For lenders: the four numbers in a charging-site credit file

A charging site arrives at a credit desk as a pro forma with one row of revenue copied across fifteen columns. The desk's job is to find the years that row is hiding. The engine's job is to put them on the page before anyone has to look, so this post is the lender's view of the default site: eight stalls, PG&E BEV-2, $0.45 per kWh, $1.21M of capex with $968,000 borrowed over ten years at 8%.

Four numbers carry most of the weight. Here is how each is built and what it says.

1. Debt service coverage ratio#

DSCR is the cash available for debt service divided by the debt service. The numerator is what the engine calls EBITDA: revenue less the variable costs (electricity, Tesla's network fee, any excise or revenue tax) less the operating fixed costs (rent, insurance, accounting, entity fees, property tax), before the loan payment. It is what would be left if the loan did not exist. The denominator is the $140,934 annual payment.

The site does not cover its payment from its own cash in any year of the loan. It comes closest in year nine, at 1.06×, and slips back in year ten when the nine-year LCFS credit line ends. The minimum over the loan life is −0.12×, the average 0.61×. Every dollar of shortfall is funded by the owner, which is the peak cash deficit in year one and the cumulative hole after that.

The ramp decides this table. On the engine's Premium curve the same site reaches 1.14× in year six and 1.64× in year nine; on the Slow-ramp curve it never passes 0.49×. The site, tariff and loan are identical across the three. The first-year utilization path is the credit risk.

2. Fixed charge coverage ratio#

FCCR widens the lens. The numerator is revenue less variable costs only; the denominator is every fixed charge the site carries, operating costs and loan payment together. It answers a different question from DSCR: not "can the site pay the bank" but "can the site pay everyone it has promised to pay".

The jump at year eleven is the loan ending: fixed charges fall from $180,623 to $40,373 and the ratio multiplies by five. The minimum over the asset life is 0.38×, the average 2.26×, and that average is mostly the last five years. A lender whose exposure ends at year ten sees only the first ten rows.

3. Peak cash deficit#

The two ratios say the site cannot carry itself through the loan. The deficit says how much that costs, and when.

In year one, month by month, the cumulative cash position bottoms at −$158,515 in December. That is the working capital the owner has to bring to closing beyond the $242,000 down payment and $46,875 of sales tax. It is the number a lender asks to see in a reserve account, and the engine reports it on the headline for that reason.

The year-one figure is not the whole exposure. On the Standard ramp the cumulative position keeps falling to −$556,572 at the end of year ten, the last year of the loan, because net operating cash after debt service is negative in every year but nine. The owner funds that gap from outside the site. Whether they can is a question about the owner, not the stalls, and it is the second thing a credit file has to establish.

4. Payback against term#

Payback on the default site is 171 months. The loan is 120. For the last fifty-one months of the loan the bank is being repaid from a site that has not yet returned its own capital, and for the first fifty-one months after the loan ends the owner is still recovering theirs.

That ordering is normal for a leveraged asset with a long ramp, and it is also the reason term matters so much on this site: the rate-and-term post shows a five-year term pushing the year-one deficit to $253,112 and five-year cumulative cash to −$678,746, while a fifteen-year term keeps debt service inside the years the site would otherwise be free. The gap between payback and term is a measure of how much of the asset's life the lender's exposure covers.

The bands, stated and not applied#

The engine carries a set of internal threshold bands, calibrated to SBA-504 and small commercial real estate conventions, which it uses only to locate the boundary its reverse-solve searches against. They are stated here because a credit desk will recognise them: minimum DSCR of 1.20× or better, 1.10× to 1.20× as the middle band, below 1.10× as the lower; minimum FCCR of 1.25×, 1.10× to 1.25×, below 1.10×; payback within 60 months, 61 to 96, beyond 96; and no cash-negative years after year two, one or two, more than two.

Against those bands the default site reads as it reads: minimum DSCR below zero, minimum FCCR 0.38×, payback at 171 months, seven cash-negative years after year two. The engine does not grade it. It reports the ratios, the deficit and the payback, year by year, and leaves the bands as reference lines for whoever is reading. What the reverse-solve adds is the distance: retail price would have to be $0.579 rather than $0.45, or year-two utilization 281 kWh per stall per day rather than 227, for the same site to cross its internal boundary.

What moves the file#

The same four numbers for the same site with 50% down instead of 20%: DSCR reaches 1.13× in year six and 1.69× in year nine, the year-one deficit falls to $108,918, and cumulative cash at year ten is −$28,069 rather than −$556,572. With more equity the site covers its smaller payment from year six; with less it never does. The loan-vs-cash post sets out the full trade, and the ramp post covers the curve that decides the first table above.

Section 7 of the engine's report carries DSCR and FCCR for every year alongside the rollups, for any address and capital structure entered in the scenario wizard. The methodology page documents the ratio definitions used here.

Engine outputs are model projections under the stated assumptions, not advice or guarantees. The threshold bands are described as the engine's internal reference values and are not a rating of any site or borrower.