A slow record of the memory-chip cycle: a small set of hard series, each point carrying its source and date, held against tripwires stated in advance. Updated when a series prints, not on a clock.
Assessmentas of 2026-10-09
The shipment, pricing and demand-side series still do not show a cycle turn, and no tripwire changed state. Micron's year-end filing shows contract liabilities of $13.0bn, up from $422mn a quarter earlier, mostly deposits customers paid to secure supply under strategic agreements. The contract price outlook remains the series that changed most: increases forecast for the fourth quarter are 10 to 15% for DRAM and 15 to 20% for NAND.
Next readings
~ inferred from the usual reporting cadence
Each tripwire is a threshold on one series, worded before the data arrives and never edited afterwards. The shaded band in each graphic is the firing side of the threshold. A fired tripwire is a recorded observation about the cycle, not a call on any security.
+29.9%
3 months to 2026-09 vs the 3 before
Fires below 0% · data to 2026-09-30
The latest three full months of Korean semiconductor exports sum to less than the three months before them.
Year-on-year growth above 200% will decelerate on base effects alone, so the test is on the dollar level. A rolling three-month window absorbs the quarter-end shipment bulge.
Worded 2026-10-02
+12.5%
2026-Q4 outlook, DRAM midpoint (the lower of the two)
Fires at or below 0% · data to 2026-09-30
TrendForce's first-published outlook for a quarter has a midpoint at or below zero for conventional DRAM or for NAND Flash.
Contract prices are what memory makers are paid. A flat-to-down outlook is the pricing leg of the cycle turning.
Worded 2026-10-02
+27.2%
combined capex, 2026-Q2 vs 2026-Q1
Fires on two straight quarters below 0% · data to 2026-06-30
Combined quarterly capex of Microsoft, Meta, Alphabet and Amazon falls from the prior quarter for two quarters in a row.
These four are the largest buyers of server memory. One down quarter is common seasonally; two is a change of direction.
Worded 2026-10-02
0 of 4
hyperscalers lowered their capex guide at the latest report
Fires at 2 lowered · data to 2026-07-30
Two or more of the four hyperscalers lower their capex guide at their most recent earnings report.
Guidance moves before realized spending does. Reclassifications that change the stated figure without changing the plan are recorded as maintained.
Worded 2026-10-02
1 of 12
cohort members with capex above operating cash flow
Fires at 3 above 1.0 · data to 2026-09-03
Three or more cohort members have trailing-twelve-month capex above operating cash flow.
Capacity committed beyond what operations fund has to be financed, and it arrives whether or not demand holds.
Worded 2026-10-02
0 of 2
balances below the prior reported balance
Fires at 1 lower · data to 2026-09-03
Contract liabilities at Micron or Sandisk fall from the prior reported balance.
Customers prepay to secure supply when it is scarce. A falling balance means prepayments are being consumed faster than new ones are signed.
Worded 2026-10-02
2 of 5
memory makers below consensus at the latest report
Fires at 2 below · data to 2026-09-30
Two or more of the five memory makers (Micron, Sandisk, SK Hynix, Samsung, Kioxia) reported below consensus in their most recent result.
In July 2026 one result below consensus coincided with a 40-55% decline in memory equities while absolute results were at records. This tripwire measures delivery against what was expected, which nothing else in the ledger does.
Worded 2026-10-02
10%
of global DRAM revenue, 2026-Q2 (highest estimate)
Fires at 15% · data to 2026-09-24
A TrendForce or Counterpoint quarterly estimate puts CXMT's share of global DRAM revenue at 15% or higher.
CXMT's share was about 4% in mid-2025 and about 10% in mid-2026. At 15% it would be a fourth supplier of comparable weight to Micron in a market that has had three.
Worded 2026-10-02
$60.3bn
September 2026 · +262.8% year on year · highest month in the record
| Month | Exports | YoY | Source |
|---|---|---|---|
| 2026-09 | $60.3bn | +262.8% | 2026-10-01 |
| 2026-08 | $46.65bn | +209.0% | 2026-09-01 |
| 2026-07 | $41.01bn | +178.8% | 2026-08-03 |
| 2026-06 | $44.82bn | +199.5% | 2026-07-01 |
| 2026-05 | $37.2bn | +169.4% | 2026-06-01 |
| 2026-04 | $31.9bn | +173.5% | 2026-05-04 |
| 2026-03 | $32.8bn | +151.4% | 2026-04-01 |
| 2026-02 | $25.16bn | +160.8% | 2026-03-01 |
| 2026-01 | $20.54bn | +102.7% | 2026-02-01 |
| 2025-12 | $20.8bn | +43.2% | 2026-01-01 |
| 2025-11 | $17.3bn | +38.6% | 2025-12-01 |
| 2025-10 | $15.7bn | +25.4% | 2025-11-03 |
| 2025-09 | $16.6bn | +22.0% | 2025-10-01 |
| 2025-08 | $15.1bn | +27.1% | 2025-09-02 |
| 2025-07 | $14.71bn | +31.6% | 2025-08-01 |
| 2025-06 | $15bn | +11.6% | 2025-07-01 |
| 2025-05 | $13.8bn | +21.2% | 2025-06-02 |
| 2025-04 | $11.7bn | +17.2% | 2025-05-02 |
| 2025-03 | $13.1bn | +11.9% | 2025-04-01 |
| 2025-02 | $9.6bn | −3.0% | 2025-03-04 |
| 2025-01 | $10.1bn | +8.1% | 2025-02-02 |
Full-month semiconductor export value and year-on-year change. Korea is home to two of the three large DRAM makers, so this is the earliest broad monthly read on memory shipments and pricing. Flash (10-day, 20-day) figures are excluded.
Newest data 2026-09-30Tripwire: Not fired
DRAM +10 to 15% · NAND +15 to 20%
2026-Q4 outlook · 2026-Q3 was DRAM +13 to 18%, NAND +10 to 15%
| Quarter | DRAM | NAND | Later revisions |
|---|---|---|---|
| 2026-Q4 | 10 to 15% | 15 to 20% | |
| 2026-Q3 | 13 to 18% | 10 to 15% | |
| 2026-Q2 | 58 to 63% | 70 to 75% | |
| 2026-Q1 | 55 to 60% | 33 to 38% | 2026-02-02: Conventional DRAM revised from +55–60% to +90–95%2026-02-02: NAND Flash revised from +33–38% to +55–60% |
| 2025-Q4 | 8 to 13% | 5 to 10% | 2025-10-29: Conventional DRAM revised from +8–13% to +18–23% |
| 2025-Q3 | 10 to 15% | 5 to 10% | |
| 2025-Q2 | -5 to 0% | no blended range published | |
| 2025-Q1 | -13 to -8% | -15 to -10% |
TrendForce's forecast quarter-on-quarter change in blended conventional DRAM and NAND Flash contract prices, recorded as first published for each quarter. Revisions are kept alongside.
Newest data 2026-09-30Tripwire: Not fired
$165.1bn
2026-Q2, four companies combined · +27.2% on the prior quarter
| Quarter | MSFT | META | GOOG | AMZN | Combined |
|---|---|---|---|---|---|
| 2026-Q2 | 35.8 | 30.1 | 44.9 | 54.2 | 165.1 |
| 2026-Q1 | 30.9 | 19.0 | 35.7 | 44.2 | 129.8 |
| 2025-Q4 | 29.9 | 21.4 | 27.9 | 39.5 | 118.6 |
| 2025-Q3 | 19.4 | 18.8 | 24.0 | 35.1 | 97.3 |
| 2025-Q2 | 17.1 | 16.5 | 22.4 | 32.2 | 88.2 |
| 2025-Q1 | 16.7 | 12.9 | 17.2 | 25.0 | 71.9 |
| 2024-Q4 | 15.8 | 14.4 | 14.3 | 27.8 | 72.3 |
| 2024-Q3 | 14.9 | 8.3 | 13.1 | 22.6 | 58.9 |
| 2024-Q2 | 13.9 | 8.2 | 13.2 | 17.6 | 52.9 |
| 2024-Q1 | 11.0 | 6.4 | 12.0 | 14.9 | 44.3 |
| 2023-Q4 | 9.7 | 7.6 | 11.0 | 14.6 | 42.9 |
| 2023-Q3 | 9.9 | 6.5 | 8.1 | 12.5 | 36.9 |
| 2023-Q2 | 8.9 | 6.1 | 6.9 | 11.5 | 33.4 |
| 2023-Q1 | 6.6 | 6.8 | 6.3 | 14.2 | 33.9 |
Quarterly purchases of property and equipment for Microsoft, Meta, Alphabet and Amazon, summed. The demand side of the memory cycle.
Newest data 2026-06-30Tripwire: Not fired
3 raised · 1 maintained · 0 lowered
latest report from each company, through 2026-07-30
| MSFT | CY2026 $175bnfrom $190bnCalendar-2026 capex expectations unchanged in substance, but the useful-life change shifts data center leases from finance to operating, adjusting the reported expectation to ~$175B; Q1 FY27 capex over $50B. | = maintained2026-07-29 |
| META | FY2026 $130 to 145bnfrom $115 to 135bn → $125 to 145bn2026 capex narrowed to $130-145B from $125-145B (low end raised, high end unchanged). | ▲ raised2026-07-29 |
| GOOG | FY2026 $195 to 205bnfrom $175 to 185bn → $180 to 190bn2026 capex range updated to $195-205B from $180-190B, primarily due to accelerated capacity delivery. | ▲ raised2026-07-22 |
| AMZN | FY2026 $220bnfrom $200bnNow expect approximately $220B cash capex in 2026, up from prior estimate of about $200B, driven by higher memory costs. | ▲ raised2026-07-30 |
| Company | Note | Following year |
|---|---|---|
| MSFT | Stated figure fell from about $190B to about $175B because data-center leases were reclassified from finance to operating leases, which are not counted as capex; Microsoft said its calendar-2026 plans were otherwise unchanged. Recorded as maintained because the ledger tracks the spending plan. IMPORTANT: the reported number fell ($190B -> ~$175B) only because of an accounting change (data center useful life extended 15 -> 25 years, moving future leases from finance to operating, which are excluded from capex). The preceding sentence says: 'Outside of this useful life impact, our calendar year 2026 CapEx investment expectations remain unchanged.' Treat as maintained on an underlying basis; not like-for-like with the April guide. | Expect FY2027 capital expenditures to grow year-over-year, given demand signals across the portfolio. |
| META | Company language is 'narrowed', not 'increased'; classified 'raised' because the low end moved up and the high end did not move. On the call (Q2 2026 transcript PDF), Susan Li: 'we aren't providing a specific outlook for 2027 CapEx at this time.' | no comment recorded |
| GOOG | CFO Anat Ashkenazi on call. Guidance is not in the Q2 2026 earnings release itself (call only). | Continue to expect CapEx to increase significantly in 2027; more details later. |
| AMZN | Stated by CEO Andy Jassy; attributed the increase to 'the higher cost of memory pushing this number up from our prior estimate of about $200 billion.' | No 2027 dollar figure; Jassy said the lion's share of 2027 capacity being added is largely reserved. |
The forward capex guide each hyperscaler gave at each earnings report, and whether it was raised, held or lowered against the prior report.
Newest data 2026-07-30Tripwire: Not fired
1 of 12 above 1.0
AMZN 1.07 · cohort median 0.25 · the four hyperscalers are charted
| Company | Role | Capex / OCF | Period end | Basis |
|---|---|---|---|---|
| AMZN | hyperscaler | 1.07 | 2026-06-30 | four quarters |
| GOOG | hyperscaler | 0.71 | 2026-06-30 | four quarters |
| META | hyperscaler | 0.69 | 2026-06-30 | four quarters |
| MSFT | hyperscaler | 0.63 | 2026-06-30 | four quarters |
| MU | memory-maker | 0.34 | 2026-09-03 | four quarters |
| AMAT | equipment | 0.33 | 2026-07-26 | four quarters |
| AMD | hbm-consumer | 0.17 | 2026-06-27 | four quarters |
| LRCX | equipment | 0.16 | 2026-06-28 | four quarters |
| ASML | equipment | 0.12 | 2025-12-31 | fiscal year |
| KLAC | equipment | 0.09 | 2026-06-30 | four quarters |
| NVDA | hbm-consumer | 0.05 | 2026-07-26 | four quarters |
| SNDK | memory-maker | 0.02 | 2026-07-03 | four quarters |
Trailing-twelve-month capital expenditure divided by operating cash flow, per US-filing cohort member. Above 1.0 a company is spending more on capacity than its operations generate. Sandisk's fab investment runs through its joint venture with Kioxia and is understated here.
Newest data 2026-09-03Tripwire: Not fired
Micron $13,000mn · Sandisk $1,242mn
latest reported balances (2026-Q3, 2026-Q2)
| Quarter | Micron | Sandisk |
|---|---|---|
| 2026-Q3 | $13,000mn | not reported |
| 2026-Q2 | $422mn | $1,242mn |
| 2026-Q1 | not reported | $511mn |
| 2025-Q4 | not reported | not reported |
| 2025-Q3 | $169mn | not reported |
| 2025-Q2 | $146mn | $25mn |
| 2025-Q1 | $555mn | not reported |
| 2024-Q4 | $614mn | not reported |
| 2024-Q3 | $907mn | not reported |
| 2024-Q2 | $756mn | not reported |
Contract liabilities (cash received from customers ahead of delivery) on the balance sheets of Micron and Sandisk. A hard proxy for long-term supply agreements with prepayment: customers paying ahead to secure supply.
Newest data 2026-09-03Tripwire: Not fired
2 of 5 below consensus at the latest report
SK Hynix, Kioxia
Whether each reported result came in above or below the consensus stated at the time. US companies are compared on earnings per share, the others on operating profit; consensus providers differ by article and are named in the source. The percentage is the gap between the reported figure and consensus.
Newest data 2026-07-31Tripwire: Fired
CXMT 10% of DRAM revenue
2026-Q2, highest published estimate · 4% in 2025-Q2
| Company | Market | Quarter | Share | Estimate by |
|---|---|---|---|---|
| CXMT | DRAM | 2025-Q2 | 4% | Counterpoint, 2026-09-04 |
| CXMT | DRAM | 2026-Q1 | 8% | Counterpoint, 2026-05-26 |
| CXMT | DRAM | 2026-Q1 | 7.6% | TrendForce, 2026-07-27 |
| CXMT | DRAM | 2026-Q2 | 10% | Counterpoint, 2026-09-04 |
| CXMT | DRAM | 2026-Q2 | 9.5% | TrendForce, 2026-09-24 |
| YMTC | NAND | 2025-Q2 | 11% | Counterpoint, 2026-09-03 |
| YMTC | NAND | 2025-Q3 | 10% | Counterpoint, 2026-09-03 |
| YMTC | NAND | 2025-Q4 | 12% | Counterpoint, 2026-09-03 |
| YMTC | NAND | 2026-Q1 | 14% | Counterpoint, 2026-09-03 |
| YMTC | NAND | 2026-Q1 | 13% | Counterpoint (via TechNode), 2026-06-22 |
| YMTC | NAND | 2026-Q2 | 14% | Counterpoint, 2026-09-03 |
| Company | Period | Capacity | Source |
|---|---|---|---|
| CXMT | 2024-12 | 100k | TechPowerUp / MyDrivers (via TrendForce news), 2024-12-30 |
| CXMT | 2026-07 | 265k | Tech Times, 2026-07-25 |
| CXMT | 2026-07 | 320k | Counterpoint, 2026-07-15 |
| CXMT | 2026-08 | 300k | Reuters (via TrendForce news), 2026-08-04 |
| CXMT | 2026-09 | 300k | Seoul Economic Daily, 2026-09-04 |
| CXMT | 2026-09 | 320k | Commercial Times (via TrendForce news), 2026-09-24 |
| CXMT | 2026-12 | 350k | Tech Times, 2026-07-25 |
| CXMT | 2026-12 | 350k | TechPowerUp (via TrendForce news), 2026-08-04 |
| CXMT | 2027 | 420k | Counterpoint, 2026-07-15 |
| YMTC | 2026-04 | 200k | Reuters (via Evertiq), 2026-04-14 |
| YMTC | 2027 | 50k | Reuters (via Evertiq), 2026-04-14 |
Quarterly revenue-share estimates from TrendForce and Counterpoint; where the two differ the higher is charted and both are in the table. CXMT makes DRAM and YMTC makes NAND, so the two lines are shares of different markets.
Newest data 2026-09-24Tripwire: Not fired
| Reported | Company | Event |
|---|---|---|
| 2026-07-28 | CXMT | First-day close roughly 466% above the RMB 8.66 offer price (opened at RMB 49.50; intraday peak ~535%). |
| 2026-07-28 | CXMT | Commerce Department reportedly paused a recommendation to add CXMT to the Entity List in June 2026. |
| 2026-07-28 | CXMT | Scheduled: from Dec 23, 2027, Section 5949 of the FY2023 NDAA bars all federal agencies from procuring CXMT semiconductor products (an indirect DoD ban on end items containing 1260H-entity goods starts June 30, 2027). |
| 2026-07-27 | CXMT | CXMT's STAR Market IPO application was accepted by the Shanghai Stock Exchange on 2025-12-20; listing committee deliberation 2026-05-27; CSRC approval 2026-06-12. |
| 2026-07-27 | CXMT | CXMT IPO passed the SSE listing committee on 2026-05-27 and received CSRC registration approval on 2026-06-12. |
| 2026-07-27 | CXMT | Market capitalization at the IPO offer price exceeded RMB 579bn; ~RMB 57.9bn raised before over-allotment. |
| 2026-07-27 | CXMT | First day (2026-07-27): shares rose as much as 471.6% to RMB 49.50 vs RMB 8.66 offer, lifting market cap to ~RMB 3.3 trillion (US$487bn), China's most valuable listed company. |
| 2026-07-25 | CXMT | MATCH Act passed the House Foreign Affairs Committee in April 2026; it would extend the DUV export ban to CXMT and bar ASML from servicing its installed tools (not enacted as of source date). |
| 2026-06-12 | CXMT | DoD's June 8, 2026 Section 1260H list kept CXMT and YMTC on the list (reversing the withdrawn February version); DoD direct procurement ban effective June 30, 2026. |
| 2026-06-12 | YMTC | DoD's June 8, 2026 Section 1260H list kept YMTC (and CXMT) on the list, reversing the withdrawn February version. YMTC has been on the Commerce Entity List since December 2022. |
| 2026-05-28 | CXMT | CXMT prospectus: 2025 revenue RMB 61.799bn, up 155.60% YoY. |
| 2026-05-28 | CXMT | Planned IPO fundraising per prospectus: RMB 29.5bn (the amount actually raised at pricing was RMB 57.9bn). |
| 2026-05-22 | CXMT | CXMT prospectus: Q1 2026 revenue RMB 50.8bn and net profit RMB 24.7bn. |
| 2026-02-13 | CXMT | A Feb 2026 Pentagon notice said CXMT (and YMTC) would be removed from the Section 1260H list (on which they appeared as of the Jan 7, 2025 list); the notice was later withdrawn. |
| 2026-02-13 | YMTC | A Feb 2026 Pentagon notice said YMTC (and CXMT) would be removed from the Section 1260H list; the notice was later withdrawn. |
| 2025-09-26 | YMTC | YMTC Holding valuation reached RMB 161.6bn (~US$22.7bn) after funding from 16 institutional investors (April 2025, per Caixin timeline); restructured into a joint-stock company Sept 2025 ahead of an IPO. |
| 2025-01-16 | CXMT | CXMT was not added to the US Entity List in the December 2024 export-control round, though DoD had identified it as military-associated. |
Dated estimates of CXMT (DRAM) and YMTC (NAND) market share and wafer capacity, plus listing and policy events. Added after July 2026.
15% to 52% below peak
five memory makers, latest close against each one's high since May 2026
| Since May 2026 | Peak to low | Now vs peak |
|---|---|---|
| Micron | −39% | −15% |
| Sandisk | −56% | −31% |
| SK Hynix | −55% | −42% |
| Samsung | −43% | −28% |
| Kioxia | −65% | −52% |
| Company | Peak | Low after peak | Peak to low | Latest | Latest vs peak |
|---|---|---|---|---|---|
| Micron | 1,213.56 (2026-06-25) | 739 (2026-07-29) | −39% | 1,034.03 (2026-10-09) | −15% |
| Sandisk | 2,335 (2026-06-25) | 1,015.89 (2026-07-29) | −56% | 1,613.86 (2026-10-09) | −31% |
| SK Hynix | 2,919,000 (2026-06-22) | 1,322,000 (2026-07-30) | −55% | 1,681,000 (2026-10-08) | −42% |
| Samsung | 362,500 (2026-06-18) | 207,000 (2026-07-30) | −43% | 262,000 (2026-10-08) | −28% |
| Kioxia | 36,233.33 (2026-06-22) | 12,793.33 (2026-07-29) | −65% | 17,505 (2026-10-09) | −52% |
Share prices are not a cycle indicator and no tripwire reads them. They are shown because the gap between what prices do and what the series above do is itself worth recording. Local-currency closes, rebased; the dashed line is a semiconductor benchmark.
Newest data 2026-10-09
2026-10-09 · Micron filed its fiscal 2026 10-K on October 9, adding its year-end balance sheet and cash flows to the prepayment and capex series
Micron's 10-K for the fiscal year ended September 3 puts contract liabilities at $13,000mn, against $422mn at the end of May. The filing describes the balance as primarily customer deposits under strategic customer agreements: the deposit is returned if the customer meets its minimum purchase commitment and may be kept, in whole or part, if it does not. Sandisk's balance was last reported at $1,242mn, up from $511mn.
The same filing lowers Micron's ratio of trailing-twelve-month capex to operating cash flow from 0.49 to 0.34: capex of $30.7bn against operating cash flow of $89.7bn. One of twelve cohort members has capex above operating cash flow (Amazon, 1.07); the capex-trap tripwire fires at three.
Neither change moves a tripwire. Seven of eight remain not fired. The fired one still records that two of the five memory makers, SK Hynix and Kioxia, reported below consensus at their most recent result. Korean chip exports in the three months to September were 29.9% above the prior three, and hyperscaler capex rose 27.2% quarter on quarter in the second quarter.
Micron's deposits are a different instrument from prepayments consumed as product ships. Because they are scheduled to be repaid years out, a later fall in Micron's balance may reflect scheduled repayment as much as weaker demand, and the customer-advances tripwire would read both the same way. The next scheduled readings are hyperscaler reports from about October 21, Korean and Japanese memory-maker results from about October 28, and October Korea exports on November 1.
Through September the shipment, pricing and demand-side series do not show a cycle turn: Korean chip exports set a record and hyperscaler capex is still rising. The one fired tripwire records SK Hynix and Kioxia reporting below consensus in July. The largest change in the record is the contract price outlook, where forecast quarterly increases fell from roughly 60% to roughly 15%.
This entry opens the ledger. Ten series are backfilled, eight tripwires are stated as of today, and one of them reads as fired at the outset. Because the tripwires were worded after July 2026, they can only be scored against data that prints from here on.
Between late June and late July 2026 the shares of every large memory maker fell sharply. On closing prices, Sandisk fell 56% from June 25 to July 29, SK Hynix 55%, Samsung 43%, Micron 39% and Kioxia 65%, against 29% for a broad semiconductor benchmark. All five have since recovered part of the move and remain 12% to 47% below their June highs.
The series that describe the industry itself did not turn over the same weeks. Korean semiconductor exports were $44.8bn in June, $41.0bn in July, $46.7bn in August and a record $60.3bn in September. Combined capex at the four hyperscalers rose 27% quarter on quarter to $165bn in the second quarter, and Meta, Alphabet and Amazon each raised their 2026 capex guide in late July. Customer prepayments on Sandisk's balance sheet rose from $511mn to $1,242mn in the quarter to early July.
One series did change character. TrendForce's contract price outlook, which had called for conventional DRAM up 58 to 63% and NAND up 70 to 75% in the second quarter, was published on July 3 at up 13 to 18% and up 10 to 15% for the third, and on September 30 at up 10 to 15% and up 15 to 20% for the fourth. Prices are still forecast to rise; the pace of increase fell by roughly three quarters in one step. That release came within a week of the equity peak.
Two things that coincided with the decline were not measured by the earlier version of this tracker, and both are series here now. First, delivery against expectations: on July 29 SK Hynix reported operating profit of 60.5 trillion won, more than six times the year before and below a 64 trillion won consensus, and Kioxia reported below consensus two days later. Second, Chinese supply: CXMT listed in Shanghai on July 27 after raising RMB 57.9bn, and research-firm estimates put its share of global DRAM revenue near 10% in the second quarter, up from about 4% a year earlier.
Assessment at baseline: through September the shipment, pricing and demand-side series do not show a cycle turn. Seven of eight tripwires are not fired. The one that is fired records that two of the five memory makers reported below consensus at their most recent result; Micron's September 30 report came in above (earnings per share of 33.42 against an estimate of 32.56). The slowing rate of contract price increases appears to be the most informative change in the record so far, and the next TrendForce outlook and the October Korea export print are the next scheduled readings on it.
Freshness is measured on the date of each series' newest data point, not on when it was last fetched. A source that stops publishing shows as stale here even if the fetch keeps succeeding. The bar is the age of the newest point against that series' limit.
| Series | Source | Newest data | Age against limit | Status |
|---|---|---|---|---|
| Korea semiconductor exports | Korea Customs Service / MOTIE monthly trade release, via same-day press | 2026-09-30 | 9 of 45 days | ○ current |
| Contract price outlook (TrendForce) | TrendForce press center | 2026-09-30 | 9 of 110 days | ○ current |
| Capex / operating cash flow | SEC XBRL companyfacts | 2026-09-03 | 36 of 150 days | ○ current |
| Hyperscaler capex, realized | SEC XBRL companyfacts | 2026-06-30 | 101 of 150 days | ○ current |
| Hyperscaler capex guidance | Company earnings releases and calls | 2026-07-30 | 71 of 150 days | ○ current |
| Customer prepayments at memory makers | SEC XBRL companyfacts | 2026-09-03 | 36 of 150 days | ○ current |
| Results vs consensus, US memory makers | Finnhub | 2026-09-30 | 9 of 150 days | ○ current |
| Results vs consensus, non-US memory makers | Company releases via same-day press | 2026-07-31 | 70 of 150 days | ○ current |
| China memory supply | Research-firm estimates and company filings via press | 2026-09-24 | 15 of 200 days | ○ current |
| Cohort equity prices | Yahoo Finance chart endpoint | 2026-10-09 | 0 of 6 days | ○ current |
The ledger tracks the industry cycle (pricing, shipments, capital commitment, demand-side spending). Equity prices are shown as context and are not treated as a cycle indicator. Every figure traces to a filing, an official release or a named estimate, linked in the tables above. Estimates from research firms differ in definition and may be revised. This is educational content documenting an industry cycle, not investment advice.