Constellation Software: The Capital-Allocation Machine at Full Price
A ROIC spread of almost 13 percentage points and still no margin of safety. Quality yes, price no: why CSU is a Hold.
Constellation Software’s 2025 consolidated financial statements show net income of $512m [1]. From the same accounts, a free cash flow of $2,664m can be derived [3]. More than five times the profit arrived as cash, and yet the share has lost around 40% from its high in 2026 [3]. The reason is not the arithmetic but a new question: for the first time, the market is seriously asking whether AI-assisted software development erodes the moats of the thousand niche providers CSU has bought up over three decades.
Whether CSU is a good business is not up for debate as far as I am concerned; the figures below prove it beyond doubt. The question is whether the fallen price now offers a margin of safety. My answer after working through fiscal year 2025: no, not yet.
CSU earns 19.4% on its invested capital, 12.9 percentage points above its cost of capital, and free cash flow has exceeded net income by a factor of three on average over nine years. The core business ties up no capital; all of the value is created at the point where surplus cash flows into new acquisitions. My weighted fair value of CAD 2,734 sits around 8% below the price of CAD 2,969, and 78–85% of it depends on the terminal value, that is, on the long-term assumption AI is currently calling into question. Quality yes, price no: Hold.
This article is part of a six-part series on vertical-market-software compounders. The other parts: Lumine Group, Topicus, Asseco Poland, CHAPTERS Group and the VMS comparison. The topic dossier Tokenisation provides the framework for the group’s AI risk.
The foundation: a business that needs almost no capital
CSU earns three quarters of its money from recurring maintenance and subscription revenue; in fiscal 2025 that was $8,700m of $11,623m in revenue [1]. The business delivers a high, predictable cash flow and ties up hardly any operating capital in doing so, because customers pay in advance: the deferred-revenue balance of $2,891m effectively finances the ongoing business [1]. That is what makes the ROIC calculation peculiar and, at the same time, revealing.
A methodological point first. For a serial acquirer, I consider the standard split “ROIC including and excluding goodwill” too narrow. CSU amortises not only goodwill but, above all, acquired technology and customer assets: $1,182m of amortisation in fiscal 2025 alone [1]. Economically, these intangibles are part of the purchase price paid. I therefore split including and excluding all acquisition intangibles, meaning goodwill, customer and technology together. That fundamentally changes the finding.
| Metric (FY2025, $m) | Value | Derivation / source |
|---|---|---|
| EBIT (after amortisation & impairment) | 1,852 | Revenue 11,623 − expenses 9,728 − impairment 43 [1] |
| Tax rate (normalised) | 27% | between 19% effective in Q1'26 [2] and the cash rate; reported 37.6% distorted by non-deductible revaluation [1] |
| NOPAT | 1,352 | 1,852 × (1 − 0.27) |
| Invested capital incl. | 6,965 | interest-bearing debt 5,787 + equity 4,267 − cash 3,089 [1] |
| Acquisition intangibles (book value) | 8,368 | goodwill 1,782 + technology 2,606 + customer 3,980 [1] |
| Invested capital excl. | −1,403 | 6,965 − 8,368 → negative |
| WACC | 6.5% | Ke 6.8% (CAPM, β 0.66) · Kd 5.5% · D/V 10.5% |
The ROIC finding: three lenses, one conclusion
| Lens | ROIC | Spread vs. WACC | What it measures |
|---|---|---|---|
| Incl. intangibles (book value) | 19.4% | +12.9pp | Return on capital currently employed |
| Leonard logic (cash NOPAT / gross IC) | 16.2% | +9.6pp | Return on all capital ever deployed in M&A ($13,703m) |
| Excl. intangibles | n/a | ∞ | Core business is financed with negative capital |
The third finding is the decisive one. Strip out the acquired intangibles and invested capital turns negative: the operating core business needs less capital than CSU carries in equity and debt. In economic terms, operating capital efficiency is effectively infinite. Every dollar of value creation arises not in operations but at the point where surplus cash is allocated to new acquisitions.
“At CSU, the source of value is not the business but the decision about what happens to the business’s money.”
Where the growth comes from, and where it does not
Breaking down the reinvestment cements this logic. Organically, the reinvestment rate is negative, around −10% of NOPAT: through its negative working capital, the business actually releases capital. The whole game runs through acquisitions, and there CSU reinvests around 80% of its free cash flow [3].
The value-creation equation, value creation ∝ (ROIC − WACC) × reinvestment rate × growth, therefore has an unusual structure at CSU: a high positive spread multiplied by a high reinvestment rate that sits entirely in the M&A channel. As long as CSU finds enough targets at reasonable multiples, the machine runs. This is where the bear takes aim.
What the earnings line conceals: nine years of cash flow
Anyone valuing CSU on its price-to-earnings ratio, LTM at around 60 [3], is measuring the wrong quantity. The cash flow statements of the past nine years show why: free cash flow exceeds reported net income by a factor of three on average. In 2025 the conversion was as high as 520% [3].
This section uses the broad “free cash flow” (CFO − CapEx, FY2025 ≈ $2,664m [3]), the measure that shows the conversion versus earnings most clearly. My DCF, by contrast, deliberately uses the narrower metric reported by the company, FCFA2S (Free Cash Flow Available to Shareholders, FY2025 = $1,683m [1]), which additionally deducts lease repayments and the non-controlling interest (NCI) — and is therefore what actually accrues to CSU shareholders. The conversion message holds under both definitions; for valuation, the more conservative figure is the right one.
| Year | FCF ($m) | Net income ($m) | FCF / NI | M&A / FCF |
|---|---|---|---|---|
| 2017 | 508 | 222 | 229% | 50% |
| 2019 | 733 | 333 | 220% | 69% |
| 2021 | 1,271 | 310 | 410% | 93% |
| 2022 | 1,256 | 512 | 245% | 125% |
| 2023 | 1,737 | 565 | 307% | 98% |
| 2024 | 2,129 | 731 | 291% | 71% |
| 2025 | 2,664 | 512 | 520% | 50% |
| Avg. 2017–25 | n/a | n/a | 302% | 75% |
Source of the series: TIKR Terminal [3]. The driver of this gap is an item most analyses overlook: the amortisation of capitalised contract costs (“deferred charges”). It amounted to $689m in 2025 and thereby even exceeded the amortisation of acquisition intangibles ($505m) [1]. Both are non-cash expenses that depress book earnings without ever costing cash. Add the two together and most of the $2.2bn by which free cash flow exceeded net income in 2025 is explained.
The P/E of 60 is an optical illusion. On the basis of free cash flow, CSU trades closer to 18x to 19x: still no bargain, but an entirely different story from “60x earnings”. Whoever measures CSU by earnings considers the share absurdly expensive; whoever measures it by cash sees a highly profitable business at an ambitious but not irrational price.
The FCF series also demonstrates the robustness of the machine. Over eight years, free cash flow grew at around 23% per year, with a single, marginal down year (2022, −1.2%) [3]. Such consistency across a full cycle including the interest-rate turn is rare and the strongest empirical argument against the AI bear: so far, nothing has slowed the cash generation.
The financing question: self-funded or on credit?
A frequent objection to serial acquirers is that their growth is debt-financed. The cash flow statements refute this for CSU, with one qualification I find instructive. Until 2020, CSU funded its acquisitions entirely from free cash flow after dividends; net use of debt was minimal [3]. With the large deals of 2021–2023 (Allscripts, Optimal Blue), meaningful leverage entered the picture for the first time: in 2022 a funding gap of around $400m was covered with debt [3].
What matters is what happened next. In 2024 and 2025, free cash flow grew faster than M&A volume, so that CSU again generated a clear surplus, around $1.2bn above its M&A requirement in 2025 [3]. The interest burden from the senior notes issued in 2024 is comfortably manageable at roughly 8% of free cash flow [1]. CSU has thus shown that it uses leverage when needed but can bring it back down on its own. I read that as proof of discipline, not as a warning sign.
Valuation: greatness is priced in
Because CSU reports a levered free cash flow after interest (FCFA2S, $1,683m in fiscal 2025 [1]), I discount it directly at the cost of equity, with no further deduction of debt. Three scenarios, each with its own discount rate and terminal growth:
Bear case (30%)
CAD 1,525
- Discount rate 11.0% · terminal growth 2.5%
- AI erodes the moats, organic growth below 4%, more expensive M&A environment
- −49% vs. price
Base case (50%)
CAD 2,693
- Discount rate 9.5% · terminal growth 3.5%
- M&A machine keeps running, growth normalises to 10–14%
- −9% vs. price
Bull case (20%)
CAD 4,648
- Discount rate 8.5% · terminal growth 4.5%
- Retirement wave of software founders delivers more targets, AI lowers costs across the portfolio
- +57% vs. price
Probability-weighted (30 / 50 / 20), the result is a fair value of CAD 2,734, around 8% below the current price of CAD 2,969 [3]. The terminal-value share is 78–85%. That is unavoidable for a compounder, but it should be read as a warning sign: almost the entire value lies beyond the forecast horizon and therefore depends on a long-term growth assumption, and that is the very thing AI is currently calling into question.
Bear versus bull
- Structurally the best capital allocation, +13pp ROIC spread, consistent for decades
- Negative working capital: the business funds its own growth
- FCF exceeds earnings by a factor of 3; the true multiple is ~18x, not 60x
- FCF CAGR of 23% over 8 years, only one marginal down year: robust through the cycle
- Retirement wave of software founders enlarges the pool of acquisition targets
- Decentralised model is built against key-person risk
- AI lowers barriers to entry in precisely the niches that make up CSU’s moat
- Organic growth of only 4% [1]; without M&A there is no growth
- If the M&A environment becomes more expensive, the value-creation equation breaks
- Mark Leonard is not standing for re-election to the board [4]: the end of an era
- 78–85% of the DCF value sits in the terminal value
Four points in the calculation rest on my own assumptions and should be checked against the original filing: the normalised tax rate of 27% (interpolated between the effective Q1 2026 rate of 19% and the cash tax rate); the EBIT definition, i.e. whether impairment and FX count as operating or one-off; the treatment of the IRGA liability as financial debt within invested capital; and the beta of 0.85 used for the discount rate instead of the reported 0.66. Anyone setting one of these assumptions differently arrives at a different fair value — the direction of the verdict does not change.
Conclusion
CSU is outstanding in quality; the ROIC finding and the nine-year cash flow series leave no doubt about that. The optically off-putting P/E of 60 is misleading: on the basis of free cash flow, which exceeds net income by three times on average, CSU trades closer to 18x to 19x. But even that is fair value, not a discount.
My weighted fair value sits around 8% below the price. Even after the 40% drawdown, the share offers no margin of safety, but at best fair value against a real, new risk: AI.
Rating: Hold / watchlist, no buy at the current price. For existing holders I see no reason to sell: the machine works, and cash flow continues to grow at a double-digit rate. As a new investor I would stagger my entry and open a first tranche only below around CAD 2,400, near the midpoint of the bear and base cases. Add only if the AI bear proves overdone and FCF generation remains stable. Patience is the position here.
Update log
- 05.09.2026 — Editorial revision to house style, list of sources added; figures and rating unchanged.
- 06.06.2026 — First publication based on the audited 2025 consolidated financial statements and the Q1 2026 interim report; price basis 05.06.2026.
Sources
- Constellation Software Inc., Results for the Fourth Quarter and Year Ended December 31, 2025 (press release with audited IFRS consolidated financial statements for FY2025; full statements and MD&A on SEDAR+), GlobeNewswire, 09.03.2026. Revenue and recurring revenue, EBIT components, net income, tax rate, FCFA2S, amortisation, balance-sheet items (deferred revenue, debt, equity, cash, goodwill and intangibles), organic growth, senior notes. → https://www.globenewswire.com/news-release/2026/03/09/3251696/0/en/constellation-software-inc-announces-results-for-the-fourth-quarter-and-year-ended-december-31-2025-and-declares-quarterly-dividend.html
- Constellation Software Inc., Results for the First Quarter Ended March 31, 2026 (unaudited Q1 2026 interim report), GlobeNewswire, 12.05.2026. Effective tax rate Q1 2026. → https://www.globenewswire.com/news-release/2026/05/12/3293525/0/en/constellation-software-inc-announces-results-for-the-first-quarter-ended-march-31-2026-and-declares-quarterly-dividend.html
- TIKR Terminal, market data for Constellation Software Inc. (TSX: CSU), as of 05.06.2026 (closing price): price CAD 2,969, decline from high, LTM P/E, beta; historical cash flow series 2017–LTM (free cash flow, net income, M&A volume, funding balance).
- Constellation Software Inc., Announces Mark Leonard’s Decision to not Stand for Re-Election to Board of Directors, GlobeNewswire, 27.03.2026. Founder’s departure from the board. → https://www.globenewswire.com/news-release/2026/03/27/3264152/0/en/constellation-software-inc-announces-mark-leonard-s-decision-to-not-stand-for-re-election-to-board-of-directors.html