%-hike-odds regime the market is discounting cash flows steeply and paying only for what is close. Breadth: Extraordinarily narrow on Aug 27 and broadly negative on Aug 28. On Aug 27 the S&P rose 0.72% and the Nasdaq 1.57% while TEN of the eleven sectors closed NEGATIVE - technology was the only gainer, with XLK +3.2% - and decliners outnumbered advancers on the NYSE by 1.02 to 1. TheStreet noted the Dow closed +0.19% with just SIX of its thirty holdings advancing. That is a one-sector index gain, not a rally. On Aug 28 the Russell 2000 fell 1.37% to 2,973.09 against the S&P's 0.25% decline, a 112-basis-point underperformance on the session the September hike probability moved to 57% - the small-cap tape is now the cleanest expression of the rate view. For the week all three major indexes still finished higher (S&P +0.5%, Nasdaq +0.9%, Dow +0.5%, the Dow's first winning week in three), which is a fair description of the headline and a poor description of the market..A · Daily Pulse
What moved — and why
▲ Rising — dailycomposite
| # | Ticker | Move | Driver | Score |
|---|---|---|---|---|
| 1 | CRM | — | The defining print of the window and, by scale, the most improbable move in the artifact. Salesforce closed +22.5% on Aug 27 on an FQ2 revenue beat, a single-session gain that 24/7 Wall St. characterised as essentially unheard of at this market capitalisation - and then added another 3.06% on Aug 28 as a Dow gainer while the index fell. The catalyst had two legs: the quarter itself, and the announcement of a Claudeforce collaboration with Anthropic, which reframed Salesforce from an AI-disruption VICTIM into an AI-distribution winner. That reframing is the window's central theme. MAGNITUDE CONFLICT, flagged not averaged: 24/7 Wall St. gives +22.5% close-basis, CNBC 'more than 22%', a Tier-3 aggregator +22.6%, and an intraday 24/7 card carried +8.71% at a $206B valuation earlier the same session. The close-basis figure is used and the spread disclosed. | 91 |
| 2 | OKTA | — | Largest single-session percentage gain of any ranked name in the window. Okta beat and raised on agentic-AI-driven identity demand; management framed the thesis quantitatively, estimating roughly 90 machine identities per human worker requiring security, with one customer's Claude agent instances reportedly going from 50 to 1,500 in weeks and 81% of surveyed CISOs conceding agents run without adequate security. That is a demand argument specific to the agent era rather than a generic security-spend argument, which is why it re-rated the whole identity cohort. MAGNITUDE CONFLICT: CNBC gives 'more than 28%', 24/7 Wall St. gives 29-30%. The lower CNBC figure is used and the spread disclosed. MARKET CAP NOT VERIFIED - the only figures retrievable were stale aggregator pages dated to late 2025 ($15.0-15.9B); the stock is described by two sources as having roughly doubled in 2026, so eligibility is not in doubt at any plausible figure, but the number is flagged rather than estimated. | 90 |
| 3 | CRWD | — | Best single day in company history: +20.5% on Aug 27 on a record FQ2 - adjusted EPS $0.31 against $0.29, revenue $1.47B against $1.44B, FY net new ARR growth guidance raised to 34% at the midpoint and FY revenue guided to $5.99-6.01B against a $5.93B estimate. Then gave back roughly 4.7% on Aug 28 in sympathy with SentinelOne's guidance cut rather than on any CrowdStrike-specific news, which is the cleanest illustration available of how tightly the cybersecurity cohort is now correlated. AUG 28 MAGNITUDE CONFLICT: a Tier-3 tracker gives -4.72% at the open, 24/7 Wall St. headlines 'CrowdStrike declines 7%' while its own body text says the sympathy move was -4%. The -4.72% open-basis figure is used and explicitly labelled as such, because no confirmed CLOSE-basis figure was obtainable for Aug 28. | 86 |
| 4 | NVDA | — | The event the entire window was organised around, and the reason both tracks moved. Nvidia reported Aug 26 AMC: Q2 revenue $96.22B against a $2.22 EPS versus $2.09 expected, with revenue more than DOUBLING year over year, and - the number that actually mattered - a projected fiscal-2028 revenue growth rate of 70% against the 44% that LSEG-surveyed analysts had modelled. Shares closed +8.7% on Aug 27, lifting the whole complex. Then fell 4.45% on Aug 28 as Warsh's Jackson Hole remarks lifted September hike odds to 57%. The two-session round trip is the point: a 26-point beat on out-year growth bought the sector roughly one day. CNBC notes NVDA rose only about 1% for the WEEK while SMH fell more than 3% - the guide did not generalise. AUG 28 MAGNITUDE CONFLICT: Trading Economics' post-close Dow list gives -4.45%, a Yahoo article sidebar gives -4.85%, and a live aggregator page carried -0.2% at midday. The close-basis Dow figure is used; the aggregator is discarded for reasons documented in data_limitations. | 81 |
| 5 | SNPS | — | The best-performing of the most-active names on Aug 27 per a stockanalysis.com feed item, which puts the single-session gain above 13% following an Aug 26 AMC report. Analytically this is the EDA read-through from the Nvidia guide: if fiscal-2028 accelerator revenue is going to grow 70%, the design-tools layer sits upstream of every one of those chips, and Synopsys is the purest listed expression of that. MAGNITUDE CONFLICT, flagged: the 13% figure is a single-session Aug 27 read, while CNBC's Aug 28 wrap describes Synopsys as up 'more than 11%' on a WEEK-TO-DATE basis in the same paragraph that gives Marvell and Applied Materials weekly declines. Those two bases are not reconcilable from free sources and are not averaged. Ranked mid-list rather than high because of that ambiguity. | 76 |
| 6 | PANW | — | Rose roughly 12.8% on Aug 27 with NO earnings of its own - pure sympathy with the CrowdStrike and Okta prints. That is precisely why momentum is the only high sub-score here and why the catalyst score reflects the FORWARD event rather than the realised one: Palo Alto reports fiscal Q4 on Sep 1, the first trading day inside the horizon, and it will be reporting into a cohort that has just been re-rated on somebody else's numbers. An intraday 24/7 Wall St. card sized a 7% move as adding roughly $20B to a $314B company; the $314B figure is used as the market-cap basis and flagged approximate since it is an intraday derivation rather than a dated statistical source. | 71 |
| 7 | AFRM | — | The clearest consumer-credit datapoint of the window, and it cut AGAINST the tightening narrative. Affirm reported FQ4 revenue up 33% to $1.17B with GMV up 36% to $14.1B, operating income up to $147.3M from $58.1M, and net income of $1.62B - though that headline number is inflated by the release of a valuation allowance on domestic deferred tax assets and should NOT be read as operating earnings. FY2027 guidance exceeded expectations. Worth isolating because it lands in the same session as a hawkish Warsh: a buy-now-pay-later lender posting 36% GMV growth and rising margins is not what a consumer under rate stress looks like. MAGNITUDE CONFLICT: two independent post-close sources give +9%, a Tier-3 recap gives +9.5%, Schwab's 9:13 ET open note gave +12%, and a Yahoo trending-ticker widget showed +5.08%. The +9% two-source consensus is used and marked approximate. | 69 |
| 8 | AVGO | — | Rose 3.79% on Aug 27 in the Nvidia read-through and ranks low on momentum for exactly that reason - it is a sympathy move, not an idiosyncratic one. It is ranked at all because of the FORWARD catalyst, which is the largest scheduled single-name event inside the horizon: Q3 FY2026 results with consensus looking for roughly 91.5% EPS growth to $3.24 on revenue up 84.1% to $29.4B, and semiconductor solutions revenue forecast to more than double from $9.2B to $18.5B. Schwab's Aug 28 note framed it precisely: Nvidia's 70% out-year guide soothed the AI-durability question, and investors now want the same tidings from Broadcom. CATALYST DATE CONFLICT, flagged rather than resolved: a CMC Markets week-ahead places the Broadcom preview under Wednesday Sep 2, while this pipeline's own 2026-07-31 artifact recorded Sep 3. Both dates are carried in the calendar. | 67 |
| 9 | AMZN | — | Top Dow gainer on Aug 28 at +4.02% on a session when the index closed lower and the Russell 2000 fell 1.37% - a striking divergence with no confirmed single-name catalyst on the day. Two things make it worth ranking despite the unverified driver. First, it is close-basis verified from a post-close source. Second, the direction is informative in context: on Aug 27 a Tier-3 AI-sector tracker recorded AMZN, META and GOOGL all DIPPING as capital rotated out of hyperscalers and into security and SaaS, and then on Aug 28 Amazon alone reversed hard while the software winners consolidated. Ranked low because the magnitude is modest and the driver is unconfirmed - stated rather than invented. unverified | 58 |
| 10 | ADBE | — | Rose 5.7% on Aug 27 with no Adobe-specific catalyst, carried by the Salesforce-led software re-rate. Flagged as a deep-dive library name and it surfaced ORGANICALLY - this is the first time in several runs that Adobe has appeared in a ranked list rather than sitting in the Track B downside runners-up. The connection to prior work is direct and uncomfortable for that runner-up thesis: Adobe has been held out of the ranked downside book on the argument that at roughly 10x forward the AI-disruption de-rating had already done its work, and a 5.7% single-session gain on a day the market decided AI-exposed software companies are BENEFICIARIES rather than victims is a small piece of evidence in the same direction. Ranked last on this list on magnitude and unverified driver. ↔ deep-diveunverified | 54 |
▼ Falling — dailycomposite
| # | Ticker | Move | Driver | Score |
|---|---|---|---|---|
| 1 | PYPL | — | The largest decline of the window and the only one driven by a pure de-rating rather than an operating result. Bloomberg reported that Advent International and Stripe had abandoned a pursuit of PayPal valued above $50B, which would have ranked among the largest leveraged buyouts ever. The stock had embedded a takeover premium; removing the possibility removed the premium. This is a cleaner lesson than an earnings miss: nothing about the underlying business changed on Aug 28, only the probability distribution over ownership. MAGNITUDE FLAGGED: Bloomberg's premarket print was -16%, Schwab's 9:13 ET note -14%, a Yahoo trending widget -12.05% at $54.07, the Yahoo quote page -13% with an internal reference to a 12.5% plunge, and two post-close recaps 'more than 12%' and '13-15%'. The post-close cluster is -12% to -13%; -12.5% is used and marked approximate. MARKET CAP CONFLICT: companiesmarketcap gives $52.44B for August and MacroTrends $52.75B as of Aug 17, while Trading Economics gives $64.09B. The lower clustered figure is used; eligibility is not in doubt at any of them. | 86 |
| 2 | MRVL | — | The most analytically important decline in the window, because Marvell BEAT AND RAISED and was punished anyway. FQ2 adjusted EPS $0.94 against $0.92, revenue up 37% year over year to $2.74B against $2.72B, data-center revenue up 46% to $2.17B, and FY27/FY28 revenue growth guidance RAISED to 45%/50% from 40%/45%. It fell roughly 10.3%. Two reasons are visible in the record: Q3 gross-margin guidance of 57.5-58.5% against a 58.5% StreetAccount consensus, and - the substantive one - investor frustration that the newly announced Google custom-chip arrangement, which could generate up to $120B of revenue through fiscal 2033, carries no near-term revenue contribution. This is the mirror image of Nvidia forty-eight hours earlier: Nvidia was paid for an out-year number, Marvell was not. The discriminator appears to be proximity, not magnitude. Note the stock entered the print up 179% year to date at roughly 58x forward. MAGNITUDE: post-close sources cluster -10.28% to -10.47% with Trading Economics at -10.3%; intraday prints ranged -6.75% at 09:17 ET to -8% at 10:55 ET. The -10.3% close-basis figure is used. Aug 27's -1.30% regular-session close at $241.93 independently corroborates it - $216.79 quoted midday Aug 28 is -10.4% off that base. | 84 |
| 3 | RBRK | — | The purest sell-the-news reaction of the window. Rubrik beat massively - FQ2 adjusted EPS $0.20 against a $0.04 estimate on revenue of $427.3M against roughly $396M, with revenue up 38% year over year, 3,084 customers above $100K of annual recurring subscription revenue, and FY revenue guidance of $1.685-1.693B against a $1.281B consensus - and fell about 8%. The single identifiable blemish was non-GAAP gross margin of 81% against an 81.7% StreetAccount consensus. That is a 70-basis-point miss against a guidance raise of more than $400M, and the 70 basis points won. Rubrik had also risen double digits on Aug 27 in the CrowdStrike/Okta sympathy move, so a meaningful part of this is simply the give-back of a gain earned on someone else's news. MAGNITUDE CONFLICT: an after-hours print of -10%, a premarket print of 'over 5%', and post-close recaps at -8%. The -8% post-close figure is used. MARKET CAP NOT VERIFIED this run; on a $1.69B FY revenue guide at any plausible software multiple, eligibility is not in doubt, but the figure is flagged rather than estimated. | 73 |
| 4 | ADSK | — | Third instance in this window of a beat punished on the guide, and the tidiest one. Autodesk delivered FQ2 adjusted EPS of $3.30 against $3.12 on revenue of $2.05B against $2.01B, with a 41% adjusted operating margin and $561M of free cash flow - then guided Q3 adjusted EPS to $3.04-3.09 against a $3.14 consensus and full-year FY2027 EPS to $12.52-12.60, below the midpoint of estimates, citing near-term commercial real estate and construction design billings. Placed on the falling list despite being UP 6.2% on Aug 27 because the Aug 27 gain was software-cohort sympathy while the Aug 28 decline was Autodesk's own catalyst - the placement rule for this run is the name's own idiosyncratic event, and both sessions are recorded so the reader can net them. The commercial-real-estate attribution is worth isolating: it is the only rate-transmission-to-real-economy datapoint in the artifact, and it landed in the same session as a hawkish Warsh. | 69 |
| 5 | MMM | — | Second-worst Dow component on Aug 28 at -2.56% with no confirmed single-name catalyst. Included because it is close-basis verified from a post-close source and because the industrial complex is where a hawkish Jackson Hole should show up first: Warsh said this summer's better PCE and CPI readings do not indicate underlying trends have meaningfully improved, September hike odds jumped to 57%, and the Russell 2000 fell 1.37% against the S&P's 0.25%. 3M and Honeywell leading the Dow lower on the same session fits that transmission rather than any company news. Ranked near the bottom precisely because the magnitude is small and the driver is inferred, not confirmed. unverified | 44 |
| 6 | HON | — | Third-worst Dow component on Aug 28 at -2.19%, no confirmed session catalyst - the same rate-transmission read as 3M and reported on the same basis. Ranked last on this list. It is here rather than omitted because the alternative was to leave the falling side at five names while a verified close-basis decline in a mega-cap industrial sat unused; it is ranked last because a small move with an inferred driver should not outrank a documented one. unverified | 43 |
B · One-Year Conviction
Asymmetric 12-month skew
NO MARKET-CAPITALIZATION VERIFICATION PASS WAS RUN. Every Track B market_cap_usd is null and flagged. All twenty positions are unambiguous large caps whose eligibility is not in question, but the pipeline's standing rule is to flag rather than estimate, and the reason for the omission is stated plainly: the research budget for this run went to disambiguating session magnitudes across a two-sided window and to reconstructing a 19-session Track B gap. Valuation figures in the theses are carried from the 2026-07-31 artifact and are now roughly four weeks stale. Tier-1 sources standing behind these theses: Micron FQ3-2026 via SEC EDGAR, NVIDIA IR, GE Vernova Q2-2026 IR, Alphabet Q2-2026 slides, FactSet Earnings Insight.
▲ Upside skew — 1yrconviction
| Ticker | Thesis | Key risk | Score |
|---|---|---|---|
| NVDA | The thesis was tested at its weakest point - out-year durability - and answered. FQ2 revenue of $96.22B more than DOUBLED year over year with adjusted EPS of $2.22 against $2.09, and guidance implied roughly 70% fiscal-2028 revenue growth where the Street had modelled 44%. That out-year figure is the whole argument: the bear case on Nvidia has never been that this quarter disappoints, it is that the capex cycle rolls over in 2027-28, and the company just guided directly at that window. Schwab's framing of the market's reaction is the honest one - Nvidia's guide soothed AI-durability fears, and investors now want the same from Broadcom. Moved to rank 1 on merit, not on drift. | The guide bought the SECTOR about one session: NVDA rose ~1% for the week while SMH fell more than 3%, and NVDA itself gave back 4.45% on Aug 28; 30-day implied volatility collapsed to 33 against a 52-week range of 32-55 immediately post-print - the options market is pricing almost no further surprise, which is a poor setup for a name whose thesis depends on repeated upside; Memory input-cost inflation persists and CXMT's H1 revenue surge (below) cuts both ways on the cost floor; Beta near 2.2 makes this the wrong vehicle if the capex thesis is wrong; Rate risk is now explicit: September hike odds moved to 57% on Aug 28 and long-duration growth is the direct casualty | 88 |
| TSM | The unavoidable chokepoint of the AI buildout and the only large-cap in the complex that converts the capex boom into cash rather than consuming it. Full-year revenue growth guided above 40% with capex lifted to $64B; pricing power sits simultaneously over Nvidia, AMD, Broadcom and Apple. Nvidia's 70% fiscal-2028 growth guide is a direct volume read-through to the foundry, and Marvell's newly disclosed Google custom-chip arrangement adds another advanced-node customer to the same queue - but neither is TSMC-specific news, which is why nothing moved this run. | Taiwan geopolitical tail risk is the entire bear case and is not diversifiable; Customer concentration in Nvidia; Rising capex intensity compresses free-cash-flow conversion; A 2027 hyperscaler capex cut hits with a two-to-three-quarter lag; No catalyst inside this horizon means no near-term falsification test | 87 |
| MSFT | Thesis confirmed on the tape in the prior window and therefore partly spent. FQ4 revenue $90.01B against $87.62B expected, Azure +43% constant currency against a 40.2% estimate, annual Azure revenue past $100B for the first time. The 12-month case no longer rests on a de-rating being unwound - that happened in the last week of July - so from here it rests on Azure holding a 40%-handle growth rate against a much larger base. This window supplied no new evidence on that question, and the honest note is that when leadership rotated, it rotated away from here. ↔ deep-dive | The cheap-entry argument is gone: this is a growth-durability call, not a re-rating call; Azure growth is materially OpenAI-linked, concentrating exposure to one counterparty's funding; Capex raised again with the last print; This window's winners were AI-exposed SOFTWARE, not the hyperscalers that own the buildout - if that rotation persists it is a relative-performance problem even if the thesis is right | 84 |
| AVGO | The custom-ASIC leg is the structural AI trade: every hyperscaler dollar spent reducing Nvidia dependence is by definition spent with Broadcom. Q1 FY2026 AI semiconductor revenue $8.4B, up 106% year over year on total revenue of $19.31B, with a $73B AI backlog deliverable over 18 months. This window added a genuinely useful comparable: Marvell disclosed a Google custom-chip arrangement worth up to $120B through fiscal 2033 and the market SOLD it 10.3% for being too far out. Broadcom reports inside this horizon and will be asked the same question about the same customers. | Customer concentration (Google, Meta, OpenAI) - and Google just publicly diversified toward Marvell; The '$100B AI revenue in 2027' figure is a CEO statement on a call, NOT formal guidance; The Marvell reaction is a live warning that this tape now discounts out-year custom-silicon revenue steeply; Options are pricing roughly a 7.6% post-earnings move with neutral positioning - no directional cushion; Earnings date itself is in conflict across sources (Sep 2 vs Sep 3) | 82 |
| GEV | Electricity, not silicon, is the binding constraint on AI. Q2 2026 (Tier 1, company release): orders $24.2B up 88% organic, backlog $176B, gas power equipment backlog from 100GW to 116GW, revenue $11.1B up 22%, free cash flow $5.1B in the quarter - more than all of 2025 - with FY26 FCF guidance raised from $6.5-7.5B to $11.5-12.5B. Still the strongest verified estimate-revision evidence on either list. | Highest headline multiple on this list at roughly 52x forward, and the long end just moved against long-duration assets again; Wind segment remains a drag; Turbine slot reservations are not firm revenue; A one-year New York State moratorium on 50MW+ data centers shows regulatory bottlenecks are real; Fell more than 7% on its own Jul 22 print on margin concerns - the market is not paying for backlog alone; No catalyst inside this horizon | 77 |
| GOOGL | Best full-stack AI position in the world - TPUs, models, distribution and cloud - and the only Mag-7 name whose revenue growth ACCELERATED on AI. Q2 2026: revenue $119.8B up 24%, operating income $40.8B up 30%, Google Cloud $24.8B up 82% with cloud operating margin expanding to 35.6% from 20.7%, and a $514B cloud backlog. This window's contribution is indirect but substantive: the size of the Marvell arrangement is a measurement of how large Google's internal accelerator programme has become, disclosed by the counterparty rather than by Alphabet. | 2026 capex guided to $195-205B, Q2 free cash flow -$5.9B (first negative quarter since the 2004 IPO), buybacks suspended; Street models FCF going from +$13B in 2026 to -$13B in 2027; EARNINGS QUALITY: Q2 diluted EPS of $9.11 was materially inflated by unrealized equity-securities gains and is deliberately excluded from the valuation sub-score; Antitrust remedies remain live; The Marvell deal is also a cost line for Alphabet, not only a validation - $120B of committed custom-silicon spend is capex the market is already refusing to fund elsewhere; GOOGL DIPPED on Aug 27 while security and SaaS rallied | 75 |
| MU | SPLIT CONVICTION, and the split WIDENED this window rather than narrowing. Bull, intact: FQ3 2026 (Tier 1, SEC-filed) revenue $41.46B against $9.30B a year earlier, GAAP diluted EPS $24.67, operating cash flow $25.39B, FQ4 guidance $50.0B plus or minus $1.0B at roughly 86% gross margin, and 16 supply agreements with 14 carrying minimum commitments of roughly $100B through 2030. Bear, materially strengthened: CXMT's first-half revenue grew more than 870% to roughly $22.36B, which is no longer a capacity-announcement story but a revenue-recognition one. An 870% revenue increase at the marginal supplier is the single most concrete piece of evidence yet that the supply response to the shortage is arriving, and it directly undercuts the Samsung-guidance upgrade applied last run. Michael Burry's disclosed short near $880 has not been retired. The honest read: the bull leg rests on contracted volume and the bear leg now rests on a competitor's audited revenue line. ↔ deep-dive | CXMT H1 revenue up more than 870% to ~$22.36B - the oversupply thesis now has a number, not just a forecast; Peak-cycle earnings at a trough multiple - the classic value trap in semis; An ~85-86% gross margin is roughly 25 points above any historically sustainable level; Customer inventories 7-9 weeks against an 8-week warning threshold and a 10-week exit trigger; Highest-variance name on either list - the 12-month distribution is bimodal, not merely uncertain; Fell 5.8% on Aug 24 and again on Aug 28; the July snapback is not holding cleanly | 75 |
| LLY | Clear structural winner of the GLP-1 duopoly. Lilly is compounding roughly 29% EPS while Novo has guided 2026 to a 5-13% constant-currency sales decline and failed the ZEUS cardiovascular outcomes trial that was its most visible attempt to build a second franchise. Oral orforglipron extends Lilly past injectable supply constraints. Carried as much for portfolio construction as for absolute upside - the only top-10 upside name with zero exposure to the hyperscaler capex question, which matters more this run than last given how concentrated the rest of the book is. | THE Q2 PRINT WAS NOT VERIFIED THIS RUN - it landed Aug 5, inside the inter-run gap, and no confirmed result was retrieved. This is the single largest known information hole in the Track B book; Consensus PT implied only ~11% upside as of the last verified read; US drug pricing / most-favored-nation policy risk; ~34x forward leaves no room for a manufacturing stumble; A rival's trial failure is a relative, not absolute, improvement | 75 |
| META | CONTRARIAN and unresolved. You are paying roughly 17.6x forward - the cheapest multiple in mega-cap tech - for a business that grew revenue 28% to $60.80B with ad impressions up 14% and price-per-ad up 12%. The core business is accelerating. But Q2 EPS of $6.18 missed $7.22, operating margin compressed to 31% from 43%, free cash flow fell to $784M from $8.55B, and 2026 capex is guided to $130-145B. The structural flaw the July window market-tested - capex with no third-party revenue offset - was not addressed in this window, and this window's lesson made it worse rather than better: the market spent two sessions paying enormous premiums for AI-exposed software companies that MONETIZE agents, and Meta is not one of them. ↔ deep-dive | THE STRUCTURAL FLAW, market-tested and unaddressed: unlike MSFT, GOOGL and AMZN, Meta has no external cloud revenue line to monetize the buildout; Reality Labs lost $4.03B in the quarter; Third consecutive run on watch_for_invalidation with no resolving evidence in either direction - a position that cannot be falsified is a position that should be sized down; No catalyst inside this horizon, so the flag will carry again; The cheap multiple has been cheap for three runs and has not been the reason anything happened | 66 |
| VRT | Pure play on data-center thermal management and power distribution - the physical bottleneck layer. Rack densities from Rubin-class systems make liquid cooling non-optional, converting a cyclical hardware business into a content-per-rack growth story. Nvidia's 70% fiscal-2028 revenue guide is the cleanest available demand input for that thesis. Ranked last because the evidence remains two-sided and because a roughly 29x forward multiple in a week when September hike odds moved to 57% is the wrong end of the duration spectrum. | Highest beta on this list to any hyperscaler capex cut; Competitive entry from Schneider and Eaton; The last observed options positioning was actively bearish (1:1.4 toward puts on 104K contracts) and has not been refreshed since Jul 30 - the absence of an update is not an improvement; The Q2 revenue miss that took the stock down 17.26% on Jul 29 is the execution slip a ~29x multiple cannot absorb repeatedly; Ranked last precisely because nothing has resolved | 64 |
▼ Downside skew — 1yrconviction
| Ticker | Thesis | Key risk | Score |
|---|---|---|---|
| NVO | The cleanest verified downward-revision case on either list, converted last run from an estimates thesis into a pipeline thesis. The company guided 2026 to a full-year constant-currency sales DECLINE of 5-13% against 10% growth in 2025 and cut US GLP-1 prices by up to 70%, defending volume at the cost of mix. On Jul 31 ziltivekimab missed its ZEUS primary endpoint outright with a hazard ratio of 0.99 (95% CI 0.88-1.11) across more than 6,300 patients - biology worked, outcomes did not - against a market Jefferies sizes above $10B annually. UBS modelled only 1.7% of NPV but expected a 2-3x reaction because the failure compounds a run of R&D setbacks. | THE Q2 PRINT WAS NOT VERIFIED THIS RUN - it landed Aug 5 and the result is unknown to this artifact. Ranked first on downside conviction while carrying an unread quarter is a real limitation, stated rather than hidden; A trailing P/E near 12 is genuinely cheap and limits downside velocity; HERMES and ARTEMIS both continue, reading out H1 2027 in distinct populations - a single failed trial is not a failed molecule; 2026 adjusted operating profit guidance was explicitly maintained; the Q3 impairment is non-cash; Roughly 25% came out over twelve months plus ~9% on the ZEUS news - a meaningful portion is priced | 77 |
| ORCL | The only name on either list where the risk is solvency-adjacent rather than valuation-driven, and the thesis has narrowed to exactly that. Oracle levered its balance sheet to fund a roughly $300B OpenAI contract and the credit market repriced it: CDS spreads hit a record 198.23bp and S&P downgraded to BBB-, one notch above junk. Total debt $167.43B against $43.06B of equity, TTM free cash flow -$23.69B, with roughly $40B of further financing planned for FY2027 and $90-95B directed into AI data-center expansion across Texas, New Mexico, Wisconsin and Michigan. What the gap established is that the DEMAND question is settled in Oracle's favour - OCI is growing 84% - so this is now purely a bet on financing cost in a regime where September hike odds moved to 57% on Aug 28. That is a narrower and more falsifiable thesis than it was, which is why it stays, and a more dangerous one to hold, which is why sentiment was cut hard. | THE SHORT IS LOSING: +16% during August, and the stock recovered from a monthly low near $114 to roughly $149; Citi's 90-day positive catalyst watch with a $330 target is an explicit, dated, contrarian bet against this position; OCI +84% to $4.89B settles the demand question against the thesis; A successful $40B raise at reasonable spreads collapses the CDS and the thesis with it; Consensus PT of $244.12 sits far above spot - the Street is emphatically not on this side; September earnings sit just OUTSIDE this horizon, so the position carries an unhedged event within weeks | 76 |
| AAPL | Thesis confirmed on the tape in the prior window. The FQ3 print delivered the predicted shape: a genuine beat (revenue $109.4B up 16%, EPS $2.02 up 29%) undermined by a forward guide of 9-11% revenue growth against 12% consensus, with Services at $30.74B and Greater China both short. Management attributed the guide to advanced-chip and MEMORY supply constraints, making Apple a price-taker to the same shortage enriching the memory complex. Barclays, Goldman and Morgan Stanley all cut targets within hours. This window added nothing, and the CEO transition to John Ternus on Sep 1 falls inside the catalyst horizon. ↔ deep-dive | $123.3B of TTM free cash flow and the largest buyback in history put a hard floor under it; Wells Fargo RAISED its target to $350 and BofA reiterated $380 on the same print - the sell-side is genuinely split; Supply constraints are a timing problem, not a demand problem; management said so explicitly; The CEO transition to John Ternus on Sep 1 is inside this horizon and is a plausible narrative reset with no fundamental content; Roughly $350-500B of value came out in one session on Jul 31 - much of the easy move is done | 76 |
| TSLA | $1.22 trillion of market capitalization resting on $3.81B of earnings. The underlying auto business is a roughly 14%-growth, 3.7%-net-margin manufacturer; the valuation is optionality on robotaxi and Optimus with no verifiable revenue line. This is the largest pure discount-rate exposure of any US mega-cap, and the discount rate moved against it again this window - decisively so on Aug 28, when a hawkish Warsh repriced the September meeting toward a hike. TSLA remained a top-three active option line on both sessions without producing a directional print, which is now a multi-run pattern: enormous positioning, no signal. | ESTIMATE-REVISION DIRECTION IS UNVERIFIED across several consecutive runs - the sub-score is neutral 50 rather than guessed, and the persistence of that hole is itself a weakness in this position; Persistent retail bid; TSLA is a top-three active option every session in this window; Any credible autonomy milestone re-rates it violently; Consensus PT of $398.30 sits well above spot - the Street is not on this side; Rates being high is a necessary but not sufficient condition; the market has ignored this before, including through two prior runs of the same argument | 73 |
| MRK | Merck rallied 57% INTO the Keytruda loss of exclusivity, and the deterioration is visible in aggregate estimates: FactSet attributed the Health Care sector's Q2 downward revision specifically to Merck and Gilead. The company announced $3B of annual cost cuts ahead of the cliff - a defensive signal. Paying roughly 21x forward for 5.6% EPS growth against a patent cliff is the worst growth-adjusted multiple among mega-cap pharma. | THE AUG 17 PDUFA OUTCOME IS UNKNOWN TO THIS ARTIFACT - subcutaneous Keytruda conversion could materially blunt the cliff and an approval would be direct evidence against this position. This is the largest unresolved information gap in the downside book; The last observed options skew was 2:1 toward CALLS into the Aug 4 print - the market was positioned for a beat; A 2.6% dividend limits downside velocity; Pipeline or M&A can re-rate it - the company is actively acquisitive; Consensus PT is still above spot | 72 |
| INTC | A 4.5x move in twelve months on a company still losing roughly $11B a year, driven by policy and partnership headlines rather than profits - $8.9B of CHIPS funding converted to a government equity stake, a $5B Nvidia investment, an Apple chip-making agreement. At roughly 54.8x forward it remains the most expensive large-cap semi on either list with the weakest competitive position and no trailing earnings. The prior run flagged the first genuinely uncomfortable datapoint - a Q2 beat-and-raise the market paid 12.53% for - and this window supplied neither confirmation nor refutation, only a 3.38% sympathy gain on someone else's earnings. A position on watch that gets a quiet window stays on watch. | THESIS UNDER PRESSURE since the Jul 30 beat-and-raise; a second rewarded beat should retire this rather than re-score it; A sovereign/strategic backstop makes a fundamental short genuinely dangerous - the US government is an equity holder; 18A/14A yield news is binary and headline-driven; Server CPU demand is genuinely strong, and Nvidia's 70% out-year guide raises the whole complex's volume outlook; Consensus PT sits well ABOVE spot; Second consecutive run on watch with no resolving evidence - the flag is doing no work until the next print | 71 |
| CRWV | The thesis has cleanly bifurcated and both halves are now measurable, which is the most useful thing that happened to this position. DEMAND LEG: dead. Q2 revenue up 112% to $2.58B, a $104B backlog, named business with Anthropic and Meta, and guidance raised - anyone still short CoreWeave on the belief that AI infrastructure demand evaporates is wrong and the numbers say so. CREDIT LEG: alive and demonstrated. Capex guidance was raised to $35-39B against roughly $35B of existing debt funding depreciating GPUs with -$10.6B of free cash flow, Altman Z-Score 0.36 and Piotroski F-Score 2. On Aug 18 the stock fell 12.10% on above-average volume specifically as the 30-year yield spiked, and it was down 13.52% over five sessions - a source described it as having become the sentiment barometer for rate-sensitive AI infrastructure. That is the position working for the right reason. Retained on the credit leg alone, with the demand leg formally abandoned. | THE DEMAND LEG IS DEAD AND IS NO LONGER PART OF THIS THESIS - stated rather than quietly dropped; Contracted revenue with investment-grade counterparties materially de-risks the near term; Anthropic and Meta as named Q2 customers make the counterparty-quality argument harder, not easier; Wall Street price targets span $74 to $250 - the dispersion means there is no consensus to fade; Borrow cost is likely punitive on a name this crowded; THIRD consecutive run on watch_for_invalidation. The standing condition was a financing event within the quarter while capex kept rising; capex rose and no financing event has occurred. If the next run produces no credit event, this should be retired rather than carried a fourth time | 71 |
| WDAY | The rarest outcome in this book: the thesis was CONFIRMED and the trade was IMPAIRED, in the same eight-hour window, and the two must be separated. The fundamental argument is that headcount-licensed HCM is the most literally AI-exposed pricing model in enterprise software, and Workday's own guidance now says so - Q3 cRPO of 11-12% against a 14% Street expectation and FY2028 subscription growth of roughly 11% against 14% in Q2 is a company telling you its billable-unit growth is decelerating, even while AI SKU ARR reached roughly $600M and AI products crossed $100M of new ACV. But on Aug 13 Reuters reported Silver Lake in talks to acquire the company at a market value around $43B; the stock rose nearly 18% for its best day in ten years and was halted for volatility. Management declined to address the talks on the Aug 27 call - the legally normal response when discussions are live - and simultaneously announced the new $4B buyback. A short position facing live take-private talks plus an open-ended buyback is uninvestable regardless of how right the fundamental thesis is. Retained because the analysis is correct and removing it would misrepresent the record; flagged for invalidation because the EXPRESSION is broken. | LIVE LBO RISK IS THE WHOLE PROBLEM: Reuters-reported Silver Lake talks at roughly $43B, with Needham flagging a possible $240-250 deal range based on recent HCM comparables. A confirmed bid ends this position at a loss regardless of fundamentals; A new $4B open-ended buyback puts a company bid under the stock; The stock is up roughly 48% in three months and has erased its year-to-date loss - the entry is far worse than when this position was opened; Q2 itself BEAT: non-GAAP EPS $2.75 against $2.61, subscription revenue $2.471B up 13.9%; Deutsche Bank and BTIG both DOWNGRADED after the buyout pop, which cuts toward the thesis - the sell-side split is genuine; WATCH_FOR_INVALIDATION: if Silver Lake talks are confirmed or a bid is tabled, retire immediately rather than re-score | 65 |
| AMD | AMD trades at a large premium to Nvidia's forward multiple while being the structurally weaker accelerator franchise - squeezed from above by Nvidia's ecosystem and from below by Broadcom, Google TPU and Meta MTIA custom silicon. This window sharpened the squeeze from both directions at once: Nvidia guided fiscal-2028 revenue growth to 70% against a 44% consensus, and Marvell disclosed a Google custom-chip arrangement worth up to $120B through fiscal 2033. Both of AMD's competitive flanks got quantitatively stronger in the same 48 hours. Remains a relative-value short against NVDA far more than an absolute one, and it is ranked last because the Aug 4 print was never read. | THE AUG 4 EARNINGS RESULT WAS NOT VERIFIED THIS RUN - it landed inside the inter-run gap and is unknown to this artifact; Genuine second-source demand from hyperscalers wanting a Nvidia alternative is real and structural; Server CPU share gains against Intel are real and margin-accretive; A 46.6% revenue growth forecast is not fantasy; Ranked last on the downside book precisely because the evidence base is thinnest here | 65 |
C · Options Heat
Where the premium is paying up
| Ticker | Signal | Read | Implication | Lag |
|---|---|---|---|---|
| PUT_SKEW_MASSACRE | pc_skew | Three pre-print books skewed to PUTS were run over by large up moves in the same 24 hours: CRM Aug-28 weekly IV 121 vs Sept 53 (52wk 25-61) at 1 call to 1.9 puts, realized +22.5%; OKTA weekly IV 209 vs Sept 80 (52wk 30-83) at 1 to 1.6 puts, realized ~+28%; SNPS weekly IV 130-140 vs Sept 63 (52wk 33-60) at 1 to 2 puts, realized ~+13%. CRWD was the lone large call-skewed book (1.7:1) and it was RIGHT at +20.5%. | The single most useful pattern in this artifact. Going into the Aug 26 AMC slate the free-tier options market was positioned bearishly on enterprise software and bullishly on security, and it had the sign backwards on three of four. The consistency matters more than any individual case: this was not one crowded trade being wrong, it was a sector-level positioning error. Treat published put skew in software as a contrarian input rather than a directional one until the pattern breaks. IMPORTANT LIMIT: a put:call VOLUME ratio is equally consistent with put buying and with put selling, and no free source tags side - so this is evidence about POSITIONING DISPERSION, not proof anyone was short. | All figures are Market Rebellion mid-session broker-desk snapshots taken Aug 25 and Aug 26, published BEFORE the events. Not exchange data. No post-print vol/OI table was obtainable from any free source: Barchart Unusual Options Activity and Market Chameleon render client-side and return page chrome only. |
| NVDA | iv_spike | Pre-print (Aug 26 mid-session): Aug-28 weekly call IV 93 vs September 43, against a 52-week range of 32 to 55; call:put 1.6:1, down from 2.8:1 on Aug 25. Post-print (Aug 28 mid-session): 30-day IV at 33 against that same 52-week range of 32-55. | A collapse from roughly 3x the annual band to within one point of the annual LOW in two sessions. The options market is now pricing almost no further Nvidia surprise, immediately after the largest positive surprise of its quarter. That is either the cheapest long-vol setup in mega-cap tech or an accurate read that the story is settled for a while - and note the call skew was FALLING into the print even as IV rose, so the crowd was reducing directional exposure while paying up for movement. NVDA heads the 'active options' line on both Aug 28 reports regardless. | Intraday broker-desk snapshots (11:00-ish ET), non-final. The 30-day IV figure is a mid-session Aug 28 read and may not equal the settled close. |
| MRVL | vol_oi | The Aug-28 expiration - the first after the Aug 27 print - carried 32,410 call contracts against 28,447 puts for a put/call ratio of 0.53, with the full chain at 0.56, as of Aug 24. Broadcom's full chain sat at 1.02 at the same moment. Realized was roughly -10.3%. | Bearish outcome against decisively bullish positioning, and the only hard contract-count datum in the artifact. The contrast with Broadcom's 1.02 full-chain ratio is the useful part: at the same moment, the options market was leaning long Marvell and neutral-to-short Broadcom into the same custom-silicon narrative. Marvell's reaction is now the base case Broadcom's own print will be measured against inside this horizon. | Contract counts come from a Tier-2 editorial note dated Aug 24, four sessions before the event, not from an exchange feed. No post-print vol/OI obtainable. MRVL tops the 'popular stocks' line on both Aug 28 reports with no figures attached. |
| ADSK | pc_skew | Skew reversed inside 24 hours: Aug 26 mid-session showed Aug-28 weekly call IV 136 vs September 64 (52wk 22-60) at 1 call to 3.1 PUTS focused on September-4 weekly puts; Aug 27 mid-session showed weekly IV 195 vs September 65 at 1.4 CALLS to 1 put. Realized approximately -4%. | Weekly implied vol at roughly 3.25x the top of the annual band, with the skew flipping from heavily bearish to modestly bullish in the final session before the print. The puts were right and got sold into the event. A single-session skew reversal of that size immediately pre-event is the clearest 'positioning, not information' signature in this window. | Two intraday broker-desk snapshots, both pre-event and non-final. No side tagging. |
| GAP | iv_spike | Aug-28 weekly call IV at 260 against September 72, versus a 52-week range of 35 to 73 - the most extreme term structure and the highest absolute weekly IV observed in this window - with call:put 1.7:1. Realized roughly +13-14%. GAP then appears on FOUR separate Aug 28 lists: increasing unusual call volume, increasing unusual put volume, popular stocks, and active options. | The call skew was directionally right and the implied move was roughly the right size. Recorded here specifically because GAP was EXCLUDED from the ranked lists on unverified market-cap eligibility - the exclusion was a universe decision, not a signal decision, and this entry exists so that is visible rather than silently dropped. | Market Rebellion mid-session Aug 27, pre-event. The Aug 28 list memberships are unranked with no volumes attached - useful for crowding, useless for direction. |
| OKTA | iv_spike | Aug-28 weekly call IV 209 against September 80, versus a 52-week range of 30 to 83 - roughly 2.5x the top of the annual band - rising from 175 the prior session. OKTA then appears on the DECREASING-IV list on Aug 27 pre-market, Aug 27 mid-session, Aug 28 pre-market and Aug 28 mid-session. | The largest event premium of the slate, followed by the most complete vol crush - four consecutive decreasing-IV list appearances. Anyone long that weekly option was right on direction and still had to beat a 209 vol print. The IV-crush risk that this pipeline routinely flags was maximal here and the underlying move was large enough to overcome it, which is rare and worth recording as a base rate. | Pre-event broker-desk snapshot Aug 26; the decreasing-IV lists are unranked membership with no levels attached. |
| PYPL | pc_skew | Aug 20 pre-market: 30-day IV at 30 against a 52-week range of 25 to 56 - near the annual LOW - with call:put at 4.9 calls to 1 put focused on August 60 calls. Eight sessions later the Advent/Stripe pursuit collapsed and the stock fell roughly 12.5%. | The most instructive stale datum in the artifact. A 4.9:1 call ratio at bottom-of-range implied volatility is what deal-anchored positioning looks like: the market was long upside optionality and was NOT paying for the risk that the deal died. That is a general lesson about takeover-premium names - the vol surface stops pricing binary risk once a bid is assumed - and it is recorded despite being eight sessions stale precisely because the staleness is the point. | Market Rebellion pre-market Aug 20, reflecting the Aug 19 close. EIGHT SESSIONS STALE relative to RUN_DATE and explicitly NOT used to score the PYPL entry. PYPL sits in the Aug 28 'popular stocks' and 'active options' lines with no figures published. |
| IREN | iv_spike | Aug-28 weekly call IV at 227 against September 100 into the Aug 27 AMC print. Realized approximately -13.35%. IREN sits in the 'active options' line on both Aug 28 reports and in 'popular stocks' on the mid-session report. | Second-highest weekly implied vol in the window behind GAP, in a name that was excluded from the ranked lists on universe eligibility. Recorded so the exclusion is visible: IREN produced one of the two largest declines of the session and it is absent from the falling list for a market-cap reason, not for lack of evidence. | Market Rebellion mid-session Aug 27, pre-event. The realized figure is from a Yahoo trending-ticker widget - a documented caching-risk source in this pipeline - and is used here only as an order-of-magnitude read, never as a ranked figure. |
| AVGO | iv_spike | Options are pricing an approximate 7.6% post-earnings move with positioning characterised as neutral; gamma positioning is described as showing strong support at the $350 strike and strong resistance at $400. Separately, the AVGO full-chain put/call ratio was 1.02 on Aug 24 against Marvell's 0.56. | The largest scheduled single-name event inside the horizon, and the market is not leaning. Neutral positioning into a print whose direct comparable (Marvell) just fell 10.3% on a beat-and-raise means there is no crowded side to squeeze in either direction - the move should be information-driven rather than positioning-driven. Directly relevant to Track B upside rank 4. | CMC Markets week-ahead commentary published Aug 28 - a broker characterisation of positioning, not an exchange measurement, and no IV level or contract count is given. The full-chain ratio is from a Tier-2 editorial note dated Aug 24. Earnings DATE itself is in conflict across sources (Sep 2 vs Sep 3). |
| CROWDING_WATCH | vol_oi | Names on Market Rebellion's 'popular stocks with increasing option volume' AND 'active options' lines across all three reports in this window (Aug 27 pre-market, Aug 28 pre-market, Aug 28 mid-session): INTC, MU, MSTR, PLTR, SOFI, AMD, SPCX. Aug 28 adds MRVL, IREN, CRM, PYPL, PCG, MARA, GAP. NVDA, TSLA and AAPL head the 'active options' line on every report. | The persistent crowding is in memory (MU), Intel, and high-variance retail vehicles (MSTR, SOFI, MARA, SPCX, PCG) - and NOT in the software and security names that actually produced the window's 20-28% moves. That divergence has now held across two consecutive runs: the retail options bid and the institutional equity bid are in different places. Note PLTR remains on every crowding list while being retired from Track B this run - crowding is not a thesis. | Unranked membership lists with no volumes, IV levels or skew ratios attached. Useful for crowding, useless for direction. No CBOE index-level put/call ratio was obtained for either session - the official daily statistics page has returned navigation-only content across multiple runs and was not retried. |
D · Catalyst Calendar
Upcoming events
| Date | Event | Tickers | IV move |
|---|---|---|---|
| 2026-08-31 | Final trading day of August - month-end rebalancing flows, which a Tier-3 source expected to add volatility to the week's biggest movers. NOTE: one Tier-3 source incorrectly described Aug 28 as August's final session and a live aggregator incorrectly referenced a Labor Day weekend; Labor Day 2026 is Sep 7 and Aug 31 is a normal full session. No major scheduled US macro release or large-cap earnings identified for this date - treat as 'none scheduled', LOW confidence. | macro | unknown |
| 2026-09-01 | Palo Alto Networks fiscal Q4 earnings. The first test of whether the agent-security re-rate survives a second print - PANW rose 12.8% on Aug 27 purely on CrowdStrike's and Okta's numbers, with no results of its own. Track A rising rank 6 and the lead name in the upside runners-up cohort. | PANW | unknown |
| 2026-09-01 | ISM Manufacturing PMI (August) and JOLTS job openings (July, consensus 7.273M from 7.359M). The JOLTS ratio of openings to unemployed workers is the specific number to watch given three prior FOMC dissents favouring a hike and September hike odds now at 57%. Also: Apple CEO transition to John Ternus takes effect (relevant to Track B downside rank 3). | macro | unknown |
| 2026-09-02 | Broadcom Q3 FY2026 earnings - DATE IN CONFLICT, carried on both Sep 2 and Sep 3. CMC Markets' week-ahead places the preview under Wednesday Sep 2; this pipeline's 2026-07-31 artifact recorded Sep 3. Consensus looks for roughly 91.5% EPS growth to $3.24 on revenue up 84.1% to $29.4B, with semiconductor solutions more than doubling from $9.2B to $18.5B and a Q4 guide of $3.88 on $35.03B. Options price an approximate 7.6% move with neutral positioning. The single most consequential scheduled event in the horizon. | AVGO | high |
| 2026-09-02 | HPE Q3, NetApp Q1, Snowflake Q2 earnings. CMC's week-ahead lists these under Sep 1 while its own day headings are internally inconsistent in the retrieved text; the cluster is placed here and the ambiguity is disclosed rather than resolved. Snowflake is the cleanest read-through to the enterprise-software re-rate that defined this window. | HPENTAPSNOW | unknown |
| 2026-09-03 | Broadcom Q3 FY2026 earnings - ALTERNATE DATE, carried per the 2026-07-31 artifact. Only one of Sep 2 and Sep 3 is correct; both are listed so the horizon covers the event under either. | AVGO | high |
| 2026-09-03 | ADP National Employment Report (August) and Challenger job cuts, plus ISM Services PMI (August). The ADP print is the leading tell for the Sep 4 payrolls report. In a 57%-hike-probability regime a hot services print is the more dangerous of the two. | macro | unknown |
| 2026-09-04 | BLS Employment Situation - August nonfarm payrolls and unemployment rate, 8:30 ET. The dominant macro event of the horizon by a wide margin. With Warsh explicitly stating that price stability is the Fed's predominant focus and market-implied September hike odds at 57%, this print plus the Sep 3 ISM Services are what actually set the September meeting. Index volatility should be pinned here, not to any earlier date. Also: PYPL dividend ex-date ($0.14). | macro | unknown |
E · Trade Ideas
What the data implies
AVGOneutral · track A/B
Rationale The highest-information event in the horizon, and deliberately framed as a volatility and read-through view rather than a direction. The setup is unusually clean because its comparable just reported: Marvell beat, RAISED FY27/FY28 growth guidance to 45%/50% from 40%/45%, disclosed a Google custom-chip arrangement worth up to $120B through fiscal 2033 - and fell 10.3%, explicitly on questions about how quickly that revenue arrives. Broadcom faces the identical question about the identical customers, with consensus looking for roughly 91.5% EPS growth on revenue up 84.1% to $29.4B. Options price an approximate 7.6% move with positioning described as neutral, and Broadcom's full chain sat at a 1.02 put/call ratio on Aug 24 against Marvell's 0.53-0.56 - so unlike Marvell there is no crowded long to unwind. The read-through matters beyond the name: if Broadcom is also sold on out-year timing, the custom-silicon de-rating is a cohort event, which is the registered promotion condition for MRVL on the downside book and a direct negative for Track B upside rank 4 and GOOGL's structural leg.
Invalidation The volatility view has no invalidation. For the underlying Track B long: a beat that is sold on custom-silicon timing, following Marvell's template, would be the second instance of the same rejection and should cut the structural sub-score materially rather than be treated as noise. Gamma positioning is described as strong support at the $350 strike and strong resistance at $400 - useful as a map, not as a thesis.
Risk Conflicting earnings dates are a genuine execution hazard here, not a footnote. The market-cap and valuation figures behind the Track B thesis are roughly four weeks stale and were not refreshed this run.
PLTRneutral · track B
Rationale Recorded as a POSITION-CLOSURE note rather than a trade, because a book that only documents its winners is not a book. This pipeline carried Palantir as a downside conviction for months on a multiple-compression argument - roughly 56x sales - and on 2026-07-31 it registered the exact condition that would falsify the position: an Aug 3 beat-and-raise in a tape that had started paying for beats. That is precisely what happened. Revenue grew roughly 93%, full-year guidance went to $7.65-7.66B, US government revenue hit $687M with federal obligations above $1B for the first time, and the stock rose 43.69% in the following month. The prior run also flagged that pre-print options were 2.4:1 call-skewed at weekly implied vol above the 52-week range - the positioning warning was correct and the position was held anyway. Two lessons are worth keeping: a valuation short with no catalyst is a bet on other people's patience, and the pre-registered invalidation condition is the single most valuable artifact this pipeline produces, because without it this would now be a re-scoring exercise instead of a closure.
Invalidation n/a. Explicitly NOT eligible for reinstatement on a pullback. Any future Palantir position requires a new thesis with a new falsification test, not a lower entry on the old one.
Risk The multiple-compression argument may still prove correct on a multi-year horizon. Being early and being wrong are indistinguishable from inside a losing position, which is the entire reason the falsification condition was written down in advance.
MRVLneutral · track A
Rationale The most instructive single name of the window and deliberately UNRANKED on Track B, so the reasoning is recorded here instead. Marvell beat on both lines, raised FY27 and FY28 revenue growth guidance to 45% and 50% from 40% and 45%, reported data-center revenue up 46% to $2.17B, and announced a Google custom-chip arrangement that could generate up to $120B of revenue through fiscal 2033. It fell 10.3%. The proximate reasons were a Q3 gross-margin guide of 57.5-58.5% against a 58.5% consensus and, per Reuters commentary, disappointment that the Google deal contributes nothing near-term. The structural reading is more useful: forty-eight hours earlier the same market paid Nvidia 8.7% for a fiscal-2028 growth number. The discriminator is not magnitude, it is credibility of timing - Nvidia has an eight-quarter record of converting guidance into revenue and Marvell is asking for trust on a 2033 horizon at roughly 58x forward after a 179% year-to-date run. In a 57%-hike-odds regime, distant cash flows discount hard.
Invalidation For the bearish read: any disclosure that quantifies near-term Google revenue contribution removes the specific objection that caused this selloff. For the bullish read: a second custom-silicon name being sold on the same timing objection - Broadcom inside this horizon - converts this from a Marvell problem into a cohort problem. That second case is the registered promotion condition for adding MRVL to the Track B downside book next run.
Risk The fundamental business is not the issue - 37% revenue growth and a raised guide are real. This is entirely a duration-and-multiple argument, which makes it a rates trade wearing a semiconductor costume.
Rationale The book's residual short exposure has quietly become a RATES position rather than a capex position, and this note makes that explicit so it can be sized deliberately. Both theses had two legs and in both cases the demand leg is now dead: CoreWeave reported Q2 revenue up 112% to $2.58B with a $104B backlog and named business with Anthropic and Meta, and Oracle's OCI revenue grew 84% to $4.89B with OpenAI, Meta and NVIDIA signing GPU-compute contracts. Anyone still short either name on the belief that AI infrastructure demand evaporates is simply wrong. What survives is the funding leg, and it is demonstrably live: on Aug 18 CoreWeave fell 12.10% specifically as the 30-year yield reached 5.32%, and one source described it as having become the sentiment barometer for rate-sensitive AI infrastructure. Oracle carries $167.43B of debt against $43.06B of equity with -$23.69B of TTM free cash flow and roughly $40B of further financing planned. On Aug 28 September hike odds moved to 57%. That is the mechanism, working.
Invalidation CRWV: a completed financing at reasonable spreads, or a quarter in which rising yields do NOT produce disproportionate downside. This is the position's THIRD consecutive run on watch_for_invalidation and the standing condition - a credit event within the quarter - has not been met; if the next run produces nothing, retire rather than carry a fourth time. ORCL: a successful $40B raise at reasonable spreads collapses the CDS and the thesis together. Citi's 90-day positive catalyst watch with a $330 target is a dated, explicit contrarian bet and should be treated as a real counterparty view, not noise.
Risk Both positions are now bets on the same macro variable, which means they are one position, not two, and should be sized as one. Consensus price targets sit far above spot on both. If September delivers a hold rather than a hike, both legs lose simultaneously.
Rationale Framed as a WATCHLIST note rather than a trade because the honest position is that this pipeline does not yet know whether the Aug 27 software re-rate was a repricing or a squeeze. What happened is not in doubt: Salesforce +22.5% at mega-cap scale, Okta +28%, CrowdStrike +20.5% on its best day ever, Palo Alto +12.8% with no earnings of its own, and the iShares Expanded Tech-Software ETF +7.6% - on a session when ten of eleven S&P sectors closed NEGATIVE. The narrative underneath is genuinely new and genuinely durable-sounding: the agent era creates identities and attack surface that must be secured and licensed, which reframes enterprise software from AI victim to AI beneficiary. Okta management quantified it at roughly 90 machine identities per human worker. But three things argue for patience. The pre-print options books were skewed to PUTS on CRM, OKTA and SNPS and were run over, so a meaningful part of the move is mechanical. Both CRWD and OKTA had already roughly doubled in 2026 before this. And CRWD gave back 4-7% the very next session on a competitor's guidance cut, which is not what a durable re-rate looks like.
Invalidation The cohort thesis gets its first real test on Sep 1 when Palo Alto reports, having risen 12.8% on nobody's numbers. A beat that holds confirms a sector re-rate; a beat that is sold - the Marvell, Rubrik and Autodesk pattern from the same window - says the market paid for a narrative and is already taking it back. That single print is worth more than any amount of further analysis here.
Risk This cohort is the strongest structural story surfaced this run and it is in NONE of the Track B book. That is a real gap, and the registered promotion conditions in upside_runners_up exist so it can be closed on evidence rather than on enthusiasm.
Methodology & limitations
- THE DEFINING PROBLEM OF THIS RUN: week-to-date figures presented in session-recap prose. CNBC's Aug 28 wrap gives Marvell down more than 8%, Applied Materials down more than 5%, Synopsys up more than 11% and Cadence up more than 6% - all WEEK-TO-DATE - inside a paragraph structured like a session recap. MRVL's actual Aug 28 close-basis move was roughly -10.3%, materially larger than the weekly figure quoted in the same article. Any figure appearing near the phrase 'this week' was treated as weekly and excluded from session scoring; AMAT and CDNS were held out of the ranked lists entirely for this reason. This generalises the premarket-block failure mode documented on 2026-07-31 into a broader rule: establish the BASIS of every percentage before using it.
- A LIVE AGGREGATOR PAGE WAS DISCARDED WHOLESALE. stockmarketwatch.com's live 'stock market today' page, crawled 25 minutes before this build, described the Aug 28 session as showing 'mixed but largely positive results' with the Nasdaq-tracking QQQ up 0.06% - against confirmed closes of S&P -0.25% and Nasdaq -0.52% - and referenced a Labor Day holiday weekend that is nine days away. Its percentage labels also contradict its own quoted prices: it showed MRVL at $216.79 labelled -6.8%, when $216.79 against the confirmed Aug 27 close of $241.93 is -10.4%. The PRICES appear sound and the LABELS do not. Recorded as a new documented failure mode.
- NVDA's AUG 28 MOVE IS SOURCE-CONFLICTED and the conflict is large: Trading Economics' post-close Dow list gives -4.45%, a Yahoo article sidebar gives -4.85%, and the discarded live aggregator gave -0.2% at midday. The two post-close sources agree within 40 basis points and CNBC's independent statement that NVDA rose only about 1% for the week while gaining 8.7% on Thursday corroborates a materially negative Friday. The -4.45% figure is used.
- NO MARKET-CAPITALIZATION VERIFICATION PASS WAS RUN. Only two ranked entries carry a cap figure and both are flagged. This is the largest methodological shortfall of this run and it is stated rather than hidden: the research budget went to magnitude disambiguation across a two-sided window and to reconstructing a 19-session Track B gap. The consequence was handled conservatively - five names with genuinely borderline eligibility (ESTC, GAP, IREN, S, ULTA) were HELD OUT rather than ranked on assumption, which cost the falling list four entries.
- A 19-TRADING-SESSION GAP SEPARATES THIS RUN FROM THE PRIOR ONE and Track B carries real information holes as a result. Specifically unverified: the LLY and NVO Q2 prints (Aug 5), the AMD Q2 print (Aug 4), and the MRK KEYTRUDA QLEX PDUFA decision (Aug 17). Each is named in the affected entry's key_risks and flagged. Sub-scores were NOT moved on unread events - four upside positions and three downside positions carry explicit 'unchanged, and here is why not' deltas rather than drift.
- NO LIVE OPTIONS TAPE. Every options figure is a Market Rebellion broker-desk snapshot, published either pre-market (reflecting the prior close) or mid-session (intraday, non-final). Barchart Unusual Options Activity and Market Chameleon render client-side and return page chrome only. No post-print vol/OI table was retrievable for any name in this window.
- NO BID/ASK SIDE TAGGING on any free source. Every call:put figure here is a VOLUME ratio. This window supplied the most emphatic demonstration yet of why that matters: three put-skewed books (CRM 1:1.9, OKTA 1:1.6, SNPS 1:2) preceded up moves of 22.5%, 28% and 13%, while one call-skewed book (MRVL 0.53 P/C) preceded a 10.3% decline. Those outcomes are consistent with the crowd being wrong AND with systematic premium selling; the data cannot distinguish them.
- NO CBOE PUT/CALL RATIO obtained for either session; the official daily statistics page has returned navigation-only content across multiple runs. No index-level hedging read is claimed.
- UNRESOLVED SOURCE CONFLICTS, published as flags rather than point-estimated: CRM's Aug 27 move (+22.5% close vs +22.6% vs 'more than 22%' vs an +8.71% intraday card); OKTA (+28% CNBC vs 29-30% 24/7); NVDA Aug 27 (+8.7% vs +8.5% vs +8.4% across three sources) and Aug 28 (above); MRVL Aug 28 (-10.3% vs -10.28% vs -10.47%, with intraday prints from -6.75% to -8%); PYPL (-12.05% to -16% across six sources); CRWD Aug 28 (-4.72% open vs -4% vs -7% within a single article); AFRM (+5.08% to +12%); ESTC (+17% to +24.8%); SNPS (13% session vs 11% weekly); FTNT (5.7% vs 'double digits' from the same publisher); Russell 2000 (2,973.09 post-close vs 2,972.37 and 2,977.04 from intraday widgets); the 30-year yield's historical framing (highest since early 2024 per one August source vs highest since 2007 per the prior run - not reconciled).
- VIX AND RUSSELL 2000 ARE SEMI-VERIFIED. Both come from a Tier-3 weekly summary that reports them as post-close figures. VIX 14.35 reconciles arithmetically against the Zacks-confirmed Aug 27 close of 14.51 (a -1.10% session change), and the Russell's 2,973.09 at -1.37% implies an Aug 27 close of 3,014.34, which matches a Yahoo sidebar's 2,972.37 less 41.97 exactly. Both are accepted on that arithmetic, not on source tier. NOTE the same Tier-3 source reports VIX as down 5.2% from 15.13 - that is a WEEK-over-week change against the prior Friday, not a session change, and conflating them would have produced a wrong number.
- THE 30-YEAR TREASURY CLOSE FOR AUG 28 WAS NOT OBTAINED. The 5.32% figure in market_regime.levels is dated Aug 18 and explicitly labelled. Given that the entire residual short book is now a rates position, this is a meaningful gap.
- GOLD AND BITCOIN LEVELS ARE INTRADAY RANGES, not settled closes, and are published as ranges accordingly.
- THE CXMT DATAPOINT - first-half revenue up more than 870% to roughly $22.36B - is the most consequential single fact in this artifact for the memory complex, and its only source is a Tier-3 morning-note aggregator. It was applied to MU's structural sub-score anyway, because the alternative was to leave a documented 8-point downgrade unmade on sourcing grounds while last run's 10-point UPGRADE rested on comparable evidence. The asymmetry is flagged so a reader can discount it.
- BLOCKED OR UNAVAILABLE SOURCES: Bloomberg and CNBC return 403 to automated fetch (their content here comes via search index); Seeking Alpha is robots-disallowed and only its headline summary was accessible; CappNotes Closing Look was not indexed for either session in this window, which removed the pipeline's most reliable single close-basis recap source; the Market Rebellion Aug 28 mid-session report was reached via search rather than direct fetch after a permissions error.
- SCORES ARE HEURISTIC AND NOT BACKTESTED. Sub-scores are analyst judgment applied to cited evidence, not a fitted model. Track A options sub-scores rest on pre-event implied-volatility snapshots and unranked activity lists, which measure expectation and crowding - not realised institutional flow.
Not financial advice. Educational/analytical only — generated by Daily Market Pulse template v1 - multi-desk research run (market regime and rates, Track A movers across two sessions, options positioning and crowding, forward catalysts, Track B conviction review with a 19-session gap reconstruction), synthesized 2026-08-29. Do your own diligence.