← Cockpit
INF_042predictionMacro/Economyshortage-to-oversupply

Current market panic over AI compute and power will follow the historical pattern 'shortage leads to oversupply' — trillions flowing into data centers, GPU manufacturing, and energy procurement will ultimately crash the marginal cost of both compute an...

Predictor: Marc Andreessen

Prior probability
62.0%
Current probability
47.8%
evolves via intake + LBP
Conviction
4/5
Signal quality
B
Resolution
pending
Window
2027-01-01 – 2032-12-31
Edges in / out
2 / 0
Tickers exposed
30

Prediction text

Current market panic over AI compute and power will follow the historical pattern 'shortage leads to oversupply' — trillions flowing into data centers, GPU manufacturing, and energy procurement will ultimately crash the marginal cost of both compute and power. | First hyperscaler capex-growth deceleration

Key catalyst: First hyperscaler capex-growth deceleration

Watch events: GPU spot-price softening; hyperscaler capex guidance revisions; IRENA / IEA deflation curves

Resolution evidence

Status: pending

Historical precedents (1990s fiber glut, 2000s VC overbuild) validate pattern; contemporary analogues visible in 2026 H2 GPU secondary-market softening and announced 2027-28 hyperscaler fab capacity additions.

Predictor: Marc Andreessen

κ + Brier as of 2026-05-22
κ (discount)
0.500
Brier
Hits / Misses
0 / 0
Hit rate

Evidence about this node from Marc Andreessen 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

Not linked

This node isn't linked to a reference class. The Bayesian update applies without outside-view blending.

Probability over time

3 prob_history rows
0%25%50%75%100%prior 62%2026-04-302026-04-302026-05-02
intake v2milestone miss sweeplbp propagationreference class assignedlegacy v1prior_prob (analyst seed)current = 47.8%

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: 1 overdue ⏱ · 7 pending
  1. 2025-09-30overdueGoldman Sachs 2027 oversupply risk warning published
    How: Goldman Sachs Equity Research formally publishes a forecast warning of long-term data center market oversupply risk starting in 2027, citing efficiency gains + supply additions.
    Source: Goldman Sachs — AI to drive 165% increase in data center power demand by 2030conf 88%
    Notes: HIT — sell-side acknowledgement of the oversupply scenario validates Andreessen's pattern call early.
  2. 2026-12-31pendingHyperscaler 2026 capex grows 36% YoY ($600B) with 2027 deceleration to ~30% growth
    How: Big-5 hyperscaler 2026 capex confirmed at >$600B (36%+ YoY); Moody's or Goldman Sachs publishes 2027 estimate showing decelerating growth rate (e.g., from 52% in 2026 to ~30% in 2027), confirming the inflection point.
    Source: CreditSights — Technology: Hyperscaler Capex 2026 Estimates; IEEE ComSoc Tech Blog — Hyperscaler capex > $600 bn in 2026conf 88%
    Notes: Documented YoY-growth deceleration is the precursor to oversupply per Andreessen's historical-pattern framing.
  3. 2026-09-01 → 2027-12-31pendingData center occupancy peaks ~95% late 2026, moderates starting 2027
    How: JLL, Synergy Research, or DataCenterMap publishes occupancy data showing global DC occupancy peaks at >95% in late 2026, then declines (moderates) starting 2027 as new supply comes online faster than demand.
    Source: Goldman Sachs — AI to drive 165% increase in data center power demand by 2030conf 70%
    Notes: Occupancy is the cleanest leading indicator of the supply/demand inflection between shortage and oversupply.
  4. 2028-01-31pendingQ1 window check-in (25%)
  5. 2027-06-01 → 2029-12-31pendingBlackwell/Rubin GPU spot prices fall >30% from 2026 peak ASP
    How: Secondary-market trackers (e.g., GPU brokers, eBay enterprise listings, OEM channel surveys) report Blackwell or Rubin GPU spot/secondary ASP falling >30% from 2026 peak (~$30K+ per GPU) — even temporarily.
    Source: Anticipated — Liftr Insights, ServeTheHome, channel survey dataconf 50%
    Notes: Direct fulfillment of Andreessen's 'crash the marginal cost of compute' clause. GPU resale price is the cleanest test.
  6. 2029-03-02pendingQ2 window check-in (50%)
  7. 2027-06-01 → 2030-12-31pendingFirst major US power-PPA gets renegotiated downward by hyperscaler
    How: A top-5 hyperscaler publicly renegotiates a major (>500MW) Power Purchase Agreement at lower price, OR a regional grid operator (PJM, ERCOT, MISO) reports declining commercial-contract prices in DC-heavy zones for the first time since 2023.
    Source: Anticipated — utility press releases, PJM/ERCOT capacity-auction resultsconf 40%
    Notes: Cascade — power oversupply mirrors compute oversupply, completing Andreessen's both-shortages-flip claim.
  8. 2030-04-01pendingQ3 window check-in (75%)

No downstream cascades — this prediction is a leaf in the dependency graph.

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: 48%)

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
metadata_milestone_miss_sweep2026-05-02T22:07:21Z47.8%-4.5pp
metadata_milestone_miss_sweep bayesian_v2 n=1 inside=0.478 blend=0.478 LLR=-0.178 κ=0.50 no_blend
Raw metadata
{
  "trf": 1,
  "kappa": 0.5,
  "base_rate": null,
  "predictor": "Marc Andreessen",
  "total_llr": -0.4054651081081644,
  "grace_days": 7,
  "bayesian_v2": true,
  "prior_logit": 0.08845716547823701,
  "bayes_factor": "1.2:1 against",
  "blend_reason": "no reference_class linked",
  "inside_prior": 0.5220998828906968,
  "kappa_source": "predictor_table",
  "n_milestones": 1,
  "blend_applied": false,
  "contributions": [
    {
      "llr": -0.4054651081081644,
      "kind": "llm_pre_event",
      "kappa": 0.44,
      "label": "Goldman Sachs 2027 oversupply risk warning published",
      "weight": 0.4,
      "strength": "weak",
      "confidence": 0.88,
      "source_url": "https://www.goldmansachs.com/insights/articles/ai-to-drive-165-increase-in-data-center-power-demand-by-2030",
      "adjusted_llr": -0.17840464756759233,
      "expected_date": "2025-09-30",
      "measurement_criterion": "Goldman Sachs Equity Research formally publishes a forecast warning of long-term data center market oversupply risk starting in 2027, citing efficiency gains + supply additions."
    }
  ],
  "evidence_kind": "metadata_milestone_miss_sweep",
  "inside_source": "history_v2",
  "inside_weight": 0.3,
  "outside_weight": 0.7,
  "posterior_prob": 0.47752827814999,
  "posterior_logit": -0.08994748208935532,
  "predictor_brier": null,
  "inside_posterior": 0.47752827814999,
  "blended_posterior": 0.47752827814999,
  "reference_class_id": null,
  "total_adjusted_llr": -0.17840464756759233,
  "predictor_n_resolved": 0
}
LBP2026-04-30T16:39:51Z52.2%-3.4pp
Network propagation: 55.6% → 52.2%
5-iter LBP, residual 0.00825 · damping 0.5, w_intrinsic 0.5 · method lbp_v2 · run 0c8a4ea3
LBP2026-04-30T02:18:57Z55.6%-6.4pp
Network propagation: 62.0% → 55.6%
5-iter LBP, residual 0.00825 · damping 0.5, w_intrinsic 0.5 · method lbp_v1 · run 592311ef

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
killerTK02
AI Compute Supply Shock (TSMC/Taiwan Disruption)
12.0%0.0500.620+0.074
killerTK09
Energy Grid Cap (Data Center Power Wall)
35.0%0.0500.620-0.057

Top outgoing (children)

Predictions THIS node influences

No outgoing edges.

Ticker exposure

30 ticker(s) linked

Beneficiaries (22)

ARMAMBADOCNNBISALABIRENTSMCEVASITMCRWVNVDAINTCAMDANETAVGOCSCODELLLNVGYMRVLNXPIQCOMSIEGY

Adverse (4)

XLYXHBHYGKRE

Prerequisites (2)

Predictions that must hit first
TypePredTitleDomainLag
killerTK09Energy Grid Cap (Data Center Power Wall)
killerTK02AI Compute Supply Shock (TSMC/Taiwan Disruption)

Dependents (0)

Predictions enabled by this
TypePredTitleDomainLag
No dependents

Linked documents (8)

Auto-generated by cosine similarity from Polymarket / Manifold / EDGAR / GDELT
SimSourceTitleMarket probPolarityReviewedPublished
0.663gdeltai fears surging crude oil 151510490.htmlmentionspending2026-04-30
0.650manifoldGOOG > AMD at market close on August 31st?33%mentionspending2026-05-26
0.633gdeltmeta shares fall on concerns over ai spending legal scrutiny ce7f58dad08ef227mentionspending2026-04-30
0.592manifoldWill somebody managram me at least 1500 mana before market close?39%mentionspending2026-04-23
0.592manifoldWill somebody managram me at least 1500 mana before market close?51%mentionspending2026-05-03
0.592manifoldWill somebody managram me at least 1500 mana before market close?47%mentionspending2026-06-02
0.574github_releasepytorch/pytorch v2.5.1mentionspending2024-10-29
0.550manifoldWill the stock market crash if the Republicans win the midterms?16%mentionspending2026-05-08

Raw metadata

From Thesis_Timeline_v1.0_FINAL workbook
{
  "nia": false,
  "mode": "FORECAST",
  "role": "Cited-VC",
  "context": "a16z 2026 Outlook frames current capex boom as classical overbuild cycle. Pairs with Huang gigawatt-factory thesis (INF_003) and Altman DC blanket (INF_006) — the oversupply on the other side is the investment thesis.",
  "to_year": 2032,
  "conv_cues": "classical economic law framing; explicit historical analogy",
  "direction": "HAPPEN",
  "from_year": 2027,
  "timeframe": "2027-2032",
  "conv_level": "HIGH",
  "milestones": [
    {
      "kind": "llm_pre_event",
      "label": "Goldman Sachs 2027 oversupply risk warning published",
      "notes": "HIT — sell-side acknowledgement of the oversupply scenario validates Andreessen's pattern call early.",
      "source": "Goldman Sachs — AI to drive 165% increase in data center power demand by 2030",
      "status": "overdue",
      "weight": 0.4,
      "ordinal": -8,
      "source_id": null,
      "confidence": 0.88,
      "source_url": "https://www.goldmansachs.com/insights/articles/ai-to-drive-165-increase-in-data-center-power-demand-by-2030",
      "expected_date": "2025-09-30",
      "miss_emitted_at": "2026-05-02T22:07:21.384228+00:00",
      "miss_emitted_by": "metadata_milestone_sweep",
      "research_origin": "deep_research",
      "measurement_criterion": "Goldman Sachs Equity Research formally publishes a forecast warning of long-term data center market oversupply risk starting in 2027, citing efficiency gains + supply additions."
    },
    {
      "kind": "llm_pre_event",
      "label": "Hyperscaler 2026 capex grows 36% YoY ($600B) with 2027 deceleration to ~30% growth",
      "notes": "Documented YoY-growth deceleration is the precursor to oversupply per Andreessen's historical-pattern framing.",
      "source": "CreditSights — Technology: Hyperscaler Capex 2026 Estimates; IEEE ComSoc Tech Blog — Hyperscaler capex > $600 bn in 2026",
      "status": "pending",
      "weight": 0.4,
      "ordinal": -7,
      "source_id": null,
      "confidence": 0.88,
      "source_url": "https://techblog.comsoc.org/2025/12/22/hyperscaler-capex-600-bn-in-2026-a-36-increase-over-2025-while-global-spending-on-cloud-infrastructure-services-skyrockets/",
      "expected_date": "2026-12-31",
      "research_origin": "deep_research",
      "measurement_criterion": "Big-5 hyperscaler 2026 capex confirmed at >$600B (36%+ YoY); Moody's or Goldman Sachs publishes 2027 estimate showing decelerating growth rate (e.g., from 52% in 2026 to ~30% in 2027), confirming the inflection point."
    },
    {
      "kind": "llm_pre_event",
      "label": "Data center occupancy peaks ~95% late 2026, moderates starting 2027",
      "notes": "Occupancy is the cleanest leading indicator of the supply/demand inflection between shortage and oversupply.",
      "source": "Goldman Sachs — AI to drive 165% increase in data center power demand by 2030",
      "status": "pending",
      "weight": 0.4,
      "ordinal": -6,
      "source_id": null,
      "confidence": 0.7,
      "expected_date": "2027-05-02",
      "research_origin": "deep_research",
      "expected_date_range": {
        "to": "2027-12-31",
        "from": "2026-09-01"
      },
      "measurement_criterion": "JLL, Synergy Research, or DataCenterMap publishes occupancy data showing global DC occupancy peaks at >95% in late 2026, then declines (moderates) starting 2027 as new supply comes online faster than demand."
    },
    {
      "kind": "quartile_checkpoint",
      "label": "Q1 window check-in (25%)",
      "status": "pending",
      "weight": 0.05,
      "ordinal": -5,
      "source_id": null,
      "expected_date": "2028-01-31",
      "observed_date": null
    },
    {
      "kind": "llm_pre_event",
      "label": "Blackwell/Rubin GPU spot prices fall >30% from 2026 peak ASP",
      "notes": "Direct fulfillment of Andreessen's 'crash the marginal cost of compute' clause. GPU resale price is the cleanest test.",
      "source": "Anticipated — Liftr Insights, ServeTheHom
... (truncated)