Systemic SRISK (Value-at-Risk)
$284B
Expected systemic capital shortfall
Insolvent / Distressed Nodes
4 / 10
Tier-1 capital depleted below buffer
Circular Counterparty Loss
$118B
Unbacked cloud-for-equity write-offs
Grid / PPA Stranded Debt
$62B
Take-or-pay capacity overhang
Hyperscaler Frontier Lab Hardware/Fab Grid Utility Debt SPV Downstream
Frontier Lab Alpha (Foundational Models)
Role: Foundation Model Architect · High Cash Burn / Cloud Credit Dependent
Capital:$45B
Health:38% (Distressed)
Direct Exposure:$85B
Eisenberg-Noe Clearing Log Round 2 / Settled
[Round 1] Shock injected at Frontier Lab Alpha: 70% burn acceleration creates $38B liquidity shortfall.
[Round 1] Titan Cloud writes down $24.5B in prepaid cloud-equity conversion rights.
[Round 2] Circular AI Venture Fund SPV margin call fails: $32B compute collateral liquidation.
[Round 2] Axiom Silicon absorbs buffer impairment; orders backlogged but fab solvency intact.
Systemically Important AI Providers (SIAP) 10 Tracked Entities
Entity Tier-1 Buffer Loss Absorp. Status
Baseline contagion model initialized. Click 'Inject Contagion Shock' to simulate clearing rounds.

The 2008 Analogy: Why AI Infrastructure Mirrors SIFI Contagion

In her New York Times essay, financial journalist Bethany McLean (author of The Smartest Guys in the Room) warned that the unchecked concentration of foundational AI labs, cloud hyperscalers, and utility grids creates the exact structural vulnerability of the 2008 banking crisis: Too Big to Fail.

Instead of subprime mortgage CDOs, today’s systemic risk takes the form of circular cloud financing—where hyperscalers inject billions into frontier labs on the contractual condition that funds are recycled immediately as compute spend, inflating valuations while concentrating tail risk.

How This Stress Model Works (Eisenberg-Noe Clearing Vector)

  • Interbank-style Bilateral Obligations: Each node represents a systemic player (Hyperscaler, Lab, Foundry, Utility, SaaS distributor) connected by debt, compute agreements, and PPA energy contracts.
  • Clearing Algorithm: When a shock occurs, losses propagate along directed edges until clearing balances settle or buffer thresholds collapse.
  • Circular Debt Spillover: When a frontier lab burns cash without enterprise software payback, cloud hyperscalers write down equity while carrying fixed long-term power grid take-or-pay commitments.

Contagion simulator: fictional write-downs and ordered propagation

Read the explanation

The startup circular-financing preset immediately shocks a fictional venture vehicle with baseline capital seventy five billion by seventy five percent. Remaining capital is eighteen point seven five. Bars encode seventy five and eighteen point seven five at five pixels per fictional billion. Increasing shock to one hundred percent leaves two due to a minimum floor, rather than zero. Insolvency is assigned when remaining capital is at most fifteen percent of baseline. A separate global bail-in buffer sets the distressed threshold; individual node buffer fields are not used by that test. These amounts and thresholds are synthetic, not audited company balance sheets, investment valuations or predictions of actual defaults. A clearing round visits the directed edges in stored order. If the venture vehicle is insolvent, its thirty-weight debt edge applies severity point thirty five, transmitting ten point five to the lab, taking lab capital sixty to forty nine point five. A merely distressed sender uses point fifteen, a four point five cut. Bars compare those two cuts at six pixels per fictional billion. Each recipient cut is capped at forty percent of its current capital, with transmission ignored when at most point five. Mutations happen immediately, so later edges can react within the same round and changing edge order can change propagation. This is an authored iterative write-down heuristic rather than a solved simultaneous interbank clearing fixed point or validated Eisenberg-Noe implementation. The displayed systemic shortfall adds all positive capital impairments plus one hundred forty times circular exposure. At eighty five percent exposure, that constant component is one hundred nineteen billion even before measured losses. Circular-loss caption separately adds twenty four regardless of node write-downs. Bars encode these constant offsets at two pixels per fictional billion. Grid caption also adds a parameter offset; a nonzero label therefore need not mean an observed loss. Centering the canvas restores node capital from baseline as well as positions but does not refresh all telemetry, unlike reset. Native test applies a full SPV shock, steps one clearing round, generates and downloads real local JSON and pauses visual pulses. Four native widths compare exact static canvas buffers and drawing after that native pause. No empirical stress calibration, financial advice or public publication is established.

Enjoy this tool? Build your own with Super