The chip boom-bust (2 to 4 years)

A 2-4 year industry-scale capex cycle driven by lagged capacity response to demand shocks. Chip-makers see demand spike, commit to multi-year fab build-outs; by the time the fabs come online, demand has adjusted and the industry faces glut. Prices crash; capex halts; next shock catches the industry under-capacity and the cycle restarts. Observables: WSTS monthly billings (demand), WFE book-to-bill (capacity commitment lead indicator), memory-price indices (first to crash in oversupply). Overlaid on this: US-China export-control events discharge surface tension at the national-security grain boundary, extending certain capex cycles while suppressing others.

Rest
The chip boom-bust (2 to 4 years) — long cycleThe chip boom-bust (2 to 4 years): Soft digestion of overcapacity leads at 45% reality support by 2028.BREAKINGPRESSUREREST45%35%20%Prior CapEx peakCOVID chip-shortage peakMemory glut troughAI accelerator surgeNow · 20262028

5 predictions in this cycle lean Perpetual AI-demand ramp on their own questions. Why the cycle still leans Soft digestion of overcapacity

Predictions in this cycle

China property

China's property crisis will deepen through 2027 — without triggering a systemic bank crisis?

Deepens Contained 85 · Spreads Or Stable 15
Reality: A 85%, B 15%. No market signal.

Quiet path · not yet modeled

No market signal

Gap 0Dec 2027
crypto markets

Bitcoin and ETH ETFs will hold $250 billion+ by end of 2027?

ETFs Hold 70 · Falls Below 30
Reality: A 70%, B 30%. No market signal.

Quiet path · not yet modeled

No market signal

Gap 0Dec 2027
geopolitics

Taiwan Strait status quo through end of 2027?

Status Quo 68 · Status Breaks 32
Reality: A 68%, B 32%. No market signal.

Quiet path · not yet modeled

No market signal

Gap 0Dec 2027
geopolitics

Taiwan: Peace or Conflict?

Peace 72 · Conflict 28 · crowd A 87
Reality: A 72%, B 28%. Crowd puts A at 87%. Gap plus 15.

Quiet path · not yet modeled

Gap +15 · crowd 87Dec 2028
tech

AI capability: breakthrough or plateau?

Breakthrough 65 · Plateau 35 · crowd A 70
Reality: A 65%, B 35%. Crowd puts A at 70%. Gap plus 5.

Quiet path · not yet modeled

Gap +5 · crowd 70Dec 2028
tech

A publicly released AI model passes 50% on FrontierMath by end of 2027?

Threshold Hit 78 · Still Short 22
Reality: A 78%, B 22%. No market signal.

Quiet path · not yet modeled

No market signal

Gap 0Dec 2027
finance

NBER declares a US recession with a start date in 2026 or 2027 by end of 2028?

Recession Called 40 · Expansion Holds 60
Reality: A 40%, B 60%. No market signal.

Quiet path · not yet modeled

No market signal

Gap 0Dec 2028
tech

An openly-released AI model holds a top-5 spot on the Chatbot Arena leaderboard at some point during 2027?

Open Model Wins 75 · Closed Dominates 25
Reality: A 75%, B 25%. No market signal.

Quiet path · not yet modeled

No market signal

Gap 0Jan 2028
geopolitics

US adds 50 or more Chinese companies to its chip-related export control list during 2027?

Fifty Plus Added 60 · Fewer Added 40 · crowd A 65
Reality: A 60%, B 40%. Crowd puts A at 65%. Gap plus 5.

Quiet path · not yet modeled

Gap +5 · crowd 65Mar 2028
tech

A publicly-released AI model achieves 85% or higher on SWE-bench Verified at some point during 2027?

Benchmark Cleared 70 · Still Short 30 · crowd A 60
Reality: A 70%, B 30%. Crowd puts A at 60%. Gap minus 10.

Quiet path · not yet modeled

Gap −10 · crowd 60Jan 2028
geopolitics

US notifies Congress of at least $5 billion in cumulative arms sales to Taiwan during calendar years 2026 and 2027?

Arms Flow 60 · Arms Slow 40 · crowd A 55
Reality: A 60%, B 40%. Crowd puts A at 55%. Gap minus 5.

Quiet path · not yet modeled

Gap −5 · crowd 55Jan 2028
geopolitics

US average effective tariff rate on imports from China stays above 30 percent through the end of 2027?

Tariff Stays High 72 · Tariff Falls 28 · crowd A 65
Reality: A 72%, B 28%. Crowd puts A at 65%. Gap minus 7.

Quiet path · not yet modeled

Gap −7 · crowd 65Mar 2028

Pressure moments

All →

Movers

All →
Model notes (framework)

The reconciliation note surfaces when this cycle’s top fork disagrees with the lean of its live child predictions — a sign the cycle’s read is tightening or breaking down.

See the method for how a cycle’s forks are set.