Geopolitical Risk Monitor
Supply-side risk premium, event log, and risk factor assessment for global AI compute markets.
Data Disclosure
The Geopolitical Risk Premium (GRP) series in this edition is an analyst estimate, not a statistically computed result. Values represent the author's calibrated judgment, informed by the documented events in the Event Log below and publicly available geopolitical risk indicators. They are not derived from the formal event-study framework described in the methodology white paper (§5).
The formal event-study GRP — with estimation windows, abnormal price changes, and statistical confidence intervals — is planned for Vol. II (May 2026), when sufficient monthly price observations will be available to estimate a credible baseline.
All GRP values carry confidence: analyst_estimate in grp_series.csv. The series from July 2024 through March 2026 is additionally labelled reconstructed, meaning it was back-filled after the fact rather than recorded contemporaneously.
Use the GRP series for qualitative trend analysis only. Do not use it as a quantitative input to regressions or models without additional validation.
Cumulative GRP Series
The chart below shows the cumulative Geopolitical Risk Premium from July 2024 to April 2026. The baseline (GRP = 0%) is set at July 2024, representing a period of relatively benign geopolitical conditions at the time of the Llama 3.1 model release.
The GRP is published as an independent series, decoupled from the GCPI headline. The GCPI headline reflects observed spot prices which may already embed a geopolitical risk premium; the GRP is an attribution module, not an additive adjustment. See methodology §5 for the planned formal estimation framework.
Risk Factor Scores
The following scores represent the author's current assessment of each risk factor's potential 12-month impact on AI compute supply. Scale: 0 = no material risk, 10 = severe supply disruption. These are qualitative assessments, not model outputs.
Risk scores are analyst judgments based on publicly available policy, regulatory, and geopolitical information. They do not reflect proprietary intelligence, classified information, or formal quantitative risk models. Scores are updated monthly alongside the GCPI release.
Event Log
Documented geopolitical events that inform the GRP estimate. Each event has been independently verified against a named public source.
| ID | Date | Type | Event | GRP Direction | Source |
|---|---|---|---|---|---|
| E001 | Jun 2025 | supply_shock | AWS cuts P5/H100 on-demand prices up to 45%, signalling GPU supply glut at hyperscaler tier | ↓ negative | aws.amazon.com/blogs |
| E002 | Dec 2025 | policy_uncertainty | Trump administration signals potential H200 sale pathway to China; short-term uncertainty spike in GPU spot markets | ~ mixed | reuters.com |
| E003 | Jan 2026 | export_control | BIS implements case-by-case H200 export licensing to China. AWS raises Capacity Block prices ~15% | ↑ positive | bis.doc.gov |
| E004 | Mar 2026 | supply_chain_integrity | DOJ indicts participants in alleged $2.5B GPU smuggling network (Supermicro supply chain) targeting China | ↑ positive | justice.gov |
GRP direction: ↑ positive = increases geopolitical risk premium (upward price pressure on supply-constrained providers); ↓ negative = reduces premium; ~ mixed = ambiguous short/long-term effect. All event source URLs link to the originating institution's domain; specific article URLs will be added as the event log is formally maintained going forward.
GRP Estimation Framework
The GRP is designed to be estimated via a standard event-study methodology (MacKinlay 1997), adapted for commodity price benchmarks rather than asset returns:
| Window | Definition | Purpose |
|---|---|---|
| Estimation window | [τ − 90, τ − 10] days | Establishes no-event price baseline |
| Event window | [τ, τ + 30] days | Measures abnormal price response |
Average Abnormal Price changes (AAR) are computed for the affected provider cohort relative to the estimation baseline. Cumulative Abnormal Returns (CAR) across events are summed to yield the headline GRP percentage.
Current status (v0.1.0): Insufficient monthly price observations exist to run this estimation with statistical validity. The formal event-study GRP will be introduced in Vol. II once at least 12 months of observed (non-reconstructed) monthly price data are available. Until then, the analyst-estimated GRP series serves as a qualitative placeholder. See methodology §5 for the full specification.
Data Downloads
Both data files are open under CC BY 4.0. Use subject to the disclosures above.