The staunchest AI evangelists claim that AI will “take over the world” any day now and that the ink on the SaaS eulogy is almost dry. While we can’t entirely discount this potentiality, the probability seems quite low. Therefore, this paper provides a horizon-stratified risk framework for investors to assess AI’s effects on SaaS companies. Risk assessment is often about timelines; therefore, we categorize impacts into three horizons: cyclical (within 12 months), secular (one to five years), and super secular (beyond five years).

We believe the risks to SaaS posed by AI proliferation, though meaningful, are much less uniformly catastrophic in the near term than the current narratives suggest. AI poses a threat of substitution to existing SaaS products and business models, but AI tools also offer SaaS firms opportunities for efficiency and growth. The industry faces genuine structural discontinuity that will separate the companies capable of fundamental reinvention from those doomed to obsolescence, but the SaaS sector as a whole is not monolithically vulnerable.

For two decades, SaaS has been one of the most dependable structural stories in technology investing. A business model with recurring revenue, low marginal costs of production, and high switching costs made it attractive to investors. However, the emergence of large language models (LLMs) for human interaction and autonomous agentic AI systems seriously challenges each foundational pillar.

The market has not only taken notice but also taken a chainsaw to software sector market capitalizations. In the first two months of 2026, excluding the Iran war’s negative effect on risk assets, roughly $2 trillion in software sector market capitalization evaporated in what analysts call the “SaaSpocalypse.” During this period, the price-to-earnings multiple of the representative S&P North American Expanded Technology Software Index fell 36%, from 71 to a still-lofty 45.

Within the sector, Workday, ServiceNow, Salesforce, Oracle, and Microsoft fell 37%, 29%, 26%, 25%, and 18%, respectively. In their latest earnings reports, Microsoft indicated slower-than-hoped Azure monetization, ServiceNow provided a decelerating growth outlook, and the others offered guidance that lacked the rosiness of the past.

While it is easy to simply declare SaaS dead, that’s an overreaction. Different threats operate on different timescales, through different mechanisms, and with different degrees of impact. All this ultimately means that potential outcomes have different probabilities of coming true, which in asset pricing leads to different discount rates.

Architectural Displacement by Agentic AI

The secular horizon features a more structurally significant threat. We start from the perspective that traditional SaaS offerings fall into three layers: the data management layer, the business logic layer, and the user interface. AI undermines this architecture.

As Microsoft CEO Satya Nadella explained, in an AI agent-driven world, business logic migrates into an AI orchestration layer (Johnson, 2026).¹ Investors are obviously skeptical of this pivot arriving in the near term, for good reason: It is a fundamental shift in how enterprises structure themselves and requires fundamental change management. However, over a one-to-five-year time frame, this seems very reasonable to expect. Firms today are not ripping out their enterprise resource planning (ERP) systems because ChatGPT exists. But the pressure to do so will increase over time as the technology matures and competitors find cost savings. Gartner estimates AI agents could replace 35% of point-product SaaS tools by 2030 (Huang, 2026).

Commoditization of Feature Sets

Perhaps the most strategically corrosive secular term risk is the commoditization of SaaS functionality. As foundation models improve and costs fall, the barriers to replicating SaaS application features steadily diminish. Vertical AI startups are building competitive products at a fraction of the cost and time required by incumbents. They are reportedly growing at rates approaching 400% year-over-year (Bennett, et al., 2025).

While the goal in competing with entrenched SaaS firms used to be feature parity, that is largely irrelevant when AI can replicate features on demand. What buyers are underwriting instead is data depth, workflow lock-in, and proprietary datasets that create durable differentiation.

Regulatory Proliferation and Compliance Architecture Costs

The medium-term regulatory picture is a genuine headwind for SaaS companies. The U.S. CLOUD Act, China’s PIPL, and India’s Digital Personal Data Protection Act, among others, create a patchwork of conflicting requirements that SaaS firms must navigate. For SaaS companies embedding AI capabilities into their platforms, the EU AI Act requires meaningful compliance architecture, including documented bias testing, audit trails, human oversight mechanisms, and risk assessments for high-impact use cases.

Simultaneously, the EU Data Act extends portability obligations to industrial and non-personal data, directly threatening the data moats that SaaS companies have built over years of customer interactions. If customers can port their data more easily, switching costs decline and erode what once were defensible moats. That this is happening at a regulatory level means that individual firms cannot use their resources to protect their current positions.

The Potential Displacement of SaaS as a Paradigm

Beyond the five year window, fundamental analysts usually subsume assumptions into what is called a valuation terminal value since unknown variables grow exponentially at that horizon. While it’s too early to price with precision, the possibility that the SaaS delivery paradigm itself could become less relevant should be considered. This is not a prediction that it will or that enterprise software will disappear entirely. Rather, it is an observation that the current SaaS architecture combining data, business logic, and interface into a managed subscription may not be the dominant software stack in the 2030s.

Analysis of the emerging software stack by Crawford, McLaughlin, Doddapaneni, and Fiore (2025) describes a three-tier architecture, with systems of record at the base, agent operating systems in the middle, and outcome interfaces at the top. In this model, traditional SaaS applications risk being compressed into the bottom tier, reduced to data repositories, and priced accordingly. The differentiation value migrates to agent orchestration platforms where hyperscalers and foundation model providers hold structural advantages.

A 2025 Accenture study projects that AI agents will be the primary users of most enterprises’ internal digital systems by 2030. Software explicitly designed and priced for human interaction will need to be reconceived for machine-to-machine consumption (Tibbetts and Jones, 2025). Just as hardware has not disappeared but become a commodity, in the most disruptive long-term scenario, SaaS could face an analogous trajectory.

Foundation Model Providers as Vertical Integrators

The final long-term risk carries a certain irony. What if foundation model providers, the very firms whose application programming interfaces (APIs) many SaaS companies are now building on, vertically integrate into the application layer and displace the SaaS intermediaries sitting between them and enterprise customers? As these providers extend from model APIs to agent frameworks, workflow automation, and domain-specific applications, they occupy value chain territory that SaaS companies have historically controlled.

SaaS companies building on third-party model APIs face an uncomfortable reality: They may be funding the competitive displacement of their own business models. While not an immediate risk, this kind of structural dynamic is difficult to reverse once it gains momentum.

The risks above are real, but they are not uniformly distributed across the SaaS landscape. Mission-critical platforms running complex, regulated, and highly domain-specific workflow systems with years of embedded logic cannot be replicated by a generalist AI model overnight or without considerable work and investment. Therefore, for the near horizon, SaaS firms will retain a durable right to earn and therefore can remain profitable components of our portfolios. But going forward, investing in SaaS firms requires renewed appreciation and analysis of the following factors:

• Data Moat Depth. Does the company possess proprietary data assets that AI-native entrants cannot reproduce? This is the single most important durability question.
• Pricing Model Adaptability. Companies still wedded to per-seat subscriptions in a world of AI-driven seat compression face a structural revenue headwind. Those transitioning toward outcome- or consumption-based pricing are better positioned.
• Workflow Depth vs. Feature Breadth. Point-product tools with wide but shallow functionality face a significantly higher risk of commoditization than deep, workflow-integrated platforms.

AI poses genuine and, in some cases, serious risks to SaaS companies. Near-term, the pricing architecture is already under pressure as seat compression materializes and shadow AI creates compliance exposure. Over the one-to-five-year horizon, agentic AI threatens workflow disintermediation, regulatory fragmentation raises compliance costs, and commoditization erodes feature-based moats. Over a longer horizon, the SaaS paradigm itself faces architectural pressures, data-sovereignty constraints, and the looming possibility that foundation model providers could become vertical integrators.

The “SaaSpocalypse” framing implies the entire sector is at risk now and in the near term. The evidence suggests something more nuanced: Roughly a third of the point-product SaaS market faces serious displacement risk over the next decade, while deeply integrated, data-rich platforms with genuine workflow ownership are more likely to coexist with than be replaced by AI.

We’ve already seen enormous multiple compression in software. Some of that compression reflects genuine fundamental risk. Some of it reflects market extrapolation of worst-case scenarios that are unlikely to materialize for every SaaS company.