Subject: Re: Topicus: Not Constellation
Topicus reported Q1 2026 on 5 May. On a first pass it looked pretty ordinary. Revenue up 23%, organic growth 5%, FCFA2S up only 2%, reported net income down 21%. The shares followed the headline, drifting from about C$100 to C$90.18, where they now sit. I don't think that it is quite right, but the simple read is that this is still a good compounder but it's starting to mature a bit faster than bulls hoped.

The operating cash flow line was much better than the headline FCFA2S number. Pre-working-capital operating cash flow was €106.5m, versus €80.9m last year, up about 32%. FCFA2S was held back by a smaller seasonal working-capital benefit and higher financing and lease costs. Those are of course real costs, but they don't say much about whether the underlying software businesses are still working. The recurring revenue side was fine, with organic growth of 7%, roughly where it has been every quarter since Q1 2024. The weak line was professional services, where organic growth was -1%, probably the thing in the quarter worth watching. The reported net income decline is mostly noise: Q1 2025 had a €31.4m fair-value gain from Asseco that flattered last year's number, and stripping that out, Q1 2026 net income was actually up about 42%.

The Asseco piece has also moved on. The Warsaw price has come back from a YE25 peak of around PLN 191 to PLN 175. Even so, the 23.14% stake is worth roughly €791m on a €523.55m cost basis, with about €267m of unrealised gain still sitting there. There is also the dividend: PLN 13.05 per share, ex-date 13 May, payable 22 May. Topicus gets about €59m of cash, which lands in Q2 FCFA2S. Useful, but not normal. It is an unusually large payout, around 92% of consolidated 2025 net profit, helped by the Sapiens disposal. I would not capitalise that Q2 boost. The next dividend, paid in 2027, will be meaningfully smaller, with the 2025 base inflated by the Sapiens disposal.

At C$90 and LTM FCFA2S of about €222m, the trailing multiple is around 21x. Strip out the Asseco stake and the dividend receivable and the operating business is closer to 19x. Still not a bargain-bin price, but there is more room for error than there was a couple of weeks ago. The bigger issue is that the pitch has changed. A year ago it was simple enough: Constellation-style VMS, small sticky vertical software acquisitions, decentralised operators, steady compounding. That is still in there, but it is no longer the whole business. Cipal Schaubroeck was bought at roughly 10x EBITDA and 2x sales, above their previous 6-8x EBITDA comfort zone. Asseco is only maybe 30-40% true VMS by revenue. This is drifting from a pure VMS compounder into a broader and more mixed capital allocator. That still may work very well, and may even be the right thing to do as Topicus gets larger. But I continue to suspect we will not really know for another couple of years. It is a different capital-allocation problem from the old Constellation/TSS model, where you buy a small sticky niche software business cheaply, leave it alone, and do it again. As such, it probably deserves a slightly lower multiple.

The near-term issue is deployment. In 2025, Topicus deployed around €775m all-in, but close to half of that was the Asseco stake, which you should not expect to repeat at that size. Strip Asseco out and ordinary bolt-on deployment was around €390m. In Q1 2026, plus committed deals since quarter-end, the bolt-on number is around €60m. That is well below last year's run-rate, and the ordinary bolt-on cadence has gone quiet. If Q2 and Q3 look the same, the bull case becomes harder to underwrite. If deployment gets back to normal, the base case still works. The balance sheet is in better shape too: gross debt is around €449m against €331m of cash, so there is room for whatever sensible deals turn up.

The services number is the bit worth watching, with the usual warning label that one quarter is not much of a sample. Recurring software grew 7%. Services shrank 1%. The most plausible explanation is not that AI is suddenly coming for the important stuff. More likely, it is nibbling first at the lower-grade implementation and customisation hours. The mission-critical compliance systems are still buried in local tax rules, payroll conventions, language oddities, municipal processes, and all the other bureaucratic plumbing. That was always a big part of the moat. The code was only part of it. AI is not going to tidy up a Belgian municipal accounting workflow over a long weekend.

If this is the start of something, I would not read it as bad news. The revenue line may look a little less pretty, but the mix should be margin-positive over time. August will tell us more. It also fits the broader point from a couple of weeks ago: in regulated verticals, the rulebook is the asset. The first thing AI eats is the services work around the edge, not the sticky core.