What a wild week on the markets. In what has been dubbed the “TACO” (Trump Always Chickens Out) trade, the markets were on the brink of some extreme volatility due to tariffs, but levelled out to end the week after some agreements were made.
It’s been a hot start to the year for me, with my portfolio up around 3.5%. At one point, it was up over 5%. Although I find it hard to imagine we can have another year of double-digit returns on the market, I would certainly take it.
I hope you liked my Foundational Stock releases. There was a bit of shuffling on both lists, however, the core usually stays the same. Any questions, feel free to put them on the Q&A, or simply reply to this email.
In this week’s newsletter, I want to go over one of the hottest stories on the market, besides the waves Trump is making, and that is the large-scale drawdown in software stocks.
Why cover this?
Well, I have been hit with a wave of questions on Foundational Stock Constellation Software, which falls into the software bucket.
This was a company that was first featured here at Stocktrades Premium back in 2019. So, it’s been a staple here for a long time, and is under pressure for the first time pretty much since its IPO back in 2006.
A lot of investors understand that AI has the potential to disrupt software. However, I think where many investors slip up is not understanding how AI can disrupt software and why I think a company like Constellation is not necessarily sheltered from this risk, but less exposed.
Let’s dig right into it.
Why are software stocks getting crushed right now?
As I mentioned, at this point, nobody’s asking if AI is the culprit. We all know this. What we’re now asking is whether or not this school of thought is overblown or justified.
This SaaS selloff isn’t just about a bad quarter. In fact, a lot of these companies are reporting fairly strong results. Look to a company like Adobe below. Free cash flow has never been higher.
If one were to look at that chart with no knowledge of what is going on, they could make the simple assumption that Adobe’s stock price has gone up. However, it’s down by 60%+.
If there is one thing you can take away from this newsletter, it is that whatever these companies have done in the past is completely irrelevant. The market is always forward-looking, and for software stocks, this is especially important.
Companies like Intuit, ServiceNow, Adobe, Constellation, Salesforce, and many others are already down double digits in 2026. For most, total drawdowns exceed 40%.
So, the real question is why? It comes down to the “seat-based model,” primarily.
The seat‑based model is under siege
The core issue is simple: AI agents don’t need user seats.
Most of these software companies make the bulk of their money on a per-seat basis. If they sell a piece of software to a company with 100 employees and that company doubles in size to 200 employees, the software company’s revenue, in theory, doubles.
I say “in theory” because not everyone at the company might need a seat. However, you get the point:
Higher headcount means more licenses, and SaaS companies count on these increases in seats to drive recurring revenue.
So, how is AI impacting this?
To sum it up in a single sentence, AI tools can now handle work that used to require dozens of paid seats.
In order to explain how, I need to speak on a piece of technology that is rocking the software world.
Claude Code & Claude Cowork
Anthropic released two products that we can attribute a large amount of the software drawdown to.
Claude Code launched in February 2025 as a developer tool that automates coding tasks. This is an advanced tool. It doesn’t have much of a user interface and is more so utilized by developers.
What does Claude Code do? Effectively, it takes away the grunt work for a lot of these companies.
Let’s say you have a mountain of customer data you need to organize. Phone numbers, contracts, whatever it may be. A traditional LLM will tell you how you should organize it.
Claude Code will take over your desktop and do it for you.
In comes, Claude Cowork
On January 12, 2026, the company released Claude Cowork.
Claude Code is complex. It would have required developer skills in order to be utilized. With Claude Cowork, however, it has a front-end interface that can be utilized by someone with zero coding experience.
The idea is simple. You give Claude Cowork access to your desktop, describe what you want done, and it executes its work.
It can read, edit, and create files. It builds Excel spreadsheets with working formulas, generates PowerPoint presentations, and handles multi-step workflows without waiting for prompts at each stage.
The most alarming thing? Anthropic used Claude Code to build Claude Cowork in 10 days. AI is building its own AI software.
An example using a company like Adobe
Historically, a large company’s media team might have consisted of 5 people who worked through any particular piece of digital media for, let’s say, an advertisement campaign.
One person creates a multitude of sizes for different platforms, the other applies branding (logos, colors, etc). Another person is the copywriter. Then, you have the asset manager, which keeps all the files logged, tracks conversion rates, etc. And finally, you’d likely have a manager who oversees the entire operation.
In theory, Claude Cowork absorbs every single person’s job on that team outside of the manager. The manager can now tell Cowork to do the job of the copywriter, the designer, the asset manager, etc. Not only can it do these things, but it does them in a fraction of the time.
So what does this mean for a company like Adobe? A company could cut their seats back from 5 to 1, which poses a substantial issue.
Claude has pulled in 18.9 million users through this no-code approach. Many are discounting it, stating that it’s nowhere near good enough.
Even if one were to agree with this, we do have to remember that 18 months ago, AI couldn’t figure out how many fingers to put on a human hand in terms of image generation. Now, it can create direct clones of people and celebrities that are indistinguishable from the real deal in both image and video format.
Whether enterprises trust AI agents for mission-critical workflows without extensive human involvement will be a key trend to watch. Right now, I’d say the answer is no. But two years from now? Who knows. Which is exactly why these software companies are taking a thrashing.
Software companies need to adapt their business models, or risk material disruption
The recurring revenue model that made software companies some of the best businesses to own on the planet isn’t cutting it anymore.
Growth has slowed across the sector, there is zero question. Look below to the chart of Adobe again. A once 20%+ top-line grower, this has declined to 10%.
The software companies that survive will likely be the ones that can provide a solution-based business model, or charge substantially more for AI-based seats.
Here is a prime example from a company like Salesforce:
Salesforce
At this point, Salesforce charges per seat. Let’s say a company has a 20-employee customer service department that answers 1,000 tickets every month. That company pays Salesforce for those 20 seats.
That same company turns around and utilizes something like Claude Cowork to train its AI model to be able to handle those 1,000 customer service requests. All of a sudden, 20 seats turn into 1 or 2 seats.
The company has reduced its overall seat count by 90%, an absolute disaster for Salesforce.
So, how does Salesforce counter this? There are two ways.
The first being, charge significantly more for AI access. Lets say out of the 2 seats remaining, one of them is an AI agent. Salesforce will say:
“The cost to utilize an AI agent is now 10x that of a normal seat.”
The second way to counter this is to say:
“You are now no longer paying us per seat. Instead, we will charge you based on every successful customer service ticket you complete.”
Problem solved, right? Well, not really.
The path forward is usage-based subscriptions, but the picture becomes cloudier
With usage-based subscriptions, revenue becomes less predictable. Recurring revenue is what made these companies so sought after. Forecasting gets harder when consumption fluctuates quarter to quarter.
If a company has a slow quarter, customer service-wise, the old model would still have generated $XX per seat. Now, revenue would decline, as fewer tickets were settled.
This is precisely why you have seen the earnings multiples of many of these software companies collapse. I’ve attached a chart of some companies below. Keep in mind, this is the price-to-earnings ratio, but in percentage decline.
Adobe’s P/E ratio has declined by effectively 50%. Meaning if a year ago the market was paying 40x earnings, they’re now paying 20x.
The main reason these companies traded at 40x+ earnings was sticky, predictable, recurring revenue and high margins.
Now, with the fears that a usage-based model will be the key to survival, plus these companies having to develop new AI tools inside their platforms, hitting margins, nothing is predictable anymore.
When this happens, valuation multiples collapse.
How is Constellation Software (TSE:CSU) impacted by all of this?
As I mentioned, most major software players like Salesforce or Adobe are built on models that scale with headcount. Their revenue is directly tied to how many humans are clicking buttons. If an AI agent like Claude Cowork replaces 10 employees, those software companies lose 10 subscriptions.
Constellation, however, dominates vertical market software, where the model is often built on site-wide maintenance or mission-critical “tolls” that a business must pay regardless of how many people are on the payroll.
Yes, Constellation does have exposure to seat-based software. However, it is a smaller portion of the business, and is more widespread across industries.
Constellation’s software acts more like a utility bill for a specific industry, such as a municipality’s water billing or a transit authority’s scheduling, than a per-person productivity tool.
In addition to this, the threat of AI disruption is often a matter of scale and incentive. Venture capitalists and startups see large disruption potential in software. And for the most part, I agree with them.
They want to disrupt CRM companies like Salesforce or image editing companies like Adobe because the prize is enormous. The total addressable markets are massive.
They have little incentive to spend millions of dollars building an AI agent specifically designed for the highly niche and small markets where Constellation thrives, such as software for independent car washes or library management in specific regions.
Some of Constellation’s pieces of software address markets that are $20-50M in size. Adobe’s TAM? It is estimated to be over $240B.
Even as tools like Claude Code make the act of writing code significantly cheaper, the actual “coding” of a program is often the least expensive part of a vertical software business.
In Constellation’s world, the real value lies in expertise, regulatory compliance, and deep integration with legacy systems. I’ll give you an example below. Here is an image of a piece of software called Constellation Justice Systems, owned by CSU.
It was developed in 1996. The Seattle City Attorney’s Office is still using this piece of software today. They have spent millions of dollars attempting to move on to a better system. However, they always fall back to this.
I have absolutely zero coding experience, and I could code a better user interface than this. That is the exact point I’m trying to make. The coding is irrelevant.
AI doesn’t know the specific legal requirements for tax filing in a remote county or the 20-year-old technical quirks of a local hospital’s database.
The bull and bear case for Constellation
I don’t want to give the impression that Constellation is immune to these software headwinds. The market is fairly efficient at pricing in what it thinks. The market doesn’t re-rate a stock’s valuation multiple by ~40% for nothing.
However, I do think in this case, there is a high probability tha the market is wrong about the impacts of AI in relation to Constellation Software.
Just like I thought the market was wrong on Bull List Stock Aritzia (TSE:ATZ) in 2023. The market thought the brand was falling out of favor and sold it off from $50 to $23. I, on the other hand, thought the issue was simply some mistiming of inventory purchases.
The market was efficient at pricing-in the brand falling out. However, the market was ultimately wrong.
The same could be said with Bull List Stock BRP Inc at $40/share (now $105) or Exchange Income Corp at $40 (now $96).
Why I think the market is wrong about Constellation
The bull case for Constellation in this environment is that they may be the greatest beneficiary of AI.
If platforms like Claude Code allow their thousands of developers to work ten times faster, Constellation’s internal costs to maintain their existing software library will plummet, allowing them to extract even higher margins from their sticky customers.
Additionally, as a world-class capital allocator, Constellation is not tied to a specific industry. If they see a segment of the software market becoming too risky due to AI, they can simply stop acquiring in that space and pivot their cash flows toward different, more resilient niches.
They are basically a diversified portfolio of essential, “boring” businesses, making them one of the few software entities that can use this SaaS drawdown as an opportunity to acquire more VMS companies at discounted prices.
How do we approach this from a situation where the market might be right?
As always, we are dealing with probabilities here. If we believe there is a 70/30 chance that Constellation thrives with AI versus being disrupted, we need to prepare for that 30%.
While their niche markets are currently protected by a lack of incentive for big AI players to enter, tools like Claude Code significantly lower the barrier to entry for small, nimble competitors.
At some point, if these AI tools do get good enough, a $50M TAM market might be attractive to enter.
The most reasonable way to limit the risk of being wrong is to set proper allocations. Constellation is roughly 4.5%-5% of my portfolio, and it will stay that way. Where people get into trouble is when they see a stock decline and decide to go from a 5% allocation to 10%.
