While the five-year chart has hardly been thrilling, the one-year chart tells a more interesting story. The shares were nearly C$200 last July and are now around C$95, so this is not simply a case of the stock marking time for years. It is also a fairly sharp de-rating even as the business kept compounding at a decent clip. Two small points though. The cash-flow growth number also depends a lot on which cash-flow metric you are using. FCFA2S is their own free cash flow available to shareholders metric, and it was up 23% in FY25; plain operating cash flow was up 19%. Of course, neither number is exactly a disaster. And the Asseco accounting loss is not quite as phantom as it sounds. There was a 221.7m expense, but there was also 119.7m of income from derivatives, fair-value related adjustments and a dilution gain associated with Asseco, so the hit to reported earnings is more like 100m than the full 222m. Still ugly, of course, just not quite as ugly as the headline makes it look.

At roughly 218.7m of FCFA2S, and converting the market cap into euros, I get something like 23-24x. That is not silly for a business like this, but it is not exactly cheap either. The bigger question, to my mind, is whether the Constellation/TSS approach of buying small, sticky niche software businesses and leaving them largely alone can still work as well in Europe at this size, particularly as the deals get larger. Organic growth was 4% in 2025, so a lot still depends on their ability to keep finding and buying good businesses. And Asseco is not the usual sort of deal in that mould. That is not the standard tuck-in. They didn't buy another small software company outright and leave it to get on with things. They ended up with close to a quarter of a large listed Polish software group, via an initial block purchase and then a larger treasury-share deal once the approvals came through. That is a different and more complicated bet. It may work very well, but I suspect we will not really know for another couple of years. Debt is also no longer trivial. At year-end, current and non-current loans came to about 692.5m before lease obligations, against 326.7m of cash. That is not alarming, but it does make the story a bit less pristine. The model is still attractive, but it is no longer quite as simple as "cash-rich serial acquirer buys endless small software companies."

The AI angle is interesting too. The easy bear case is that one decent engineer can now build a workable replacement for some creaky niche VMS product over a long weekend. But I think this is where your European regulation point matters. If these products are embedded in tax, accounting, healthcare, local government, payroll and the rest of it, then the moat was probably never just the code. It was the compliance burden, the local-language quirks, the ugly integrations, and a customer base with very little desire to entrust anything mission-critical to a new AI-native vendor. In that sort of market, AI may matter less than people think. More bureaucracy is a nuisance, of course, but for a company like this it may not be entirely bad news: in these regulated niches, the rulebook can be part of the moat.