The entire global installed base of data centers must be ripped out and replaced — legacy DCs unfit for AI factory workloads.
Predictor: Jensen Huang
Prediction text
The entire global installed base of data centers must be ripped out and replaced — legacy DCs unfit for AI factory workloads. | Hyperscaler DC retirement + greenfield announcements
Key catalyst: Hyperscaler DC retirement + greenfield announcements
Watch events: Hyperscaler DC retirement announcements; liquid-cooling penetration rates.
Resolution evidence
Hyperscalers retiring >10-year-old DC capacity aggressively 2025-2026; liquid-cooling retrofits and high-density rack deployments accelerating.
Predictor: Jensen Huang
Calibration plot (stated vs observed)
Evidence about this node from Jensen Huang 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
This node isn't linked to a reference class. The Bayesian update applies without outside-view blending.
Probability over time
Milestone chain
- 2026-03-18overdueNvidia Kyber rack architecture announced (144 GPUs, 800VDC, vertically oriented trays)How: At GTC 2026, Nvidia formally announces Kyber rack architecture: 144 GPUs in vertically oriented compute trays, 800VDC power distribution replacing legacy 208VAC, designed for 150-300kW per rack — explicitly incompatible with legacy data center electrical infrastructure.Source: DCD — GTC: Nvidia's Jensen Huang, Ian Buck, and Charlie Boyle on the future of data center rack densityconf 95%Notes: HIT — direct technical proof that legacy DCs (208VAC racks) cannot support next-gen architectures, supporting the rip-and-replace thesis.
- 2026-06-01 → 2026-12-31pendingNvidia Vera Rubin NVL72 ships at 35x token throughput vs Hopper at equivalent powerHow: Nvidia begins shipping Vera Rubin NVL72 (72 Rubin GPUs + 36 Vera CPUs via NVLink) with documented 35x token-throughput improvement over Hopper at equivalent power; first hyperscaler customer publicly confirms deployment.Source: Tom's Hardware — Nvidia GTC 2026 keynote live blog — Vera Rubin GPUsconf 85%Notes: Hardware shipment is the trigger that forces hyperscalers to rebuild legacy DC capacity — economic obsolescence becomes physical.
- 2026-12-31pendingHyperscaler total capex hits $600B in 2026 (36% YoY), with explicit DC retirement disclosuresHow: Combined Big-5 hyperscaler (AMZN/MSFT/GOOG/META/ORCL) capital expenditure crosses $600B in 2026 (36%+ YoY growth), with at least 2 hyperscalers disclosing legacy DC retirement / decommission programs in 10-Ks.Source: Introl Blog — Hyperscaler CapEx Hits $600B in 2026; CreditSights — Hyperscaler Capex 2026 Estimatesconf 88%Notes: Capex magnitude + retirement disclosures together demonstrate Jensen's 'rip-out and replace' framing materializing in financials.
- 2027-05-02pendingQ1 window check-in (25%)
- 2026-09-01 → 2029-06-30pendingMajor hyperscaler (top-5) publicly announces decommissioning/retirement of >1GW pre-2024 DC capacityHow: A top-5 hyperscaler (AWS, Azure, GCP, Meta, Oracle) publicly announces decommissioning or wholesale rebuild of >1GW of pre-2024 (Hopper-or-older era) data center capacity in a single press release or 10-K disclosure.Source: Anticipated — hyperscaler 10-K filings, earnings calls, press releasesconf 55%Notes: Direct test of the 'entire global installed base ripped out and replaced' clause at single-firm scale.
- 2028-08-30pendingQ2 window check-in (50%)
- 2027-06-01 → 2031-12-31pendingIndustry analyst firm reports legacy CPU-era DC stranded-asset writedowns >$50B aggregateHow: A major industry analyst (Gartner, IDC, JLL, Synergy Research) publishes a report documenting cumulative stranded-asset writedowns or impairment charges on legacy CPU-era data center capacity exceeding $50B aggregate across the industry.Source: Anticipated — Gartner, IDC, Synergy Research, JLL Data Center Outlookconf 45%Notes: Cascade — financial confirmation that legacy DCs were structurally obsoleted, not merely supplemented.
- 2029-12-29pendingQ3 window check-in (75%)
No downstream cascades — this prediction is a leaf in the dependency graph.
What if this resolves?
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
Raw metadata
{
"trf": 0.9522997749294145,
"kappa": 0.8077,
"base_rate": null,
"predictor": "Jensen Huang",
"total_llr": -0.4054651081081644,
"grace_days": 7,
"bayesian_v2": true,
"prior_logit": 0.4084959905205715,
"bayes_factor": "1.4:1 against",
"blend_reason": "no reference_class linked",
"inside_prior": 0.6007271908193117,
"kappa_source": "predictor_table",
"n_milestones": 1,
"blend_applied": false,
"contributions": [
{
"llr": -0.4054651081081644,
"kind": "llm_pre_event",
"kappa": 0.767315,
"label": "Nvidia Kyber rack architecture announced (144 GPUs, 800VDC, vertically oriented trays)",
"weight": 0.4,
"strength": "weak",
"confidence": 0.95,
"source_url": "https://www.datacenterdynamics.com/en/analysis/nvidia-gtc-jensen-huang-data-center-rack-density/",
"adjusted_llr": -0.31111945942801617,
"expected_date": "2026-03-18",
"measurement_criterion": "At GTC 2026, Nvidia formally announces Kyber rack architecture: 144 GPUs in vertically oriented compute trays, 800VDC power distribution replacing legacy 208VAC, designed for 150-300kW per rack — explicitly incompatible with legacy data center electrical infrastructure."
}
],
"evidence_kind": "metadata_milestone_miss_sweep",
"inside_source": "history_v2",
"inside_weight": 0.3333901575494098,
"outside_weight": 0.6666098424505902,
"posterior_prob": 0.5243249146906058,
"posterior_logit": 0.09737653109255534,
"predictor_brier": 0.01276,
"inside_posterior": 0.5243249146906058,
"blended_posterior": 0.5243249146906058,
"reference_class_id": null,
"total_adjusted_llr": -0.31111945942801617,
"predictor_n_resolved": 8
}Network propagation neighbors
Top incoming (parents)
Edges that influence THIS node's belief
| Kind | Node | Their prob | P(c|s=T) | P(c|s=F) | Δ implied |
|---|---|---|---|---|---|
| killer | TK09 Energy Grid Cap (Data Center Power Wall) | 35.0% | 0.050 | 0.780 | +0.034 |
Top outgoing (children)
Predictions THIS node influences
No outgoing edges.
Prerequisites (2)
| Type | Pred | Title | Domain | Lag |
|---|---|---|---|---|
| correlate | S_COMPUTE_100GW_2030 | Compute: 100GW national-scale by Dec 2030 | compute_scale | — |
| killer | TK09 | Energy Grid Cap (Data Center Power Wall) | — | — |
Dependents (0)
| Type | Pred | Title | Domain | Lag |
|---|---|---|---|---|
| No dependents | ||||
Validations (1)
| Observed at | Status | By | Notes |
|---|---|---|---|
| 2026-04-29 | partial | thesis_timeline_v1.0_import | Hyperscalers retiring >10-year-old DC capacity aggressively 2025-2026; liquid-cooling retrofits and high-density rack deployments accelerating. |
Linked documents (8)
| Sim | Source | Title | Market prob | Polarity | Reviewed | Published |
|---|---|---|---|---|---|---|
| 0.610 | github_release | tensorflow/tensorflow v2.20.0 | — | mentions | pending | 2025-08-13 |
| 0.607 | github_release | tensorflow/tensorflow v2.20.0-rc0 | — | mentions | pending | 2025-07-28 |
| 0.598 | gdelt | article f2a47492 8d06 57cd 8c0a 7492330da42b.html | — | mentions | pending | 2026-04-30 |
| 0.586 | gdelt | article ddcab03b 2b7f 5303 8341 e1e8f4863883.html | — | mentions | pending | 2026-04-30 |
| 0.573 | gdelt | article ca552487 4402 5a81 b6d9 dc82aa416d0a.html | — | mentions | pending | 2026-04-30 |
| 0.570 | github_release | tensorflow/tensorflow v2.17.1 | — | mentions | pending | 2024-10-24 |
| 0.565 | github_release | tensorflow/tensorflow v2.16.2 | — | mentions | pending | 2024-06-28 |
| 0.545 | edgar_8k | GoDaddy Inc. (GDDY) (CIK 0001609711) | — | mentions | pending | 2026-06-05 |
Raw metadata
{
"nia": false,
"qty": "full DC-fleet refresh",
"mode": "FORECAST",
"role": "Cited-Executive",
"context": "Implies multi-trillion-dollar replacement cycle for ~500GW of global DC capacity.",
"to_year": 2032,
"conv_cues": "must be ripped out; CEO",
"direction": "HAPPEN",
"from_year": 2026,
"timeframe": "this decade",
"conv_level": "HIGH",
"milestones": [
{
"kind": "llm_pre_event",
"label": "Nvidia Kyber rack architecture announced (144 GPUs, 800VDC, vertically oriented trays)",
"notes": "HIT — direct technical proof that legacy DCs (208VAC racks) cannot support next-gen architectures, supporting the rip-and-replace thesis.",
"source": "DCD — GTC: Nvidia's Jensen Huang, Ian Buck, and Charlie Boyle on the future of data center rack density",
"status": "overdue",
"weight": 0.4,
"ordinal": -9,
"source_id": null,
"confidence": 0.95,
"source_url": "https://www.datacenterdynamics.com/en/analysis/nvidia-gtc-jensen-huang-data-center-rack-density/",
"expected_date": "2026-03-18",
"miss_emitted_at": "2026-05-02T22:07:21.384228+00:00",
"miss_emitted_by": "metadata_milestone_sweep",
"research_origin": "deep_research",
"measurement_criterion": "At GTC 2026, Nvidia formally announces Kyber rack architecture: 144 GPUs in vertically oriented compute trays, 800VDC power distribution replacing legacy 208VAC, designed for 150-300kW per rack — explicitly incompatible with legacy data center electrical infrastructure."
},
{
"kind": "llm_pre_event",
"label": "Nvidia Vera Rubin NVL72 ships at 35x token throughput vs Hopper at equivalent power",
"notes": "Hardware shipment is the trigger that forces hyperscalers to rebuild legacy DC capacity — economic obsolescence becomes physical.",
"source": "Tom's Hardware — Nvidia GTC 2026 keynote live blog — Vera Rubin GPUs",
"status": "pending",
"weight": 0.4,
"ordinal": -8,
"source_id": null,
"confidence": 0.85,
"source_url": "https://www.tomshardware.com/news/live/nvidia-gtc-2026-keynote-live-blog-jensen-huang",
"expected_date": "2026-09-15",
"research_origin": "deep_research",
"expected_date_range": {
"to": "2026-12-31",
"from": "2026-06-01"
},
"measurement_criterion": "Nvidia begins shipping Vera Rubin NVL72 (72 Rubin GPUs + 36 Vera CPUs via NVLink) with documented 35x token-throughput improvement over Hopper at equivalent power; first hyperscaler customer publicly confirms deployment."
},
{
"kind": "llm_pre_event",
"label": "Hyperscaler total capex hits $600B in 2026 (36% YoY), with explicit DC retirement disclosures",
"notes": "Capex magnitude + retirement disclosures together demonstrate Jensen's 'rip-out and replace' framing materializing in financials.",
"source": "Introl Blog — Hyperscaler CapEx Hits $600B in 2026; CreditSights — Hyperscaler Capex 2026 Estimates",
"status": "pending",
"weight": 0.4,
"ordinal": -7,
"source_id": null,
"confidence": 0.88,
"source_url": "https://introl.com/blog/hyperscaler-capex-600b-2026-ai-infrastructure-debt-january-2026",
"expected_date": "2026-12-31",
"research_origin": "deep_research",
"measurement_criterion": "Combined Big-5 hyperscaler (AMZN/MSFT/GOOG/META/ORCL) capital expenditure crosses $600B in 2026 (36%+ YoY growth), with at least 2 hyperscalers disclosing legacy DC retirement / decommission programs in 10-Ks."
},
{
"kind": "quartile_checkpoint",
"label": "Q1 window check-in (25%)",
"status": "pending",
"weight": 0.05,
"ordinal": -6,
"source_id": null,
"expected_date": "2027-05-02",
"observed_date": null
},
{
"kind": "llm_pre_event",
"label": "Major hyperscaler (top-5) publicly announces decommissioning/retirement of >1GW pre-2024 DC capacity",
"notes": "Direct test o
... (truncated)