tkal · The Record · every dated call, graded

Posted	Claim · horizon · falsifier	Confidence	Status
Oct 19, 2026	The book stays defensive and its one buy, 2027 HBM from Micron, is now a single-supplier call, carried provisionally; the October 22 Micron Taiwan bonus deadline is the first test (Issue №022). Horizon by October 22, 2026 (Micron Taiwan bonus deadline) · Falsifier A strike notice, work stoppage or disclosed output loss at Micron Taiwan fires the reversal.	LOW <40%	OPEN
Oct 16, 2026	YMTC's Q1 warehouse was held for price, not unsold: in the next financial update in its STAR review, revenue per quarter after Q1 2026 averages above Q1's RMB 47.04B (YMTC pre-IPO analysis). Horizon at YMTC's next prospectus financial update (undated; estimated by December 31, 2026) · Falsifier Quarters after Q1 2026 average below RMB 47.04B of revenue in the next update (the inventory built in the best quarter was product nobody took).	MEDIUM 40-70%	OPEN
Oct 15, 2026	TSMC's pricing keeps outrunning dilution: 4Q26 revenue per wafer shipped stays above US$10,438 (TSMC Q3 2026 deep dive). Horizon by January 14, 2027 · Falsifier 4Q26 revenue per wafer shipped falls below US$10,438.	MEDIUM 40-70%	OPEN
Oct 15, 2026	The memory staircase rises in October: the five Taiwan memory makers file October 2026 revenue at least 10% above September's NT$90.36B, i.e. NT$99.40B or more (Taiwan monthly revenue, September 2026 edition). Horizon by November 10, 2026 · Falsifier The five memory makers file October revenue less than 10% above September's NT$90.36B (below NT$99.40B).	MEDIUM 40-70%	OPEN
Oct 14, 2026	ASML's customer advances hold: non-current contract liabilities on the Q4 2026 balance sheet stay above EUR 5.63B, the end-2024 high (ASML Q3 2026 deep dive). Horizon by January 27, 2027 · Falsifier Non-current contract liabilities on the Q4 2026 balance sheet fall below EUR 5.63B.	MEDIUM 40-70%	OPEN
Oct 13, 2026	Applied Materials' missing WFE outlook was about fab supply, not demand: Q4 FY26 revenue prints at or above $10.25B with non-GAAP gross margin at or above 50.4%, and the fiscal 2027 guide steps up from there (Applied Materials SEMICON West investor briefing). Horizon by November 12, 2026 · Falsifier Q4 FY26 revenue below $9.75B (bottom of the guided range), or the first fiscal 2027 revenue guide below the $10.25B Q4 midpoint.	MEDIUM 40-70%	OPEN
Oct 12, 2026	The book is defensive and its only buy is 2027 HBM4 allocation, carried as provisional and low conviction until the 2027 HBM4 settlements due around October 31 confirm it or fire the reversal (Issue №021). Horizon by October 31, 2026 (2027 HBM4 settlements) · Falsifier The 2027 HBM4 settlements from SK hynix, Samsung and Micron come in well below asking, or 8-high becomes the main Rubin stack.	LOW <40%	OPEN
Oct 12, 2026	Nanya's contract price holds up: October-December 2026 monthly revenue averages above NT$44.55B (Nanya Q3 2026 deep dive). Horizon by January 10, 2027 · Falsifier October-December monthly revenue averages below NT$44.55B.	MEDIUM 40-70%	OPEN
Oct 6, 2026	CrowdStrike's fourth quarter did not get easier: Q3 FY27 net new ARR lands inside $343M-347M and the fiscal 2027 ending-ARR range is held, leaving a roughly $421M fourth quarter in place (CrowdStrike Fal.Con briefing). Horizon by early December 2026 (Q3 FY27 print) · Falsifier The Q3 FY27 print raises the fiscal 2027 ending-ARR range above $6,611.9M, meaning the fourth quarter got easier rather than harder.	MEDIUM 40-70%	OPEN
Oct 6, 2026	Sandisk's contract model holds at the first print after Investor Day: Q1 FY27 non-GAAP gross margin at or above 83% with the eight-customer agreement count holding or growing (Sandisk Investor Day). Horizon by October 29, 2026 · Falsifier Q1 FY27 non-GAAP gross margin below 80%, or fiscal 2027 bit coverage under the agreements restated below 50%.	MEDIUM 40-70%	OPEN
Oct 5, 2026	The durable lane is 2027 HBM, repriced by Micron and reportedly by Samsung, while consumer DRAM and channel NAND roll over; Samsung's 2027 HBM4 settlement, due by end-October, confirms or breaks it (Issue №020). Horizon by October 31, 2026 (Samsung 2027 HBM4 settlement) · Falsifier Samsung's 2027 HBM4 settlement comes in low against the reported quote of more than three times HBM3E; a settlement near the quote keeps HBM the cleanest lane.	MEDIUM 40-70%	OPEN
Oct 1, 2026	WT Microelectronics' supplier float holds through Q3: net working capital (receivables plus inventory minus payables) on the September 30, 2026 balance sheet at or below the March high of NT$127.97B (WT Microelectronics Q2 2026 deep dive). Horizon by November 14, 2026 · Falsifier Net working capital above NT$127.97B on September 30 (the float has started to turn and growth is back on borrowed money).	MEDIUM 40-70%	OPEN
Sep 30, 2026	Jabil's AI ramp is still being financed through net-booked components: prepaid expenses and other current assets on the November 30, 2026 balance sheet stay above $4,559M (Jabil Q4 FY26 deep dive). Horizon by mid-December 2026 (Q1 FY27 release) · Falsifier Prepaid expenses and other current assets fall below May's $3,925M at November 30 (the FY26 build was a timing lump that shipped); between $3,925M and $4,559M is partial.	MEDIUM 40-70%	OPEN
Sep 30, 2026	Micron's customers keep paying into the build: noncurrent customer contract liabilities on the Q1 FY27 balance sheet rise above $18.0B (Micron Q4 FY26 deep dive). Horizon by December 16, 2026 · Falsifier Noncurrent customer contract liabilities no higher than $12,895M at the Q1 FY27 balance sheet (no new money, Q4 was a signing lump); in between is partial.	MEDIUM 40-70%	OPEN
Sep 30, 2026	Phison's holding period catches up with price: Q3 2026 gross margin prints at or below 65.3% (Phison Q2 FY26 deep dive). Horizon by early November 2026 · Falsifier Q3 gross margin above 65.3% (price still outrunning the cost layer).	MEDIUM 40-70%	OPEN
Sep 30, 2026	WPG's third quarter lands in the bottom half of its own NT$460B-500B guide, because June was a pull-in: September 2026 revenue comes in below NT$191.67B (WPG Holdings Q2 2026 deep dive). Horizon by October 10, 2026 · Falsifier September revenue at or above NT$191.67B, the June level held without a pull-in to explain it. Resolved Oct 10, 2026. Wrong. The call needed WPG's September revenue below NT$191.67B, the bar its Q3 guide midpoint required after a flat July (NT$146.10B) and August (NT$142.24B), on the view that June had been a pull-in. September came in at NT$219.59B, up 54.4% on August, on the TWSE/MOPS monthly filing, which cites technology and energy facility expansion rather than any pull-in. That puts the three-month sum near NT$507.92B, above the top of the NT$460B to NT$500B guide, not in its bottom half. The quarter-end ratio the call leaned on (1.15 times the first two months) was blown through at about 1.52 times. Graded 2026-10-20, 10 days after its own horizon.	MEDIUM 40-70%	MISS
Sep 29, 2026	CDW's June working-capital build is the new price of the hardware mix, not a lump: the three-month cash conversion cycle for the quarter ending September 30, 2026 is 16 days or more (CDW Q2 2026 deep dive). Horizon by early November 2026 · Falsifier Three-month cash conversion cycle at 12 days or fewer for the September quarter (the June build converted); 13-15 days is partial.	MEDIUM 40-70%	OPEN
Sep 29, 2026	Insight's fee layer, not the server cycle, drives its earnings: Q3 cloud gross profit at or above $150M with services gross profit growing faster than product gross profit (Insight Enterprises Q2 2026 deep dive). Horizon by November 5, 2026 · Falsifier Product gross profit outgrows services gross profit in Q3 (hardware has become the engine); cloud below $150M with services still ahead is partial.	MEDIUM 40-70%	OPEN
Sep 29, 2026	Seagate's June volume step was the outlier and the September quarter is met mostly on price: HDD exabytes shipped for the quarter ending October 2, 2026 come in below 228 (Seagate Q4 FY26 deep dive). Horizon by October 27, 2026 · Falsifier Fiscal Q1 exabytes reach 237 or more (HAMR capacity arriving on schedule, volume still carrying more than half the growth); 228-237 is partial.	MEDIUM 40-70%	OPEN
Sep 28, 2026	Enterprise and data-center SSD demand survives Micron's fiscal fourth-quarter print: the pipeline's only buy, carried as provisional because a single research house supported it (Issue №019). Horizon by September 30, 2026 (Micron FQ4 FY26 print) · Falsifier Micron's FQ4 print shows data-center NAND softening or customer inventory builds. Resolved Sep 30, 2026. Micron Q4 FY26 (14 weeks ended Sep 3, 2026; reported Sep 30, 2026): NAND revenue $14.1B, up 42% sequentially on bit shipments up about 10% and prices up about 30%. Data-center SSD revenue was "nearly $10 billion", over two-thirds of NAND revenue, and Core Data Center ($18.0B, 90% gross margin) became Micron's largest business unit. Management disclosed no customer inventory build and said it has no line of sight to supply and demand returning to balance. Neither leg of the falsifier fired. Two caveats on the record: the quarter ran one week longer than the one before, which flatters the sequential growth, and Micron's own inventory rose nine days to 129 days, which is the supplier's stock, not the customers' build the falsifier named. Graded blind against the written falsifier; first-party confirmation of a read that had rested on one publisher. Detail: tkal.news/earnings/micron-q4-fy26.	LOW <40%	HIT
Sep 28, 2026	Arrow's supplier float goes back: receivables minus payables on the October 3, 2026 balance sheet rise above the April level of $1.22B (Arrow Electronics Q2 2026 deep dive). Horizon by October 29, 2026 · Falsifier Receivables minus payables on the October 3, 2026 balance sheet at or below $0.90B (the narrow gap is the design); between $0.90B and $1.22B is partial.	MEDIUM 40-70%	OPEN
Sep 28, 2026	Avnet's June payables were timing: on the October 3, 2026 balance sheet payables fall back below 90% of inventory while total debt stays above June's $3.21B (Avnet Q4 FY26 deep dive). Horizon by October 28, 2026 · Falsifier Payables hold near inventory and Q1 FY27 operating cash flow is positive, meaning suppliers are structurally carrying the book; any other combination is partial.	MEDIUM 40-70%	OPEN
Sep 28, 2026	Ingram Micro's Q2 working-capital build was seasonal and priced, and December releases it: fiscal year-end working capital dollars at or below $4.14B (Ingram Micro Q2 2026 deep dive). Horizon by early March 2027 (FY2026 year-end report) · Falsifier Fiscal year-end working capital above $4.95B (the build has become a floor, as at TD SYNNEX); between $4.14B and $4.95B is partial.	MEDIUM 40-70%	OPEN
Sep 24, 2026	The working capital TD SYNNEX has tied up since November is the new price of growing Hyve, not a one-quarter lump: trade working capital ends fiscal 2026 above the $5.71B it reached in May (TD SYNNEX Q3 FY26 deep dive). Horizon by January 2027 (Q4 FY26 results, November 30, 2026 balance sheet) · Falsifier Trade working capital (receivables plus inventory minus payables) at or below $4.75B on the November 30, 2026 balance sheet; between $4.75B and $5.71B is a partial result.	MEDIUM 40-70%	OPEN
Sep 21, 2026	Reversal risk at its ceiling with zero triggers fired means the pipeline could not test itself, not that something broke: the split holds, with server DDR5 and enterprise TLC still positive while client DDR5, client DDR4, all three mobile DRAM lanes and channel QLC read down, carried as the prior week's finding surviving rather than new evidence (Issue №018). Horizon by mid-January 2027 (1Q27 memory contract prints) · Falsifier A 1Q27 contract print that moves the enterprise and channel tiers the same way.	LOW <40%	OPEN
Sep 15, 2026	Taiwan's memory makers are climbing a staircase, not sitting on a plateau: October revenue for the five memory makers files at least 10% above September (Taiwan monthly revenue, August 2026 edition). Horizon by November 10, 2026 (October monthly revenue filings) · Falsifier The five memory makers file October revenue less than 10% above September.	MEDIUM 40-70%	OPEN
Sep 14, 2026	With credit stable, no stress alert active and an 8-to-32-week lead-lag, the binding constraint this quarter is measurement rather than financing, and the Q3 supplier inventory print due by October 31 is the nearest hard read on whether the value growth carried any bits (Issue №017). Horizon by October 31, 2026 (Q3 supplier inventory print) · Falsifier The page writes no explicit wrong-if; the test it names is the Q3 supplier inventory print due by October 31. The read it defends is that data-centre memory value growth (Dell'Oro +182%) carried no measurable bits and that financing is not the constraint; a print showing bit growth moving with the value growth, or a credit stress event inside the quarter, would break it.	MEDIUM 40-70%	OPEN
Sep 11, 2026	Oracle stopped funding the AI build-out and started intermediating it, and the customer prepayment is a borrowing with a real cost: in Q2 FY27 the prepayment does not repeat, or it repeats with a visible interest cost (Oracle Q1 FY27 deep dive). Horizon by December 9, 2026 (Oracle Q2 FY27 print) · Falsifier Q2 FY27 brings a second financing-component prepayment of roughly $10B or more with imputed interest still immaterial, and gross capex steps down into the $20,500M-$22,167M the guide implies.	MEDIUM 40-70%	OPEN
Sep 7, 2026	DDR4, the one commodity leg still calling up, is price without volume: a series-record spot on falling transaction volume, surging vendor inventories and a 70% inversion over DDR5 reads as a 0.50 phantom ratio rather than demand, with the Taiwan distributor August revenue print around September 10 as the named resolver (Issue №016). Horizon by September 10, 2026 (Taiwan distributor August revenue print) · Falsifier The Taiwan distributor August revenue print shows volume behind the DDR4 price, meaning the memory-weighted distributors keep growing sequentially alongside the record spot and the bid reads as demand rather than phantom. Resolved Sep 10, 2026. The call needed the distributor tape to show DDR4's record price without the volume to back it. The named resolver landed on time. Supreme Electronics, the memory-weighted distributor, reported August revenue of NT$40.741B, down 47.0% from July's NT$76.874B, and credited its growth to memory price increases; WPG reported NT$142.238B, down 2.64% on July; only WT Microelectronics rose, up 8.5%, on communications, industrial, automotive and data-centre demand. Two of the three largest names shrank on the month while prices held, which is the price-without-volume shape the call described, and the read held afterwards: Nanya reported third-quarter selling prices up more than 60% on bit shipments down mid-single digits, and by Issue №020 on October 5 DDR4 spot was in a second correction. Two caveats on the record: no distributor broke out DDR4 or filed a bit count, so the print separates price from volume only by inference; and the DDR4 premium over DDR5 narrowed gradually rather than breaking. Graded PARTIAL rather than HIT: the price-without-volume shape showed up in the totals, but the print could not isolate DDR4, and the pipeline's own week-after read said the DDR4 premium narrowed instead of breaking. Graded 2026-10-20, 40 days after its own horizon.	MEDIUM 40-70%	PARTIAL
Sep 6, 2026	Lenovo's record infrastructure margin is substantially a timing win, earned on components bought before they repriced: by the December quarter ISG operating margin falls toward the low single digits while ISG revenue still grows (Lenovo Q1 FY26/27 deep dive). Horizon by February 2027 (Lenovo Q3 FY26/27 print) · Falsifier ISG operating margin holds above 6% in Q3 FY26/27 while raw materials and work-in-progress stop outgrowing revenue.	MEDIUM 40-70%	OPEN
Sep 3, 2026	Broadcom's binding constraint has moved from chip demand to how fast signed gigawatts become shipped ones: at the Q4 print the ~$115B fiscal 2027 AI figure holds or is cut while Q4 AI revenue meets the $21.7B guide (Broadcom Q3 FY26 deep dive). Horizon by December 9, 2026 (Broadcom Q4 FY26 print) · Falsifier Broadcom raises fiscal 2027 AI revenue above ~$115B with the same six customers, meaning supply, not site readiness, was the constraint.	MEDIUM 40-70%	OPEN
Sep 3, 2026	HPE's record margin is a holding-period margin, not an operating one: the standalone Q4 inventory outflow holds at or above Q3's $2,891M with suppliers still funding about four-fifths of it (HPE Q3 FY26 deep dive). Horizon by early December 2026 (HPE Q4 FY26 print and full-year cash flow statement) · Falsifier The standalone Q4 inventory outflow narrows below roughly $1.5B while revenue lands inside the $13.9B-$14.8B guide.	MEDIUM 40-70%	OPEN
Sep 3, 2026	Western Digital's growth changed ingredients, with price doing the work volume used to: first-quarter fiscal 2027 exabyte shipments grow less than 25% year on year even as revenue lands near the $4.1B guide (WDC Q4 FY26 deep dive). Horizon by October 22, 2026 (WDC Q1 FY27 print) · Falsifier Q1 FY27 exabytes shipped up 25% or more year on year with revenue at or near the $4.1B guide.	MEDIUM 40-70%	OPEN
Sep 2, 2026	Dell is now funding its own AI backlog: in Q3 FY27 the financing-receivables add-back stays above roughly $2B, free cash flow again lands below the year-ago $506M, and the financing book keeps growing faster than revenue (Dell Q2 FY27 deep dive). Horizon by late November 2026 (Dell Q3 FY27 print) · Falsifier The financing-receivables add-back falls back to roughly $1.2B or below while revenue tracks the $49.0B guide and gross margin holds near 20%.	MEDIUM 40-70%	OPEN
Sep 2, 2026	Palo Alto's acquisitions bought a bigger number, not a faster one: lapped NGS ARR growth holds at or below 23%, adjusted free-cash-flow margin sits at 38.0% or lower, and the FY28 40% target gets restated rather than approached (Palo Alto Networks Q4 FY26 deep dive). Horizon by February 2027 (Palo Alto Networks Q2 FY27 print; first read at Q1 in November) · Falsifier NGS ARR growth reaccelerates above roughly 26% on the lapped base while non-GAAP operating margin holds at or above 29.5% and adjusted FCF margin returns toward 38.4%.	MEDIUM 40-70%	OPEN
Aug 31, 2026	Credit stress is near term: four lenders (Blue Owl, Ares, Blackstone, Golub) sit in the non-accrual registry on a 2.8% five-year-high default cohort, Blue Owl sits on both sides of the transmission through the $2.4B IREN financing, and the cascade onto HBM, server DDR5 and enterprise SSD demand is read at twelve weeks rather than sixteen to thirty-two (Issue №015). Horizon by November 23, 2026 (twelve weeks; Blue Owl BDC filing November 15) · Falsifier The page writes no explicit wrong-if. Its named test is whether the non-accrual number stays an aggregate or acquires a borrower's name by the November 15 Blue Owl BDC filing; no named AI borrower and no stress transmission into HBM, server DDR5 or enterprise SSD demand inside the twelve-week horizon would mean the near-term call was wrong.	MEDIUM 40-70%	OPEN
Aug 31, 2026	CXMT's first-half windfall was earned on a plant finished before the cycle turned, and the half sits in the trough between two builds: full-year 2026 capex stays near the RMB 43.19B the half annualises to, commitments sit or are revised, and gross margin compresses toward the cost base (CXMT H1 2026 deep dive). Horizon by April 30, 2027 (FY2026 annual report; partial read in the Q3 report) · Falsifier Full-year 2026 capex at or above roughly RMB 50B, construction in progress climbing past RMB 36.78B as the RMB 52.44B of commitments draws down, and main-business gross margin still above roughly 70%.	MEDIUM 40-70%	OPEN
Aug 28, 2026	Marvell's custom-silicon ramp is a revenue story that is not an earnings story: Q3 FY27 non-GAAP gross margin prints at or below 57.5%, or the fourth quarter is guided below 57.5% (Marvell Q2 FY27 deep dive). Horizon by early December 2026 (Marvell Q3 FY27 print) · Falsifier Q3 FY27 non-GAAP gross margin at or above 58.5% on revenue of at least $3,150M, with Q4 guided inside 57.5-58.5% on revenue of at least $3,692.9M.	MEDIUM 40-70%	OPEN
Aug 27, 2026	CrowdStrike's Q2 record was a Flex conversion event booked in one quarter and the implied $421M fourth quarter is the aggressive part of the guide: Q3 FY27 net new ARR does not beat the top of the guide while the full-year ending-ARR midpoint is raised again (CrowdStrike Q2 FY27 deep dive). Horizon by early December 2026 (CrowdStrike Q3 FY27 print) · Falsifier Q3 FY27 net new ARR at or above $347M and the FY27 ending-ARR midpoint raised above $6,607.45M again.	MEDIUM 40-70%	OPEN
Aug 27, 2026	NVIDIA moved from selling into the AI build-out to financing it: Q3 days sales outstanding holds near 60 days and free cash flow again lands well below net income (NVIDIA Q2 FY27 deep dive). Horizon by November 18, 2026 (NVIDIA Q3 FY27 print) · Falsifier Q3 days sales outstanding returns toward 45 days and operating cash flow re-converges with net income, meaning the extra fifteen days were a one-time accommodation.	MEDIUM 40-70%	OPEN
Aug 25, 2026	ADATA's second half is a volume story, not a margin story: the 12.6% incremental gross margin governs it rather than the 42.11% average, so Q3 gross margin does not hold at 40% or above while revenue grows more than 30% (ADATA Q2 2026 deep dive). Horizon by November 14, 2026 (ADATA Q3 2026 report) · Falsifier Q3 gross margin holds at or above 40.0% while revenue grows more than 30% QoQ.	MEDIUM 40-70%	OPEN
Aug 25, 2026	Winbond's revenue is a staircase that steps once a quarter on contract repricing, not a volume expansion: monthly revenue holds at or above July's level through September, Q3 operating margin is not below Q2's 48.4%, and CMS bit shipments do not jump (Winbond Q2 2026 deep dive). Horizon by November 14, 2026 (Q3 2026 report; monthly legs due Sep 10 and Oct 10) · Falsifier August or September monthly revenue below July's NT$26.773B; or Q3 operating margin below Q2's 48.4%; or Q3 CMS bit shipments up more than 10% sequentially.	MEDIUM 40-70%	OPEN
Aug 24, 2026	Client DRAM contract price gains keep decelerating while server and HBM allocation stays oversubscribed, so the two lanes diverge rather than converge through year-end (Issue №014, corrected edition). Horizon by December 31, 2026 · Falsifier Q4 client DRAM contract gains print at or above the Q3 13 to 18% band, or 2027 server allocation fill rates rise above 80%.	MEDIUM 40-70%	OPEN
Aug 20, 2026	Corsair's memory pass-through spread is closing: the Q3 memory gross margin does not hold above 23% while components revenue declines less than low double digits (Corsair Q2 2026 deep dive). Horizon by November 5, 2026 (Corsair Q3 2026 print and 10-Q) · Falsifier Q3 memory gross margin holds above 23% alongside components revenue declining less than low double digits.	MEDIUM 40-70%	OPEN
Aug 20, 2026	Longsys's warehouse is converting into cash: H2 2026 operating cash flow turns positive as the 152-day holding period matures into shipments (Longsys H1 2026 deep dive). Horizon by April 30, 2027 (FY2026 annual report) · Falsifier Another multi-billion-RMB H2 operating cash outflow with inventory days extending past 152, which would mean the stockpile is not converting.	MEDIUM 40-70%	OPEN
Aug 20, 2026	Team Group's record Q2 was a repricing of inventory it already owned, not a share gain: August and September revenue will not rise materially above NT$2.55B and Q3 revenue will not inflect up double digits, a second straight quarter of contracting bit volume (Team Group Q2 2026 deep dive). Horizon by early November 2026 (Q3 statement; August and September monthly revenue due Sep 10 and Oct 10) · Falsifier August and September monthly revenue print materially above NT$2.55B and Q3 revenue inflects up double digits, meaning Team Group is converting a mix shift into growth. Resolved Oct 10, 2026. Wrong, and the monthly filings settled it before the Q3 statement. The call needed August and September revenue to stay near July's NT$2.55B, with no double-digit Q3 inflection. August came in at NT$3.51B (+38.0% on July) and September at NT$4.08B (+16.2% on August), on the TWSE monthly filings, so Q3 sums to about NT$10.14B, up roughly 33.6% on Q2's NT$7.59B. The August filing credits higher shipments as well as higher prices, which cuts against the contracting-bit-volume read the call rested on. Graded 2026-10-20 on the September filing; the November Q3 statement can confirm the margin but cannot reverse the revenue legs.	MEDIUM 40-70%	MISS
Aug 17, 2026	20 to 40 percent of HBM and server DDR5 order pull is phantom, sourced from 2027 allocation acceptance and Foxconn pull-in language rather than committed builds, and the deficit narrative is extending rather than repricing until the Q3 CY26 hyperscaler capex commentary says otherwise (Issue №013). Horizon by October 30, 2026 (Amazon, Meta, Microsoft and Alphabet Q3 CY26 capex commentary) · Falsifier Amazon, Meta, Microsoft and Alphabet Q3 CY26 capex commentary around October 30 echoes Alphabet's August sustainability pushback rather than reaffirming the plan, which would mean the deficit narrative is repricing, not extending.	MEDIUM 40-70%	OPEN
Aug 10, 2026	Apple's answer to the bipartisan Schumer letter, due by August 21, decides the China-substitution math: a hard commit to reject CXMT and YMTC memory preserves tracked-vendor LPDDR5X capture at Apple globally, while silence or a soft answer re-prices it and hands the China suppliers the opening (Issue №012). Horizon by August 21, 2026 · Falsifier The page set a binary rather than a direction. The read it was defending is that CXMT/YMTC substitution has moved from projection to volume; a hard public Apple commitment to reject CXMT and YMTC would cap that read and preserve tracked-vendor LPDDR5X capture.	MEDIUM 40-70%	OPEN
Aug 3, 2026	Memory contract pricing is decelerating, not reversing: TrendForce's Q3 benchmark still points up (DRAM +13-18%, NAND +10-15% QoQ), and the Q4 refresh due around October 15 is the topping validator, where a flatline or reversal would be the first real crack (Issue №011). Horizon by October 15, 2026 (TrendForce Q4 contract refresh) · Falsifier TrendForce's Q4 contract refresh shows a flatline or reversal. Resolved Sep 30, 2026. The call needed TrendForce's Q4 contract outlook to keep pointing up, more slowly, with no flatline and no reversal. TrendForce published its 4Q26 forecast early, on September 30: conventional DRAM contracts up 10-15% quarter on quarter and overall NAND up 15-20%. Neither a flatline nor a reversal appeared. DRAM decelerated from the Q3 band of 13-18%, as the call said it would. NAND did not: it re-accelerated from 10-15%, led by enterprise SSD, the only line where increases were speeding up. The no-reversal half was right everywhere; the deceleration half was right for DRAM and wrong for NAND. Graded 2026-10-20, 5 days after its own horizon.	MEDIUM 40-70%	HIT
Jul 27, 2026	Credit stress is HIGH, with the near-term flag on and a six-week cascade into HBM: a non-accrual naming a specific GPU-cloud borrower (CoreWeave, Lambda, Nebius, Applied Digital) before September 7 confirms it, and absent that, hyperscaler self-funded capex keeps the deficit intact (Issue №010). Horizon by September 7, 2026 · Falsifier No non-accrual naming a specific GPU-cloud borrower appears before September 7, and the cascade into HBM does not show. Resolved Sep 7, 2026. The call's headline was a six-week credit cascade into HBM, to be confirmed by a non-accrual naming CoreWeave, Lambda, Nebius or Applied Digital before September 7. No such non-accrual appeared in the Q2 BDC filing season. Octus counted BDC non-accrual debt at cost down 5% in Q2, to $9.5B from $10B across 173 BDCs, and named no AI or GPU-cloud borrower. The pipeline's own credit file on August 31 (Issue №015) still showed non-accruals only as an aggregate with no borrower's name. By September 14 (Issue №017) credit read stable with no stress alert. HBM did not compress either: it stayed sold out, and by October 5 (Issue №020) Micron had raised its 2027 HBM pricing. What played out was the fallback clause, hyperscaler capex keeping the deficit intact. That was the branch the call said would follow if its own thesis did not confirm, not the call itself. Graded 2026-10-20, 43 days after its own horizon.	MEDIUM 40-70%	MISS
Jul 20, 2026	SK hynix's Q2 print confirms the memory tightness read, and an HBM4 timing reset or DRAM guidance below +13-18% QoQ would break it (Issue №009). Horizon by July 29, 2026 (SK hynix Q2 2026 print; the issue dated it July 22) · Falsifier SK hynix resets HBM4 timing, or guides DRAM ASPs below +13-18% QoQ. Resolved Jul 29, 2026. The call needed SK hynix's Q2 print to show no HBM4 slip and DRAM pricing no weaker than +13-18% quarter on quarter. SK hynix reported on July 29: revenue of KRW 79.3T, up 257% year on year, and operating profit of KRW 60.5T at a record 76% margin. HBM4 began mass shipment in the quarter, with yields described as approaching HBM3E maturity, so there was no timing reset. DRAM average selling prices rose about 30% quarter on quarter, above the band. Neither leg of the falsifier fired. Two caveats on the record. SK hynix does not publish ASP guidance, so the DRAM leg is graded on the realized quarter, and the ~30% figure is a secondary estimate. The rate of increase also halved from about 60%, and the shares fell about 10% on a consensus miss: the tightness held, but the acceleration did not. The issue had the print dated July 22; it came a week later. Graded 2026-10-20, 83 days after its own horizon.	MEDIUM 40-70%	HIT
Jul 13, 2026	Meta's capex language, which put 2026 at $125-145B partly because of memory pricing, holds on its July 29 print; if it turns into deferral, the single anchor holding contract-lane pricing weakens with it (Issue №008). Horizon by July 29, 2026 (Meta Q2 2026 print) · Falsifier Meta's July 29 capex language flinches into deferral. Resolved Jul 29, 2026. The call needed Meta's capex language to hold on July 29 rather than turn to deferral. It held, and tightened upward. Meta now expects 2026 capital expenditures of $130-145B, "narrowed from our prior outlook of $125-145 billion", which raises the low end. Q2 capex including finance-lease principal was $31.08B against $31.86B of operating cash flow, leaving $784M of free cash flow, and Meta issued $24.91B of long-term debt in the quarter to help fund the build. There was no deferral. The release did not restate memory pricing as a reason; that attribution came from the earlier update, and it was Amazon the next day that tied its own raise to $220B to higher memory costs. The stock fell about 10% after hours on the cash line, not on any retreat in spend. Graded 2026-10-20, 83 days after its own horizon.	MEDIUM 40-70%	HIT
Jul 6, 2026	Blue Owl's Q2 2026 10-Q, due around July 31, is the credit test: a second non-accrual or a redemption honor rate under 30% opens an 8-12 week cascade to enterprise SSD and server DDR5, and until then the credit risk stays a near-term flag, not an active alert (Issue №007). Horizon by October 23, 2026 (12 weeks after the Blue Owl Q2 10-Q) · Falsifier The page's own trigger rule fails: a second non-accrual or a sub-30% redemption honor rate appears, and no cascade into enterprise SSD and server DDR5 demand follows within 8-12 weeks.	MEDIUM 40-70%	OPEN
Jun 29, 2026	Section 232 phase-2 lands July 1 as the binary swing for the commodity client DDR5 lane: implementation without a memory carve-out deepens the client rollover, and a carve-out relieves the front-running and starts unwinding the 35% phantom ratio (Issue №006). Horizon by July 1, 2026 (Section 232 phase-2) · Falsifier The page named no falsifier beyond the July 1 event: phase-2 implementation without a memory carve-out deepens the commodity client rollover, and a carve-out unwinds the front-running. Resolved Jul 1, 2026. The call rested on a dated event that did not happen. July 1 was the deadline for a Commerce Department report under the January semiconductor proclamation, not an implementation date. No phase-2 tariff took effect and no memory carve-out was announced. The report was never published, and Commerce Secretary Lutnick only confirmed on September 2 that phase two is coming, with no rates, product list or effective date. With neither branch firing, the call is graded on the read it defended, a commodity client rollover, and that did not show in prices either. TrendForce's mainstream DDR4 spot rose 2.88% in the week to July 7, and its September 30 forecast had conventional DRAM contracts still rising 10-15% in 4Q26, with PC DRAM held up by OEMs buying ahead of 2027. End demand did weaken, but the price rollover the call named did not arrive. Graded 2026-10-20, 111 days after its own horizon.	MEDIUM 40-70%	MISS
Jun 26, 2026	Dell's non-GAAP gross margin holds at or above 18% in the Q2 FY27 print and sets no new trailing-twelve-month low — operating leverage keeps outrunning AI-mix margin erosion. Horizon by September 15, 2026 · Falsifier Q2 FY27 non-GAAP gross margin prints below ~18% AND below the Q1 FY27 level (18.1%). Resolved Sep 1, 2026. Q2 FY27 (quarter ended Jul 31, 2026; reported Sep 1, 2026): non-GAAP gross margin 21.1%, against 18.7% a year earlier and 18.1% in Q1 FY27; GAAP 20.9% against 18.3%. Neither leg of the falsifier fired - the print was not below ~18%, and not below Q1's 18.1% - and it set no new trailing-twelve-month low, which remains Q1 FY27's 18.1%. Graded blind against the written falsifier, which it passed. The mechanism named in the call was only half right: operating margin did expand to 11.5% from 6.0%, but gross margin rose mainly because the AI share of revenue FELL about 190bps to 34.92%, not because operating leverage absorbed a rising AI mix. Right answer, partly wrong reason - recorded here rather than quietly claimed, and the reasoning is corrected in the Dell Q2 FY27 deep dive.	MEDIUM 40-70%	HIT
Jun 26, 2026	Lenovo's Infrastructure Solutions Group reports a positive operating profit again in its Q1 FY26/27 print, confirming the FY25/26 server turn is structural rather than a single-quarter peak (Lenovo Q4 FY25/26 deep dive). Horizon by August 31, 2026 · Falsifier ISG reports an operating loss or breakeven (at or below $0) for Q1 FY26/27. Resolved Aug 13, 2026. The call needed one thing: ISG operating profit above zero in the next quarter. Lenovo's Q1 FY26/27 (three months ended Jun 30, 2026; reported Aug 13, 2026) showed ISG operating profit of $776.9M on record ISG revenue of $8.5B, up 98%, against an operating loss of $85.5M a year earlier and $73M for the whole of FY25/26. The falsifier did not come close to firing. The Lenovo Q1 FY26/27 dive agrees the margin turn stuck, and adds a caveat: the balance sheet suggests much of the 9.1% ISG margin came from components bought before prices rose, a timing win that pays once. That question is now its own dated test (Q3 FY26/27, February 2027). Graded 2026-10-20, 50 days after its own horizon.	MEDIUM 40-70%	HIT
Jun 24, 2026	Micron's strategic customer agreements are real take-or-pay economics rather than a peak-cycle headline: the next filings confirm the roughly $100B minimum-price backlog and customers post roughly $10B of cash deposits in fiscal Q4, on a margin floor above any prior peak (Micron Q3 FY26 deep dive). Horizon by September 30, 2026 (Micron FQ4 FY26 print) · Falsifier Customer deposits fail to arrive at the promised ~$10B in fiscal Q4, or the filings do not confirm the ~$100B of SCA backlog, which would mean the commitments were not real. Resolved Sep 30, 2026. The June dive said the proof the agreements were real was cash: about $10B of customer deposits in fiscal Q4, a ~$100B remaining-performance-obligation figure confirmed in the filings, and a margin that held as price increases moderated. Micron's Q4 FY26 (quarter ended Sep 3, 2026; reported Sep 30, 2026) brought $12.3B of customer deposits in the quarter per management, and noncurrent customer contract liabilities on the 8-K balance sheet rose to $12,895M from $568M in May. Management put remaining performance obligations at approximately $150B across 26 agreements, against about $100B across 16 in June. Non-GAAP gross margin was 87.0% against the ~86% guide, while DRAM price increases moderated to high-teens percent from low-60s. Neither leg of the falsifier fired, and the Micron Q4 FY26 dive graded the June read right on all four of its watch items. Caveat on the record: the ~$150B RPO figure comes from the earnings call, not yet from a filed number. Graded 2026-10-20, 20 days after its own horizon.	MEDIUM 40-70%	HIT
Jun 22, 2026	DDR4 spot has fallen three weeks running, and a fourth decline on the June 26 TrendForce print escalates the segment to dumping and confirms the commodity rollover (Issue №005). Horizon by June 26, 2026 (TrendForce weekly spot print) · Falsifier DDR4 spot does not post a fourth straight weekly decline on the late-June TrendForce print. Resolved Jun 24, 2026. The call needed a fourth straight weekly fall in DDR4 spot. TrendForce's weekly spot update of June 24 had the mainstream DDR4 chip (1Gx8 3200MT/s) up 0.28% to US$35.90. The following weeks kept rising, to US$36.10 on July 1 and US$37.14 on July 7 (+2.88%). There was no fourth decline and no dumping. The premise was wrong as well: TrendForce's June prints showed DDR4 rising 3.57% and 2.22% in the first two weeks, with only one -0.28% week (June 16). The 'third weekly decline' the pipeline carried was a TrendForce spot update dated April 8. The issue had warned that its evidence was 21 days stale, and this was the item that broke. Graded 2026-10-20, 116 days after its own horizon.	MEDIUM 40-70%	MISS
Jun 14, 2026	Micron's fiscal Q3 print on June 24 is the pivot for the hold-with-upside-bias memory dashboard; softening HBM 2027 allocation visibility or weaker DRAM contract guidance resets it (Issue №004). Horizon by June 24, 2026 (Micron FQ3 FY26 print) · Falsifier Micron's FQ3 print softens HBM 2027 allocation visibility or guides DRAM contract pricing weaker. Resolved Jun 24, 2026. The call needed Micron's June 24 print to hold, not soften, HBM 2027 visibility and DRAM pricing. Micron reported fiscal Q3 revenue of $41.5B, up 346% year on year, a record 84.9% gross margin, and DRAM prices up in the low 60s percent, then guided the next quarter to about $50B at about 86% gross margin. Management said HBM3E and HBM4 were fully booked through calendar 2027 with demand running into 2028, and disclosed 16 long-term customer agreements carrying about $100B of performance obligations. Neither leg of the falsifier fired, and the next issue (№006) moved every memory lane to UP off this print. One caveat on the record: Micron flagged that the rate of DRAM price increases would moderate, and it did (high-teens percent in the September quarter against low-60s in June). That is a slower rise, not weaker contract guidance. Graded 2026-10-20, 118 days after its own horizon.	MEDIUM 40-70%	HIT
Jun 10, 2026	Oracle's backlog is real but the funding bill is the binding constraint, not demand: the September Q1 FY27 print grades whether the $638B of RPO converts on the +58-64% cloud growth guide while capex and the $20B equity program show how tight the funding math is (Oracle Q4 FY26 deep dive). Horizon by September 10, 2026 (Oracle Q1 FY27 print) · Falsifier Q1 FY27 total cloud growth undershoots the +58-64% guide while RPO keeps inflating, reopening the 'booked but unbuildable' question. Resolved Sep 10, 2026. The June call said the backlog was real and the funding, not the demand, was the constraint, with the Q1 FY27 cloud guide of +58-64% as the conversion test. Q1 FY27 (quarter ended Aug 31, 2026; reported Sep 10, 2026): total cloud revenue $11.6B, up 62% in dollars, inside the guide; RPO $664B, up from $638B. The falsifier did not fire. The funding bill also arrived: the $20B ATM was fully used in the quarter, 141 million shares at about $141.20, and capex for the year was guided to $90-95B against $55.7B in FY26. Right direction, wrong mechanism, and the Oracle Q1 FY27 dive says so: the next increment of funding came from customers, not from more equity or debt, through $11,363M of 'customer prepayments with significant financing component', and total borrowings fell $4,204M. Graded 2026-10-20, 40 days after its own horizon.	MEDIUM 40-70%	HIT
Jun 8, 2026	Cloud capex holds its $830B 2026 pace through the Q2 hyperscaler earnings cluster around July 30; a first crack below that trajectory invalidates the AI-lane pull-through thesis (Issue №003). Horizon by July 30, 2026 (Q2 hyperscaler earnings cluster) · Falsifier A Q2 hyperscaler print shows the first crack below the $830B 2026 cloud-capex trajectory. Resolved Jul 30, 2026. The call needed the Q2 hyperscaler prints to show no crack below the 2026 capex trajectory. None did: all four US builders raised or tightened their plans upward. Alphabet (July 22) lifted 2026 capex to $195-205B from $180-190B and said 2027 goes higher; Microsoft (July 29) raised its outlook and guided fiscal Q1 FY27 capex above $50B; Meta (July 29) narrowed its range to $130-145B from $125-145B, lifting the low end; Amazon (July 30) raised its 2026 plan to $220B from about $200B, citing higher memory costs. The $830B figure is TrendForce's top-nine aggregate and was not re-published inside the window, so the call is graded on the direction of the four prints that make up most of it. Graded 2026-10-20, 82 days after its own horizon.	MEDIUM 40-70%	HIT
Jun 5, 2026	Phison's record quarter is real, but what decides the next leg is not EPS: NAND pricing has to hold through the next settlement and the NT$72B inventory build has to convert to cash rather than turn into an inventory trap (Phison Q1 FY26 deep dive). Horizon by August 13, 2026 (Phison Q2 FY26 print) · Falsifier NAND pricing breaks before the next settlement, or the NT$72B inventory sits rather than selling through into cash as the cycle cools. Resolved Aug 13, 2026. The June call had two legs: NAND pricing holds, and the NT$72B inventory converts to cash. Q2 FY26 (quarter ended Jun 30, 2026; reported Aug 13, 2026) landed the first: gross margin rose to 65.3% from 61.3%, inventory write-downs shrank to about 1.10% of margin from 1.50%, and every monthly revenue filing cited higher shipments and higher average prices. It missed the second: inventory grew again to NT$91.79B from NT$72.20B, and operating cash flow was negative NT$1.02B for the quarter and negative NT$14.87B for the half, so the stock did not turn into cash. The trap did not spring, because prices kept rising, but the cash conversion the page named did not happen either. The AI-ecosystem share the page also watched held at 38% rather than moving toward the 'past 50%' target. The Phison Q2 FY26 dive called the frame right; graded blind against the two named legs, one of two landed. Graded 2026-10-20, 68 days after its own horizon.	MEDIUM 40-70%	PARTIAL
Jun 3, 2026	2 of {MasTec, EMCOR, Comfort Systems, Sterling} report DSO extension >5 days QoQ on Q2 prints Horizon by August 31, 2026 · Falsifier All four names maintain or compress DSO on their Q2 prints Resolved Aug 31, 2026. All four names had reported Q2 2026 by the Aug 31 horizon. DSO computed as period-end net receivables divided by quarterly revenue, times days in quarter: MasTec 91.0 -> 88.0 (-3.1 days), EMCOR 95.8 -> 97.1 (+1.3), Comfort Systems 99.9 -> 94.8 (-5.1), Sterling 71.2 -> 72.8 (+1.6). Zero of four extended by more than 5 days and two compressed; the call needed 2 of 4 and got none. The construction pipeline's own Q2 sweep reached the same verdict independently and retired the instrument: no Tier-1 working-capital stress event appeared anywhere in the 90-day evidence stream, and the strain that did surface was a $337M execution charge at AECOM. The binding constraint in this cycle is labour productivity, not customer payment behaviour, so cash conversion was the wrong place to look. Graded 2026-09-06, 6 days after its own horizon.	MEDIUM 40-70%	MISS
May 27, 2026	Q2 2026 DRAM settlement comes in below the published guidance band Horizon by July 15, 2026 · Falsifier Q2 DRAM settlement prints inside the published guidance band Resolved Jul 15, 2026. Wrong, and in the opposite direction. The call was that the Q2 2026 DRAM settlement would print BELOW the published guidance band as phantom distributor demand unwound. Q2 contract prices instead rose roughly 60% quarter on quarter, the largest single-quarter increase of the cycle, and suppliers carried it into Q3: Samsung sought as much as 20% more on average DRAM selling prices, SK Hynix removed the ceilings from its long-term supply agreements, and both described HBM, DRAM and NAND capacity as essentially sold out for 2026. Graded MISS rather than PARTIAL: the settlement did not print below the band, and it did not print inside it either. The demand-refusing-a-price mechanism the call was built on did eventually appear, but in the RATE of increase in 3Q26 and attributed to consumer-electronics makers refusing another rise - not to the phantom-inventory unwind, and not inside the window. Graded 2026-09-06, 53 days after its own horizon; that delay is a failure of this scorecard, not of the call, and it is written up in Machine Room No.006.	MEDIUM 40-70%	MISS

Oct 19, 2026

The book stays defensive and its one buy, 2027 HBM from Micron, is now a single-supplier call, carried provisionally; the October 22 Micron Taiwan bonus deadline is the first test (Issue №022).

Horizon by October 22, 2026 (Micron Taiwan bonus deadline) · Falsifier A strike notice, work stoppage or disclosed output loss at Micron Taiwan fires the reversal.

LOW
<40%

OPEN

Oct 16, 2026

YMTC's Q1 warehouse was held for price, not unsold: in the next financial update in its STAR review, revenue per quarter after Q1 2026 averages above Q1's RMB 47.04B (YMTC pre-IPO analysis).

Horizon at YMTC's next prospectus financial update (undated; estimated by December 31, 2026) · Falsifier Quarters after Q1 2026 average below RMB 47.04B of revenue in the next update (the inventory built in the best quarter was product nobody took).

MEDIUM
40-70%

OPEN

Oct 15, 2026

TSMC's pricing keeps outrunning dilution: 4Q26 revenue per wafer shipped stays above US$10,438 (TSMC Q3 2026 deep dive).

Horizon by January 14, 2027 · Falsifier 4Q26 revenue per wafer shipped falls below US$10,438.

MEDIUM
40-70%

OPEN

Oct 15, 2026

The memory staircase rises in October: the five Taiwan memory makers file October 2026 revenue at least 10% above September's NT$90.36B, i.e. NT$99.40B or more (Taiwan monthly revenue, September 2026 edition).

Horizon by November 10, 2026 · Falsifier The five memory makers file October revenue less than 10% above September's NT$90.36B (below NT$99.40B).

MEDIUM
40-70%

OPEN

Oct 14, 2026

ASML's customer advances hold: non-current contract liabilities on the Q4 2026 balance sheet stay above EUR 5.63B, the end-2024 high (ASML Q3 2026 deep dive).

Horizon by January 27, 2027 · Falsifier Non-current contract liabilities on the Q4 2026 balance sheet fall below EUR 5.63B.

MEDIUM
40-70%

OPEN

Oct 13, 2026

Applied Materials' missing WFE outlook was about fab supply, not demand: Q4 FY26 revenue prints at or above $10.25B with non-GAAP gross margin at or above 50.4%, and the fiscal 2027 guide steps up from there (Applied Materials SEMICON West investor briefing).

Horizon by November 12, 2026 · Falsifier Q4 FY26 revenue below $9.75B (bottom of the guided range), or the first fiscal 2027 revenue guide below the $10.25B Q4 midpoint.

MEDIUM
40-70%

OPEN

Oct 12, 2026

The book is defensive and its only buy is 2027 HBM4 allocation, carried as provisional and low conviction until the 2027 HBM4 settlements due around October 31 confirm it or fire the reversal (Issue №021).

Horizon by October 31, 2026 (2027 HBM4 settlements) · Falsifier The 2027 HBM4 settlements from SK hynix, Samsung and Micron come in well below asking, or 8-high becomes the main Rubin stack.

LOW
<40%

OPEN

Oct 12, 2026

Nanya's contract price holds up: October-December 2026 monthly revenue averages above NT$44.55B (Nanya Q3 2026 deep dive).

Horizon by January 10, 2027 · Falsifier October-December monthly revenue averages below NT$44.55B.

MEDIUM
40-70%

OPEN

Oct 6, 2026

CrowdStrike's fourth quarter did not get easier: Q3 FY27 net new ARR lands inside $343M-347M and the fiscal 2027 ending-ARR range is held, leaving a roughly $421M fourth quarter in place (CrowdStrike Fal.Con briefing).

Horizon by early December 2026 (Q3 FY27 print) · Falsifier The Q3 FY27 print raises the fiscal 2027 ending-ARR range above $6,611.9M, meaning the fourth quarter got easier rather than harder.

MEDIUM
40-70%

OPEN

Oct 6, 2026

Sandisk's contract model holds at the first print after Investor Day: Q1 FY27 non-GAAP gross margin at or above 83% with the eight-customer agreement count holding or growing (Sandisk Investor Day).

Horizon by October 29, 2026 · Falsifier Q1 FY27 non-GAAP gross margin below 80%, or fiscal 2027 bit coverage under the agreements restated below 50%.

MEDIUM
40-70%

OPEN

Oct 5, 2026

The durable lane is 2027 HBM, repriced by Micron and reportedly by Samsung, while consumer DRAM and channel NAND roll over; Samsung's 2027 HBM4 settlement, due by end-October, confirms or breaks it (Issue №020).

Horizon by October 31, 2026 (Samsung 2027 HBM4 settlement) · Falsifier Samsung's 2027 HBM4 settlement comes in low against the reported quote of more than three times HBM3E; a settlement near the quote keeps HBM the cleanest lane.

MEDIUM
40-70%

OPEN

Oct 1, 2026

WT Microelectronics' supplier float holds through Q3: net working capital (receivables plus inventory minus payables) on the September 30, 2026 balance sheet at or below the March high of NT$127.97B (WT Microelectronics Q2 2026 deep dive).

Horizon by November 14, 2026 · Falsifier Net working capital above NT$127.97B on September 30 (the float has started to turn and growth is back on borrowed money).

MEDIUM
40-70%

OPEN

Sep 30, 2026

Jabil's AI ramp is still being financed through net-booked components: prepaid expenses and other current assets on the November 30, 2026 balance sheet stay above $4,559M (Jabil Q4 FY26 deep dive).

Horizon by mid-December 2026 (Q1 FY27 release) · Falsifier Prepaid expenses and other current assets fall below May's $3,925M at November 30 (the FY26 build was a timing lump that shipped); between $3,925M and $4,559M is partial.

MEDIUM
40-70%

OPEN

Sep 30, 2026

Micron's customers keep paying into the build: noncurrent customer contract liabilities on the Q1 FY27 balance sheet rise above $18.0B (Micron Q4 FY26 deep dive).

Horizon by December 16, 2026 · Falsifier Noncurrent customer contract liabilities no higher than $12,895M at the Q1 FY27 balance sheet (no new money, Q4 was a signing lump); in between is partial.

MEDIUM
40-70%

OPEN

Sep 30, 2026

Phison's holding period catches up with price: Q3 2026 gross margin prints at or below 65.3% (Phison Q2 FY26 deep dive).

Horizon by early November 2026 · Falsifier Q3 gross margin above 65.3% (price still outrunning the cost layer).

MEDIUM
40-70%

OPEN

Sep 30, 2026

WPG's third quarter lands in the bottom half of its own NT$460B-500B guide, because June was a pull-in: September 2026 revenue comes in below NT$191.67B (WPG Holdings Q2 2026 deep dive).

Horizon by October 10, 2026 · Falsifier September revenue at or above NT$191.67B, the June level held without a pull-in to explain it.

Resolved Oct 10, 2026. Wrong. The call needed WPG's September revenue below NT$191.67B, the bar its Q3 guide midpoint required after a flat July (NT$146.10B) and August (NT$142.24B), on the view that June had been a pull-in. September came in at NT$219.59B, up 54.4% on August, on the TWSE/MOPS monthly filing, which cites technology and energy facility expansion rather than any pull-in. That puts the three-month sum near NT$507.92B, above the top of the NT$460B to NT$500B guide, not in its bottom half. The quarter-end ratio the call leaned on (1.15 times the first two months) was blown through at about 1.52 times. Graded 2026-10-20, 10 days after its own horizon.

MEDIUM
40-70%

MISS

Sep 29, 2026

CDW's June working-capital build is the new price of the hardware mix, not a lump: the three-month cash conversion cycle for the quarter ending September 30, 2026 is 16 days or more (CDW Q2 2026 deep dive).

Horizon by early November 2026 · Falsifier Three-month cash conversion cycle at 12 days or fewer for the September quarter (the June build converted); 13-15 days is partial.

MEDIUM
40-70%

OPEN

Sep 29, 2026

Insight's fee layer, not the server cycle, drives its earnings: Q3 cloud gross profit at or above $150M with services gross profit growing faster than product gross profit (Insight Enterprises Q2 2026 deep dive).

Horizon by November 5, 2026 · Falsifier Product gross profit outgrows services gross profit in Q3 (hardware has become the engine); cloud below $150M with services still ahead is partial.

MEDIUM
40-70%

OPEN

Sep 29, 2026

Seagate's June volume step was the outlier and the September quarter is met mostly on price: HDD exabytes shipped for the quarter ending October 2, 2026 come in below 228 (Seagate Q4 FY26 deep dive).

Horizon by October 27, 2026 · Falsifier Fiscal Q1 exabytes reach 237 or more (HAMR capacity arriving on schedule, volume still carrying more than half the growth); 228-237 is partial.

MEDIUM
40-70%

OPEN

Sep 28, 2026

Enterprise and data-center SSD demand survives Micron's fiscal fourth-quarter print: the pipeline's only buy, carried as provisional because a single research house supported it (Issue №019).

Horizon by September 30, 2026 (Micron FQ4 FY26 print) · Falsifier Micron's FQ4 print shows data-center NAND softening or customer inventory builds.

Resolved Sep 30, 2026. Micron Q4 FY26 (14 weeks ended Sep 3, 2026; reported Sep 30, 2026): NAND revenue $14.1B, up 42% sequentially on bit shipments up about 10% and prices up about 30%. Data-center SSD revenue was "nearly $10 billion", over two-thirds of NAND revenue, and Core Data Center ($18.0B, 90% gross margin) became Micron's largest business unit. Management disclosed no customer inventory build and said it has no line of sight to supply and demand returning to balance. Neither leg of the falsifier fired. Two caveats on the record: the quarter ran one week longer than the one before, which flatters the sequential growth, and Micron's own inventory rose nine days to 129 days, which is the supplier's stock, not the customers' build the falsifier named. Graded blind against the written falsifier; first-party confirmation of a read that had rested on one publisher. Detail: tkal.news/earnings/micron-q4-fy26.

LOW
<40%

HIT

Sep 28, 2026

Arrow's supplier float goes back: receivables minus payables on the October 3, 2026 balance sheet rise above the April level of $1.22B (Arrow Electronics Q2 2026 deep dive).

Horizon by October 29, 2026 · Falsifier Receivables minus payables on the October 3, 2026 balance sheet at or below $0.90B (the narrow gap is the design); between $0.90B and $1.22B is partial.

MEDIUM
40-70%

OPEN

Sep 28, 2026

Avnet's June payables were timing: on the October 3, 2026 balance sheet payables fall back below 90% of inventory while total debt stays above June's $3.21B (Avnet Q4 FY26 deep dive).

Horizon by October 28, 2026 · Falsifier Payables hold near inventory and Q1 FY27 operating cash flow is positive, meaning suppliers are structurally carrying the book; any other combination is partial.

MEDIUM
40-70%

OPEN

Sep 28, 2026

Ingram Micro's Q2 working-capital build was seasonal and priced, and December releases it: fiscal year-end working capital dollars at or below $4.14B (Ingram Micro Q2 2026 deep dive).

Horizon by early March 2027 (FY2026 year-end report) · Falsifier Fiscal year-end working capital above $4.95B (the build has become a floor, as at TD SYNNEX); between $4.14B and $4.95B is partial.

MEDIUM
40-70%

OPEN

Sep 24, 2026

The working capital TD SYNNEX has tied up since November is the new price of growing Hyve, not a one-quarter lump: trade working capital ends fiscal 2026 above the $5.71B it reached in May (TD SYNNEX Q3 FY26 deep dive).

Horizon by January 2027 (Q4 FY26 results, November 30, 2026 balance sheet) · Falsifier Trade working capital (receivables plus inventory minus payables) at or below $4.75B on the November 30, 2026 balance sheet; between $4.75B and $5.71B is a partial result.

MEDIUM
40-70%

OPEN

Sep 21, 2026

Reversal risk at its ceiling with zero triggers fired means the pipeline could not test itself, not that something broke: the split holds, with server DDR5 and enterprise TLC still positive while client DDR5, client DDR4, all three mobile DRAM lanes and channel QLC read down, carried as the prior week's finding surviving rather than new evidence (Issue №018).

Horizon by mid-January 2027 (1Q27 memory contract prints) · Falsifier A 1Q27 contract print that moves the enterprise and channel tiers the same way.

LOW
<40%

OPEN

Sep 15, 2026

Taiwan's memory makers are climbing a staircase, not sitting on a plateau: October revenue for the five memory makers files at least 10% above September (Taiwan monthly revenue, August 2026 edition).

Horizon by November 10, 2026 (October monthly revenue filings) · Falsifier The five memory makers file October revenue less than 10% above September.

MEDIUM
40-70%

OPEN

Sep 14, 2026

With credit stable, no stress alert active and an 8-to-32-week lead-lag, the binding constraint this quarter is measurement rather than financing, and the Q3 supplier inventory print due by October 31 is the nearest hard read on whether the value growth carried any bits (Issue №017).

Horizon by October 31, 2026 (Q3 supplier inventory print) · Falsifier The page writes no explicit wrong-if; the test it names is the Q3 supplier inventory print due by October 31. The read it defends is that data-centre memory value growth (Dell'Oro +182%) carried no measurable bits and that financing is not the constraint; a print showing bit growth moving with the value growth, or a credit stress event inside the quarter, would break it.

MEDIUM
40-70%

OPEN

Sep 11, 2026

Oracle stopped funding the AI build-out and started intermediating it, and the customer prepayment is a borrowing with a real cost: in Q2 FY27 the prepayment does not repeat, or it repeats with a visible interest cost (Oracle Q1 FY27 deep dive).

Horizon by December 9, 2026 (Oracle Q2 FY27 print) · Falsifier Q2 FY27 brings a second financing-component prepayment of roughly $10B or more with imputed interest still immaterial, and gross capex steps down into the $20,500M-$22,167M the guide implies.

MEDIUM
40-70%

OPEN

Sep 7, 2026

DDR4, the one commodity leg still calling up, is price without volume: a series-record spot on falling transaction volume, surging vendor inventories and a 70% inversion over DDR5 reads as a 0.50 phantom ratio rather than demand, with the Taiwan distributor August revenue print around September 10 as the named resolver (Issue №016).

Horizon by September 10, 2026 (Taiwan distributor August revenue print) · Falsifier The Taiwan distributor August revenue print shows volume behind the DDR4 price, meaning the memory-weighted distributors keep growing sequentially alongside the record spot and the bid reads as demand rather than phantom.

Resolved Sep 10, 2026. The call needed the distributor tape to show DDR4's record price without the volume to back it. The named resolver landed on time. Supreme Electronics, the memory-weighted distributor, reported August revenue of NT$40.741B, down 47.0% from July's NT$76.874B, and credited its growth to memory price increases; WPG reported NT$142.238B, down 2.64% on July; only WT Microelectronics rose, up 8.5%, on communications, industrial, automotive and data-centre demand. Two of the three largest names shrank on the month while prices held, which is the price-without-volume shape the call described, and the read held afterwards: Nanya reported third-quarter selling prices up more than 60% on bit shipments down mid-single digits, and by Issue №020 on October 5 DDR4 spot was in a second correction. Two caveats on the record: no distributor broke out DDR4 or filed a bit count, so the print separates price from volume only by inference; and the DDR4 premium over DDR5 narrowed gradually rather than breaking. Graded PARTIAL rather than HIT: the price-without-volume shape showed up in the totals, but the print could not isolate DDR4, and the pipeline's own week-after read said the DDR4 premium narrowed instead of breaking. Graded 2026-10-20, 40 days after its own horizon.

MEDIUM
40-70%

PARTIAL

Sep 6, 2026

Lenovo's record infrastructure margin is substantially a timing win, earned on components bought before they repriced: by the December quarter ISG operating margin falls toward the low single digits while ISG revenue still grows (Lenovo Q1 FY26/27 deep dive).

Horizon by February 2027 (Lenovo Q3 FY26/27 print) · Falsifier ISG operating margin holds above 6% in Q3 FY26/27 while raw materials and work-in-progress stop outgrowing revenue.

MEDIUM
40-70%

OPEN

Sep 3, 2026

Broadcom's binding constraint has moved from chip demand to how fast signed gigawatts become shipped ones: at the Q4 print the ~$115B fiscal 2027 AI figure holds or is cut while Q4 AI revenue meets the $21.7B guide (Broadcom Q3 FY26 deep dive).

Horizon by December 9, 2026 (Broadcom Q4 FY26 print) · Falsifier Broadcom raises fiscal 2027 AI revenue above ~$115B with the same six customers, meaning supply, not site readiness, was the constraint.

MEDIUM
40-70%

OPEN

Sep 3, 2026

HPE's record margin is a holding-period margin, not an operating one: the standalone Q4 inventory outflow holds at or above Q3's $2,891M with suppliers still funding about four-fifths of it (HPE Q3 FY26 deep dive).

Horizon by early December 2026 (HPE Q4 FY26 print and full-year cash flow statement) · Falsifier The standalone Q4 inventory outflow narrows below roughly $1.5B while revenue lands inside the $13.9B-$14.8B guide.

MEDIUM
40-70%

OPEN

Sep 3, 2026

Western Digital's growth changed ingredients, with price doing the work volume used to: first-quarter fiscal 2027 exabyte shipments grow less than 25% year on year even as revenue lands near the $4.1B guide (WDC Q4 FY26 deep dive).

Horizon by October 22, 2026 (WDC Q1 FY27 print) · Falsifier Q1 FY27 exabytes shipped up 25% or more year on year with revenue at or near the $4.1B guide.

MEDIUM
40-70%

OPEN

Sep 2, 2026

Dell is now funding its own AI backlog: in Q3 FY27 the financing-receivables add-back stays above roughly $2B, free cash flow again lands below the year-ago $506M, and the financing book keeps growing faster than revenue (Dell Q2 FY27 deep dive).

Horizon by late November 2026 (Dell Q3 FY27 print) · Falsifier The financing-receivables add-back falls back to roughly $1.2B or below while revenue tracks the $49.0B guide and gross margin holds near 20%.

MEDIUM
40-70%

OPEN

Sep 2, 2026

Palo Alto's acquisitions bought a bigger number, not a faster one: lapped NGS ARR growth holds at or below 23%, adjusted free-cash-flow margin sits at 38.0% or lower, and the FY28 40% target gets restated rather than approached (Palo Alto Networks Q4 FY26 deep dive).

Horizon by February 2027 (Palo Alto Networks Q2 FY27 print; first read at Q1 in November) · Falsifier NGS ARR growth reaccelerates above roughly 26% on the lapped base while non-GAAP operating margin holds at or above 29.5% and adjusted FCF margin returns toward 38.4%.

MEDIUM
40-70%

OPEN

Aug 31, 2026

Credit stress is near term: four lenders (Blue Owl, Ares, Blackstone, Golub) sit in the non-accrual registry on a 2.8% five-year-high default cohort, Blue Owl sits on both sides of the transmission through the $2.4B IREN financing, and the cascade onto HBM, server DDR5 and enterprise SSD demand is read at twelve weeks rather than sixteen to thirty-two (Issue №015).

Horizon by November 23, 2026 (twelve weeks; Blue Owl BDC filing November 15) · Falsifier The page writes no explicit wrong-if. Its named test is whether the non-accrual number stays an aggregate or acquires a borrower's name by the November 15 Blue Owl BDC filing; no named AI borrower and no stress transmission into HBM, server DDR5 or enterprise SSD demand inside the twelve-week horizon would mean the near-term call was wrong.

MEDIUM
40-70%

OPEN

Aug 31, 2026

CXMT's first-half windfall was earned on a plant finished before the cycle turned, and the half sits in the trough between two builds: full-year 2026 capex stays near the RMB 43.19B the half annualises to, commitments sit or are revised, and gross margin compresses toward the cost base (CXMT H1 2026 deep dive).

Horizon by April 30, 2027 (FY2026 annual report; partial read in the Q3 report) · Falsifier Full-year 2026 capex at or above roughly RMB 50B, construction in progress climbing past RMB 36.78B as the RMB 52.44B of commitments draws down, and main-business gross margin still above roughly 70%.

MEDIUM
40-70%

OPEN

Aug 28, 2026

Marvell's custom-silicon ramp is a revenue story that is not an earnings story: Q3 FY27 non-GAAP gross margin prints at or below 57.5%, or the fourth quarter is guided below 57.5% (Marvell Q2 FY27 deep dive).

Horizon by early December 2026 (Marvell Q3 FY27 print) · Falsifier Q3 FY27 non-GAAP gross margin at or above 58.5% on revenue of at least $3,150M, with Q4 guided inside 57.5-58.5% on revenue of at least $3,692.9M.

MEDIUM
40-70%

OPEN

Aug 27, 2026

CrowdStrike's Q2 record was a Flex conversion event booked in one quarter and the implied $421M fourth quarter is the aggressive part of the guide: Q3 FY27 net new ARR does not beat the top of the guide while the full-year ending-ARR midpoint is raised again (CrowdStrike Q2 FY27 deep dive).

Horizon by early December 2026 (CrowdStrike Q3 FY27 print) · Falsifier Q3 FY27 net new ARR at or above $347M and the FY27 ending-ARR midpoint raised above $6,607.45M again.

MEDIUM
40-70%

OPEN

Aug 27, 2026

NVIDIA moved from selling into the AI build-out to financing it: Q3 days sales outstanding holds near 60 days and free cash flow again lands well below net income (NVIDIA Q2 FY27 deep dive).

Horizon by November 18, 2026 (NVIDIA Q3 FY27 print) · Falsifier Q3 days sales outstanding returns toward 45 days and operating cash flow re-converges with net income, meaning the extra fifteen days were a one-time accommodation.

MEDIUM
40-70%

OPEN

Aug 25, 2026

ADATA's second half is a volume story, not a margin story: the 12.6% incremental gross margin governs it rather than the 42.11% average, so Q3 gross margin does not hold at 40% or above while revenue grows more than 30% (ADATA Q2 2026 deep dive).

Horizon by November 14, 2026 (ADATA Q3 2026 report) · Falsifier Q3 gross margin holds at or above 40.0% while revenue grows more than 30% QoQ.

MEDIUM
40-70%

OPEN

Aug 25, 2026

Winbond's revenue is a staircase that steps once a quarter on contract repricing, not a volume expansion: monthly revenue holds at or above July's level through September, Q3 operating margin is not below Q2's 48.4%, and CMS bit shipments do not jump (Winbond Q2 2026 deep dive).

Horizon by November 14, 2026 (Q3 2026 report; monthly legs due Sep 10 and Oct 10) · Falsifier August or September monthly revenue below July's NT$26.773B; or Q3 operating margin below Q2's 48.4%; or Q3 CMS bit shipments up more than 10% sequentially.

MEDIUM
40-70%

OPEN

Aug 24, 2026

Client DRAM contract price gains keep decelerating while server and HBM allocation stays oversubscribed, so the two lanes diverge rather than converge through year-end (Issue №014, corrected edition).

Horizon by December 31, 2026 · Falsifier Q4 client DRAM contract gains print at or above the Q3 13 to 18% band, or 2027 server allocation fill rates rise above 80%.

MEDIUM
40-70%

OPEN

Aug 20, 2026

Corsair's memory pass-through spread is closing: the Q3 memory gross margin does not hold above 23% while components revenue declines less than low double digits (Corsair Q2 2026 deep dive).

Horizon by November 5, 2026 (Corsair Q3 2026 print and 10-Q) · Falsifier Q3 memory gross margin holds above 23% alongside components revenue declining less than low double digits.

MEDIUM
40-70%

OPEN

Aug 20, 2026

Longsys's warehouse is converting into cash: H2 2026 operating cash flow turns positive as the 152-day holding period matures into shipments (Longsys H1 2026 deep dive).

Horizon by April 30, 2027 (FY2026 annual report) · Falsifier Another multi-billion-RMB H2 operating cash outflow with inventory days extending past 152, which would mean the stockpile is not converting.

MEDIUM
40-70%

OPEN

Aug 20, 2026

Team Group's record Q2 was a repricing of inventory it already owned, not a share gain: August and September revenue will not rise materially above NT$2.55B and Q3 revenue will not inflect up double digits, a second straight quarter of contracting bit volume (Team Group Q2 2026 deep dive).

Horizon by early November 2026 (Q3 statement; August and September monthly revenue due Sep 10 and Oct 10) · Falsifier August and September monthly revenue print materially above NT$2.55B and Q3 revenue inflects up double digits, meaning Team Group is converting a mix shift into growth.

Resolved Oct 10, 2026. Wrong, and the monthly filings settled it before the Q3 statement. The call needed August and September revenue to stay near July's NT$2.55B, with no double-digit Q3 inflection. August came in at NT$3.51B (+38.0% on July) and September at NT$4.08B (+16.2% on August), on the TWSE monthly filings, so Q3 sums to about NT$10.14B, up roughly 33.6% on Q2's NT$7.59B. The August filing credits higher shipments as well as higher prices, which cuts against the contracting-bit-volume read the call rested on. Graded 2026-10-20 on the September filing; the November Q3 statement can confirm the margin but cannot reverse the revenue legs.

MEDIUM
40-70%

MISS

Aug 17, 2026

20 to 40 percent of HBM and server DDR5 order pull is phantom, sourced from 2027 allocation acceptance and Foxconn pull-in language rather than committed builds, and the deficit narrative is extending rather than repricing until the Q3 CY26 hyperscaler capex commentary says otherwise (Issue №013).

Horizon by October 30, 2026 (Amazon, Meta, Microsoft and Alphabet Q3 CY26 capex commentary) · Falsifier Amazon, Meta, Microsoft and Alphabet Q3 CY26 capex commentary around October 30 echoes Alphabet's August sustainability pushback rather than reaffirming the plan, which would mean the deficit narrative is repricing, not extending.

MEDIUM
40-70%

OPEN

Aug 10, 2026

Apple's answer to the bipartisan Schumer letter, due by August 21, decides the China-substitution math: a hard commit to reject CXMT and YMTC memory preserves tracked-vendor LPDDR5X capture at Apple globally, while silence or a soft answer re-prices it and hands the China suppliers the opening (Issue №012).

Horizon by August 21, 2026 · Falsifier The page set a binary rather than a direction. The read it was defending is that CXMT/YMTC substitution has moved from projection to volume; a hard public Apple commitment to reject CXMT and YMTC would cap that read and preserve tracked-vendor LPDDR5X capture.

MEDIUM
40-70%

OPEN

Aug 3, 2026

Memory contract pricing is decelerating, not reversing: TrendForce's Q3 benchmark still points up (DRAM +13-18%, NAND +10-15% QoQ), and the Q4 refresh due around October 15 is the topping validator, where a flatline or reversal would be the first real crack (Issue №011).

Horizon by October 15, 2026 (TrendForce Q4 contract refresh) · Falsifier TrendForce's Q4 contract refresh shows a flatline or reversal.

Resolved Sep 30, 2026. The call needed TrendForce's Q4 contract outlook to keep pointing up, more slowly, with no flatline and no reversal. TrendForce published its 4Q26 forecast early, on September 30: conventional DRAM contracts up 10-15% quarter on quarter and overall NAND up 15-20%. Neither a flatline nor a reversal appeared. DRAM decelerated from the Q3 band of 13-18%, as the call said it would. NAND did not: it re-accelerated from 10-15%, led by enterprise SSD, the only line where increases were speeding up. The no-reversal half was right everywhere; the deceleration half was right for DRAM and wrong for NAND. Graded 2026-10-20, 5 days after its own horizon.

MEDIUM
40-70%

HIT

Jul 27, 2026

Credit stress is HIGH, with the near-term flag on and a six-week cascade into HBM: a non-accrual naming a specific GPU-cloud borrower (CoreWeave, Lambda, Nebius, Applied Digital) before September 7 confirms it, and absent that, hyperscaler self-funded capex keeps the deficit intact (Issue №010).

Horizon by September 7, 2026 · Falsifier No non-accrual naming a specific GPU-cloud borrower appears before September 7, and the cascade into HBM does not show.

Resolved Sep 7, 2026. The call's headline was a six-week credit cascade into HBM, to be confirmed by a non-accrual naming CoreWeave, Lambda, Nebius or Applied Digital before September 7. No such non-accrual appeared in the Q2 BDC filing season. Octus counted BDC non-accrual debt at cost down 5% in Q2, to $9.5B from $10B across 173 BDCs, and named no AI or GPU-cloud borrower. The pipeline's own credit file on August 31 (Issue №015) still showed non-accruals only as an aggregate with no borrower's name. By September 14 (Issue №017) credit read stable with no stress alert. HBM did not compress either: it stayed sold out, and by October 5 (Issue №020) Micron had raised its 2027 HBM pricing. What played out was the fallback clause, hyperscaler capex keeping the deficit intact. That was the branch the call said would follow if its own thesis did not confirm, not the call itself. Graded 2026-10-20, 43 days after its own horizon.

MEDIUM
40-70%

MISS

Jul 20, 2026

SK hynix's Q2 print confirms the memory tightness read, and an HBM4 timing reset or DRAM guidance below +13-18% QoQ would break it (Issue №009).

Horizon by July 29, 2026 (SK hynix Q2 2026 print; the issue dated it July 22) · Falsifier SK hynix resets HBM4 timing, or guides DRAM ASPs below +13-18% QoQ.

Resolved Jul 29, 2026. The call needed SK hynix's Q2 print to show no HBM4 slip and DRAM pricing no weaker than +13-18% quarter on quarter. SK hynix reported on July 29: revenue of KRW 79.3T, up 257% year on year, and operating profit of KRW 60.5T at a record 76% margin. HBM4 began mass shipment in the quarter, with yields described as approaching HBM3E maturity, so there was no timing reset. DRAM average selling prices rose about 30% quarter on quarter, above the band. Neither leg of the falsifier fired. Two caveats on the record. SK hynix does not publish ASP guidance, so the DRAM leg is graded on the realized quarter, and the ~30% figure is a secondary estimate. The rate of increase also halved from about 60%, and the shares fell about 10% on a consensus miss: the tightness held, but the acceleration did not. The issue had the print dated July 22; it came a week later. Graded 2026-10-20, 83 days after its own horizon.

MEDIUM
40-70%

HIT

Jul 13, 2026

Meta's capex language, which put 2026 at $125-145B partly because of memory pricing, holds on its July 29 print; if it turns into deferral, the single anchor holding contract-lane pricing weakens with it (Issue №008).

Horizon by July 29, 2026 (Meta Q2 2026 print) · Falsifier Meta's July 29 capex language flinches into deferral.

Resolved Jul 29, 2026. The call needed Meta's capex language to hold on July 29 rather than turn to deferral. It held, and tightened upward. Meta now expects 2026 capital expenditures of $130-145B, "narrowed from our prior outlook of $125-145 billion", which raises the low end. Q2 capex including finance-lease principal was $31.08B against $31.86B of operating cash flow, leaving $784M of free cash flow, and Meta issued $24.91B of long-term debt in the quarter to help fund the build. There was no deferral. The release did not restate memory pricing as a reason; that attribution came from the earlier update, and it was Amazon the next day that tied its own raise to $220B to higher memory costs. The stock fell about 10% after hours on the cash line, not on any retreat in spend. Graded 2026-10-20, 83 days after its own horizon.

MEDIUM
40-70%

HIT

Jul 6, 2026

Blue Owl's Q2 2026 10-Q, due around July 31, is the credit test: a second non-accrual or a redemption honor rate under 30% opens an 8-12 week cascade to enterprise SSD and server DDR5, and until then the credit risk stays a near-term flag, not an active alert (Issue №007).

Horizon by October 23, 2026 (12 weeks after the Blue Owl Q2 10-Q) · Falsifier The page's own trigger rule fails: a second non-accrual or a sub-30% redemption honor rate appears, and no cascade into enterprise SSD and server DDR5 demand follows within 8-12 weeks.

MEDIUM
40-70%

OPEN

Jun 29, 2026

Section 232 phase-2 lands July 1 as the binary swing for the commodity client DDR5 lane: implementation without a memory carve-out deepens the client rollover, and a carve-out relieves the front-running and starts unwinding the 35% phantom ratio (Issue №006).

Horizon by July 1, 2026 (Section 232 phase-2) · Falsifier The page named no falsifier beyond the July 1 event: phase-2 implementation without a memory carve-out deepens the commodity client rollover, and a carve-out unwinds the front-running.

Resolved Jul 1, 2026. The call rested on a dated event that did not happen. July 1 was the deadline for a Commerce Department report under the January semiconductor proclamation, not an implementation date. No phase-2 tariff took effect and no memory carve-out was announced. The report was never published, and Commerce Secretary Lutnick only confirmed on September 2 that phase two is coming, with no rates, product list or effective date. With neither branch firing, the call is graded on the read it defended, a commodity client rollover, and that did not show in prices either. TrendForce's mainstream DDR4 spot rose 2.88% in the week to July 7, and its September 30 forecast had conventional DRAM contracts still rising 10-15% in 4Q26, with PC DRAM held up by OEMs buying ahead of 2027. End demand did weaken, but the price rollover the call named did not arrive. Graded 2026-10-20, 111 days after its own horizon.

MEDIUM
40-70%

MISS

Jun 26, 2026

Dell's non-GAAP gross margin holds at or above 18% in the Q2 FY27 print and sets no new trailing-twelve-month low — operating leverage keeps outrunning AI-mix margin erosion.

Horizon by September 15, 2026 · Falsifier Q2 FY27 non-GAAP gross margin prints below ~18% AND below the Q1 FY27 level (18.1%).

Resolved Sep 1, 2026. Q2 FY27 (quarter ended Jul 31, 2026; reported Sep 1, 2026): non-GAAP gross margin 21.1%, against 18.7% a year earlier and 18.1% in Q1 FY27; GAAP 20.9% against 18.3%. Neither leg of the falsifier fired - the print was not below ~18%, and not below Q1's 18.1% - and it set no new trailing-twelve-month low, which remains Q1 FY27's 18.1%. Graded blind against the written falsifier, which it passed. The mechanism named in the call was only half right: operating margin did expand to 11.5% from 6.0%, but gross margin rose mainly because the AI share of revenue FELL about 190bps to 34.92%, not because operating leverage absorbed a rising AI mix. Right answer, partly wrong reason - recorded here rather than quietly claimed, and the reasoning is corrected in the Dell Q2 FY27 deep dive.

MEDIUM
40-70%

HIT

Jun 26, 2026

Lenovo's Infrastructure Solutions Group reports a positive operating profit again in its Q1 FY26/27 print, confirming the FY25/26 server turn is structural rather than a single-quarter peak (Lenovo Q4 FY25/26 deep dive).

Horizon by August 31, 2026 · Falsifier ISG reports an operating loss or breakeven (at or below $0) for Q1 FY26/27.

Resolved Aug 13, 2026. The call needed one thing: ISG operating profit above zero in the next quarter. Lenovo's Q1 FY26/27 (three months ended Jun 30, 2026; reported Aug 13, 2026) showed ISG operating profit of $776.9M on record ISG revenue of $8.5B, up 98%, against an operating loss of $85.5M a year earlier and $73M for the whole of FY25/26. The falsifier did not come close to firing. The Lenovo Q1 FY26/27 dive agrees the margin turn stuck, and adds a caveat: the balance sheet suggests much of the 9.1% ISG margin came from components bought before prices rose, a timing win that pays once. That question is now its own dated test (Q3 FY26/27, February 2027). Graded 2026-10-20, 50 days after its own horizon.

MEDIUM
40-70%

HIT

Jun 24, 2026

Micron's strategic customer agreements are real take-or-pay economics rather than a peak-cycle headline: the next filings confirm the roughly $100B minimum-price backlog and customers post roughly $10B of cash deposits in fiscal Q4, on a margin floor above any prior peak (Micron Q3 FY26 deep dive).

Horizon by September 30, 2026 (Micron FQ4 FY26 print) · Falsifier Customer deposits fail to arrive at the promised ~$10B in fiscal Q4, or the filings do not confirm the ~$100B of SCA backlog, which would mean the commitments were not real.

Resolved Sep 30, 2026. The June dive said the proof the agreements were real was cash: about $10B of customer deposits in fiscal Q4, a ~$100B remaining-performance-obligation figure confirmed in the filings, and a margin that held as price increases moderated. Micron's Q4 FY26 (quarter ended Sep 3, 2026; reported Sep 30, 2026) brought $12.3B of customer deposits in the quarter per management, and noncurrent customer contract liabilities on the 8-K balance sheet rose to $12,895M from $568M in May. Management put remaining performance obligations at approximately $150B across 26 agreements, against about $100B across 16 in June. Non-GAAP gross margin was 87.0% against the ~86% guide, while DRAM price increases moderated to high-teens percent from low-60s. Neither leg of the falsifier fired, and the Micron Q4 FY26 dive graded the June read right on all four of its watch items. Caveat on the record: the ~$150B RPO figure comes from the earnings call, not yet from a filed number. Graded 2026-10-20, 20 days after its own horizon.

MEDIUM
40-70%

HIT

Jun 22, 2026

DDR4 spot has fallen three weeks running, and a fourth decline on the June 26 TrendForce print escalates the segment to dumping and confirms the commodity rollover (Issue №005).

Horizon by June 26, 2026 (TrendForce weekly spot print) · Falsifier DDR4 spot does not post a fourth straight weekly decline on the late-June TrendForce print.

Resolved Jun 24, 2026. The call needed a fourth straight weekly fall in DDR4 spot. TrendForce's weekly spot update of June 24 had the mainstream DDR4 chip (1Gx8 3200MT/s) up 0.28% to US$35.90. The following weeks kept rising, to US$36.10 on July 1 and US$37.14 on July 7 (+2.88%). There was no fourth decline and no dumping. The premise was wrong as well: TrendForce's June prints showed DDR4 rising 3.57% and 2.22% in the first two weeks, with only one -0.28% week (June 16). The 'third weekly decline' the pipeline carried was a TrendForce spot update dated April 8. The issue had warned that its evidence was 21 days stale, and this was the item that broke. Graded 2026-10-20, 116 days after its own horizon.

MEDIUM
40-70%

MISS

Jun 14, 2026

Micron's fiscal Q3 print on June 24 is the pivot for the hold-with-upside-bias memory dashboard; softening HBM 2027 allocation visibility or weaker DRAM contract guidance resets it (Issue №004).

Horizon by June 24, 2026 (Micron FQ3 FY26 print) · Falsifier Micron's FQ3 print softens HBM 2027 allocation visibility or guides DRAM contract pricing weaker.

Resolved Jun 24, 2026. The call needed Micron's June 24 print to hold, not soften, HBM 2027 visibility and DRAM pricing. Micron reported fiscal Q3 revenue of $41.5B, up 346% year on year, a record 84.9% gross margin, and DRAM prices up in the low 60s percent, then guided the next quarter to about $50B at about 86% gross margin. Management said HBM3E and HBM4 were fully booked through calendar 2027 with demand running into 2028, and disclosed 16 long-term customer agreements carrying about $100B of performance obligations. Neither leg of the falsifier fired, and the next issue (№006) moved every memory lane to UP off this print. One caveat on the record: Micron flagged that the rate of DRAM price increases would moderate, and it did (high-teens percent in the September quarter against low-60s in June). That is a slower rise, not weaker contract guidance. Graded 2026-10-20, 118 days after its own horizon.

MEDIUM
40-70%

HIT

Jun 10, 2026

Oracle's backlog is real but the funding bill is the binding constraint, not demand: the September Q1 FY27 print grades whether the $638B of RPO converts on the +58-64% cloud growth guide while capex and the $20B equity program show how tight the funding math is (Oracle Q4 FY26 deep dive).

Horizon by September 10, 2026 (Oracle Q1 FY27 print) · Falsifier Q1 FY27 total cloud growth undershoots the +58-64% guide while RPO keeps inflating, reopening the 'booked but unbuildable' question.

Resolved Sep 10, 2026. The June call said the backlog was real and the funding, not the demand, was the constraint, with the Q1 FY27 cloud guide of +58-64% as the conversion test. Q1 FY27 (quarter ended Aug 31, 2026; reported Sep 10, 2026): total cloud revenue $11.6B, up 62% in dollars, inside the guide; RPO $664B, up from $638B. The falsifier did not fire. The funding bill also arrived: the $20B ATM was fully used in the quarter, 141 million shares at about $141.20, and capex for the year was guided to $90-95B against $55.7B in FY26. Right direction, wrong mechanism, and the Oracle Q1 FY27 dive says so: the next increment of funding came from customers, not from more equity or debt, through $11,363M of 'customer prepayments with significant financing component', and total borrowings fell $4,204M. Graded 2026-10-20, 40 days after its own horizon.

MEDIUM
40-70%

HIT

Jun 8, 2026

Cloud capex holds its $830B 2026 pace through the Q2 hyperscaler earnings cluster around July 30; a first crack below that trajectory invalidates the AI-lane pull-through thesis (Issue №003).

Horizon by July 30, 2026 (Q2 hyperscaler earnings cluster) · Falsifier A Q2 hyperscaler print shows the first crack below the $830B 2026 cloud-capex trajectory.

Resolved Jul 30, 2026. The call needed the Q2 hyperscaler prints to show no crack below the 2026 capex trajectory. None did: all four US builders raised or tightened their plans upward. Alphabet (July 22) lifted 2026 capex to $195-205B from $180-190B and said 2027 goes higher; Microsoft (July 29) raised its outlook and guided fiscal Q1 FY27 capex above $50B; Meta (July 29) narrowed its range to $130-145B from $125-145B, lifting the low end; Amazon (July 30) raised its 2026 plan to $220B from about $200B, citing higher memory costs. The $830B figure is TrendForce's top-nine aggregate and was not re-published inside the window, so the call is graded on the direction of the four prints that make up most of it. Graded 2026-10-20, 82 days after its own horizon.

MEDIUM
40-70%

HIT

Jun 5, 2026

Phison's record quarter is real, but what decides the next leg is not EPS: NAND pricing has to hold through the next settlement and the NT$72B inventory build has to convert to cash rather than turn into an inventory trap (Phison Q1 FY26 deep dive).

Horizon by August 13, 2026 (Phison Q2 FY26 print) · Falsifier NAND pricing breaks before the next settlement, or the NT$72B inventory sits rather than selling through into cash as the cycle cools.

Resolved Aug 13, 2026. The June call had two legs: NAND pricing holds, and the NT$72B inventory converts to cash. Q2 FY26 (quarter ended Jun 30, 2026; reported Aug 13, 2026) landed the first: gross margin rose to 65.3% from 61.3%, inventory write-downs shrank to about 1.10% of margin from 1.50%, and every monthly revenue filing cited higher shipments and higher average prices. It missed the second: inventory grew again to NT$91.79B from NT$72.20B, and operating cash flow was negative NT$1.02B for the quarter and negative NT$14.87B for the half, so the stock did not turn into cash. The trap did not spring, because prices kept rising, but the cash conversion the page named did not happen either. The AI-ecosystem share the page also watched held at 38% rather than moving toward the 'past 50%' target. The Phison Q2 FY26 dive called the frame right; graded blind against the two named legs, one of two landed. Graded 2026-10-20, 68 days after its own horizon.

MEDIUM
40-70%

PARTIAL

Jun 3, 2026

2 of {MasTec, EMCOR, Comfort Systems, Sterling} report DSO extension >5 days QoQ on Q2 prints

Horizon by August 31, 2026 · Falsifier All four names maintain or compress DSO on their Q2 prints

Resolved Aug 31, 2026. All four names had reported Q2 2026 by the Aug 31 horizon. DSO computed as period-end net receivables divided by quarterly revenue, times days in quarter: MasTec 91.0 -> 88.0 (-3.1 days), EMCOR 95.8 -> 97.1 (+1.3), Comfort Systems 99.9 -> 94.8 (-5.1), Sterling 71.2 -> 72.8 (+1.6). Zero of four extended by more than 5 days and two compressed; the call needed 2 of 4 and got none. The construction pipeline's own Q2 sweep reached the same verdict independently and retired the instrument: no Tier-1 working-capital stress event appeared anywhere in the 90-day evidence stream, and the strain that did surface was a $337M execution charge at AECOM. The binding constraint in this cycle is labour productivity, not customer payment behaviour, so cash conversion was the wrong place to look. Graded 2026-09-06, 6 days after its own horizon.

MEDIUM
40-70%

MISS

May 27, 2026

Q2 2026 DRAM settlement comes in below the published guidance band

Horizon by July 15, 2026 · Falsifier Q2 DRAM settlement prints inside the published guidance band

Resolved Jul 15, 2026. Wrong, and in the opposite direction. The call was that the Q2 2026 DRAM settlement would print BELOW the published guidance band as phantom distributor demand unwound. Q2 contract prices instead rose roughly 60% quarter on quarter, the largest single-quarter increase of the cycle, and suppliers carried it into Q3: Samsung sought as much as 20% more on average DRAM selling prices, SK Hynix removed the ceilings from its long-term supply agreements, and both described HBM, DRAM and NAND capacity as essentially sold out for 2026. Graded MISS rather than PARTIAL: the settlement did not print below the band, and it did not print inside it either. The demand-refusing-a-price mechanism the call was built on did eventually appear, but in the RATE of increase in 3Q26 and attributed to consumer-electronics makers refusing another rise - not to the phantom-inventory unwind, and not inside the window. Graded 2026-09-06, 53 days after its own horizon; that delay is a failure of this scorecard, not of the call, and it is written up in Machine Room No.006.

MEDIUM
40-70%

MISS

The Record

Calibration · resolved only

All claims · newest first