Where supply lands: 2027–2029 — scenarios, not forecasts.
Our statistical model ends at Q+3; beyond that horizon we publish no levels. What we do publish: three named scenarios, each with one driving assumption and dated, capital-committed anchors — cleanroom openings, capex commitments and edition cycles of WSTS and SEMI. A scenario without an anchor per year is not shown.
OUTLOOK, not a forecast. Capacity-anchored scenarios, 2027–2029. The statistical model (Holt damped ETS) stops at Q+3; this layer takes over from there.
Three futures, each with one assumption
The three cards below are not forecasts but possibility-spaces in the tradition of the IEA World Energy Outlook: exploratory scenarios flowing from explicit assumptions. The direction per year (tight / easing / loose) changes only through trigger events from primary sources, never through model residuals.
Loading from data/chart-data.json…
Where the model stops, the scenarios begin
The damped Holt model mechanically can keep computing beyond Q+3 — the damped trend flattens toward a level. That mechanical extrapolation has zero evidence status and is therefore switched off by default. Switch it on to see where the model would “want” to go, and watch the divergence from the anchor-driven scenario bands.
DRAM contract QoQ: model tail handing off into the scenario bands
Dated events on which each scenario rests
An anchor is a dated, publicly announced, capital-committed event whose deferral or delivery can later be tested against a press release. Every anchor links to its source with a retrieval date. Anchor statuses change only on primary sources (company newsroom, earnings, WSTS/SEMI edition).
Loading from data/chart-data.json…
Why there are no numbers here
The evidence that 2029 cannot be forecast
WSTS (T1) revised its own 2026 growth from “+26%, $975 bn” (Nov 2025 edition) to “+26% to $1.51 tn” (Jun 2026 edition) — same calendar year, a more-than-doubled revision in seven months. After such evidence, a point-level forecast for 2029 is unpublishable.
The two design parents
This page follows the IEA WEO scenario framing (named exploratory scenarios under explicit assumptions) and the EIA reference-case-with-side-cases pattern. The Q+1..Q+3 model horizon follows our own method scan (forecasting-methods, 2026-09-12).
Trigger accounting
Scenario status changes only on: a reported cleanroom opening or delay (company newsroom), a capex-guidance change (earnings), demand-side LTA/capital news, or a fresh WSTS/SEMI edition. Between triggers this page stays literally the same — stability is the argument.
Schema: outlook[] in data/chart-data.json (contract: forecast-2029-extension.md §4). Anchors carry planned_date, status, source_url and retrieved; numbers live only in anchor data, never in a value field.