Is the SaaSpocalypse actually happening?

To cut to the chase, the 'SaaSpocalypse (SaaS-pocalypse)' is roughly halfway through right now. No sweeping collapse has hit the software industry as a whole, but the per-seat pricing model is clearly facing change.

ChatGPT's 'dots' feature, released on September 29, pushed this trend along even further. Even so, software stocks fell only about 1% on the day of the announcement.

• The software ETF (IGV) dropped about 37% from its April peak, but by the end of August it had recovered to within 7% of that peak.
• Salesforce's revenue grew 11%, so it raised its full-year earnings forecast.
• The share of revenue retained from existing customers fell from 88% to 84%.
• Roughly 65–75% of enterprise software spending is still spent on user-based license contracts.

The risk is confined to a subset of companies where 'the screen is the product,' not to the software industry as a whole, and the pace of change is more gradual than the market fears.

What is the SaaSpocalypse?

First, let's define the terms. SaaS (Software as a Service) is business software you pay a recurring subscription fee to use over the internet; Salesforce, Slack, and Notion are typical examples. SaaSpocalypse is a blend of SaaS and Apocalypse, referring to market concern that AI will replace existing software markets. The term was first proposed by the investment bank Jefferies.

After Anthropic launched Claude Cowork and 11 work plugins, the software ETF (IGV) fell 32% from its peak within a week, wiping out roughly $2 trillion in market capitalization.

There are two reasons investors got scared.

Headcount cuts directly lead to revenue declines. SaaS typically uses a 'seat-based pricing' system. If 10 AI agents take over the work of 100 people, the number of paid seats drops even if the same software is still used. The consulting firm AlixPartners estimates the revenue at risk from this effect at up to $500 billion.

Also, the screen layouts employees have mastered have worked as a strong lock-in that prevents customers from leaving. But if AI can handle tasks through voice commands alone, the importance of screen mastery is likely to decline.

In short, the SaaSpocalypse does not mean the end of software. It signals that user-based pricing models and screen-centric product value are about to be re-evaluated, and the following sections should be read with that perspective in mind.

OpenAI and other AI companies have mainly aimed at offering AI model APIs as a service. However, at its developer event (DevDay) held on September 29, OpenAI unveiled 'dots' and 'ChatGPT Space,' making a full-scale push into the business software market. In fact, most of the one-hour keynote that day focused on these two products.

The dots feature is included by default in the ChatGPT Pro and Business Premium plans. On Enterprise plans, admins enable it separately as a beta version.

There are two things worth keeping in mind here.

Dots don't replace existing software; they run on top of it. Connecting to more than 4,000 apps is their core strength. So what is fundamentally changing is not the data accumulated inside software, but the work areas where people used to operate screens directly.

Not just OpenAI: Anthropic's Claude Cowork and Meta's Muse, released in September, are heading in the same direction. TechCrunch pointed out that many of dots' main features could already be built with existing tools. In other words, dots are evaluated less as a completely new innovation and more as existing technical capabilities repackaged into a user-friendly form.

The intensity of the shock is gradually easing.

Software stocks have dropped each time an AI company launched a new product, but the size of the decline is shrinking. On the day dots was announced, Salesforce and ServiceNow fell only 0.9% and 1.2%, respectively, and two days later they actually rebounded.

This table shows two things.

First, the market already treats AI companies launching business solutions as a normal occurrence. In January, the announcement of 11 plugins caused a $2 trillion swing, but in September, a far larger product launch moved stock prices by only about 1%.

Second, an index recovery does not mean overall corporate earnings have improved. The August rebound was driven by Palantir, which has succeeded in AI; its Q2 revenue jumped 93%, and its stock rose about 51% over the month. By contrast, Adobe, which relies heavily on per-seat pricing, was down about 17% year-to-date as of late August. This suggests the market is not shunning the software sector across the board, but is making selective investments based on growth potential.

Evidence that change is underway

Evidence that it isn't here yet

Three things can be read from these numbers.

Per-seat pricing has not been abolished entirely, but its share has shrunk. Salesforce, HubSpot, Docusign, monday.com, GitHub and others have introduced pay-as-you-go pricing based on 'usage' or 'number of transactions.' Docusign is the only company that has fully dropped per-seat pricing; the rest add new pricing items alongside their existing plans.

The view that the fear is overblown

There are four reasons to think this is not the "end of software," and all are persuasive.

Bank of America pointed out that the market is worried both about whether the "investment in AI" will pay off and about "AI replacing all software." If AI is that powerful, the investment will pay off; if it doesn't pay off, it won't replace software either. Both cannot be true at once.

Between May and July, Anthropic, OpenAI, AWS and Microsoft invested about $9 billion in organizations that help customers adopt AI. OpenAI Presence also cannot be bought online; OpenAI engineers install it directly. That means it is hard for companies to adopt AI agents on their own, and existing software is being replaced more slowly as a result.

dots is not finished yet either. For large enterprises, the trial is off by default. OpenAI also advises that "dots can make mistakes, so be sure to check important work." Reports say that on the eve of the event, the release of the next model (GPT-6.1 Astra) was put on hold over safety concerns (Dow Jones). For businesses, this may be read as a signal that it is hard to trust them with important work.

Existing companies are also applying AI to their own products. Salesforce's AI agent product (Agentforce) has surpassed $1.5 billion in annual subscription revenue, up more than 240% from a year earlier. ServiceNow's annual contract value for AI has passed $1 billion. Research firm Gartner expects more than 40% of enterprise AI agent projects to be canceled by the end of 2027.

Of course, this counterargument has weaknesses too. The stock rebound on October 1 happened without any special good news, as investment money moved between sectors, and it could easily reverse. Also, even as its results improve, Salesforce has lost more than 40% of its market value since the end of 2024. In other words, good results do not immediately bring back a lowered valuation.

Who is at risk, and who can hold on

It depends on whether the product's value lies in the "screen" or in "data and compliance." AI can replace screens. But storing a company's core records and handling them in line with regulations is not easy to replace.

These show what surviving companies have in common. Rather than being pushed out by AI agents, they are changing into "tools" that agents use when carrying out work. Salesforce said in its earnings call that it would open the data and workflows accumulated on its platform to "all agents and models." The fact that dots connects to 4,000 apps is a threat, but it is also evidence that those apps are still needed.

Conclusion and what to watch next

The SaaSpocalypse is progressing not as an "end" but as a "split." Products that depend only on the screen are losing value. Meanwhile, companies that hold core data are shifting strategy toward providing that data to AI agents. dots takes this trend one step further, but it is not an event that flips the board.

To the question "Is it actually happening yet?", the answer is "Yes, but slowly." Price lists changed within four months. But contracts change at renewal, and budgets change yearly. The industry sees this lag as 2 to 3 years. What we saw this year is closer to a preview, and whether it becomes the main feature can be seen in the five points below.