AI is not a future disruption. It is already pricing in today
The idea of a SaaSpocalypse has quickly moved from theory to market reality. Large language models such as ChatGPT and Claude are no longer just productivity tools. They are beginning to reshape how software businesses are valued, sold and, ultimately, structured.
In this second part of the PCV Forum series, Doug Lawson of MarktoMarket sets out the market backdrop and explores what happens when AI starts to replace core software functionality rather than simply enhance it. The key question is no longer whether AI will impact software companies, but which ones are most exposed or insulated and what that means for valuation models today.
The release of Anthropic’s Claude Codework demo marked a turning point. The demonstration showed AI agents performing complex workflows including financial analysis and project management. The market reaction was immediate and severe. Roughly $285 billion was wiped from SaaS market capitalisation within 48 hours, driven by concerns that entire categories of software could become redundant.
At the heart of this reaction is a structural fear: the potential collapse of per seat pricing. If AI agents can perform the work of multiple employees, the traditional model of charging per user begins to break down. A company that once needed 100 licences may only need 10 to 20 human users, with agents filling the gap. That shift has direct implications for revenue, pricing power, and ultimately valuation.
A useful framework discussed in the session outlines four potential outcomes for how software buying may evolve:
Most incumbents are betting on the fourth outcome. However, it is not yet clear how evenly these scenarios will play out across sectors.
Despite the disruption narrative, deal activity is still happening. However, the composition of those deals is shifting. Rather than traditional SaaS businesses, much of the current activity is concentrated in:
This suggests that capital is flowing towards categories perceived as AI adjacent or AI enabling, rather than legacy software models.
Where transactions are occurring, valuation multiples are significantly compressed. Reported deals include revenue multiples in the range of:
This reflects a more selective buyer market, where only specific assets are attracting competitive pricing.
The funding environment tells a similar story. In Q1 2026:
Capital concentration at the foundation model layer is reshaping the entire ecosystem beneath it.
There is also a useful reminder in looking at industries previously assumed to be disrupted. Publications such as The Economist and The Telegraph were widely expected to decline with the rise of the internet. Instead, both have continued to operate and, in some cases, thrive. The implication is that technological disruption does not always eliminate incumbents. In some cases, it forces adaptation rather than replacement.
The framework of software exposure to AI substitution is becoming one of the most practical ways to understand current market behaviour. The final part of this series will explore how this is affecting M&A processes in real time, including:
Catch up on Part 1 – Market backdrop and valuation context
Watch Part 3 – M&A markets and buyer behaviour
Watch Part 4 AIM, PISCES and the UK capital markets problem: what will make investors actually commit?
Larissa Adams (00:05)
Welcome to Part Two. If you have not watched Part One, Doug Lawson from MarktoMarket sets the market backdrop there. We have included the link below.
The SaaS Apocalypse is a striking name for a live issue, describing what large language models (LLMs) are doing to the valuations of traditional software businesses. Examples such as Claude and ChatGPT illustrate that the impact of AI is already under way. The key question is not whether AI will affect these businesses, but which categories are most at risk, which are more insulated and what this means for valuations today.
Now over to Doug Lawson (00:40)
The SaaS Apocalypse came into focus with Anthropic, the company behind the Claude large language model. It launched Claude Codework earlier this year. The demo showed AI agents handling sophisticated tasks such as financial analysis and project management. This spooked markets because it suggested that large parts of the software industry could become obsolete.
The reaction was significant, with approximately $285 billion wiped from SaaS market capitalisation within 48 hours. A key concern is the potential collapse of per seat pricing. Traditionally, enterprise software is priced per user but, if AI agents replace human users, the number of licences required could fall dramatically.
A useful framework suggests four possible futures for software buying. First, companies may bypass software entirely and rely on large language models. Second, businesses may build their own solutions on top of foundation models. In both of these scenarios, much of traditional software could disappear. Third, AI native software companies may emerge as winners. Fourth, existing software vendors may successfully integrate AI into their products. Most incumbents are hoping for this fourth outcome, which we also consider the most likely.
Deal activity is still taking place, but the nature of that activity is changing. Instead of traditional SaaS, we are seeing more deals in quantum computing, crypto, infrastructure software and AI drug discovery. This suggests a shift in where value is being assigned.
Recent transactions show compressed valuation multiples. Examples include revenue multiples of around 1.2x, 2.3x and up to 4x upfront. Many of these businesses were either loss making or in revenue decline, which may explain the lower valuations.
On fundraising, Q1 2026 saw approximately $330 billion raised globally. Around 81% went into AI model infrastructure and related companies. OpenAI accounted for roughly $122 billion of this total. Anthropic is also preparing a major raise at a reported valuation of around $900 billion.
There is also a historical perspective. Businesses such as The Economist and The Telegraph were expected to be disrupted by the internet but continue to operate and, in some cases, remain successful. This highlights that disruption does not always equal disappearance.
That framework for understanding software exposure to AI substitution is becoming a practical tool for assessing markets. In Part Three, Doug Lawson and Richard Angus will explore how this is affecting M&A markets, including buyer behaviour, deal timelines and valuation modelling challenges when AI impact is difficult to quantify.
