The cross-cycle DRAM lag is 11-16 months at n=2, but the 2021-22 rollover was demand-driven, so this is not n=2 on a supply lag
The question
Verbatim: "Re-key the capacity-announcement to ASP-rollover lag on the 2021-2022 memory cycle, where the vault holds monthly TrendForce data from 2021-02, to take the lag calibration from n=1 to n=2 and state it as a cross-cycle range."
Context: the lag between a capacity announcement and the subsequent ASP rollover is what the memory-cycle thesis uses to time the Phase 2 to Phase 3 transition. The 2018 re-key landed at n=1 and had to bracket to a quarter because the vault's ASP anchor holds no pre-2021 data. 2021-2022 is the one prior cycle where the anchor data is already on hand.
What we already know (from the vault)
- The parent brief measured 2018 keyed to Samsung's 2017-07-04 Pyeongtaek P1 announcement: DRAM 12-16 months, NAND 6-7 months, with NAND leading DRAM by roughly three quarters. It called that number a floor, not a transferable estimate, because the 2017 money was tool-fill on a shell that was already standing, and it recommended watching the NAND-leads-DRAM spread rather than the calendar ([[2026-09-09-2018-capacity-announce-asp-rollover-lag]]).
- I re-verified the anchor's coverage rather than trusting the filename.
trendforce-historical-2017-2026.csvholds 252 rows, 2021-02-25 to 2026-04-24, zero pre-2021. The sibling brief's finding holds. But per-year density is thin at exactly the window this question needs: 24 rows in 2021 (~2.2/month) and 8 rows in 2022 (~0.7/month), against 63-73 rows/year in 2024-2025. The 2022 sparsity is documented as real publication behaviour, not scraper failure ([[2026-05-17-trendforce-backfill-notes]]). "Monthly data from 2021-02" is generous: 2021 is roughly bi-weekly-to-monthly, 2022 has multi-month holes. - The vault's own phase-history has no
capacity-announcementrow for the 2022 cycle. Its 2022-cycle sequence runs2021-02 demand-recognitionto2021-04 demand-recognitionto2021-12 capacity-onlineto2022-03 down-cycle. The labelling methodology appliescapacity-announcementto 2017-06 and 2024-09 only. The anchor the question asks me to re-key against does not exist in the cycle it asks me to re-key ([[2026-05-18-memory-cycle-v1-1-phase-history-notes]]). - The live tripwire set has the announcement-to-rollover lag nowhere in its composite decision rule. Phase 3 fires on
(inventory days rising >2Q) AND (ASP rolling over 4-6 months), or on(greenfield bit-supply online) AND (ASP -10% QoQ). Inventory days is the tripwire furthest from triggering and highest-value ([[2026-07-07-dram-hbm-phase2-phase3-early-signals]]). - First greenfield wafer capacity in the current cycle is dated to H2 2027 as the first real step, with the Korea mega-commitment's bits landing 2029-2030 ([[2026-09-15-dram-greenfield-fab-online-dates]]).
What the web says
- Samsung Pyeongtaek P2 is the structural twin of the 2018 keying event, and it dates to 2020-08-30. Samsung announced mass production of 16Gb LPDDR5 at "the world's largest semiconductor line" and began shipping DRAM from P2 that day (Samsung Newsroom / Businesswire, 2020-08-30). Like 2017-07-04, it is simultaneously an announcement and a production start on a shell already built.
- SK Hynix M16 Icheon held its completion ceremony 2021-02-01, with mass production from July 2021 at the earliest and first DRAM output in 2H21; it was the company's first EUV fab (KED Global, 2021-02-01).
- The biggest headline number in the window is not a memory announcement at all. Samsung's 2021-05-13 raise to 171 trillion won through 2030 covers system semiconductors / foundry and System LSI, and reporting is explicit that it is separate from its memory investment (KED Global, 2021-05-13). An analyst keying to the loudest 2021 capex press release would key to the wrong product line.
- TrendForce dated the DRAM rollover to 4Q21 and attributed it to inventory, not capacity. 2021-09-22: overall DRAM ASP "will likely decline by about 3-8% QoQ in 4Q21"; PC DRAM "5-10% QoQ"; server DRAM declining "for the first time this year." Named causes: clients "carrying a higher level of DRAM inventory," North American and Chinese CSPs holding "more than eight weeks' worth of server DRAM inventory," and "global production of notebooks is expected to decline in 4Q21" (TrendForce, 2021-09-22).
- The supply side was explicitly described as restrained. TrendForce 2021-10-12 puts 2022 DRAM bit supply growth at 17.9% against bit demand growth of 16.3% - a 1.6pp surplus gap - and forecasts "a shift in the DRAM market next year from shortage to surplus." Its demand-side attributions are "inventory levels at the moment are relatively high," "the declining procurement activities of DRAM buyers going forward," and "the drop in DRAM spot prices ahead of contract prices." It does not attribute the 4Q21 decline to new capacity, and says supplier expansion plans "are expected to remain conservative." Samsung P3L was only "fully built by mid-2022" with limited 2022 contribution (TrendForce, 2021-10-12).
- 2022 notebook shipments were forecast down 7-8% YoY to ~220M units, with PC OEMs holding 8-10 weeks of DRAM inventory after pre-stocking against a shortage that did not arrive. The anchor CSV's own 2022 headlines carry the same attribution: "Owing to Seasonal Demand Downturn" (2021-12-13), "Pincered by Russian-Ukrainian War and Inflation" (2022-03-28), "Demand for Consumer Electronics Sluggish" (2022-05-10).
- NAND's 2022 path was distorted by an industrial accident, not by capacity. The anchor holds TrendForce calling NAND +5-10% for 2Q22 "Due to Material Contamination at WDC and Kioxia" (2022-02-10, repeated 2022-03-30) while DRAM was still falling. NAND resumed declining in May 2022.
Convergences and contradictions
- Convergence, and it is the one clean result: the DRAM lag reproduces. Keyed to the matched Samsung Pyeongtaek capacity event in each cycle, DRAM runs 12-16 months in 2018 and 11-15 months in 2021-22. Two independent cycles, two different fabs, overlapping brackets. That is a real and slightly surprising replication.
- Contradiction that breaks the parent brief's headline recommendation: NAND did not lead DRAM. In 2018 NAND rolled over roughly three quarters ahead. In 2021-22 both rolled over in the same quarter (4Q21) - TrendForce called NAND 0-5% QoQ declines for 4Q21 on 2021-09-23, DRAM 3-8% on 2021-09-22 - and then the sign flipped, with DRAM continuing down through 2Q22 while NAND rose on the Kioxia/WD contamination. At n=2 the "NAND leads DRAM by ~3 quarters" rule is 1-for-2 with a sign reversal in the second case. It should be demoted, not promoted to a tripwire.
- Contradiction on mechanism, and this is the load-bearing one. The vault frames the lag as supply-mechanical. The contemporaneous primary sources for 2021-22 attribute the rollover to client inventory, notebook production declines, spot leading contract, seasonality, and then war and inflation. The measured supply surplus was 1.6pp of bit growth, and the analyst who published the forecast called supplier expansion "conservative." Six of the eight decline-signalling rows in the anchor CSV between 2021-08 and 2022-05 name demand, inventory or macro in the title; two name oversupply. The 2021-22 rollover was predominantly demand-driven.
- A structural mismatch the matching lag conceals. The 2018 up-leg ran nine consecutive quarters of DRAM contract-price growth. The 2021-22 up-leg ran three (1Q21-3Q21). Two up-legs differing threefold in length produced near-identical announcement-to-rollover lags. If the lag were measuring a supply mechanism, that should not happen.
Synthesis for RDCO
The measurement, with every keying choice named. The 2021-22 DRAM rollover dates cleanly from the anchor: contract prices rose 1Q21 through 3Q21, the first forward decline call published 2021-08-10 (PC DRAM 0-5% for 4Q21), the first actual QoQ decline was 4Q21 at -3 to -8% overall ASP, and the vault's own phase-history independently marks 2021-12 and 2022-03. Keyed to Samsung Pyeongtaek P2, 2020-08-30, that is 11 months to the first published decline call, 13 months to the last up-quarter ending, and ~15 months to mid-decline-quarter: DRAM 11-15 months. NAND, keyed to the same event, is 13-15 months. Keying instead to SK Hynix M16's 2021-02-01 completion gives 6-10 months for the same rollover, and keying to Samsung's 2021-05-13 capex raise gives ~6 months while being the wrong product line entirely. Announcement selection moves the answer by about five months, which is as wide as the range itself. I keyed to Pyeongtaek P2 because it is the only event in the 2020-21 window structurally matched to the 2018 keying event: a Samsung Pyeongtaek line announcing mass production on an already-standing shell. That is a defensible choice and it is still a choice.
The cross-cycle range at n=2: DRAM 11-16 months, NAND 6-15 months. The DRAM union is narrow and the brackets overlap at 12-15 months. The NAND union does not overlap at all, so there is no NAND range worth quoting. Two observations are two observations. Do not read 11-16 as a distribution, do not take a midpoint, and do not let it set a date. What it can bear: nudging a prior, and sanity-checking a lag someone else proposes. What it cannot bear: triggering a position change, setting the shell_share threshold in [[2026-09-06-vendor-capex-composition-memory-anchor]], or being quoted as "the" lag. An honest band on an n=2 estimate where keying discretion is ±5 months is at least ±3 quarters.
The re-key does not get the thesis to a clean n=2 on supply lag, and that is the finding. The 2018 interval was already contaminated: the parent brief flagged a simultaneous iPhone-pricing and trade-war demand air pocket. The 2021-22 interval is worse - the proximate cause per every contemporaneous primary source is an inventory unwind and a post-COVID consumer-electronics demand normalization, against a supply surplus of 1.6pp that the forecasting analyst described as arising from conservative expansion. So the supply-only calibration stands at n=1 contaminated, not n=2, and arguably n=0 clean. The more parsimonious reading of the matching 11-16 month numbers is uncomfortable: memory demand wobbles every four to six quarters, so "roughly a year after a splashy capacity press release" will land on a demand wobble in almost any cycle regardless of whether the capacity mattered. The threefold difference in up-leg length across the two cycles, with no corresponding difference in lag, is the tell. The charitable counter-reading is that announcement-to-incremental-wafers really is ~12-15 months for tool-fill on a standing shell, and the demand wobble is what converts latent supply into a visible price decline - a supply-readiness plus demand-trigger composite. Both readings fit the data; neither supports using the lag as a clock.
What this does to the Phase 2 to Phase 3 call. Applying 11-16 months to the current window's three candidate keying events gives: the 2025-11-13 TrendForce capex print to Oct 2026 - Mar 2027; Micron's 10-Q period ended 2026-05-28 to Apr 2027 - Sep 2027; the Jul 2026 Korea commitment and SK Hynix's 2026-07-28 capex raise to Jun 2027 - Nov 2027. The union is Q4 2026 through Q4 2027, essentially unchanged from the parent brief - the re-key tightens the input bracket by one month and changes no date. The shell-first correction still applies and now looks larger, not smaller: both cycles I keyed were tool-fill on standing shells, whereas the current window is buildings-first with the first greenfield step dated H2 2027 and the mega-commitment's bits dated 2029-2030. So the raw range remains a floor. The actionable change is a demotion, not a date. Both measured rollovers were triggered by an inventory and demand event, and the live tripwire set already has supplier inventory days as its highest-value signal, currently at single-digit weeks against a ~5-6 week trigger. This brief says: keep weight on inventory days, strip the announcement-to-ASP lag back from a timing instrument to a prior, and retire the NAND-leads-DRAM ordering that the 2018 brief proposed as a tripwire - it failed its first out-of-sample test.
Why this is in the vault
This closes open follow-up #2 of [[2026-09-09-2018-capacity-announce-asp-rollover-lag]] ("repeat this re-keying on the 2021-2022 cycle... take the lag calibration from n=1 to n=2") and its follow-up #4 (the supply-versus-demand decomposition), and it returns a negative result on both: the DRAM number replicates while the mechanism does not, so the calibration that [[2026-09-06-vendor-capex-composition-memory-anchor]] wanted for its 45% shell_share threshold still does not exist. It also supplies the out-of-sample test that demotes the "NAND leads DRAM by ~3 quarters" rule the 2018 brief recommended adding to the Phase-3 tripwire set in [[2026-07-07-dram-hbm-phase2-phase3-early-signals]], and it documents that phase-history.csv carries no capacity-announcement row for the 2022 cycle, which /investing:label-historical-phases should either fill with Samsung P2 (2020-08-30) or leave empty deliberately.
Open follow-ups
- Quantify the 2018 cycle's supply-demand gap the same way 2021-22 now is (contemporaneous 2018-2019 DRAM bit-supply versus bit-demand growth from TrendForce or WSTS), so the two cycles' surplus gaps are comparable. 2021-22 has a measured 1.6pp; 2018 has no number at all, which is why its supply attribution is still an assumption.
- Test whether a memory capacity announcement has ever been followed by an ASP rollover without a coincident demand or inventory event. If no clean supply-only case exists across the five labelled cycles, the supply-only lag is unidentifiable from announcement data and the anchor should be retired rather than refined.
- Measure up-leg length against announcement-to-rollover lag across 2008 and 2024-current. 2018 ran nine up-quarters and 2021-22 ran three for near-identical lags; if that invariance holds at n=4, the lag is measuring demand periodicity rather than capacity.
- Settle the NAND-versus-DRAM lead ordering on the 2008 and 2024-current cycles. It is 1-for-2 with a sign reversal; two more observations decide whether 2018 was coincidence.
- Estimate the effect size of the Kioxia and Western Digital February 2022 material-contamination event on NAND ASP, so the 2022 NAND rollover can be read counterfactually - would NAND have led DRAM absent the accident?
- Recover a real 2017-2020 ASP series (Wayback CDX crawl of the pre-2021
/presscenter/news/scheme). Carried unresolved from the parent brief; it remains the binding constraint on every pre-2021 dating in this thesis, and it would also let the 2021-22 up-leg be compared to 2018's on the same instrument rather than on press-release language.
Related
- [[2026-09-09-2018-capacity-announce-asp-rollover-lag]]
- [[2026-06-28-chip-memory-cycle-phase2-phase3-indicators]]
- [[2026-07-07-dram-hbm-phase2-phase3-early-signals]]
- [[2026-09-06-vendor-capex-composition-memory-anchor]]
- [[2026-09-07-memory-supplier-inventory-days-dio-anchor]]
- [[2026-09-15-dram-greenfield-fab-online-dates]]
- [[2026-05-18-memory-cycle-v1.1]]
- [[2026-05-18-memory-cycle-v1-1-phase-history-notes]]
- [[2026-05-27-markov-equities-pipeline-spec]]
- [[2026-05-17-trendforce-backfill-notes]]
Sources
Vault
~/rdco-vault/01-projects/investing/anchors/dram-spot/trendforce-historical-2017-2026.csv- re-verified: 252 rows, 2021-02-25 to 2026-04-24, zero pre-2021; 24 rows in 2021, 8 in 2022. All 2021-22 rollover dates in this brief read off rows 2021-08-10, 2021-09-22, 2021-09-23, 2021-10-12, 2021-10-20, 2021-12-13, 2021-12-15, 2022-01-24, 2022-02-10, 2022-03-28, 2022-03-30, 2022-05-10, 2022-11-09.~/rdco-vault/01-projects/investing/anchors/dram-spot/2026-05-17-trendforce-backfill-notes.md- per-year density table; 2022 sparsity is real publication behaviour~/rdco-vault/01-projects/investing/anchors/memory-cycle-v1.1/phase-history.csv- 2022-cycle rows (2021-02, 2021-04, 2021-12, 2022-03, 2022-09, 2023-03); nocapacity-announcementrow for this cycle~/rdco-vault/01-projects/investing/anchors/memory-cycle-v1.1/phase-history-notes.md- mechanical rule definitions and the pre-2021 gap statement~/rdco-vault/06-reference/research/2026-09-09-2018-capacity-announce-asp-rollover-lag.md~/rdco-vault/06-reference/research/2026-07-07-dram-hbm-phase2-phase3-early-signals.md~/rdco-vault/06-reference/research/2026-06-28-chip-memory-cycle-phase2-phase3-indicators.md~/rdco-vault/06-reference/research/2026-09-06-vendor-capex-composition-memory-anchor.md~/rdco-vault/06-reference/research/2026-09-07-memory-supplier-inventory-days-dio-anchor.md~/rdco-vault/06-reference/research/2026-09-15-dram-greenfield-fab-online-dates.md~/rdco-vault/01-projects/investing/theses/2026-05-18-memory-cycle-v1.1.md
Web (fetched and quoted)
- TrendForce, 2021-09-22 - "DRAM Prices Projected to Decline by 3-8% QoQ in 4Q21 Due to Rising Level of Client Inventory": https://www.trendforce.com/presscenter/news/20210922-10947.html
- TrendForce, 2021-10-12 - "DRAM Prices Projected to Enter Period of Downswing in 2022 as Demand Lags Behind Supply": https://www.trendforce.com/news/2021/10/12/dram-prices-projected-to-enter-period-of-downswing-in-2022-as-demand-lags-behind-supply-says-trendforce/ (source of the 17.9% supply / 16.3% demand figures, the "remain conservative" line, and the P2L/P3L/M16 fab detail)
Web (cited from search results, body not fetched)
- Businesswire / Samsung Newsroom, 2020-08-30 - "Samsung Begins Mass Production of 16Gb LPDDR5 DRAM at World's Largest Semiconductor Line" (Pyeongtaek P2): https://www.businesswire.com/news/home/20200829005010/en/Samsung-Begins-Mass-Production-of-16Gb-LPDDR5-DRAM-at-Worlds-Largest-Semiconductor-Line
- KED Global, 2021-02-01 - SK Hynix M16 Icheon completion ceremony, mass production from July 2021: https://www.kedglobal.com/newsView/ked202102010006?lang=us
- KED Global, 2021-05-13 - Samsung raises system-semiconductor plan to 171 trillion won through 2030; described as separate from memory investment: https://amp.kedglobal.com/newsAmp/ked202105130008
- TrendForce, 2021-08-10 - PC DRAM contract prices expected to decline 0-5% in 4Q21 (row in the vault anchor CSV; first forward decline call of the cycle)
- 2022 notebook shipment forecast of -7-8% YoY to ~220M units and the 8-10 week PC OEM DRAM inventory figure come from TrendForce press material surfaced via aggregators, not from a TrendForce page I opened. Treat as semi-primary; confirm against a TrendForce original before either number is load-bearing.
Not retrieved (flagged, not retried)
- TrendForce gated
priceChartmember API and DRAMeXchange/IDC/Gartner paid feeds - still the only known route to an actual $/Gb contract ASP series. Every rollover date in this brief is read off directional and QoQ-percentage press-release language, not off a price series, in 2021-22 exactly as in 2018.