Constellation Software

IThe business

Constellation Software buys software companies that serve one narrow industry each, such as bus scheduling, hospital records, property registers, school administration or hotel booking, and then keeps them. Its website says it now has more than 1,100 businesses in nine operating groups.⁶ It looks for companies that sell their own software, have a spread of customers, low customer loss and mostly recurring income; it sets no size limit, and it says it keeps what it buys indefinitely and leaves local managers in charge.¹³ Mark Leonard started the company in 1995, when it bought Trapeze Group, which Mark Miller had co-founded in 1988.¹¹¹⁴

Everything on this page is in US dollars unless it says C$, because Constellation reports in US dollars under IFRS. The shares trade in Toronto in Canadian dollars, so the price and my per-share values are in C$, with the US$ figure beside them. The exchange rate is 1.3978, the Bank of Canada’s average on the price date.² The year ends on 31 December.

How the money is made

In 2025 revenue was US$11.62B, up 15%. Maintenance and other recurring fees were US$8.70B, or 75%. Professional services were US$2.13B, hardware US$382M and new licences only US$415M.³ Recurring revenue is what customers pay each year to keep using the software, have it supported and, in some cases, have it hosted. It is the reason the businesses are worth owning. Customers also pay ahead: deferred revenue, which is fees collected for service not yet delivered, was US$2.2B at the end of 2025.⁷

Costs are mostly people. Staff cost US$5.92B in 2025, which is 51% of revenue, and of that US$1.62B was research and development, US$1.22B maintenance, US$1.20B professional services, US$702M sales and marketing and US$1.18B general and administration.³ Total operating expenses, including depreciation but before the amortization of acquired software, were 74% of revenue, which leaves about 26%. Amortization of the intangible assets that come with acquisitions was another US$1.18B, a non-cash charge, and the main reason reported net income of US$512M is so far below the cash the business makes.³

Three layers: the core, Topicus and Lumine

The group has three layers, and a reader needs to keep them apart. First is the core, run by operating groups named Volaris, Harris, Jonas, Perseus, Vela and Modaxo, among others. Second is Topicus.com, a European group, and third is Lumine Group, which buys communications and media software. Both Topicus and Lumine were separated by a dividend in kind: Constellation gave its own shareholders shares in them and kept control. After the Topicus reorganization on 4 January 2021 Constellation distributed 39.4 million Topicus shares, and on 23 February 2023 it distributed 63.6 million Lumine shares.¹⁹²¹ Both trade on the TSX Venture Exchange. Constellation keeps 60.65% of Topicus and 61.40% of Lumine, and still reports 100% of their sales and costs, so their outside holders appear as non-controlling interests.⁷ In 2025 Topicus had revenue of US$1.76B and Lumine US$766M, 15% and 7% of the group’s US$11.62B.⁷ At the latest closes Topicus was C$96.94 and Lumine C$26.54.²³

The reason this matters to a shareholder in Constellation is that the group’s reported revenue overstates what belongs to Constellation’s own shareholders. The company’s own cash measure, free cash flow available to shareholders, takes the outside holders’ share off, US$243M of it in 2025.³ I work from that measure.

The size of the buying

Constellation spent US$1.34B net of cash acquired on businesses in 2025 and US$1.44B in the first half of 2026 alone. After 30 June 2026 it completed or committed to a further US$669M of cash purchases plus US$149M of deferred payments.³⁴ Deals in the last year include DerbySoft, a hotel-technology company, bought as a majority interest through the Juniper Group of Vela and closed on 1 June 2026, and Synchronoss, a cloud business for telephone companies that Lumine agreed to buy for an enterprise value of US$258M.²⁴²⁵ Topicus has also built a 23.14% stake in Asseco, a Polish software company, which Constellation accounts for under the equity method.²⁶⁵ Most deals are far smaller than these and the company does not list them.

The state of play now

The second quarter of 2026 was good on cash and thin on growth. Revenue rose 17% to US$3.34B, cash flow from operations was US$477M, up 10%, and free cash flow available to shareholders was US$345M, up 57%.²⁷ Almost all the revenue growth came from purchases. Organic growth, which is the change in sales of businesses already owned or bought a year ago, was 3%, and 1% once currency is removed.²⁷ Q3 results come out after the close on Friday 6 November 2026.²⁸

IIThe moat

A software company is hard to leave when it runs a customer’s daily work, is cheap next to what it controls, and has been in place for years. A bus agency scheduling routes, a city clerk filing permits and a hospital laboratory sending results are examples. Constellation’s businesses are mostly of this kind. I can’t prove it from outside, because the company doesn’t publish customer loss by business, but I can check the shape of the revenue.

The check is in the revenue types. In 2025 maintenance and other recurring revenue grew 6% from existing and recently bought businesses, at reported exchange rates. Professional services fell 2%, licences fell 6% and hardware rose 6%.³ The line that customers pay for year after year held up; the lines tied to new projects and new sales did not. That is what a durable base looks like. In 2020, the one hard year, organic growth was minus 3%, and revenue still grew 14% because purchases continued.¹⁸

Second, the businesses sit in many small markets. Constellation says it has more than 1,100 of them, and in 2025 it spent US$1.62B on research and development, a pace that matches 14% of revenue.⁶³ A competitor has to beat a specialist in each market and at that market’s scale, which is usually small enough that a venture-funded rival has little reason to try. This is the part that AI puts at risk, and I come to it below.

Where it is thin

Organic growth is low. At constant currency it was 3% in 2025 and 1% in the first half of 2026, and the 10-year average of what the company reports, on a mix of definitions, is about 2.4%.^{1617181920212234} A durable edge lets the owner raise prices above inflation each year. A growth rate of 1% to 3% suggests the company is mostly holding customers, not raising prices on them. The protection holds the cash flow in place; it does not expand it.

One business in the group is clearly shrinking. Altera, the hospital software bought from Allscripts in May 2022, had revenue of US$153M in the second quarter of 2026, down 22% at constant currency from a year before, partly because a year earlier it booked US$17M for two contracts and an amendment. The rest of the group grew 2%.⁴ It matters because Altera is a large, mature business of the sort Constellation buys, and shows what the downside looks like when the customers leave: revenue falls by a fifth in a year. It is also a reminder that Constellation’s own letters call shrinking businesses its “next best” use of capital when growing ones can’t be found at a fair price.²⁹

Third, the protection is a collection of small ones. Each business is only as safe as its own customers’ habits, and there is no shared brand, network or patent that stops a rival from picking off a niche. What the group shares is a method for buying and running them. That is the advantage, and it is a different thing from protection for any one product.

IIIThe competitive advantage

The advantage is in how Constellation buys. Mark Leonard wrote in 2021 that most of its competitors use as much borrowed money as they can and sell what they buy within three to seven years.¹⁰ A founder who wants his staff and customers looked after, and doesn’t want to be sold on again, may take a lower price from a buyer that keeps. That is the company’s pitch, and its criteria page makes it: it buys and holds, preserves the culture and leaves leaders in charge.¹³

Second, it has a large group of people who do this for a living. In 2017 Leonard said 26 operating group and portfolio managers spent more than half their time on acquisitions, with 60 more full-time acquisition staff across the company.²⁹ I don’t have a current count. The structure is now nine operating groups, so the number of people looking at deals has risen.⁶

Third, it has patient capital and a record. The 2018 letter set the policy: invest all retained capital when it can reach its target return, and if it can’t find enough deals, keep the cash and wait.³⁰ Cash was US$2.89B at 30 June 2026, which shows the patience is real.⁴

What the advantage has to fight

An advantage in buying shrinks as the buyer grows. US$1.2B of net purchases in 2021 was a large number of small deals; US$1.4B in the first half of 2026 alone means larger deals, more brokers and more competition. The February 2021 letter said so. It stopped special dividends, said the company would pursue much larger software deals at lower hurdle rates and said “I anticipate that CSI’s return on investors’ capital will decrease.”¹⁰ The company made three large software deals in its first 26 years and said it had put under 10% of its free cash flow into that segment.¹⁰ The brokers who sell large businesses invited it to only 16% of the sales processes it knew about.¹⁰

That is a plain description of the trade-off. The company can buy small businesses at high returns and can’t spend all its cash that way, or it can spend all its cash on larger businesses at lower returns. It chose the second. The question for the price is how far the return falls.

Rivals, and how it compares with the closest names on this site

I haven’t found a rival’s audited figures that I can set beside these in a table without guessing at definitions, so I leave out peer multiples. Two comparisons on this site are closer. Tiny Ltd and Bending Spoons are also serial buyers of small businesses; my base value for each implies about 9% a year, against about 8% here. Intuit, Adobe, GoDaddy and Booking Holdings are the names where the market is afraid of AI, and they share that one idea with Constellation, so owning two or three of them is not three separate bets. The valuation section sets the numbers side by side.

IVManagement

Who runs it

Mark Leonard resigned as President on 25 September 2025 for health reasons and stayed on the board. Mark Miller, the Chief Operating Officer, was appointed President the same day.¹¹ Miller joined the board on 9 December 2025.³¹ On 27 March 2026 the company said Leonard would not stand for re-election at the 15 May annual meeting, and that he would stay as an adviser on a plan for “Permanent Engaged Minority Shareholder” investments, which are long-term minority stakes alongside the company’s usual purchases of whole businesses.⁸ The shares closed at C$4,137.62 on 24 September 2025, the day before the resignation, and C$3,660.00 two days later.¹

Miller has worked with the group for more than 30 years, and the company says he co-founded Trapeze in 1988, the first company it bought in 1995. His role has been to develop leaders and buy more companies.¹⁴ Jamal Baksh is chief financial officer and Bernard Anzarouth the chief investment officer.²⁸ On 21 October 2026 the company named Daan Dijkhuizen, chief executive of the Topicus operating group and head of its venture arm, as Chief AI Officer, to widen that venture arm into AI-native software companies and to help the operating groups use AI.³² A director, Claire Kennedy, resigned on 31 August 2026 to become Canada’s Consul General in Chicago.³³

How they are paid, and what that pays them to do

The pay plan is plain and it is aimed at the right thing. Senior bonuses depend on return on invested capital and net revenue growth, where return is bonus net income divided by the equity capital shareholders have put in. For most operating groups the board nets 5% off the return as a “risk free” charge, and if the return does not beat it there is no bonus.¹² Executives are required to put 75% of their after-tax bonus into shares, held in escrow for an average of at least four years, and the same executive can take a bonus fully in cash once in five years.¹² The circular says the company has no formal rule against hedging that stake.¹²

Amounts are modest for a company this size. Mark Miller was paid US$1.04M in 2025, the chief financial officer US$866K, and the three operating group heads named in the circular between US$1.5M and US$2.6M.¹² Leonard took no salary or bonus in 2023, 2024 or 2025, and Miller has waived both from 1 January 2026.¹² So the people at the top are paid in the shares they must buy, and the pay for the managers below them rewards earning a good return on the capital that was spent to buy their business. This pays them to keep the return up, which is the right thing, and it pays them to avoid deals that look large and thin, which is the risk the 2021 letter itself raised.

I did not find a current figure for how many shares Leonard or Miller own. The circular says no one owns or controls more than 10% of the shares.¹² I won’t guess a number. It means the leaders are heavily invested through bonus rules, but I can’t say how much of their own wealth is in the stock.

What they did with capital over a full cycle

In 2018 to 2025 the company made US$9.3B of owner cash by my measure and spent US$8.8B of it, net of cash acquired, on purchases. It paid a regular dividend of US$4 a share each year, and a US$20 special in 2019, which the 2021 letter ended.^{16171819202122310} The share count has stayed at 21.2 million since 2016: no buybacks and no issuing.⁹³ So per-share owner cash grew from about US$29 in 2018 to about US$98 in 2025.¹⁶³ That is a good record, built at returns of 15% to 20% a year on the purchases, by my arithmetic.

The reading I can’t settle is what happens after the founder. Leonard’s ability, by the evidence of his letters, was to say no. The 2017 letter describes a large deal that “didn’t quite meet our hurdle rate” and let go.²⁹ The 2021 letter says he “stopped arguing” and accepted lower hurdle rates for larger deals.¹⁰ The buying has since grown from US$1.2B in 2021 to US$1.4B in the first half of 2026 alone.¹⁹⁴ What I can’t judge is whether Miller and Anzarouth will say no as often.

VFinancials

The reported numbers need translating, and the translation changes the picture. Reported net income was US$512M in 2025, down 30% from US$731M, because of currency losses of US$154M, a US$440M revaluation charge and a US$260M charge for revaluing an equity-method investment to cost.³ Cash flow from operations was US$2.73B, up 24%. I look at cash.

The company’s own cash measure is free cash flow available to shareholders, which it calls FCFA2S. It is operating cash flow after interest, lease payments and equipment, with the outside holders’ share of Topicus and Lumine taken off, and with the non-cash IRGA charge taken off as well.³ The IRGA is the agreement under which the Joday Group, the former owners of Topicus’s main business, can sell their units to Constellation at a formula price; the liability is US$1.19B.⁵ I treat that as debt, so I add the charge back, and I take out interest received because I count the cash separately. The bridge shows both steps.

Owner cash was US$2.07B in 2025 and about US$2.12B over the 12 months to June 2026. That is four times net income. There is no stock-based pay to take out, because bonuses are cash and the shares executives must buy come from the market, not new issues. The money needed to stand still, which is equipment (US$68M in 2025) and lease principal (US$131M), is already inside the number.³ What is not inside it is the cost of buying growth.

The buying, and what it earns

From 2018 to 2025 Constellation put US$8.8B, net of cash acquired, into acquisitions and made US$9.3B of owner cash. The figure for 2025 was US$1.34B and for the first half of 2026 US$1.44B.^{1617181920212234} On top of that, deferred payments owed on recent deals were US$224M current and US$208M later at the end of 2025.⁷

The company used to publish a return on invested capital. Adjusted net income divided by average invested capital was 24% in 2009, peaked at 38% in 2015 and was 29% in 2018, the last year it was reported. Average invested capital kept appearing until 2022, when it was US$4.64B.¹⁶²⁰ The fall from 38% to 29% was what Leonard’s 2021 letter predicted when larger deals came in.¹⁰

Since 2018 I have to build the return myself. If the 2018 businesses had grown owner cash by 2% a year with no purchases, then the extra owner cash since is what the purchases earned. Divided by the net spend of earlier years, that was about 14% in 2022, 17% in 2023, 17% in 2024 and 20% in 2025. It is a rough measure: it ignores the debt used and the timing of deals, and the 2% path is my assumption. But it is a measure of real cash, not an accounting profit, and it says the machine still earns well above the 9% I discount it at.

The balance sheet

The company has more cash than a buyer of small businesses needs for a year or two. Cash was US$3.09B at the end of 2025 and US$2.89B at 30 June 2026, against US$4.2B of debt; about US$1.3B of the cash is inside Topicus, Lumine and other subsidiaries that have their own lenders, and the company warns that there are limits on moving it up.⁴ The debt with recourse to Constellation itself is US$1.47B, of which senior notes at 5.158% and 5.461% make US$1.0B.⁵ Net debt is small against owner cash of about US$2.1B. The balance sheet is not the risk.

The share price tells its own story. It rose from C$610 at the end of 2016 to C$4,445 at the end of 2024, then fell to C$3,301 at the end of 2025. From the high of C$5,181.96 on 8 May 2025 to the low of C$2,258.18 on 17 February 2026 the shares fell 56%; they have since regained 53% of the low.¹ The fall began before the AI call: the stock dropped 6.0% on 25 September 2025, the day of Leonard’s resignation, and 5.9% the day after.¹

VIValuation

First, how this goes badly wrong

Before any numbers, I ask what would hurt an owner most. Three things, in order of how much they would cost.

AI makes the small markets contestable. Constellation’s businesses are safe partly because each market is too small to attract a well-funded rival. If AI lets a handful of engineers build a credible rival for any one of them, customers get a choice and price pressure follows. The company says as much in its own risk factor: AI “may also reduce barriers to entry” and bring “price pressure.”³ Two things would show it: maintenance revenue stops growing, and the price of the businesses it buys stops falling to fit the risk. Organic growth is already 1% at constant currency in the first half of 2026.⁴

The return on new money falls as the cheque gets bigger. The company has said its return on investors’ capital will fall as it does larger deals.¹⁰ If new purchases earn 8% instead of 15%, the engine that turns cash into more cash is only a little better than a bond. My yield scale below shows what that costs.

The leader who said no is gone. Discipline is what the figures depend on, and I can’t measure it from outside.

These three are not independent. If AI makes sellers want more for their businesses, or owners less able to sell, the second follows the first, and the third is how it would show in the numbers. That is why I do not multiply them as separate odds; the bear case moves them together.

What an owner can really take out

Owner cash was about US$2.07B in 2025 and US$2.12B in the 12 months to June 2026, or about US$100 a share, C$140. At the price, the market value is US$52.3B, so an owner buying the whole company receives a cash yield of about 4.2% on US$2.19B, my estimate for 2026, before anything is spent on growth. Reported net income was US$512M, under a quarter of that. The gap is mostly amortization of acquired software, which costs no cash, and the non-cash charge on the Joday liability, which I treat as debt.³ The cash number is higher than the earnings number for a real reason, but amortization is also a record of what was paid for the businesses, so I do not think the income statement is wrong to count it. The honest summary is that the cash is real and the price paid for the growth is not in it.

What I can’t judge

I can’t know what the next US$50B of purchases will earn. That depends on prices that sellers want, which I can’t see, and on AI, which no one can. The company says it doesn’t know either: Leonard wrote that management “don’t purport to know the future of AI.”³⁵ So I show a range. If every year’s purchases earn 20% less than my base, the value is C$2,122; 40% less, C$1,210; 20% more, C$4,268.

How the cases are built

I start with owner cash of US$2.19B for 2026, my estimate from the first half. Each year the existing businesses grow by an organic rate, and the year’s purchases, which absorb all the owner cash, earn their yield from the next year on. I give credit for half of the roughly US$2.1B of purchases made or committed in 2026, since that part is not yet in this year’s cash.⁴ After 2036 the business keeps growing at a stated rate, and pays out what it does not need for that growth. All cash is discounted at 9%: the 10-year Canada bond at 3.78%, plus a 5.2-point premium for equity risk.³⁶ Essentially all of the value sits after 2036, because I assume everything is reinvested until then. That is how the machine works, and it is also why any model of it is sensitive.

Three comparisons help. My base grows owner cash 14.7% a year over ten years, less than the 19% a year it managed from 2018 to 2025. The base spends US$50B on purchases over the decade, against US$8.8B in the last eight years, which means the purchases must be larger and the buyer has to find them. And the value of the existing businesses alone, with no purchases and 2% growth, is C$2,020, about two-thirds of my base.

Topicus, Lumine and Asseco

I do not value Topicus and Lumine as separate lines. Constellation consolidates them, and the owner cash I use already takes out the outside holders’ 39% and 39% shares.⁷ Their shares trade at C$96.94 and C$26.54. To check my work against those prices I would need their share counts, which I did not take from a filing, so I did not. The Asseco stake is in the claims at its carrying value, about US$0.7B.²³⁵ The Joday liability is a debt of US$1.19B.⁵ These net to about US$1.3B with the rest of the net debt, under 3% of the market value.

The price needs one of three things. Purchase yields have to be 6% better each year than my base. Or organic growth has to run 1.1 points above it for ten years. Or the discount rate has to be 8.65%. Any one is plausible, which is why the price is only 11% above the base. The three are not independent, though: faster organic growth and better yields both come from healthy small markets, and both are what AI puts at risk.

The margin of safety, and the price against the range

My margin-of-safety price is C$2,180, US$1,560. It is 70% of my base value. I pick 70% because the main input, the return on future purchases, is the one thing I can’t observe, and a 30% cushion covers the lower half of my yield scale. At that price the existing businesses alone are worth C$2,020, which is close to that figure, so the purchases are nearly free. The shares last closed below it on 5 January 2023.¹ There is no margin of safety at C$3,447.02.

The price is 11% above my base and sits between base and bull, about one seventh of the way across. The weighted value of C$3,300 (25% bear, 50% base, 25% bull) is 4% under the price. The market’s case is that the machine has not been broken by the departure of its founder or the arrival of AI, and the stock fell 56% and has since risen 53% from its low. My case against it is that organic growth is 1%, the buying has to grow into larger deals at lower returns, and the shares are not cheap compared with owner cash of 4.2%. Both views fit the evidence so far. The market may well be right.

The next best use of the same dollar

The closest comparables on this site are Tiny and Bending Spoons, which also buy small software businesses. My base for each gives about 9% a year to year 10, against 8.2% for Constellation. The pages for Intuit, Adobe and GoDaddy, where the market is afraid of AI too, give 13% to 16%, with prices below base. Constellation has the stronger cash record and balance sheet of the group, but those pages show a lower price against their base. Business quality and price are two questions, and Constellation answers the first well and the second only at a lower price.

Those AI-fear pages share an idea with this one: that software that does a narrow job is being repriced for AI. So holding Constellation, Adobe and Intuit is one bet and not three.

If the bull case needs many things at once, I multiply them. For the bull value of C$5,550 the company must keep organic growth near 4% for ten years (about one in five on the history: it did that in two years out of ten), earn 14% to 19% on US$64B of purchases (about one in three, given the 14% to 20% so far and the lower hurdle rates for larger deals), and see no real damage from AI (about two in three). Multiplied: 20% x 33% x 67%, which is about 4%. I put the bull at about 1 in 20.

Where this could be in 5 and 10 years

To reach C$4,911 at the end of 2031, owner cash has to reach about US$4.4B, up from US$2.2B, which means US$17B of purchases in five years at 15% yields. The most likely way my base is wrong is the yield on purchases. A 20% shortfall takes about C$1,000 a share off the value.

VIIThe case for the price

What most people miss

Three things cut in Constellation’s favour and get little attention. First, the cash is four times the earnings: US$2.07B against US$512M of net income in 2025, because most of the gap is non-cash amortization.³ Second, the people at the top take no cash pay. Leonard waived salary and bonus for three years, Miller has waived both from January 2026, and managers below are paid on return on capital and must put 75% of their after-tax bonus into the shares.¹² Third, the businesses without Altera still grow: maintenance and other recurring revenue rose 4% at constant currency in the second quarter of 2026 without it, and the whole group grew 1%, not 2%, because of Altera’s decline.⁴

The bull case

In plain words: the VMS businesses are slow, sticky and cheap for their customers, AI helps them more than it threatens them because the software holds the customer’s data and workflow, and the buying machine will keep finding deals that earn well above the cost of capital. The company supplies the pieces. Leonard’s 2021 letter says the business can put its cash to work at high returns and will pursue larger deals.¹⁰ Miller said in March 2026 that Leonard’s “philosophy and operating discipline” will endure.⁸ In September 2025 Leonard said management does not claim to know the future of AI, and in October 2026 the company named a Chief AI Officer to push AI-native investments through its venture arm.³⁵³² I found no named analyst or fund manager whose published view on this stock I could confirm from a primary document, and a web search was not available to me, so I name no one.

Against reality

If purchases rise from US$2.4B in 2027 to US$8.6B in 2036 and every year’s purchases earn 16% falling to 11%, owner cash reaches US$8.6B in 2036 and the shares are worth C$3,110. The price needs yields about 6% better than that. Today’s numbers are close to the line in one respect and far from it in another. The first-half 2026 purchases of US$1.44B are on pace for about US$2.9B this year, already above the base figure for 2027, and the returns on 2019 to 2024 spend were 14% to 20%. But organic growth is 1% and the first-year returns on the new, larger deals can’t be seen yet.

I multiplied out the bull odds in Valuation: about 4%. The weighted value is C$3,300, and the price is 4% above it.

What a 10x would take. Ten times C$3,447.02 is C$34,470, a market value of about US$522B. At 25 times owner cash, a multiple near today’s 24, that needs US$20.9B of owner cash, 9.5 times the US$2.2B I estimate for 2026. At today’s ratio of owner cash to revenue, 17%, that is about US$120B of revenue, against US$12.6B over the last twelve months. In ten years that is 26% a year compounding. Constellation’s owner cash grew 19% a year from 2018 to 2025, and its share price rose 5.5 times in the last ten years, 18.6% a year. So it would have to beat its own record at ten times the size, while small deals at 15% yields run into the size of the deal pool. That is close to impossible, and I rest no case on it.¹¹⁶³

Where I come down

What moved me was the cash and the record on purchases: 14% to 20% on my arithmetic, and no buybacks or new shares since 2016. What didn’t was the price, which already pays for the machine to work a bit better than I assume while organic growth is 1% and AI and succession are open. What would raise my values: organic growth of 3% or more at constant currency for four quarters, yields above 15% on the larger deals, and maintenance growth of 4% without Altera.

VIIIRisks, and what would change my view

The three ways this goes badly wrong are at the start of Valuation. These are the others.

• Currency. The company reports in US dollars and the shares are priced in Canadian dollars. If the Canadian dollar rises 10% against the US dollar to about 1.27, the same US-dollar value is worth 10% fewer Canadian dollars, and my base falls from C$3,110 to about C$2,799.²
• Altera and the old healthcare pools. Shrinking businesses can run for years, as the company says, but revenue fell 22% in a quarter.⁴
• Government and health customers. Harris serves public sector, healthcare and utilities customers, and budgets and rules matter more there than in private markets.⁶
• Subsidiaries and the Joday liability. Topicus and Lumine have their own boards and lenders, and US$1.3B of the group’s cash sits there.⁴ The Joday liability of US$1.19B can be called in cash.⁵
• Disclosure. The company stopped publishing ROIC after 2018 and average invested capital after 2022, so the main number, the yield on purchases, has to be rebuilt from cash flows.

AI, the fear this shares with a few other pages

The fear is that AI lets small teams build credible rivals to narrow software and lets customers do the work with general tools. Adobe, Intuit, GoDaddy and Booking Holdings are the pages on this site where the market has priced the same idea, so they and Constellation are one bet. The evidence so far is mixed. Maintenance revenue still grew 6% in 2025 across the group; the core without Altera grew 2% at constant currency in the second quarter of 2026, against 3% for the whole group in 2025.³⁴ The company’s appointment of a Chief AI Officer and its venture arm are a hedge: buying AI-native businesses with the cash.³² That hedge costs money and comes with the same price risk.

The two-point and 30% cases together land at C$1,372, a little under my bear of C$1,440. My bear is therefore a serious AI case and not a worst case. A real break, with customers leaving and sellers unable to find buyers, would take the stock below it.