Why critical industries hold their value when the market turns

The figure that ran in the headlines this spring was €5.9 billion of European growth debt deployed in the first quarter, the strongest quarterly number in years. The figure that actually tells you something is the deal count, down for the fourth quarter in a row. Three or four large AI infrastructure raises dragged the average ticket up to around €90 million, roughly three times the prior year. Take those out and what is left is a smaller, more capital-efficient borrower, with fewer lenders willing to do the work on a €1 to 3 million facility.

A €15 million loan and a €2 million loan take about the same amount of credit work to underwrite. As growth debt funds raise bigger vehicles they need bigger deals to put the money out, so the small end stops being worth their time, even though that is exactly where late-Seed and Series A companies need help.

Equity is doing the same thing. European venture had its strongest quarter since 2022, but a few companies absorbed most of it and the overall count fell sharply. Capital is piling up at the top and getting scarce below. The headline figure and what it actually contains have stopped being the same thing.

This is usually the kind of market a specialist gets built for. Not because the headlines flatter anyone, they do not, but because the conditions underneath reward careful underwriting. Spreads are wider. Fewer lenders are competing for the good credits. And the companies that need €1 to 3 million right now tend to be the ones sitting on real assets and real revenue, with good reason not to hand over more equity than they have to.

The question every credit decision comes back to

Strip away the growth rate and the pitch and every growth debt decision lands on one question. If the next equity round takes longer than expected, what is actually behind the loan? Not how fast the company grew last quarter. What is still standing if the story stalls for a year.

That question is why I keep coming back to companies in what I would loosely call critical industries: defence, space, robotics, industrial automation, software built into regulated sectors. They are rarely the fastest growers in any given quarter and they almost never lead a fundraising headline. But they share the thing that matters more than growth once conditions tighten. They hold their value.

A certification does not lapse because a funding round slipped. Hard assets do not vanish because the market repriced. Revenue under contract with institutional customers does not disappear in a quarter, and intellectual property that took years to build does not become a commodity overnight. These companies are slower to underwrite, which is part of why fewer lenders bother, and also why the part of the credit that has to carry the loan if growth stalls tends to be more honest than in most of the venture-backed economy.

Where the worst outcomes actually come from

The worst growth debt outcomes I have seen across cycles almost never came from companies that simply grew slower than expected. They came from companies whose underlying value evaporated the moment the equity story did. A consumer software business with no real switching costs. A marketplace whose unit economics only held if every round priced higher than the last. Once the story changed, there was nothing structurally underneath the loan.

The credits that held up best were not the fastest scaling. They were the ones where, even in the bad scenario, you could draw a clear line from the assets and contracts on the balance sheet to a recovery. That is the whole job at this end of the market. You are not betting on the best case. You are underwriting the worst one and making sure the company comes out the other side.

There is a shift in software credit worth naming, because it sharpens all of this. AI is changing the economics that made software attractive to lenders in the first place. Agentic tools are compressing seat counts, and pricing is drifting toward usage-based models, which makes revenue harder to predict. So for a software borrower the question is no longer only how efficiently it grows. It is how deeply the product lives inside the customer’s workflow. The deeper it is embedded, the more durable the credit. The more it looks like a nice-to-have charged per seat, the harder you have to stress-test it.

A credit choice, not a theme

When I say GPI is built around critical industries, I am not describing a marketing theme. It is a credit choice. We pick companies whose worst case is recoverable, in parts of the market where capital is structurally scarce and the pricing reflects it. Slower to underwrite, less glamorous, harder to put in a headline, and in my experience where growth debt actually performs.

The averages will keep telling a flattering story driven by a few big deals at the top. Underneath them the work is what it has always been. Find the companies with something real behind the loan, structure around the downside, and stay close enough to see trouble early. That does not change with the cycle. It just matters more in a quarter where the headline numbers and the reality underneath them have come apart.