Loan vs cash: the same site modeled both ways

Financing is the input most people set once and never revisit, and it is the one whose effects are the least intuitive. A loan does not make a site better or worse. It changes when the money leaves and arrives, who bears the ramp years, and what the headline metrics mean.

So here is the engine's default eight-stall site, PG&E BEV-2, $0.45 per kWh on the Standard ramp, $1.21M of capex, built four ways: 20% down, 35% down, 50% down and all cash. The loan, where there is one, is ten years at 8%. Nothing else changes.

The cheque at closing#

The third row is the one a day-one budget usually misses. The ramp year loses money, and a leveraged site loses more of it because the loan payment starts in month one while the stalls are still finding their drivers. At 20% down the site absorbs $158,515 before its first cash- positive month; for cash it is $53,010. Working capital is not a line beside the financing decision. It is part of it.

The years that follow#

The year-eleven row is the same in every column because that is the year the ten-year loan ends. Until then, the 20%-down site is paying $140,934 a year to the bank out of a site that generates between −$158,515 and +$8,179 of net operating cash. Its cumulative position keeps sinking for a decade, to −$556,572 in year ten, and only turns upward when the debt service stops. The all-cash site is cumulatively positive from year three.

Net operating cash, throughout, is revenue less electricity, the network fee, rent, insurance, entity fees and the loan payment: what is actually left in the account.

What the headline metrics say#

Three things stand out.

NPV barely moves. Across a $1.1M difference in borrowing, NPV varies by $11,000. The engine discounts at 10% and the loan costs 8%, so borrowing is close to a wash at the discount rate: the interest the site pays is roughly what the capital it kept would have been expected to earn. A cheaper loan tips that one way, a higher discount rate the other, but on these defaults leverage is not where the value is.

Leverage lowers IRR here, slightly. Textbook leverage raises equity return when the asset earns more than the loan costs, and the all-cash site does return 9.2% against an 8% loan. But the spread is thin, and the leveraged site front-loads the ramp losses into the capital at risk: the working-capital line above is three times larger with a loan. On these numbers the two effects roughly cancel, and the small residual runs against the loan.

The multiple tells the opposite story, and that is the point of the multiple. Cash-on-cash is total return divided by upfront capital, so a smaller cheque mechanically produces a larger multiple: 3.98× on $447,390 against 2.43× on $1.31M. Both numbers are correct. One describes the site; the other describes the cheque. The payback vs NPV vs IRR post is about exactly this kind of disagreement.

What leverage actually buys#

Put plainly: on this site, borrowing 80% does not change what the asset is worth. It changes the shape of the exposure.

The owner who finances writes a cheque one-third the size and keeps $862,000 of capital for something else. In exchange, they carry a decade in which the site's own cash does not cover its debt service, a cumulative hole that reaches $556,572 before it turns, and a dependence on year eleven arriving on schedule. The owner who pays cash has none of that exposure and all of the capital tied up.

Neither is the right answer. The engine does not render one, and the numbers above are why: the two structures have nearly the same NPV and very different years. What the model does is make the trade explicit, so the decision is about the owner's balance sheet and horizon rather than about which multiple looks better on a slide.

The next post in this series looks at the loan's own terms, since a percentage point of rate and a few years of term move these figures more than the down payment does. The scenario wizard takes down payment, rate and term as inputs for any address; the methodology page documents how the returns stream is built.

Engine outputs are model projections under the stated assumptions, not advice or guarantees. Loan terms, discount rate and site are the engine's documented defaults; the 15-year total includes the year-fifteen terminal value at the default 5× multiple and the year-one tax shield.