Ongoing · Technology / Energy
These are algorithmically-created hypotheses — not forecasts.
The central question is how long the compute and power bottleneck constrains AI buildout. The branches suggest a sustained supply crunch — extended GPU lead times and power-grid lag — is the most plausible path, with stricter regulatory constraint as a secondary brake and an efficiency breakthrough that relieves the bottleneck as the lower-probability scenario. Resolution likely depends on fab and grid capacity additions and on whether a step-change in model efficiency materialises.
Authored 2026-05-21 · OpenWatch editorial
Set at 80% anchored to public hyperscaler capex commitments: Microsoft, Alphabet, Amazon, and Meta collectively guided $250B+ in 2025 infrastructure spending, with NVIDIA H100/H200/B200 lead times extending 6–12 months as of Q1 2025. SemiAnalysis GPU supply models project HBM memory and CoWoS packaging as the binding constraints through at least mid-2026. Set high — not at 90%+ — because demand pull-forward from inference efficiency improvements (Deepseek-class innovations) remains the principal downside path.
Two consecutive Nvidia datacenter-segment earnings prints below consensus by more than 15%, paired with public hyperscaler capex guidance cuts of 10%+ — would refute the "sustained supply crunch" framing and signal the AI-infrastructure cycle has peaked.
Each branch below shows the most likely ways this plays out — with its own winners, losers, and supporting signals.
View possible paths ↓Not investment advice. Always verify independently with a qualified financial advisor.
Public prediction markets matched by AI to this scenario — agree or disagree, the bet is yours. OpenWatch does not recommend any position.
Frontier model capability milestone (Chatbot Arena score ≥1600) directly measures advancement in frontier AI models, a core component of infrastructure-driven model race dynamics.
AI industry downturn triggered by multiple events within 90 days—directly captures capex overshoot and crash scenario. NVIDIA underperformance is one of the specified resolution criteria.
EU AI Act enforcement action against frontier AI labs directly measures European regulatory implementation and compliance mechanisms central to US-EU regulatory divergence on AI infrastructure.
NVIDIA's AI accelerator revenue share exceeding 80% in H2 2026 directly quantifies monopoly deepening in GPU/accelerator markets, the core infrastructure constraint.
Quarterly capex spend among AI hyperscalers decreases before 2028, directly capturing the capex-cycle contraction phase following infrastructure overbuild in the ai-infrastructure scenario.
Frontier models and compute allocation post-April 2026 directly tracks the infrastructure race outcome and model deployment strategy within the frontier model race.
Market prices are raw values. Political contracts may exhibit favourite-longshot bias.
If this scenario occurs — possible paths
Signal counts measure media attention over the last 7 days — not the likelihood of an outcome.
Branch % = conditional on this scenario occurring · Path % = joint probability of this exact path from today
Trade lens —NVDA and ASML sustain pricing power; data-center and nuclear-PPA names bid on baseload scarcity; INTC AI-chip share compresses. · meaningful · slow
Policy lens —FERC opens an emergency rulemaking on data-centre power interconnection priority; the Commerce Department publishes updated AI Export Administration Regulations restricting advanced-node GPU exports; the EU AI Office activates its first systemic-risk designation under the AI Act.
AI-matched prediction markets — agree or disagree, the decision is yours. Clicking opens the provider's site.
NVIDIA FY2028 revenue exceeding $400B depends directly on sustained AI infrastructure demand and GPU capex cycles. Supply constraints or chip shortages would materially impact revenue trajectory.
AI bubble pop scenario aligns with sustained infrastructure supply crunch limiting scaling, depressing capex ROI, and triggering sector contraction. GPU/data center shortages would constrain AI company growth and investm
Supply-side constraints on chip availability directly intersect NVIDIA export policy and China access disputes. Sustained crunch amplifies geopolitical leverage over chip allocation and increases likelihood of policy fli
NVIDIA share price momentum through 2028 depends on sustained AI capex cycles and GPU shipment growth; supply constraints would compress valuations and near-term returns.
Supply constraints on GPU and data center components would directly reduce quarterly capex spend among hyperscalers dependent on chip availability.
For entertainment and research purposes only. OpenWatch tracks trends and signals — not real-time prices. Data updates every 4 hours. The forecasting algorithm is currently undergoing back testing, and we do not recommend any position. All trading decisions are solely your responsibility.
Markets are matched to OpenWatch scenarios by an AI worker that runs every 4 hours. New markets and price changes may not be reflected immediately.
Trade lens —AI-compliance names (PANW) and incumbents (MSFT) capture moat; NVDA capex pace decelerates at the margin; European industrial-AI adoption slows. · meaningful · slow
Policy lens —The EU AI Office activates systemic-risk obligations for GPAI model providers under the AI Act; the White House issues an updated Executive Order on AI safety incorporating mandatory model-evaluation thresholds; NIST publishes binding AI Risk Management Framework standards for critical-sector deployment.
AI-matched prediction markets — agree or disagree, the decision is yours. Clicking opens the provider's site.
EU AI Act enforcement action against frontier AI lab directly operationalizes regulatory constraints on AI infrastructure through compliance mechanisms and legal consequences.
Presidential executive order on open-source AI model security risks establishes regulatory compliance requirements affecting AI infrastructure access and deployment.
U.S. legislation or executive order addressing AI risks establishes regulatory constraints governing AI infrastructure operations and compliance obligations.
U.S. AI regulation for Mythos+ models in 2026 imposes compliance requirements on advanced AI infrastructure systems.
U.S. AI safety statute or executive order creates regulatory constraints on domestic AI infrastructure deployment and development.
For entertainment and research purposes only. OpenWatch tracks trends and signals — not real-time prices. Data updates every 4 hours. The forecasting algorithm is currently undergoing back testing, and we do not recommend any position. All trading decisions are solely your responsibility.
Markets are matched to OpenWatch scenarios by an AI worker that runs every 4 hours. New markets and price changes may not be reflected immediately.
Trade lens —NVDA and ASML repriced lower on demand-shock; application-layer names (MSFT) capture value; Taiwan/Netherlands semis-capex multiple compresses. · structural · slow
Policy lens —FERC cancels priority interconnection orders for seven large datacentre projects citing changed demand assumptions; the Commerce Department revisits GPU export-control thresholds as the compute-per-dollar bottleneck softens; Congressional appropriators request a revised national AI-compute infrastructure assessment from OSTP.
AI-matched prediction markets — agree or disagree, the decision is yours. Clicking opens the provider's site.
Global AI inference token usage growth measures direct computational demand, reflecting efficiency gains in model execution and GPU throughput optimization across data centers.
Space-based data center deployment with H100-class GPUs represents a major infrastructure efficiency breakthrough, addressing power and thermal constraints through alternative deployment models.
Orbital data center with ≥1 MW operational capacity signals breakthrough in distributed GPU/compute infrastructure efficiency, core to next-generation AI deployment.
Running frontier-level AI inference on consumer gaming GPUs represents a direct efficiency breakthrough, requiring architectural and algorithmic optimizations that compress model inference into constrained hardware.
AI solving Millennium Prize Problems requires sustained inference efficiency and computational breakthroughs that cascade into data center architecture improvements.
For entertainment and research purposes only. OpenWatch tracks trends and signals — not real-time prices. Data updates every 4 hours. The forecasting algorithm is currently undergoing back testing, and we do not recommend any position. All trading decisions are solely your responsibility.
Markets are matched to OpenWatch scenarios by an AI worker that runs every 4 hours. New markets and price changes may not be reflected immediately.
Editorial framing — events outside our X→Y→Z partition. Authored as paired 'what if positive' / 'what if negative' to capture asymmetric tail outcomes. No probability is assigned; the lean indicator is directional only.
An open-weight model with 10-100x lower inference cost matches frontier capability; the compute-supply premium collapses, AI capex shifts from "more GPUs" to "more applications", and AI-deployment economics broaden dramatically.
A high-profile AI-system failure (autonomous vehicle pile-up, medical-AI mass-misdose, or AI-orchestrated cyber incident with casualties) prompts emergency regulation that halts frontier deployment for 6-12 months across multiple jurisdictions.
Low-probability outcomes that do not belong to the conditional partition above. Surfaced alongside, never ranked, never given a probability. See the card for the trigger mechanism and the names that move if it materializes.
Mechanism: Cluster-scale training pauses or sharply slows across the three jurisdictions simultaneously, collapsing the leading-edge GPU pricing power and shifting capital toward inference / specialized accelerators and on-prem deployments.
Within a 6-month window, the US, EU, and China each independently impose binding restrictions on frontier-model training or deployment — different motivations (US security review, EU AI Act enforcement, China data-sovereignty), same effective constraint on hyperscaler training-cluster economics. Outside the modeled partition because the existing branches assume asymmetric regulation; a synchronized one would invalidate the GPU-bottleneck thesis itself.
Contingency note — Watch for any reciprocal-precedent language in regulatory proposals across the three blocs in the same quarter. Synchronization risk is usually telegraphed in advance via trade-policy committee minutes.
Mechanism: Emergency executive action across multiple jurisdictions imposes deployment restrictions on frontier models in 30-90 days, accelerates licensing requirements, and shifts capital sharply toward safety / monitoring / specialist providers rather than scale.
A high-confidence, attributed AI-driven attack on critical infrastructure (power grid, financial-market microstructure, election integrity, or hospital systems) creates a forcing function — different from the synchronized regulatory crackdown branch because this is reactive, faster, and produces emergency executive orders, not multi-year rule-making. The partition assumes orderly buildout; this assumes a Three-Mile-Island moment for AI deployment.
Contingency note — Watch for any confirmed-attribution AI-attack disclosure from CISA, NCSC, or ENISA. Emergency-rule timelines telescope when an incident is publicly attributed.
Fewer than 5 historical episodes — tilts are indicative only. Use with extra caution.
Based on 4 distinct escalation episodes 2018–2019 and 2025 (Section 301 tariffs, Huawei entity list, chip export controls, 2025 tariff spiral); sector returns measured over the 3-month window following each escalation announcement.
Countries and companies most at risk or with most upside across this scenario overall
Information cutoff: 2026-05-21 · Authored: AI-generated, council-reviewed · Live signal counts updated hourly