On the timing of a violent unwind — certain vs uncertain

2026-06-09. Structured: spec-unwind-timing.json. Direct answer to "when does this come to a head?" with absolute-certainty-only discipline: only proven/definitional statements are called certain; every forecast/date is called uncertain.

The short answer

The timing is UNCERTAIN — and unforecastable in principle. No one can give the date with certainty, and anyone who gives you a precise date is wrong or selling something. What is certain is the structure; what you get instead of a date is an instrument panel.

CERTAIN (proven — not timing)

• The core is solvent only while external capital keeps flowing; at zero inflow it is insolvent (Z3 T4, UNSAT).
• OpenAI needs ≥$1.03T external capital it doesn't have (Z3 T3, UNSAT).
• Self-set marks (AI fair-value, bank HTM, private-credit NAV) must reverse when a forcing event prices them — an IPO below the mark forces a writedown (reflexive_marks M3; MarkUnwind invariant violated). Certain that it reverses if the event occurs — not when.
• The capital-stop cascade path exists (OpenAI→CoreWeave→Oracle, TLA+).
• There is no shock-safe configuration (every leg's UNSAT + the contagion matrix).

UNCERTAIN (stated plainly: the date is not knowable)

Three reasons it's unforecastable in principle:

1. Reflexivity (Soros): the trigger is confidence/coordination, which responds to its own forecast — a variable that reacts to being predicted can't be reliably predicted.
2. Minsky: stability breeds the leverage that destabilizes; the "Minsky moment" is only datable in hindsight. The model gives fragility, not the day.
3. Keynes: "markets can remain irrational longer than you can remain solvent" — bubbles persist past fundamentals, then break suddenly on an exogenous catalyst.

Empirical check: dot-com, 2008, SVB-2023 each had the structure visible for years; the break was sudden and essentially no one called the date. The fragility was forecastable; the date was not.

Also uncertain: whether it's violent or a slow/bailout-cushioned deflate; which trigger fires first; whether policy defers it again; the loss magnitude and who's made whole.

OBSERVABLE (watch these, not the calendar)

Capital-tap triggers: a yen/BOJ carry unwind; a failed or down OpenAI/Anthropic mega-round; an IPO that prices below the last private mark (fires the M3 reversal); an Oracle/CoreWeave debt-service miss. Credit/rate triggers: a long-rate spike reopening bank HTM (already turned up, $306B→$325B in Q1-2026); a private-credit default cluster beyond First Brands/Tricolor (bad-PIK ~6.4%, ~40% of borrowers FCF-negative, defaults ~5.8%); a stablecoin/Treasury wobble or sovereign (MGX/PIF/SoftBank) pullback. The dated inflections (contractual, staged — not forecasts): (a) Google may terminate the $30B deal as early as Sep 30 2026 if SpaceX misses GPU delivery; (b) either party (Google and Anthropic) can give 90 days' notice after Dec 31 2026 — so the first effective exits could land ~Q1 2027. Renewal pulls SpaceX into the structural core; exercise unwinds the cancelable slice.

Live trigger panel

The watch-list is operationalized with current FRED readings on the charts page: the carry differential (US−Japan 10Y, down from ~3.85pp peak to ~2pp as the JGB rose to ~2.5% — arming), the 2Y−funds gap (~neutral), HY OAS (~cycle lows — complacent), long rates (30Y ~5%, pressuring bank HTM — already reopened to $325B in Q1-26), the AI mark reversal (pending an IPO below the private mark), and the SpaceX deal cliffs (Sep 30 2026 / 90-day notice from Dec 31 2026). The read: fragile and arming, not yet firing — the panel sizes the kindling, it does not date the spark.

Verdict

Structure: certain. Date: unknowable. The honest deliverable is the trigger panel above (including the staged SpaceX cliffs — Sep 30 2026 delivery-miss, 90-day notice from Dec 31 2026) plus the falsifiers in UNMASKING §8 — watch the indicators, not the calendar.

← Research index · structured data: spec-unwind-timing.json · spec-unwind-timing.md