substack.com - SaaS, widely misunderstood ($CSU.TO, $TOI.V, $ADBE, $UBER, $CRM)
A Contrarian Take on the AI Disruption Narrative
Section titled “A Contrarian Take on the AI Disruption Narrative”***TLDR: Follow the economics
Another day, another sell-off in software-as-a-service (SaaS). What I find most interesting about these days is the number of bears coming out, patting themselves on the back, and asserting that their convictions are, in fact, correct. Stock price movement is not indicative of a business’s underlying fundamentals and is meaningless in the short term. It’s possible to have an incorrect read on a company and make money, and vice versa. This should be known, but it’s an incredibly common mistake among investors. The Adobe bears are some of the most vocal right now.
For the past decade, Adobe has posted record revenue and ARR on an absolute basis in 39 of 40 quarters. Last month, Adobe posted its Q4 results, and guess what? Another record print. Management went ahead and gave guidance for the calendar year 2026. Every quarter will be a record quarter for Adobe this year. Despite this, the market didn’t seem to care. The business is down 10%. I’ve come to realize that SaaS, as things stand, is deeply misunderstood. And I can only imagine this is because the average investor is not well-versed in the discipline of software engineering.
I have been a software engineer for 10 years now. My experience ranges from companies as small as five-person startups to large Fortune 500 companies (AWS). I’ve seen the full range of product life cycles, from building software to support millions of transactions per second to deploying MVPs in just a couple of weeks. I know how to scale to support a million users. I know when to prioritize product quality vs speed. I know what fires to let burn and trade-offs to make. I understand the performance, security, and scaling implications of every decision. Do You? It’s called “engineering” for a reason. Building production-grade software at scale is exceptionally challenging. There’s a reason the average Bay Area engineer comfortably clears multi-six figures; AI changes none of that. My Point? Everything I say here today comes from personal experience.
Like every software engineer, I use AI every day in my development process. As a result, I’m intimately familiar with both its strengths and weaknesses. Vibe-coded software, while impressive, is still very far from being a finished product. I don’t think this is obvious to a retail investor, but it is obvious to any competent software engineer. It’s like copying a Picasso and calling yourself a master. Nobody is paying top dollar for a cheap knockoff.
This doesn’t mean I’m an AI skeptic; in fact, I’m extremely bullish on AI. It’s simply clear to me that we are at least a decade+ away from replacing any SaaS in a meaningful capacity. My focus today isn’t going to be on any single business. But rather, why are these businesses selling off in the first place? Why do people even use SaaS? And what companies I feel have attractive valuations here.
Why is SaaS Selling off?
For the better part of a decade, SaaS has been seen as the holy grail of business: exceptional margins, relatively fixed costs, and the ability to grow 10%+ annually without fail. Typically, when these businesses start to hit profitability, every incremental dollar drops to the bottom line, and earnings grow faster than revenue. Without fail, earnings were up and to the right year after year. Investors poured in and priced these businesses to perfection.
The introduction of AI created uncertainty. If a user can build products faster and leverage AI, what is stopping them from re-creating and disrupting an existing business? This poked a small hole in the SaaS thesis, and a business price for perfection could not support this narrative change. SaaS, for the first time in years, has since corrected back to what I feel are much more appropriate levels.
My personal thesis is that 90%+ of businesses are in good standing, and this rests on a simple yet compelling argument: Replacing SaaS with AI makes no sense from an economic or product perspective. Replacing SaaS would cost more and deliver a lower-quality product in almost all scenarios. This is best understood through a simple example that covers the basic economics of a SaaS business.
Slack
2,545 employees at the time of it’s acquistion is 2021
500 engineers at 150k each = $75m R&D annually
The 75M in R&D is grotesquely sandbagged, but we’ll pretend it’s not for simplicity. In reality, it’s easily 500m+ anuually.
The Financials of replacing SaaS do not Make Sense
SaaS is used because you get an exceptional product at extremely compelling financials; building it internally makes ZERO sense. The corporate version of Slack costs 220,000 a year for Slack. 75M in R&D effort annually. In other words, you’re earning a 340x amplifier for your money. SaaS has historically been an exceptional business because of how compelling the financials are for everyone. Slack is able to build an exceptional product because it focuses on one thing, their messaging app. Then they offer it at very compelling financial terms because they can leverage scale. It’s cheap for their customers while still highly profitable for them due to margins and scale.
Now, let’s examine the case of building this internally. Having one engineer plus the AI and cloud expenses will run you easily 100K? **One engineer + AI vs. 500 engineers + AI.**Who do you think builds a better product?
***The value proposition of software as a service has always been a superior product at a fraction of the cost of building it internally. AI only amplifies this.
***Which leads me to the greatest irony of all: the companies that spend the most time and effort replacing SaaS with internal tools are the most likely to be disrupted. In business, nothing matters more than focus. The second you start expending engineering, product teams, and designers to reinvent the wheel, you’re distracting from your core competency, slowing everything and everyone down. Why do you think Netflix won the streaming wars? Why does Spotify lead in streaming music? Why does TSMC make all the chips in the world and not Intel? Focus. The best thing any business can do is invest time and effort in amplifying its own product offerings with AI. Replacing existing products to save negligible costs is a waste of time. “I can AI vibe code x product in a weekend now!”
Facebook was built in two weeks. Dropbox was built over a weekend. Airbnb was built and launched over a weekend. Part of what’s made software such a compelling business is the ability to build and deploy rapidly; that has always been the case. AI doesn’t change this fact only amplifies an existing truth. Coding is the easiest part, always has been. It’s everything else that goes into software engineering that’s really challenging. For example
- You need to make sure the product is mobile-compatible, which requires both a mobile app and a mobile-compatible web app.
- You need to ensure your product is secure, so you must undergo extensive penetration testing and address any security issues that arise. If you’ve added authorization, hopefully you understand how to manage credentials safely
- You want to set up a deployment that can scale regardless of usage. It should be a full end-to-end deployment pipeline, including staging and production, secret management, auto-scaling, etc. I hope you understand AWS
- You also have to add monitoring to your product. You’ll probably want PagerDuty when your product goes down, so your team is aware. You should also probably set p95 and p99 targets and stuff
- You need to make sure your product is accessible and localized
- You need to make sure your database is:
- Consitent
- Low-latency
- Durable
This is heavily watered down and simplified. We haven’t even got to product management, life cycle management, bureaucracy, working with designers and stakeholders, etc. When you start putting all these pieces together, you’ll quickly understand how software engineering becomes complex so quickly. Let’s view this from a different light with a real-world example: Driving.
Driving a task is so simple that 90% of the U.S. adult population is capable of obtaining a driver’s license. Yet there’s not an autonomous vehicle that can safely drive from point A to point B anywhere in the continental United States. I would also argue we are farther away from this than many folks appreciate. At least another five years, and probably more.
So you take something like software engineering, which is exponentially more complicated than driving. And you’ll quickly understand the ask people are making of AI. Driving is supposed to be one of the simplest problems; it’s only uphill from here. The advancement of AI is, of course, impressive, but it still has a long way to go before it’s a genuine threat to SaaS.
“AI will replace SaaS”
SaaS, in its simplest form, is saving, manipulating, and moving data from point A to point B. Its strength is its reliability, predictability, and consistency. You always know what will happen.
This strength is also its greatest weakness, as you’re limited by SaaS inherent lack of flexibility. Of course, this is where AI shines, as it’s dynamic and flexible. This leads me to a simple question.
Would you prefer SaaS + AI, or pure-play AI?
I think the answer to this is fairly obvious: the former. Combining them allows you to play to their strengths while cancelling out their weaknesses. The businesses that are under threat from AI, for the most part, will be the ones that fail to adopt and incorporate next-generation technology into their products.
One final point
Before SaaS became popularized, many businesses built products and tools internally. Most companies already had their own internal messaging, sales software, workflow management software, etc., and those tools were typicallygarbage. Software as a Service became popularized because these businesses realized they wouldn’t have to maintain their own internal tools anymore. They pay less and get a better product. It has never been about the ability to build it. It was always about optimizing for speed, quality, and price. The introduction of AI just makes all these tools more powerful, which makes their value proposition even higher. It also makes recreating them even more difficult because you now have to learn how to customize AI for that specific use case. This brings me to my final point: which businesses are cheap and which I think are worth following here.
Constellation Software ($CSU.TO)
The business sold off for several reasons:
- Fears of AI
- The CEO stepped down
- Trading at dumb multiples
- Concerns over maintaining current growth rates
I find it funny that people can correctly identify that, with the aid of AI, you can build products faster and better. However, no one has yet considered that the number of VMS SaaS products will increase, and the total addressable market will grow. It’s actually far more likely this is the case, and I’ll explain why.
Take a second and think about how most of these companies are started to begin with: they’re not VC-founded. Often, it is an existing business trying to solve a problem, or someone building in their own free time. Then they realize that they can actually sell this product, right? Now think about how many small pains every business has, but they’re not willing to address because of the time commitment.
If you could put something together, hack it together quickly without a preexisting solution, you are far more incentivized to do so. What’s more, if you can accomplish more with fewer people, it opens up the TAM considerably to a wider range of businesses that can support a smaller operating overhead. VMS and SaaS will accelerate.
Topicus ($TOI.V) See the constellation deep dive
Topicus was spun off from Constellation Software. They operate the same business model, but they focus solely on Europe. The company is smaller, trades at higher multiples, but it also has higher growth and potential. Personally, I would probably prefer Topicus, just because it’s more in line with my investment style. I like smaller and high growth. As to why it’s sold off, you can literally copy and paste everything I wrote above about $CSU.TO. I believe Topicus will trade essentially in line with Constellation Software and will also put up solid numbers.
Adobe ($ADBE)
I’d recommend just reading the deep dive. I think Adobe has sold off more in AI than anything, but the relative revenue growth has also slowed. Still, I personally believe, given how high-quality and predictable their business is, it’s trading at very attractive multiples here.
Uber ($UBER) - Honorable mention, not really SaaS TBH
Deep Dive - by HatedMoats
Uber’s strength and moat come from its network effects. Your goal as a driver is to transport as many people as possible, because you make more money. Your goal as a rider is to travel from point A to point B as fast as possible for a reasonable price. The more drivers you have, and the more passengers you have, the better you can solve both of these core problems. For a new entrant, you struggle on both sides of the equation. Even for an autonomous product like Waymo/Tesla.
This also assumes that self-driving doesn’t become commoditized (which it will). In which case, the value will come from whoever has access to the riders, which, of course, will be Uber. As AVs come online, most people will stick with Uber simply because it’s convenient. I pull out Uber to find a ride from A to B as fast as possible. The only thing I care about in terms of price is that the p’s fair. Why would I switch to a product with fewer drivers and longer wait times to get a ride?
What I also find interesting about AVs is that if they do scale, fewer people will want to own a car, as using AVs will become more economical than ownership. I would love for AVs to get economically viable enough that I wouldn’t have to own a car. I save on both money and stress.
SalesForce ($CRM)
I don’t find Salesforce’s valuation particularly compelling here, especially given Adobe, Uber, Constellation Software, and Topicus, which are all better values. That being said, the business has also started this year with a combination of slow growth and concerns about AI. I believe it’s worth monitoring. If it sells off at multiples similar to Adobe’s, I might be interested. **
Conclusion** I firmly believe AI is more likely to be an accelerant than a disruptor. LONG ADOBE.