DRIFT SYSTEMSHIDDEN DEPENDENCIES.
DISCOVERED OPPORTUNITIES.
Repeated gigawatt-scale load-withdrawal events are observed. Wider system propagation is modelled. The possibility that separately insured assets or books could become correlated through the same grid-side dependency is inferred. Portfolio-scale insured consequence remains unproven.
Grid operators now treat computational load as an active system variable. NERC has moved from incident review to a Level 3 alert and a large-loads action plan within two years.
Upstream Domain · WORKINGEnergy Systems (secondary: Technology Infrastructure). Upstream Domain provenance. Compute / Cloud and Energy / Grid are surface territories, not Domains.
Separately sited, separately insured computational assets can share one grid-side protection contingency, so exposures managed as independent may not be.
04 AI · SCALE TEST PLANNEDOperators of dense compute clusters carry a shared protection-behaviour dependency on the bulk grid that sits outside individual site resilience assumptions.
Decision boundary: Does site resilience planning stop at the facility fence while protection behaviour is set cluster-wide? · Expression: CREATE-led (assurance, ride-through design, siting), GUARD on availability commitments Open surface →
05 Capital · SCALE TEST PLANNEDInfrastructure and data-centre capital may be priced as diversified while sharing one grid-side failure mode.
Decision boundary: Does the diversification assumption across digital-infrastructure holdings survive a shared grid dependency? · Expression: BOTH (concentration guard, dependency-aware allocation) Open surface →
09 Industry · SCALE TEST PLANNEDIndustrial loads co-located with computational clusters inherit a contingency they did not contract for.
Decision boundary: Does the plant's power-quality and continuity boundary include neighbouring large loads? · Expression: GUARD-led (continuity, contractual exposure), CREATE on grid services Open surface →
10 Reinsurance · MAP PUBLISHEDSeparately insured assets and books may share one grid-side contingency, so accumulation, correlation, underwriting and capital boundaries drawn at the data centre may not hold.
Decision boundary: Does the accumulation boundary stop at the data centre? · Expression: GUARD / CREATE / BOTH Open surface →
The mechanism, its evidence states and its sources are identical on every surface. Only the institutional meaning, decision boundary, role and GUARD / CREATE / BOTH expression change.
Accumulation / Exposure ManagementDoes the accumulation boundary stop at the data centre?
Portfolio / Aggregate ManagementDoes the diversification assumption survive a shared grid dependency?
Treaty / Specialty UnderwritingDoes the underwriting boundary survive when separate risks share the same grid dependency?
Technical Risk / EngineeringWhat physical mechanism, operational dependency and mitigation would have to exist for the risk to be real?
CRO / Enterprise RiskDoes the risk boundary stop where the business lines do?
Actuarial / AnalyticsDoes the correlation assumption survive a shared grid dependency?
Claims / Loss ScenarioDoes the event definition survive when one grid disturbance reaches several policies at once?
Cross-Class Dependency OwnerIf no existing desk owns the whole dependency, where should the cross-class RDM-03 decision live?
Each institution owns its decision boundaries. Ownership is discovered from public role evidence in the order Decision Boundary → Function → Role. Institution profiles against this dependency are shared with the institutions concerned; public listing by name follows the publication gate. Institution classes →
RDM-03 is a ratified external grid dependency mechanism. Its reinsurance consequence and opportunity remain institution-specific tests. Public evidence establishes the mechanism. The commissioned report develops what it means for the specific institution. Ratification applies to the mechanism, not to insured consequence or buyer-specific relevance.
Synchronized large-load protection behaviour can cause exposures still treated as separate to begin sharing the same contingency.
Our public-evidence review has found no evidence that RDM-03 is currently treated as a distinct reinsurance dependency class. Initiating events are observed. Wider propagation is modelled. Cross-exposure correlation is inferred. Insured consequence remains unproven.
July 2024 load loss
February 2025 load loss
NERC alert, 4 May 2026
Portfolio-scale insured consequence is not established.
| Status | Public evidence | What it supports | What it does not support |
|---|---|---|---|
| OBSERVED | NERC's incident review records approximately 1,500 MW of simultaneous voltage-sensitive load reduction in the Eastern Interconnection on 10 July 2024 following a transmission fault. | Large computational / power-electronic loads can disconnect or transfer simultaneously at scale. | It does not establish a catastrophic outage or insured loss. |
| OBSERVED | NERC's 2026 filing materials identify a second Virginia event in February 2025 involving approximately 1,800 MW of load loss. | The mechanism is not a one-off observation. | Frequency alone does not establish reinsurance relevance. |
| REGULATORY RESPONSE | NERC issued a Level 3 Computational Load Alert on 4 May 2026, calling for changes to modelling, studies, monitoring, commissioning, operations, protection and control. | The reliability concern is consequential enough to drive industry action. | A reliability alert is not evidence of insured accumulation. |
| INFERRED | Reinsurers publicly write multiple lines and treaty relationships that can share geography and critical infrastructure dependencies. | A common external trigger could, in principle, connect exposures otherwise managed separately. | This is not evidence that a specific reinsurer actually has such a concentration. |
| HYPOTHESIS | The reliability response around computational loads creates new operational, modelling, assurance and risk-transfer needs across multiple institutional surfaces. | A structural dependency can create commercially testable intervention points. | It does not establish that any particular product, service, venture or market will succeed. |
Recorded. Held in the governed register.
Governed dependency register (read-only) · state since 2026-08-31. A falsifier / re-RED trigger is recorded against this mechanism and reviewed before any state change. The text is held in the governed register.
Generate candidate dependency mechanisms from public evidence. Look for changing relationships between things still treated as separate.
Attack the candidate with contradiction, prior art, separability and falsifiability tests. Ask what evidence would kill it.
Promote only a surviving mechanism. Freeze what is OBSERVED, MODELLED, INFERRED and UNPROVEN so later work cannot silently upgrade the evidence.
Map the ratified dependency mechanism onto the public surface, its functions and decision boundaries without claiming knowledge of internal portfolios or models.
Choose the consequential question to develop. GUARD asks what the dependency could threaten. CREATE asks what recognising it could make newly valuable.
The method permits NULL at both stages. A mechanism that cannot be separated from prior art or survive contradiction does not ratify.
Develops the institution-specific consequence position: what the dependency puts under pressure, which exposures or decisions it connects, and what the institution should guard.
Develops the institution-specific opportunity position: what recognising the dependency could help create across decisions, products, services, structures, capabilities or market position.
Two complete positions from one dependency and shared evidence base: what the institution should guard, and what it could create.