A company-chosen discount rate below its own traded cost of capital inflates the asset metric the bull case rests on

Claim

The trap is not that the company lies. It discloses the rate. The trap is that the rate is an input the company selects, sitting inside a metric the market quotes as though it were observed.

The Sunrun case, 2026-08-26

Metric	Rate	Chosen or observed?
Subscriber Value (new business, the flow)	7.3%	"based on observed project-level capital costs" — floats each period
Gross/Net Earning Assets (back book, the stock)	6.0%, fixed	Company-selected, does not move
Its own senior ABS Class A notes	6.33% yield (200bp spread, traded 2026-08-04)	Observed, in cash, by third parties

The company marks its new business to market and its back book to a fixed number 130bp lower — and that fixed number is 33bp below what real buyers paid for the senior-most claim on the same collateral. The senior tranche has none of the residual's risk and still yields more than the rate used to value the residual.

Why the sensitivity is violent, not linear

The metric is a levered residual: assets are discounted, then debt is subtracted at face. Sunrun's ~$17B of contracted gross earning assets sits over ~$14.6B of nominal-fixed debt, so the entire present-value reduction lands on a ~$2.6B strip. Re-discounting moves the residual with roughly 4.5x leverage.

Discount rate	Contracted net earning assets	Market cap / metric
6.0% (company)	$3.70B	0.59x — looks like a 41% discount
6.7%	$2.79B	1.00x — breakeven
7.3% (company's own market rate)	$1.93B	1.66x — a premium
9.0%	$0.09B	—
10.0%	negative	back book no longer covers project debt

A 70bp move flips "cheap" to "expensive." A 200bp move erases the equity. Any thesis quoting "trades at 0.24x Net Earning Assets" is quoting a number whose sign is set by an assumption.

The second axis nobody quotes

The rate is rarely the only chosen input. Sunrun's base case also assumes 0.00% annualised net defaults and 90% renewal on 30-year customer relationships. Moving rate and default and renewal together to still-moderate values (7% / 0.75% / 80%) takes the metric from $9.0B to $5.0B — the headline "80% discount" becomes ~57%.

How to apply

Find the rate before you use the metric. It is in the non-GAAP definitions or the glossary, never in the headline. If a company publishes a DCF-based asset value and you cannot find the rate, do not use the metric.
Benchmark it against the company's own traded debt — the most recent securitization, bond or term-loan print. principle-primary-source-beats-vendor applies: a cash transaction in the company's own paper beats any assumed rate. A residual discounted below the yield on its own senior claim is prima facie wrong.
Look for a second, different rate elsewhere in the same document. Companies that mark one metric to market and fix another have told you which one they trust.
Re-discount and show the zero-crossing, not just a sensitivity band. For a levered residual the useful output is "equity value is zero at r = X," because X is often uncomfortably close.
Do not apply a punitive rate to the gross asset. Blend: senior at its traded yield times the advance rate, plus a 10-15% residual strip. For Sunrun that gives ~7.5-8.0%, not 12%.
Check the other chosen inputs — default rate, renewal rate, terminal assumptions — and move them together, because they are usually all set to the optimistic end at once.

Where else this fires

Any structure where the issuer publishes its own asset value: REIT NAV disclosures · BDC portfolio fair-value marks · yieldco and infrastructure CAFD models · insurance embedded value · mortgage REIT book value · any "adjusted NAV" or "earning assets" construct. Octus's analysis of Sunnova reached the identical conclusion on the identical instrument - securitization equity recoveries spanning 0% to 89% on discount-rate assumptions alone.

What would falsify this

A company whose published rate sits at or above its own traded senior yield, and which moves that rate as market conditions change. That is a company using the metric to inform rather than to flatter, and its number can be taken at face value.

[[principle-primary-source-beats-vendor]] · [[principle-down-a-lot-is-not-cheap]] · [[pitfall-fcf-definition-diverges-with-jv-partner-funds]] · [[principle-reit-wacc-has-no-tax-shield]] · [[pitfall-vendor-ev-inverts-net-cash]]

History

2026-08-26 — Established from the RUN analysis. The finding surfaced when two agents reported different discount rates (6% and 7.3%) for what was assumed to be one metric; resolving the conflict against the 10-Q sensitivity table showed both were right, and that the gap between them was the finding rather than either number.