What a wind farm taught me about certainty

You build a model. You run ten thousand sensitivities. You still need to decide.

The spreadsheet I had open in my third week at Columbia was a discounted cash-flow model for a 200 megawatt wind farm. Twenty-five years of PPA revenue, a 60/40 debt-equity stack, tax equity flipping in year six, four coverage covenants tracking against each other like leashed dogs. I had built it from a lender's perspective, which meant the margin of safety sat in the debt service coverage ratio (DSCR), and the DSCR was behaving badly.

Every time I tightened the PPA floor — reduced the price per MWh the project would guarantee the counterparty — the coverage ratio drifted south by 0.04x. Every time. It was not a bug. It was the model telling me what I already suspected: no amount of tax equity structuring was going to out-argue a bad resource year. The project's economics, at the margin, depended on the wind actually blowing.

This is the thing I keep learning. You can model the wind. You cannot model whether it blows.

Certainty as a deliverable

Finance is the business of selling certainty. Not actual certainty — that is the weather's department — but manufactured certainty. A PPA with a twenty-five year term is a bet that the offtaker will still be solvent in 2050. A tax equity flip is a bet that the tax code won't change materially in six years. A P90 production estimate is a bet that the wind resource consultant was being conservative, and that the turbines will degrade in the way their spec sheets claim they will degrade.

None of these bets is crazy. All of them are bets.

What a good model does, at its best, is make the bets legible. It does not make them disappear. Ten thousand Monte Carlo runs tell you the shape of the uncertainty, not its absence. If your job is to convince a bank, you build the model to show that even under the bottom decile the coverage ratio holds. If your job is to decide, you build it to show where it doesn't.

Where it doesn't

For the project I was working on, the coverage ratio broke under three joint conditions: a below-P99 wind year in year three, coinciding with the start of the tax equity flip, coinciding with a 50 bp rise in the cost of replacement debt. Each individual event had a rough probability. The joint probability was low — maybe 1.2% per year, maybe lower. But it was not zero. And if it happened, the equity returns went from middling to negative.

I included it in the memo. The lender's counsel circled it in red. The principal looked at the red circle for about ninety seconds and then asked me a question I still think about.

"What would make you comfortable?"

The question was not about the model. It was about me. And I realized — maybe for the first time — that half the job of a financier is to be the person whose discomfort is worth respecting. Not because I was uniquely insightful. I wasn't. But because I had sat with the model long enough to know where it broke, and the lender's time was too scarce to do that work from scratch.

The after-math

I wrote a paragraph on two mitigants — a one-year DSRA top-up and a soft cap on interest rate step-ups — and we moved on. The deal closed. The wind, as of this writing, is blowing.

But the question stays. What would make you comfortable? I do not have a general answer. I have the specific one I gave that day, which was: a slightly larger reserve, a slightly narrower hedge window, and the honesty to say the project has a skinny third year. That's what ten thousand sensitivities will get you, eventually. Not certainty. A well-considered opinion, offered on purpose, under your own name.