Artificial Intelligence agents are moving beyond assistance and into execution. The emergence of AI agents such as Claude Cowork—capable of autonomously managing workflows across legal, marketing and analytical functions—has intensified a debate among investors: could parts of the SaaS ecosystem risk structural disintermediation?
Software valuation multiples have compressed meaningfully since August 2025 as a result of perceived disruption risk from AI— particularly relative to AI infrastructure beneficiaries. The market appears to be reassessing not only growth expectations but also the durability of software business models.
We examine whether this repricing reflects a fundamental impairment of SaaS economics and where we believe the market may be mispricing both disruption risk and resilience.
Over the past decade, software and SaaS (software-as-a-service) companies have enjoyed a structural valuation premium relative to other tech areas, and more generally the broader market. The rationale was compelling: subscription-based business models provided recurring revenue streams and high earnings visibility. They offered high gross margins, low capital intensity; as companies scaled, incremental margins expanded rapidly, translating into strong free cash flow generation.
In essence, software was viewed as “defensive growth” within technology. Many solutions were deeply embedded in mission-critical workflows, creating high switching costs and durable pricing power. Compared to more cyclical segments such as semiconductors or hardware, software offered smoother earnings trajectories and lower sensitivity to economic cycles.
That premium, however, has narrowed –in some segments, sharply.
Part of the compression reflects capital rotation rather than operational deterioration.
Since the launch of ChatGPT in late 2022, investors have increasingly focused on the compute-intensive nature of large AI models. As hyperscalers have embarked on unprecedented capital expenditure cycles to expand their datacenter capacity, investors have re-directed their focus to hardware markets which offered multi-year visibility on upward revised earnings.
In contrast, software in general has yet to demonstrate comparable monetisation from AI. While embedding AI in existing software products has enhanced some of these offerings, the immediate impact on revenue and profitability has been modest. Investors have shifted their exposure from software to hardware and semiconductors, compressing software multiples in relative terms.
The recent product announcements from OpenAI and Anthropic investors question the long-term durability of revenue that has long characterised the software business model, as AI tools will represent increased competition. If AI systems, such as Claude Cowork, can autonomously execute tasks that previously required human interaction within multiple software tools, that opens the door for disintermediation risk across existing SaaS tools.
With AI disruption risk plaguing software, the market is increasingly pricing in near-term profit erosion – reflecting concerns about increased AI-investment by software vendors, pricing pressure and competitive displacement. In stark contrast, profit expectations for important submarkets in the semiconductor and hardware space keep going up as pricing continues to improve and order books are consistently breaking record-highs.
While we acknowledge these near-term headwinds and adopt a more measured view of the software sector in general, we believe disruption risk varies significantly across business models. Not all software segments are equally exposed.
Vulnerability can be high for software tools that primarily provide a UI (user interface) for relatively standard tasks. Autonomous AI agents may bypass that traditional user interface, orchestrating tasks across systems without requiring human input. Similarly, software vendors primarily serving SMBs could be more exposed, as these customers tend to be more price-sensitive and more agile in adopting AI-native tools. In such cases, AI may compress pricing power and intensify competition.
However, we believe certain segments of the software market have more resilient business models. Cybersecurity is an example, or complex software platforms that are deeply embedded in enterprise operations and very costly to integrate; or software tools serving highly regulated industries such as healthcare, financial services or government. In these segments, data security and compliance are paramount; besides, data infrastructure software such as databases will actually see near-term benefits as AI systems require structured, secure and well-governed data inputs.
The distinction is critical. The decline in software valuations and the narrative of a “SaaSpocalypse” reflect legitimate questions around margin durability, competitive dynamics and the defensibility of long-term growth in the industry. But it does not imply that all software business models are suddenly impaired.
AI is unlikely to eliminate software service economics, but is redistributing value within the ecosystem. For investors, the opportunity lies in identifying which segments are structurally vulnerable, and which remain positioned to capture value in an AI-enabled economy. The current broad-based sell-off in software has increased dispersion within the sector and offers opportunities for active stock pickers focused on selectivity and fundamental analysis.
Get information faster with a single click