The question of whether AI will disrupt incumbent software companies has become a focal point of debate in recent months in response to broader sector underperformance:
Whilst I’m usually hesitant to wade into topics that are already heavily covered and highly contentious, this very question recently became more than theoretical when my firm had a proprietary opportunity to acquire a SaaS business.
This was a classic vertical market software business with dominant share in a niche market powering more than 3,000 customers with 45%+ EBITDA margins to boot. We undertook extensive due diligence as a potential 100% owner with no exit horizon. Hence near-term market sentiment mattered much less than whether the business would remain a great business that we wanted to own over the long run.
What follows isn’t a prediction of how this will all play out – just the framing that helped me think through the problem, which I hope can be a useful contribution to the conversation.
The SaaS End State: From Product to Instrument
SaaS investment cases tend to follow a well-worn playbook, with a specific end state in mind.
There is an investment phase at the outset where capital is invested aggressively to build the product, find distribution and entrench the product into as many customers’ workflows as possible. The implicit promise is that once growth slows, this gives way to a harvest phase where the business can be run less like a product company (cashflow consuming) and more like a financial instrument (cashflow generative).
The “rule of 40” can be best understood as a shorthand for software investor expectations: where the sum of a company’s revenue growth and operating margin is expected to be around 40%. You can focus on growth at the outset without profits but the grand bargain is that once growth slows you are expected to generate profits at high margins (40% margins if 0% growth) with limited reinvestment requirements.
In other words, the business should start to resemble an annuity – a stable “coupon” stream.
VMS at Maturity: The Bond Analogy in Practice
If the ideal mature end state of SaaS is meant to look bond-like, Vertical Market Software (VMS) at maturity is the clearest expression of that analogy.
VMS refers to software built for a specific industry – enabling workflows such as dispatch, compliance reporting, billing, scheduling, or customer records management. These products don’t win on elegance or speed of innovation; they win by being embedded in business critical workflows with tight integration into customer data and processes.
Switching is risky, painful, and expensive. Customers keep using the system not because it’s beautiful, but because it is the system.
At the same time, the vertical markets themselves tend to be niche and small and therefore can only naturally accommodate one or few competitors (e.g. how many software companies to coordinate orchestras can exist profitably concurrently?).
Putting it all together, you get the ideal conditions for an annuity-like end state: the software stays installed, revenue is predictable, and the business can be run to optimize for free cashflow. That doesn’t mean neglecting the product – it means maintaining retention with bounded and disciplined spend: incremental, defensive work and just enough investment to avoid revolt and keep the machine running.
Constellation Software has famously mastered this playbook at scale for decades: acquiring mature, sticky VMS businesses with no organic growth and then maintaining – or even increasing – the effective “coupon” over time through pricing discipline and cost-out. A key part of their edge is recognizing that prior owners often underestimate how long, and how hard, those levers can be pulled without breaking retention, and applying that model in an unemotional, repeatable way across hundreds of businesses (with a massive dataset to support their decisions).
Lower Entry Costs, Higher Reinvestment
In VMS, mathematically the “R” in ROI was capped by the small market size, while the “I” required was simply too high to generate a sufficiently attractive return. As a result, even when customers complained, potential challengers couldn’t make the entry math work.
However, if the amount of investment required to build credible software falls meaningfully, that equation shifts. When the “I” in ROI comes down, returns improve mechanically – so long as the “R” remains broadly intact. Markets that were previously too small to justify the upfront effort for entry can start to look viable.
And crucially, competition doesn’t need to arrive as a clean “rip and replace.” It can arrive as unbundling, better point solutions, or simply reducing friction when switching. Often, even the credible presence of alternatives can change buyer behaviour and, in turn, the economics of incumbency.
The moment you’re forced into sustained reinvestment – modernizing architecture, matching feature velocity, and meeting new customer expectations for AI functionalities (with additional costs accruing to your foundation model suppliers) – the shape of the annuity-like end state changes. Price hikes may become harder to push through and a “just good enough” product becomes more costly to maintain. The “coupon” starts to look less stable.
Industries Are Not Static
Software has been a phenomenal industry for decades, and that run has spanned the careers of most investors active today.
For a long time, local newspapers were a classic example of what a “superior industry” looked like and were famously admired by Warren Buffett. Owning the local paper meant owning distribution, attention, and the advertising toll booth in a geographic monopoly. The model also had powerful operating leverage: once the newsroom and printing/distribution infrastructure were in place, incremental ad dollars dropped through at very high margins.
The “media tycoon” was a classic archetype during the era for a reason.
In many cases, the newspapers themselves owned the digital versions of their products. But the internet still rewired the industry’s fundamentals by changing the forces around them: it democratized distribution and lowered barriers to entry, expanded the set of substitutes and shifted consumer attention, and ultimately moved the advertising profit pool elsewhere.
Underwriting the End State
The current debate over whether AI will disrupt software is often framed around first order questions such as: Will the customer vibe-code a replacement system themselves?
But the AI puck is moving too fast and unpredictably for that to be a useful frame. Investing is about underwriting the future, not litigating what’s possible this quarter. And “software” or “SaaS” is too broad a category for sweeping conclusions.
The more relevant question is – does software still converge to something bond-like – highly cashflow generative, capital-light and predictable – or does it converge to something closer to a normal operating business, where sustaining the cash stream requires materially more reinvestment and a permanently higher operating tempo?
This is all and well but how do we think about the flip-side of this – re-pricing the bond in foundry and semicap equipment? Does the 42x multiple on KLA NTM earnings mean that the durability of the bond in that business has gone up significantly? i.e., that the users of compute (software/LLMs et al) are going to endlessly need to acquire GPUs such that semicap bonds are now far more durable and predictable than they ever were?