"Wrongly Sold Off" or "Justifiably Corrected"?
Last week, the US stock market's AI software sector put on quite a show: a collective plunge followed by a rapid rebound. Analysts have started using the term "wrongly sold off"—implying the market overreacted, these companies were undervalued, and the rebound represents a "valuation revenge."
It sounds satisfying. But I want to pour a bucket of cold water on that notion.
This isn't a case of being "wrongly sold off." Rather, the market is finally confronting a question it has long avoided: AI is not a growth engine for SaaS; AI could be the terminator of SaaS.
Why Is SaaS So Awkward in the AI Era?
The SaaS (Software as a Service) business model is built on three assumptions:
Assumption 1: Software is scarce. Enterprises need to spend hundreds of thousands or even millions on CRM, ERP, and project management software because the cost of developing and maintaining these systems is high.
Assumption 2: Switching costs are high. Once an enterprise deploys a SaaS system, its data, workflows, and training are locked in. The cost of migrating to a new system is too steep, resulting in strong customer stickiness.
Assumption 3: Feature iteration is controlled by vendors. Users wait for SaaS companies to release new versions and features, paying annual fees for upgrades.
AI is simultaneously dismantling all three assumptions:
First, software is no longer scarce. AI coding tools have drastically reduced software development costs. Many core features of traditional SaaS products can now be replicated by an AI agent in just a few days. As the marginal cost of "software" approaches zero, the pricing foundation of SaaS is shaken.
Second, switching costs are falling. AI-driven data migration tools, automatic API adaptation layers, and intelligent workflow migration technologies have significantly reduced the cost for enterprises to switch from one SaaS platform to another. Customer stickiness is no longer unbreakable.
Third, feature iteration is no longer monopolized by vendors. Enterprises can use AI to customize features themselves, eliminating the need to wait for a SaaS vendor's product roadmap. When users can "develop" features on their own, the vendor's pricing power vanishes.
Is "AI-Enabled SaaS" a False Premise?
Almost every SaaS company is telling the same story: "We've integrated AI into our products, so our value has increased."
But this is a logical trap.
If AI makes software easier to develop, customize, and replace, then SaaS companies integrating AI are actually helping their own competitors—because it lowers the technical barrier for users to build alternative solutions themselves.
It's like a taxi company saying, "We've installed GPS navigation in every cab, so our service is better." True, the service has improved—but ride-hailing platforms are using the exact same technology, and their underlying model is fundamentally more efficient.
The Truth Behind the Valuation "Revenge"
The rebound in the SaaS sector is largely driven by short-term capital factors:
- Oversold bounce. The drop was too steep, making a technical rebound a natural phenomenon.
- The "AI narrative" during earnings season. Many SaaS companies heavily emphasized their AI strategies on earnings calls, temporarily stabilizing investor sentiment.
- Improved macro liquidity. Shifts in interest rate expectations have driven a broader rebound in growth stocks.
But none of this changes a fundamental fact: the long-term growth rate of the SaaS industry is declining. AI isn't making SaaS stronger; it's creating an entirely new software consumption model—pay-per-task, on-demand access, without the need for annual subscriptions.
When an AI agent can directly call APIs to complete tasks, why would an enterprise still pay for an entire SaaS platform?
Who Will Survive?
Not all SaaS companies will die. But those that survive must meet one condition: they provide not just software functionality, but network effects.
Salesforce might survive not because its CRM software is exceptional, but because its entire ecosystem—tens of thousands of ISVs (Independent Software Vendors), hundreds of thousands of developers, and millions of users—forms an irreplaceable network.
However, most mid-tier SaaS companies—those that merely sell tools without network effects—are facing a harsh reality: their moats are being filled in, shovel by shovel, by AI.
A Moment of Clarity for Investors
The phrase "valuation revenge" is tempting because it implies the market made a mistake that savvy investors can profit from.
But perhaps the market didn't make a mistake. Perhaps it simply recognized an uncomfortable truth earlier than most: AI's impact on the software industry is not about growth, but about restructuring. In a restructured industry, old valuation logic no longer applies.
Before buying into the "SaaS rebound," ask one question: In three years, can an AI agent directly replace this company's core functionality? If the answer is "yes," then the current valuation might still be too high.