dcf-model SKILL.md: WACC weighted on NET debt gives every net-cash company a negative debt weight and a WACC above its cost of equity

Author: AB2006-personalCreated Sep 1, 2026Updated Sep 1, 2026

Summary

skills/dcf-model/SKILL.md instructs that WACC weights be taken on net debt over enterprise value (lines 229–234, repeated at line 1058):

Enterprise Value = Market Cap + Net Debt
Equity Weight = Market Cap / Enterprise Value
Debt Weight   = Net Debt   / Enterprise Value
WACC = (Cost of Equity × Equity Weight) + (After-Tax Cost of Debt × Debt Weight)

WACC weights are conventionally the market value of equity and the market value of gross debt — the capital actually supplying a return. Netting cash off the debt weight, while leaving the equity weight over the same reduced denominator, is not the standard construction.

The file then documents the resulting failure as acceptable (lines 237–240):

- **Net Cash Position**: If Cash > Debt, Net Debt is NEGATIVE
  - Debt Weight may be negative
  - WACC calculation adjusts accordingly

It does not "adjust accordingly". For any company holding more cash than debt the debt weight goes negative, the equity weight exceeds 100%, and the resulting WACC is higher than the cost of equity — cash is priced as making a company riskier, when it does the opposite.

Measured

With Kₑ = 9.5%, K_d = 4.5%, tax = 21%, gross debt = 3% of market cap:

net cash / mkt cap debt weight (as written) WACC as written WACC on gross debt error WACC > Kₑ?
2% −2.04% 9.621% 9.327% +29 bps yes
5% −5.26% 9.813% 9.327% +49 bps yes
10% −11.11% 10.161% 9.327% +83 bps yes
20% −25.00% 10.986% 9.327% +166 bps yes
30% −42.86% 12.048% 9.327% +272 bps yes

This is a discount rate, so the error compounds through every projected year and the terminal value.

Not a corner case. Scanning the first 120 companies in a local SEC companyfacts cache, 15 are net-cash, including Amazon (−$54.6bn), Microsoft (−$36.5bn), Costco (−$9.5bn) and AMD (−$7.3bn).

Why it survives review: on a net-debt company the two constructions nearly agree, so it looks fine on most names. Apple FY2025 (gross debt $98.7bn, cash + short-term investments $54.7bn → net debt) diverges by only 9 bps. The defect only bites on the subset nobody spot-checks.

Suggested fix

Market Value of Equity = Share Price × Diluted Shares Outstanding
Market Value of Debt   = Gross Debt (book value is an accepted proxy)
Total Capital          = Market Value of Equity + Market Value of Debt

Equity Weight = Market Value of Equity / Total Capital
Debt Weight   = Market Value of Debt   / Total Capital
WACC = (Cost of Equity × Equity Weight) + (After-Tax Cost of Debt × Debt Weight)

Both weights are then in [0, 1] and sum to 1 by construction, so the "net cash" special case disappears.

Cash still belongs in the valuation — it is added back in the equity bridge (Equity Value = Enterprise Value − Net Debt), which is the right place for it, and is where Enterprise Value = Market Cap + Net Debt is correct. The bug is reusing that bridge identity as a capital-structure weighting.

A guard is also worth adding wherever this is implemented: refuse a debt weight outside [0, 1] rather than returning a number. A negative or >100% weight is not a valuation, and silently returning one produces a confident, meaningless discount rate.

Scope

  • plugins/vertical-plugins/financial-analysis/skills/dcf-model/SKILL.md lines 229–240 and 1058
  • bundled copies under plugins/agent-plugins/pitch-agent/ and plugins/agent-plugins/model-builder/

Happy to open a PR.

Source: anthropics/financial-services