← Cockpit
CMQ_039predictionSemis/PackagingTSMC-CoWoS

TSMC CoWoS (Chip-on-Wafer-on-Substrate) capacity must expand to 125,000 wafers/month by 2026 to meet baseline demand from NVIDIA, AMD, and hyperscaler ASICs.

Predictor: Morgan Stanley

Prior probability
80.0%
Current probability
70.9%
evolves via intake + LBP
Conviction
5/5
Signal quality
A
Resolution
pending
Window
2026-01-01 – 2026-12-31
Edges in / out
2 / 0
Tickers exposed
28

Prediction text

TSMC CoWoS (Chip-on-Wafer-on-Substrate) capacity must expand to 125,000 wafers/month by 2026 to meet baseline demand from NVIDIA, AMD, and hyperscaler ASICs. | TSMC CoWoS monthly capacity disclosures

Key catalyst: TSMC CoWoS monthly capacity disclosures

Watch events: TSMC CoWoS monthly capacity; advanced-packaging equipment shipments.

Resolution evidence

Status: pending

TSMC CoWoS capacity ~70k WPM end-2025; on-track expansion toward 120-130k WPM end-2026 per TSMC capex guidance.

Predictor: Morgan Stanley

κ + Brier as of 2026-05-22
κ (discount)
0.633
Brier
0.0442
excellent
Hits / Misses
1 / 0
of 2 resolved
Hit rate
50.0%
Calibration plot (stated vs observed)

Evidence about this node from Morgan Stanley is multiplied by κ in /api/intake. Lower κ = less weight; floors at 0.10 (effectively silenced) and caps at 1.00 (full weight).

Reference class: fab_construction_3y

Linked via embedding similarity 0.592

Greenfield semiconductor fab construction within 3 years of announce

Base rate
58.0%
7/12 historical
Inside weight
0.571
TRF=0.61
Outside weight
0.429
pulling toward base rate
inside 78.9% → blend 70.9% -8.0pp)

Tetlock-style outside view: at TRF=1 (just predicted), outside view dominates (w_in=0.3). At TRF=0 (deadline), inside view dominates (w_in=1.0). The blend regularizes overconfident inside views toward the historical base rate.

Probability over time

5 prob_history rows
0%25%50%75%100%prior 80%2026-04-302026-05-022026-05-21
intake v2milestone miss sweeplbp propagationreference class assignedlegacy v1prior_prob (analyst seed)current = 70.9%

Milestone chain

Pre-event signals (upstream prereqs + window checkpoints) → resolution event → downstream cascades. Status/dates update from linked nodes; re-derive nightly via scripts/ops/derive_milestones.py.
Leading chain: 4 fired ✓ · 2 overdue ⏱ · 1 pending
  1. 2025-12-10hitNVIDIA secures ≥50% of TSMC CoWoS capacity for 2026-27
    How: DigiTimes / Wedbush / Astute Group reporting confirms NVIDIA holds 50%+ of TSMC's projected CoWoS capacity for 2026-27
    Source: DigiTimes — TSMC expands CoWoS capacity with Nvidia booking over half for 2026-27conf 97%
  2. 2026-01-01hitTSMC pivots toward CoWoS-L using Local Silicon Interconnect bridge architecture
    How: TSMC publicly confirms CoWoS-L (LSI Bridge) as core packaging technology to enable larger AI chips beyond reticle limit
    Source: TokenRing/FinancialContent — The Great Packaging Pivot (Jan 2026)conf 95%
  3. 2026-01-20hitAMD allocated ~11% of CoWoS capacity for MI350/MI400 ramp
    How: AMD's CoWoS allocation publicly disclosed at ~11% (vs NVIDIA 50%+, Broadcom remainder)
    Source: Wedbush — The Great AI Packaging Squeezeconf 85%
  4. 2026-03-07overdueQ1 window check-in (25%)
  5. 2026-04-08hitTSMC US-Arizona advanced packaging capacity expansion announced
    How: TSMC announces US-based advanced packaging line expansion specifically for AI customers
    Source: CNBC — Nvidia snaps up AI chip packaging capacity as TSMC expands in U.S.conf 85%
  6. 2026-05-11overdueQ2 window check-in (50%)
  7. 2026-07-15pendingQ3 window check-in (75%)
  8. 2026-10-01 → 2026-12-31pendingTSMC monthly CoWoS capacity reaches 127,000-130,000 wafers by end-2026
    How: TSMC monthly disclosures or DigiTimes/Tiger Brokers reporting confirm CoWoS monthly capacity ≥125,000 wafers by year-end 2026
    Source: FinancialContent — TSMC to Quadruple Advanced Packaging Capacity to 130,000 CoWoS Wafers Monthly by Late 2026 (Feb 2026)conf 90%
    Notes: Direct measurement of 125K wafers/month claim. Reuters/DigiTimes specifically cite 127,000 figure.
  9. 2026-06-01 → 2027-12-31pendingCoWoS capacity remains binding constraint causing AI accelerator shortages through 2027
    How: Industry analyst reports continue to cite CoWoS as primary AI accelerator capacity constraint through 2027
    Source: FusionWW — Inside the AI Bottleneck: CoWoS, HBM, and 2-3nm Capacity Constraints Through 2027conf 70%

What if this resolves?

Clamp this prediction TRUE or FALSE and run a counterfactual Gibbs sample. Surfaces the predictions whose marginals shift most under that assumption.
(live posterior: 71%)

Click a button to clamp this prediction and run a Gibbs sample. Returns the predictions whose marginals shift most. ~30s per run; ideal for stress-testing "if X resolves, what else moves?"

Evidence chain

Every probability update with full Bayesian provenance — chronological, latest first
intake_event_update2026-05-21T23:15:16Z70.9%-0.5pp
intake:7afeeb9a-f217-4dd2-b910-24ff14bdfc39 bayesian_v2 inside=0.789 blend=0.709 LLR=0.404 κ=0.58 w_in=0.57 fab_construction_3y
Raw metadata
{
  "trf": 0.6127227050582519,
  "kappa": 0.5833,
  "base_rate": 0.58,
  "predictor": "Morgan Stanley",
  "total_llr": 0.6931471805599453,
  "bayesian_v2": true,
  "prior_logit": 0.9132787660009067,
  "bayes_factor": "1.5:1 favoring",
  "blend_reason": "blend 57% inside / 43% outside (TRF=0.613, base_rate=0.580 from fab_construction_3y)",
  "inside_prior": 0.713670630849881,
  "kappa_source": "predictor_table",
  "blend_applied": true,
  "contributions": [
    {
      "llr": 0.6931471805599453,
      "kappa": 0.5833,
      "label": "CoWoS expansion remains the binding packaging constraint; lead times still extreme.",
      "adjusted_llr": 0.4043127504206161
    }
  ],
  "evidence_kind": "intake_event_update",
  "inside_source": "history_v2",
  "inside_weight": 0.5710941064592236,
  "outside_weight": 0.4289058935407764,
  "posterior_prob": 0.7090775493441308,
  "evidence_origin": "daily_intake",
  "llm_suggestions": [
    {
      "polarity": "corroborates",
      "status_change": "in_progress",
      "evidence_strength": "moderate",
      "delta_prob_suggestion": 0.03
    }
  ],
  "posterior_logit": 1.317591516421523,
  "predictor_brier": 0.01,
  "evidence_doc_ids": [],
  "inside_posterior": 0.7887807185283536,
  "blended_posterior": 0.7090775493441308,
  "reference_class_id": "fab_construction_3y",
  "total_adjusted_llr": 0.4043127504206161,
  "predictor_n_resolved": 1
}
metadata_milestone_miss_sweep2026-05-19T22:16:06Z65.9%-2.4pp
metadata_milestone_miss_sweep bayesian_v2 n=1 inside=0.714 blend=0.659 LLR=-0.237 κ=0.58 w_in=0.57 fab_construction_3y
Raw metadata
{
  "trf": 0.6183300713972515,
  "kappa": 0.5833,
  "base_rate": 0.58,
  "predictor": "Morgan Stanley",
  "total_llr": -0.4054651081081644,
  "grace_days": 7,
  "bayesian_v2": true,
  "prior_logit": 1.149786563560399,
  "bayes_factor": "1.3:1 against",
  "blend_reason": "blend 56% inside / 43% outside (TRF=0.618, base_rate=0.580 from fab_construction_3y)",
  "inside_prior": 0.7594719297524297,
  "kappa_source": "predictor_table",
  "n_milestones": 1,
  "blend_applied": true,
  "contributions": [
    {
      "llr": -0.4054651081081644,
      "kind": "quartile_checkpoint",
      "kappa": 0.5833,
      "label": "Q2 window check-in (50%)",
      "weight": 0.05,
      "strength": "weak",
      "confidence": null,
      "source_url": null,
      "adjusted_llr": -0.2365077975594923,
      "expected_date": "2026-05-11",
      "measurement_criterion": null
    }
  ],
  "evidence_kind": "metadata_milestone_miss_sweep",
  "inside_source": "history_v2",
  "inside_weight": 0.5671689500219239,
  "outside_weight": 0.4328310499780761,
  "posterior_prob": 0.6587412219514239,
  "posterior_logit": 0.9132787660009067,
  "predictor_brier": 0.01,
  "inside_posterior": 0.713670630849881,
  "blended_posterior": 0.6587412219514239,
  "reference_class_id": "fab_construction_3y",
  "total_adjusted_llr": -0.2365077975594923,
  "predictor_n_resolved": 1
}
metadata_milestone_miss_sweep2026-05-02T22:07:21Z68.2%-2.6pp
metadata_milestone_miss_sweep bayesian_v2 n=1 inside=0.759 blend=0.682 LLR=-0.237 κ=0.58 w_in=0.53 fab_construction_3y
Raw metadata
{
  "trf": 0.6650500679109432,
  "kappa": 0.5833,
  "base_rate": 0.58,
  "predictor": "Morgan Stanley",
  "total_llr": -0.4054651081081644,
  "grace_days": 7,
  "bayesian_v2": true,
  "prior_logit": 1.3862943611198908,
  "bayes_factor": "1.3:1 against",
  "blend_reason": "blend 53% inside / 46% outside (TRF=0.665, base_rate=0.580 from fab_construction_3y)",
  "inside_prior": 0.8,
  "kappa_source": "predictor_table",
  "n_milestones": 1,
  "blend_applied": true,
  "contributions": [
    {
      "llr": -0.4054651081081644,
      "kind": "quartile_checkpoint",
      "kappa": 0.5833,
      "label": "Q1 window check-in (25%)",
      "weight": 0.05,
      "strength": "weak",
      "confidence": null,
      "source_url": null,
      "adjusted_llr": -0.2365077975594923,
      "expected_date": "2026-03-07",
      "measurement_criterion": null
    }
  ],
  "evidence_kind": "metadata_milestone_miss_sweep",
  "inside_source": "history_v2",
  "inside_weight": 0.5344649524623397,
  "outside_weight": 0.4655350475376603,
  "posterior_prob": 0.6823912615332146,
  "posterior_logit": 1.1497865635603985,
  "predictor_brier": 0.01,
  "inside_posterior": 0.7594719297524297,
  "blended_posterior": 0.6823912615332146,
  "reference_class_id": "fab_construction_3y",
  "total_adjusted_llr": -0.2365077975594923,
  "predictor_n_resolved": 1
}
legacy v12026-04-30T16:13:50Z70.8%+0.0pp
reference_class_assigned bayesian_v2 inside=0.800 blend=0.708 w_in=0.53 fab_construction_3y
legacy v12026-04-30T01:56:50Z70.8%-9.2pp
reference_class_assigned bayesian_v2 inside=0.800 blend=0.708 w_in=0.53 fab_construction_3y

Network propagation neighbors

Top edges sorted by latest LBP cross-impact
All propagation →

Top incoming (parents)

Edges that influence THIS node's belief

KindNodeTheir probP(c|s=T)P(c|s=F)Δ implied
killerTK01
AGI Capability Plateau (2026-27 Training Stall)
15.0%0.0500.800-0.022
killerTK02
AI Compute Supply Shock (TSMC/Taiwan Disruption)
12.0%0.0500.800+0.001

Top outgoing (children)

Predictions THIS node influences

No outgoing edges.

Ticker exposure

28 ticker(s) linked

Beneficiaries (24)

PLABASMIYASMLIRENALABCAMTAEHRNBISNVMINVDATSMCRWVTXNAMATAMDANETATEYYDELLINTCKLACLNVGYLRCXTERTOELY

Prerequisites (2)

Predictions that must hit first
TypePredTitleDomainLag
killerTK01AGI Capability Plateau (2026-27 Training Stall)
killerTK02AI Compute Supply Shock (TSMC/Taiwan Disruption)

Dependents (0)

Predictions enabled by this
TypePredTitleDomainLag
No dependents

Linked documents (2)

Auto-generated by cosine similarity from Polymarket / Manifold / EDGAR / GDELT
SimSourceTitleMarket probPolarityReviewedPublished
0.647manifoldWhen will Codex hit 6M WAUs?mentionspending2026-06-01
0.578edgar_8kMOSAIC CO (MOS) (CIK 0001285785)mentionspending2026-05-29

Raw metadata

From Thesis_Timeline_v1.0_FINAL workbook
{
  "nia": false,
  "qty": "125k wafers/mo",
  "mode": "FORECAST",
  "role": "Cited-Firm",
  "context": "Advanced packaging is the second binding constraint behind HBM; CoWoS is the architectural glue for AI silicon.",
  "to_year": 2026,
  "cited_by": "Synthesis report",
  "conv_cues": "specific capacity; MS research",
  "direction": "NUMERIC_TARGET",
  "from_year": 2026,
  "timeframe": "by 2026",
  "conv_level": "HIGH",
  "milestones": [
    {
      "kind": "llm_pre_event",
      "label": "NVIDIA secures ≥50% of TSMC CoWoS capacity for 2026-27",
      "source": "DigiTimes — TSMC expands CoWoS capacity with Nvidia booking over half for 2026-27",
      "status": "hit",
      "weight": 0.4,
      "ordinal": -7,
      "source_id": null,
      "confidence": 0.97,
      "source_url": "https://www.digitimes.com/news/a20251210PD218/tsmc-cowos-capacity-nvidia-equipment.html",
      "expected_date": "2025-12-10",
      "observed_date": "2025-12-10",
      "research_origin": "deep_research",
      "measurement_criterion": "DigiTimes / Wedbush / Astute Group reporting confirms NVIDIA holds 50%+ of TSMC's projected CoWoS capacity for 2026-27"
    },
    {
      "kind": "llm_pre_event",
      "label": "TSMC pivots toward CoWoS-L using Local Silicon Interconnect bridge architecture",
      "source": "TokenRing/FinancialContent — The Great Packaging Pivot (Jan 2026)",
      "status": "hit",
      "weight": 0.4,
      "ordinal": -6,
      "source_id": null,
      "confidence": 0.95,
      "source_url": "https://markets.financialcontent.com/wral/article/tokenring-2026-1-1-the-great-packaging-pivot-how-tsmc-is-doubling-cowos-capacity-to-break-the-ai-supply-bottleneck-through-2026",
      "expected_date": "2026-01-01",
      "observed_date": "2026-01-01",
      "research_origin": "deep_research",
      "measurement_criterion": "TSMC publicly confirms CoWoS-L (LSI Bridge) as core packaging technology to enable larger AI chips beyond reticle limit"
    },
    {
      "kind": "llm_pre_event",
      "label": "AMD allocated ~11% of CoWoS capacity for MI350/MI400 ramp",
      "source": "Wedbush — The Great AI Packaging Squeeze",
      "status": "hit",
      "weight": 0.4,
      "ordinal": -5,
      "source_id": null,
      "confidence": 0.85,
      "source_url": "https://investor.wedbush.com/wedbush/article/tokenring-2026-1-20-the-great-ai-packaging-squeeze-nvidia-secures-50-of-tsmc-capacity-as-sk-hynix-breaks-ground-on-p-and-t7",
      "expected_date": "2026-01-20",
      "observed_date": "2026-01-20",
      "research_origin": "deep_research",
      "measurement_criterion": "AMD's CoWoS allocation publicly disclosed at ~11% (vs NVIDIA 50%+, Broadcom remainder)"
    },
    {
      "kind": "quartile_checkpoint",
      "label": "Q1 window check-in (25%)",
      "status": "overdue",
      "weight": 0.05,
      "ordinal": -4,
      "source_id": null,
      "expected_date": "2026-03-07",
      "observed_date": null,
      "miss_emitted_at": "2026-05-02T22:07:21.384228+00:00",
      "miss_emitted_by": "metadata_milestone_sweep"
    },
    {
      "kind": "llm_pre_event",
      "label": "TSMC US-Arizona advanced packaging capacity expansion announced",
      "source": "CNBC — Nvidia snaps up AI chip packaging capacity as TSMC expands in U.S.",
      "status": "hit",
      "weight": 0.4,
      "ordinal": -3,
      "source_id": null,
      "confidence": 0.85,
      "source_url": "https://www.cnbc.com/2026/04/08/tsmc-nvidia-advanced-packaging-intel.html",
      "expected_date": "2026-04-08",
      "observed_date": "2026-04-08",
      "research_origin": "deep_research",
      "measurement_criterion": "TSMC announces US-based advanced packaging line expansion specifically for AI customers"
    },
    {
      "kind": "quartile_checkpoint",
      "label": "Q2 window check-in (50%)",
      "status": "overdue",
      "weight": 0.05,
      "ordinal": -2,
      "source_id": null,
      "expected_date": "2026-05-11",
      "observed_date": null,
      "miss_emitted_at":
... (truncated)