Why I Can't Buy Constellation Software

Investing only has one core principle. You have to predict the future. If you can do that, you’ll do well. If you can’t, you won’t. Listen to Buffett a little closer, and you’ll see this is what he is saying. He doesn’t invest in companies where he doesn’t think he knows the economics in ten years. So, in essence, he doesn’t know the future.

This is my “first principle” of investing. I’ve spent a lot of time trying to come up with more principles, but any other one is a derivative of knowing the future. Diversification? That’s just protection from the first principle. How about Buffett’s “Don’t lose money”? That’s only possible if you know the future. If you hold cash and inflation is horrible, you can lose real purchasing power. How about watching the fees? That’s an interesting one, but it is a mechanical detail, not a principle. How about “own what you know”? This still derives from my first principle. It’s hard to predict the future about something you don’t know anything about.

How does this apply to Constellation Software you might be asking? Well, when you buy an individual company you are making a more specific prediction about the future. And the good news is you don’t have to be right all the time; outsized winners can pay for a lot of mistakes. But you want to make bets only when you think you can predict the future!

So what is the future of Constellation Software (CSU)? Well, let’s start with the basics of what CSU does. CSU is a holding company that buys, manages and builds vertical market software (VMS) businesses. As of today it owns over 1,000 smaller companies that specialize in niches like public transit, libraries, utilities, auto repair shops, etc. Let’s look at an example of one piece of software from the auto shop division.

R.O. Writer

AI Summary: Key Features of R.O. Writer

• Shop Management: It tracks every car sitting in a repair bay, logs which technician is assigned to it, and details the specific work required. [1]

• Smart Estimates & Invoicing: Mechanics use the system to quickly calculate parts and labor costs, building accurate quotes for customers before turning a wrench.

• Inventory Control: The software monitors the shop’s supply of oil filters, brake pads, and fluids. It can automatically order more parts from suppliers when stock runs low.

• Customer Communications: It sends automated text alerts or emails to vehicle owners when their vehicle is ready for pickup or needs immediate attention.

R.O. Writer is vertical market software. It’s only useful to someone running a car repair shop. People running these shops need software to handle these parts of the business, and that is not going away. Constellation Software goes around buying software like R.O. Writer, absorbing it into the collective and sharing learnings/infrastructure across 1000+ pieces of software that small businesses and governments find indispensable. In 2020, this was a wonderful business model. Hiring competent coders and creating your own software made no sense for your local car repair shop. They were going to buy these features from someone, and Constellation Software wants to be that someone.

But how about today? Despite what Google Finance says, CSU trades somewhere around 20 times earnings. They are always spending money to buy more companies, which distorts the “E” in their PE. Even if you disagree with me, let’s say it is 20 times earnings. When someone buys CSU today, they are implicitly saying that CSU is either going to make the same amount of money for 20+ years, or that CSU is going to earn a lot more than it does now in the future, before it goes bankrupt. How likely is that?

Well, in 2020 I would have said that’s very likely. In fact, sign me up. Of course, in 2020 no one would sell me a share of CSU at 20 times earnings. But in 2026, I don’t like CSU at 20 times earnings. In fact, I can’t predict the future four years out. But I will give you my opinion on what’s likely to happen.

Constellation Software’s vertical market strategy depends on your local repair shop not having access to shop management, invoicing, inventory and customer communications at a cheaper price. In the near term, I think today’s agentic coding is going to be a tailwind for this. As of today, your local repair shop isn’t going to use Claude Code to replace R.O. Writer. That’d be insanity, and those calling that out are right. Claude Code will instead be used by Constellation Software to make R.O. Writer even easier to build features for and maintain. And if they don’t do it, another software shop will! But it will be a software shop. Claude Code is a wonderful helper, but it isn’t…. a software shop. So in the meantime, I expect Constellation Software to earn more money, not less, because of Claude Code!

But let’s go back to the 20 times earnings being a 20 year prediction. Or as Buffett says, he needs to be able to predict the economics 10 years out. My prediction is that while Claude Code isn’t a software shop right now, it doesn’t mean an agentic one won’t be built. In fact, I am predicting one WILL be built.

I do not think the future is your local repair shop vibe-coding shop management software. I think the future is your local repair shop outsourcing shop management to an entirely agentic software shop, that has fundamentally lower margins than Constellation Software. For discussion purposes let’s call the future company AGentic Software Shop (AGSS).

How will this agentic software shop (AGSS) look? I don’t know. I’d guess it is going to evolve in phases. At first it’ll naively build a mimic of whatever you currently have. You’ll talk to AGSS and say “I’m currently using R.O. Writer, it’s doing pretty well for us, but I want lower cost and I want these feature tweaks that CSU never got around to”. Perhaps you have to install AGSS on your computer. It watches you use R.O. Writer for a month. It exports data. Then it asks you some questions and then builds your own version of R.O. Writer.

To be clear, this probably won’t be a 100% on the spot agentic coded solution. I would expect AGSS to have a plentiful backend of source code to pull from. Most likely, there is already a base repair shop repository that’s been agentically built, but human tested. What the end user ends up with is 98% shared foundational product with 2% in customizations and add ons. And when learnings, security fixes, bug updates do come along, AGSS will of course seamlessly update all the clients in the background.

The charge? $100 for the software build-out, and then it is web hosting fees plus 5%. That’s AGSS’ model, a 5% fee based on hosting. Something so low that no human-built software company could ever compete with. Could R.O. Writer compete in this environment? Of course it could, just not at today’s margins.

To predict even further into the future, I think we get to the point that all of the features of R.O. Writer are agentic. Shop management, smart invoicing, inventory control, customer communications, an agent handles all of this and the software portion itself grows ever smaller. You talk to your agent who doesn’t need a pretty dashboard. The agent orders more parts, not you, etc. Again, not great for R.O. Writer’s margins.

To me, this future seems highly likely. The only question is when it will be built. When will we get to an agentic software shop? I don’t know, but 20 years is far too long for me to bet against this business model showing up. I can’t blame CSU for continuing to do what it has always done. I don’t blame anyone who thinks agentic software shops are never going to show up and that my predicted future doesn’t materialize. But I do know I’m not willing to make this bet.

My finger in the air guess is 2033 for a real, functioning Agentic Software Shop. My first guess? Amazon. Amazon Web Services could one day announce a service that will build and host many VMS solutions. Your local vet, your local repair shop, your local needlepoint store, all of them, if they want, will be able to use Amazon to build and host software that runs their business at a fraction of the current price.

So for me, I’m willing to buy CSU at a PE of 7, as I don’t think it’ll be dead that quickly. In fact, the next seven years might be amazing for it. But someday, I think its margins get crushed.