Business
Bottom Line Up Front
Constellation is not a software company; it is a permanent-hold capital-allocation machine that happens to own ~1,000 niche vertical software businesses. The thing to underwrite is whether the compounding engine — disciplined hurdle-rate acquisitions of small VMS companies funded by maintenance-fee annuities — can keep deploying $1.5–2.0B per year at attractive returns now that the base is $11.6B. Two distortions worth separating before any judgment on the business: a large chunk of FY2025 earnings noise comes from a single accounting item (the IRGA liability and the Asseco mark) rather than from operating performance, and the AI threat to installed VMS bases has not yet shown up in the maintenance line.
Read CSU as a holdco with three components: (1) a wholly-owned VMS engine, (2) controlling stakes in two separately listed VMS roll-ups (Topicus and Lumine), and (3) a growing minority stake in Asseco Poland. The consolidated income statement blends all of this, plus a Euro-denominated derivative-like liability (the IRGA) that has nothing to do with operating performance. Free cash flow available to shareholders, not GAAP net income, is the signal.
1. How This Business Actually Works
The revenue engine is a stack of small annuities. Each acquired business unit sells maintenance and other recurring fees to a captive installed base — utilities, public-sector agencies, banks, dealers, clinics, libraries — and adds smaller, lumpier streams from new licenses, professional services, and hardware resale. In FY2025, maintenance and recurring revenue was $8.7B of $11.6B total (75%); the other three lines exist mostly to win and retain that maintenance stream.
FY2025 actuals from the management discussion and analysis. Maintenance is the only line that grew organically at the same pace as the cycle (5%) and is also the only line CSU acquires for. The other lines exist to defend it.
The economic engine in five steps.
Where the profit really comes from. Headline gross margin (37.6% in FY2025) understates VMS economics because each acquired business carries its own services, support, and hardware-resale headcount. The real return shows up further down: capex is 0.6% of revenue ($68M on $11.6B), working capital is structurally negative (customers prepay maintenance), and free cash flow conversion is 81% of EBITDA. The combination of low capital intensity and a permanent reinvestment runway is what turns a 4% organic-growth business into a 20%+ compounder.
Two cost lines deserve attention. Hardware gross margin slipped from 44% to 43% in FY2025 and to 40% in Q4 — a small but visible signal of mix pressure as hardware revenue grew 27%. The "IRGA / TSS membership liability revaluation charge" was $440M in FY2025 and $204M in Q4 alone; it is a Euro-denominated obligation linked to Topicus governance, is not an operating cost, but sits inside reported expenses and crushes optical net income.
2. The Playing Field
CSU is the inventor and scale leader of the listed VMS roll-up model. The five most apples-to-apples public comparables are one US-listed roll-up (ROP), two pure-play vertical SaaS organic operators (TYL in US government, JKHY in US community banking), and two direct CSU spinoffs running the same playbook in different verticals (TOI in Europe, LMN in telecom/media).
Source: FY2025 reported financials for revenue, FCF margin, ROIC, ROE, P/FCF, EV/EBITDA. Organic growth is FX-adjusted constant-currency where reported (CSU 4%, TOI ~9%, TYL ~9-10%); LMN is approximate. Market caps as of mid-May 2026 (TOI converted EUR→USD).
Two things jump out. CSU has the lowest FCF margin in the set — by design, because consolidating ~1,000 services-heavy business units dilutes the rate but expands the absolute pool. CSU and ROP trade at roughly identical 19x FCF, the lowest multiples in the group; TYL pays a single-vertical premium at 31x. CSU is being priced like ROP (a mature compounder) even though its acquisition cadence and organic growth profile differ.
What the peer set reveals. ROP is the closest valuation comparable but runs a different shape (fewer, larger deals; software-led pivot from industrial; reported gross margins near 70%). TYL and JKHY are the pure-play "what good looks like in one vertical" benchmarks — JKHY's 22.9% ROE shows the moat economics of a single sticky vertical. TOI and LMN are the cleanest read-throughs to CSU itself: same playbook, smaller scale, faster organic growth (compounding off smaller bases), and they trade at lower P/FCF than CSU even though they execute the same model. That is partly the rate-of-deployment story (CSU has $11.6B of revenue to fund growth from) and partly that CSU consolidates both, so a slice of CSU's market cap is just their listed values, marked daily.
3. Is This Business Cyclical?
Operationally, no. Reported revenue has risen every single year for 20 consecutive years, from $165M in FY2005 to $11.6B in FY2025 — through the GFC, the European sovereign crisis, the 2015-16 industrial recession, COVID, the 2022 SaaS drawdown, and the 2025-26 software re-rating. Customer software budgets in mission-critical operations do not get cut: a county still needs property-tax billing in a recession, a credit union still needs core processing, a transit agency still needs scheduling.
Two decades, zero down years for revenue or for FCF. The flat-to-down spots are in net income (heavily distorted by acquisition accounting and the IRGA mark), not in cash generation.
Where the cycle actually bites. Three places, in order of speed:
- License and professional-services revenue. Licenses were down 8% and professional services down 4% organically (FX-adjusted) in FY2025 as customers defer net-new IT projects. Maintenance kept growing 5% organic (and 6% in Q4), more than offsetting. This is the only place the operating cycle shows up in CSU's results.
- Acquisition pricing. The biggest threat to compounding is not customer churn; it is multiple expansion in the private VMS market that crushes ROIIC on incremental deals. Management noted in Q1 FY2026 that "there's a real disconnect between SaaSpocalypse publicly-traded stuff and private markets," meaning private comps stayed stubbornly elevated even as public software re-rated down. If that gap closes, CSU's deployment math gets easier.
- FX translation. CSU transacts in 100+ countries and reports in USD; FY2025 saw a 1-2 percentage-point drag on reported revenue and a $154M foreign-exchange loss line — presentation noise rather than economic cyclicality.
4. The Metrics That Actually Matter
Standard SaaS scoreboards (ARR, NRR, magic number) miss what matters here. The right scorecard is built around the compounding engine — does each incremental dollar of FCF buy a productive new VMS business at a price that earns the hurdle rate?
CSU's signature is lowest gross/FCF margin but highest absolute scale of cash deployment in the cohort. JKHY shows what a single-vertical compounder looks like when the engine is fully mature (16.5% ROIC, balance-sheet light). ROP sits at the opposite end — heavy goodwill, low headline ROIC, lots of leverage. CSU is closer to ROP than to JKHY structurally; the difference is CSU has more disciplined small-deal pricing and a deeper deal funnel.
The single number to watch quarter-over-quarter is capital deployed on acquisitions. FY2025 deployment was $1.34B, down from $1.52B in FY2024 and $1.70B in FY2023, while cash on the balance sheet jumped from $2.0B to $3.1B. That combination — building cash, slowing deployment — is the earliest signal that either (a) the deal supply has thinned or (b) management is holding the hurdle rate firm even as private multiples stay elevated. Either explanation matters for the thesis.
5. What Is This Business Worth?
The right lens is price-to-FCFA2S for the underlying compounding engine, supplemented by a partial sum-of-the-parts for the two listed spinoffs and the Asseco minority. Pure consolidated metrics (P/E, EV/EBITDA) miss two things: the IRGA mark contaminates net income, and the listed values of Topicus and Lumine are observable separately from CSU's wholly-owned VMS engine.
A compact sum-of-the-parts. Not a precise valuation; a frame that shows where the value sits and why headline multiples look the way they do.
Roughly $4B of CSU's ~$40B market cap is already publicly priced as Topicus and Lumine stakes. The other ~$36B is the bet on the wholly-owned VMS engine, which generated $1.59B of FCFA2S in FY2025 — roughly 23x FCFA2S for the core compounder. That is in line with where ROP trades, slightly below TYL's pure-play premium, and well above what TOI and LMN themselves trade for.
What would make this cheap or expensive. Cheap setup: capital deployment re-accelerates to $2B+/year at maintained hurdle rates, organic growth stays at or above 4%, and the AI threat does not show up in maintenance churn. Expensive setup: cash continues to build because deal supply has run out at acceptable multiples, organic growth deteriorates below 2% for two consecutive years, or succession after Mark Leonard introduces visible discipline drift. The valuation is much less about the multiple and much more about whether the engine can keep finding deals.
6. What I'd Tell a Young Analyst
Three things matter for this stock, in this order.
First, ignore reported net income — track FCFA2S and capital deployed. GAAP net income in FY2025 was distorted by a $440M IRGA revaluation, a $260M Asseco reversal, and rising acquisition amortization. CSU's own non-IFRS reconciliation (operating cash flow less interest, lease costs, capex, and the IRGA mark, less non-controlling interests) is the cleanest measure of what the engine produced for common shareholders. It grew 14% in FY2025 to $1.68B and is the only metric senior management uses internally to gauge whether the model is working.
Second, the bear case is real but is not what the headlines say. The risk is not "AI replaces niche software" — installed bases of mission-critical software with regulatory accreditation and customer-data lock-in do not get rebuilt over a weekend by a startup with Cursor. The risk is deployment crowding: at $11.6B of revenue, CSU needs to find ~$1.5-2.0B of acceptable deals per year just to stand still on the compounding rate. The early indicators are already partly visible — cash building on the balance sheet (up from $2.0B to $3.1B in FY2025), deployment slowing (down 12% YoY), and average acquisition multiples drifting above 1.5x revenue (not visible yet but the metric to back-solve every quarter).
Third, three things would change the thesis materially. (i) Two consecutive quarters of negative organic growth in the maintenance line — the first signal that the AI threat is real. (ii) A visible departure from the hurdle-rate discipline (a transformative large deal, a buyback funded by debt, a re-segmentation of operating groups to chase growth) — that would mean post-Leonard governance is drifting. (iii) Sustained capital deployment above $2B per year at organic growth above 5% — that would be evidence the engine still runs at scale and would support a re-rating closer to a single-vertical premium than to a mature-compounder discount.
The single most useful frame: read each quarter not as "did revenue beat?" but as "did the company convert this quarter's cash into next quarter's revenue at the historical IRR?" Everything else is noise.