AI capex contagion model — hyperscalers, energy, fiber and the cost of money

Move capex, costs, power prices, Treasury yields and lab solvency. Watch the money propagate from AI labs → hyperscaler balance sheets → grid equipment, utilities and fiber. Every coefficient is visible and editable. Live values are the sliders' defaults, not locks.
Methodology — how the model works →

Treasury curve

The risk-free anchor. Every capex dollar not funded from operating cash is priced off this curve plus a credit spread.
Treasury yield curve, base versus scenario.

All-in new-issue cost

Blended benchmark plus issuer credit spread. This is what each builder actually pays for marginal capex.

Where the money flows

Band width = annual dollars. Left column is compute demand; middle is the balance sheets that carry it; right is who gets paid.

Revenue impact by company

Change in annual revenue vs. the 2026 base case, in billions.
Revenue change by company.

Capex beta

Revenue elasticity to system capex, computed by perturbing capex ±10%.
Capex beta by company.

Rate beta

Revenue elasticity to a 100bp parallel shift in the Treasury curve.
Rate beta by company.

Contagion path

Company detail

Revenue above baseRevenue below baseClick any header to sort

Assumptions

Every number below is an input, not a result. Edit any cell and the model re-solves. Base revenues are 2026 guidance or run-rate where disclosed; AI exposure, operating leverage and credit spreads are estimates — override them freely. An override flips that field's provenance to User.
Suppliers — base revenue, AI exposure, margin, leverage, credit
Hyperscalers — capex, AI revenue, lab concentration, debt funding, spread
Model coefficients — how the propagation works
Sources and known limitations