New model architecture (sub-quadratic attention or similar) reduces inference compute needs 10x; existing datacenter capacity becomes oversupplied; GPU commodity prices fall.
Signal counts measure media attention over the last 7 days — not the likelihood of an outcome.
If this branch plays out and you weren't positioned, here's what you'd miss or take. AI-generated estimates, not forecasts.
▲ Missed gains if not positioned
▼ Realized losses if not hedged
Magnitudes assume — IF the branch materialises — the moves described. Actual moves depend on timing, prior positioning, and intervening events.
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.
Trade lens —NVDA and ASML repriced lower on demand-shock; application-layer names (MSFT) capture value; Taiwan/Netherlands semis-capex multiple compresses. · structural · slow
Outcomes below — each % shown is the overall probability of that full chain occurring
If this path occurs — possible outcomes
Outcome % = conditional on this path occurring · Path % = joint probability of this exact chain from today
Policy lens —FERC cancels reserved interconnection capacity for Tier-1 hyperscaler datacentre projects; the Commerce Department withdraws draft advanced GPU export-control tightening pending demand reassessment; Congressional AI caucus members request GAO audit of federal AI-infrastructure investment plans.
Trade lens —Tech hyperscalers (MSFT, GOOG) FCF surges drive multiple expansion; NVDA revenue compresses 40-50% on cancellations; datacenter REIT and PPA demand re-rates lower. · structural · slow
Policy lens —FCC opens a proceeding on edge-AI device privacy standards; the EU AI Act delegated act on embedded AI is fast-tracked; NIST releases an on-device AI risk management supplemental framework applicable to consumer devices across G7 markets.
Trade lens —AAPL on-device-AI franchise captures the value; NVDA cloud-GPU demand plateaus; QCOM and Korean memory bid on on-device silicon. · meaningful · slow
Policy lens —The Commerce Department maintains strict export controls on H20-class and above GPUs as frontier training demand sustains; the White House AI Council reviews export-licence backlogs and issues strategic guidance on allied-country access; OECD updates frontier-AI governance recommendations to address Jevons-paradox compute scaling.
Trade lens —NVDA and ASML demand sustained on Jevons-paradox dynamics; TSM advanced-node demand holds; challengers still trail in frontier training. · small move · fast
Information cutoff: 2026-05-21 · Authored: AI-generated, council-reviewed · Live signal counts updated hourly