The Last Tab
2026-02-16  ·  16 min  ·  Human Pace, The Audio Edition
----------------------------------------------------------------

Hi, I'm T. I translate risk for people making capital decisions in clean energy and artificial intelligence infrastructure. This is Human Pace, the audio edition.

Today's piece is called The Last Tab.

It's about a decision that can look almost trivial in a financial model: how much insurance to buy. Change that one number, and the effects can travel through the entire capital stack. In project finance, revenue projections get scrutinized cell by cell. Debt sculpting gets debated. Tax equity sizing gets weighed from every angle.

Insurance often gets a much quicker review. Someone looks at the last tab and says, "Looks fine. Next."

But that last tab may be where the capital structure is most exposed and least examined.

I wanted to make that exposure visible, so I built a model around a simple question: What happens when we keep the project exactly the same and change only the natural catastrophe insurance limit?

Here's the project.

A 150-megawatt solar facility, paired with a 75-megawatt, 300-megawatt-hour battery.

The model follows the full cash-flow waterfall, from construction through operations. It includes the debt engine, the tax equity partnership flip, and the covenant tests. Everything stays fixed except one number.

In one version, the natural catastrophe insurance limit is 25 million dollars. In the other, it's 50 million.

Now put each version through two loss scenarios: a 250-year event and a 500-year event. What changes?

Spoiler alert: The financial impact can be much more than the amount the insurer pays.

Let me explain. When the limit is low, the uninsured loss creates a capital gap. If the project's reserves can't cover that gap, the Debt Service Reserve Account, or DSRA, starts to drain. With less cash available for debt service, the debt service coverage ratio, or DSCR, comes under pressure. A covenant breach can follow. Sponsor distributions get trapped. And if the project still can't close the gap, the sponsor has to put in more equity.

That's the cascade.

So, one insurance decision becomes a liquidity problem, a covenant problem, and eventually an equity problem for the same project. Same power purchase agreement. Same operating assumptions. One structure absorbs the loss. The other needs a rescue.

Let's start with the 25-million-dollar limit.

Under the 250-year event, based on the current set of assumptions in this scenario, the project survives, barely. The DSRA absorbs the shortfall. DSCR compresses, although it stays above the covenant threshold. Sponsor distributions are blocked for one to two years while the reserves rebuild.

On paper, that's a near miss.

In practice, the project's safety net is gone. A second event, even a modest one, would hit a structure with very little protection left. And as catastrophe events become more frequent in many regions, a subsequent loss isn't something we can casually dismiss. It belongs in the planning conversation.

The 500-year event is a different story.

In this scenario, the DSRA is fully depleted. DSCR falls below the covenant threshold. Distributions are locked up for at least three years. To keep the structure intact, the sponsor has to contribute another 17.2 million dollars.

The sponsor's original commitment was 20 million. So the equity cure is equal to 86 percent of the original investment. That's more than a bad year. It's close to a wipeout of the equity position.

Now take the same project and raise the catastrophe limit to 50 million dollars. Nothing else changes. Same asset. Same losses. Same assumptions.

Under the 250-year event, the capital gap falls to 1.2 million dollars. The reserves remain intact, and sponsor distributions continue. The structure barely notices the event.

Even under the 500-year scenario, the capital gap is only 4.5 million dollars. There's no equity cure. No default risk. DSCR remains above the covenant threshold.

The contrast is dramatic. But before we turn that result into a recommendation, we need to be clear about what the model can and cannot tell us.

First, these numbers come from a fixed set of assumptions at a particular point in time. Change the assumptions, and the outputs change. That's what models do.

Second, the value isn't in predicting that the sponsor will need exactly 17.2 million dollars. The value is in tracing the consequences of a decision that might otherwise receive very little attention. The model invites the deal team to finish a sentence: Under these assumptions, this is the amount of risk we're willing to absorb.

That's the real deliverable. The 17.2 million dollars is simply how we get there.

At this point, the obvious question is: Why not just buy the higher limit?

In the model, the additional 25 million dollars of coverage costs an estimated 150 to 200 thousand dollars a year. Over 20 years, that's roughly 3 to 5 million dollars in additional premium. Compare that with a potential 17-million-dollar capital call, and the higher limit looks less like an expense and more like capital preservation.

If every decision were that clean, of course, everyone would buy more insurance. But they don't, because the real market introduces constraints.

For example, sometimes the capacity simply isn't available. A project may not be able to secure 50 million dollars of catastrophe coverage for a particular site or technology. Also, building the insurance tower may require several carriers, and only a few may be willing to participate.

Pricing is another constraint. Insurance doesn't become proportionally cheaper as the tower gets higher. When capacity is scarce, the upper layers can carry the highest rate on line, even though they're the least likely to be used.

Lastly, every premium dollar has another possible use. The project might spend it on higher business-income limits, broader exclusion buybacks, or a lower deductible. The catastrophe limit is only one lever in the program.

So no, the model doesn't prove that every project should buy 50 million dollars of coverage. What it does is isolate the structural consequence of choosing 25 million.

Real insurance placements involve capacity, pricing, deductibles, exclusions, and competing priorities all at once. So, whatever the final decision, the team should understand what it has chosen, even when market conditions make that choice feel involuntary.

The discipline that makes those decisions consistent is risk tolerance. Not a vague statement that the organization is "conservative" or "comfortable with risk." A real threshold. A shared view of what the project can absorb, what would require intervention, and where the line should be drawn.

That leads to a harder question. Whose risk tolerance are we talking about? Because the same insurance limit looks very different depending on where you sit in the capital stack.

For executives and the investment committee, the question is straightforward: What do we lose, and under what conditions?

They need a downside scenario they can weigh against the expected return. The insurance tab becomes useful only when someone translates policy limits into equity loss and the impact on internal rate of return, or IRR.

For lenders, they view the same decision through their collateral and the debt covenants. Is there enough insurance to rebuild the asset? Can the project continue servicing its debt? If the proceeds aren't enough to do both, who decides whether the money is used for reconstruction or debt repayment?

That answer is usually buried in the loan documents, and it usually favors the lender.

For tax equity investors, they have another definition of downside. Those investors need to know whether the tax benefits are protected. A catastrophe that delays commercial operation, interrupts production, or triggers investment-tax-credit recapture can threaten the economics of the entire position. But much of their return is concentrated in the first 5 to 7 years post Place in Service or recapture period, and that changes the instinct around recovery. Tax equity may care less about rebuilding the asset for another 20 years than it does about securing the tax benefits and protecting its exit. Their question isn't merely, "Is the project insured?" It's, "Is my particular exposure insured?"

Lastly, for the sponsor, the model brings the equity risk into focus. With a 25-million-dollar natural catastrophe insurance limit, almost the entire original investment is at risk in the 500-year scenario. With a 50-million-dollar limit, the structure absorbs the loss without an equity cure.

The sponsor also faces a consequence that doesn't appear neatly in the waterfall: reputation. A capital call doesn't only damage the project's IRR. It can weaken investor confidence in the next fundraise.

Now, I want you to consider the risk advisor.

The advisor's job is to translate probable maximum loss, or PML, into consequences the client can act on. PML is a modeling output. The deal team needs to understand liquidity, covenant pressure, and equity exposure. Insurance often lives in the gap between those two conversations.

Put all five perspectives around one table, and the conflict becomes clear.

The lender wants low deductibles and high limits because it isn't paying the premium. The sponsor wants to protect the asset without consuming unnecessary cash flow. Tax equity wants certainty around the tax benefits.

The insurance program is where those competing incentives get resolved, whether the team describes it that way or not. There are two more points worth making.

The first is about the peril itself: Why did I pick earthquake?

For solar, earthquake isn't currently a market-stressed natural catastrophe peril. It doesn't come with the same recent loss history, reactive pricing, or anxious underwriting cycle as wildfire, wind, hail, or flood. That relative quiet is useful because it lets us isolate the relationship between an insurance limit and the rest of the capital structure.

In short, earthquake helps us see the principle without much of the noise from recent events.

And I want to highlight the principle because climate change makes the principle urgent. Here is why.

Traditional return-period estimates assume stationary risk. In other words, a 250-year event is assumed to have the same probability this year as it did a few years ago. That may be defensible for earthquake. It's much harder to defend for climate-related perils.

For wildfire, wind, hail, and flood, historical return periods may no longer describe current risk very well. When those events are becoming more frequent or severe, this analysis may actually understate the exposure.

So, what can you do? The useful exercise is to substitute the peril that matters most to your portfolio and see what the framework reveals.

The second point I want to highlight is about insurance claims.

The financial model assumes a clean sequence: a loss occurs, a claim is filed, and the insurer pays the amount modeled. Real claims rarely move that neatly.

After a regional catastrophe, appointing an adjuster can take weeks or months because many policyholders are filing at the same time. Proof-of-loss requirements create a heavy documentation burden. Business-interruption losses become a negotiation because the carrier and the policyholder may disagree about projected revenue. Meanwhile, the lender and sponsor may have conflicting views about whether the proceeds should rebuild the project or repay the debt. That's a second cascade running alongside the financial one.

Imagine the 500-year event again, but add an 18-month claims process and a lender that directs the proceeds toward debt repayment.

Now the project doesn't merely face a depleted DSRA. It faces a liquidity crisis the financial model may never have contemplated. So when you define risk tolerance, don't consider only how much the policy is expected to pay. Consider when the money may arrive, what must happen before it is released, and who controls it once it does.

So where does this leave us?

One insurance limit can become a capital gap, a cash flow problem, a covenant breach, trapped distributions, and eventually an equity cure. At 25 million dollars of natural catastrophe limit, the project in this model survives the smaller loss and nearly loses the sponsor's entire position in the larger one. At 50 million dollars of natural catastrophe limit, the same structure absorbs both events without a capital call.

The real decision still has to account for available capacity, marginal pricing, competing coverage priorities, and the different interests around the table. Claims timing adds another layer, because the amount of coverage matters alongside when the proceeds arrive and who controls them.

A useful model makes those tradeoffs visible. It gives the deal team a shared language for risk tolerance and connects the insurance decision to the capital it is meant to protect.

So, go ahead and try the Insurance Capital Model, or ask your favorite AI to build you one. Explore the model by opening up the spreadsheet, and starting with the Insurance and Probable Maximum Loss tab. That's where insurance limits become capital consequences.

That's the piece. You can download the model at cleanpowerwhisperer.ai.

If this topic interests you, and you'd like to think it through with someone, you know where to find me.

Stay curious. Be safe. Be well.