Preface:

Overnight, the landscape in Silicon Valley has shifted. When Anthropic's AI agents can autonomously handle contract reviews and compliance tracking, and when Google's AI can strike at the core of game engines...

As AI evolves from a supporting tool into a direct replacement, the SaaS faith that Silicon Valley has upheld for two decades seems to be crumbling.

The market has made its harshest judgment with its feet: AI is completely disrupting the traditional SaaS model.

One AI plugin sparks a butterfly effect, causing a global software stock meltdown

The release of Anthropic's Claude Cowork legal plugin became the final straw that broke the back of global software stocks.

Legal tech leader LegalZoom plummeted nearly 20%, Thomson Reuters tumbled 15%, and the combined single-day market cap loss across software, financial services, and asset management industries reached $285 billion.

Following closely, Google DeepMind's Project Genie triggered a collective dive in gaming stocks, with Unity plunging 24.22%, and Take-Two and Roblox also declining, resulting in a combined single-day market cap shrinkage of $19.5 billion for the three companies.

This market plunge is not an isolated incident but a global software stock tsunami.

Goldman Sachs data reveals that the software industry has seen the highest net selling among all sub-sectors so far this year, with net exposure hitting a historic low of 4.2%. The software sector has evaporated $2 trillion in market value from its peak, a 30% drop.

Panic in the U.S. stock market swiftly spread to Europe and Asia, with European stocks losing $300 billion and Indian IT stocks shedding $201 billion in a single day.

Wall Street has even coined a new term for this storm: [SaaSpocalypse], or the SaaS apocalypse.

This collective software stock collapse, seemingly triggered by a minor update from Anthropic, is actually the culmination of long-held market fears about AI disrupting SaaS.

In February 2026, Anthropic launched 11 industry-specific plugins for Claude Cowork covering sales, finance, legal, and data, with the legal plugin's capabilities shaking the market.

It can directly access corporate legal databases, autonomously map entity relationships in thousand-page merger contracts, extract key dates, identify risk clauses, and even generate neatly formatted compliance reports.

From [research-drafting-review-archiving], the entire legal documentation workflow can be completed independently by AI without human intervention.

This capability directly breaches the core defenses of traditional legal tech companies. The core value of firms like Thomson Reuters and LexisNexis is built on professional legal databases and manually packaged software services.

Anthropic's AI plugin achieves—or even surpasses—these functions at a lower cost.

What terrifies the market even more is that this is not a brand-new AI model but merely 11 [research preview] plugins that have already planted the roots of AI automation in the industries on which SaaS depends.

Meanwhile, Anthropic's Claude Code, with its powerful programming capabilities, disrupts the underlying logic of coding, making the market realize that large models are no longer satisfied with being bottom layer tools (bottom-tier tools) but are now infiltrating the SaaS application layer to directly take over core enterprise workflows.

The storm's spread has far exceeded expectations. Beyond U.S. stocks, European advertising giants WPP and Omnicom collectively dropped over 10%, while Britain's Relx tumbled 14.4%.

Indian IT giants TCS and Infosys face $300 billion in revenue risk, impacting 1.6 million Indian IT workers.

Keep in mind that the core of India's IT industry is software outsourcing, and the maturation of AI programming tools is rendering this [labor cost arbitrage] business obsolete.

More brutally, this market plunge stems not from macroeconomic factors or corporate earnings falling short of expectations but from fundamental doubts about the sustainability of the SaaS business model.

Triple Blow: AI Strikes at SaaS's Core Vulnerabilities

Microsoft CEO Satya Nadella predicted as early as 2025: [The existing form of SaaS applications or commercial applications is likely to crumble in the age of intelligent agents.]

When an AI agent can understand a business's operational descriptions, process differences, and internal rules through natural language and then dynamically generate logic, the concept of [modules] becomes redundant.

First Blow: The [artificial premium] of software is disappearing

Traditional SaaS's high gross margins are built on two foundations:

① High R&D costs: Programmers translate human needs into code, with R&D expenses typically accounting for 25%-40% of annual recurring revenue.

② [Artificial packaging] of services: Enterprises must pay a hefty premium for software brands, UIs, and services to obtain standardized functions.

AI is systematically compressing both cost components.

① AI programming tools enable developers to generate runnable code directly through natural language, weakening the complexity of traditional [tech stacks].

② AI-native companies are reconstructing old workflows at lower costs.

In finance, law, and other fields, the marginal value of traditional high-priced software from Reuters, FactSet, etc., is rapidly declining. AI can handle analytical, organizational, and modeling tasks that once required [professional services], pushing software's marginal costs to the extreme.

Second Blow: Business logic migrates from the software layer to the AI agent layer

Nadella pointedly noted that the essence of most software is a [CRUD database with business logic].

When enterprises pay for SaaS, they're not buying a data container but a set of business judgments solidify (solidified) into the system.

In the past, data collection, interaction, and process triggering were all determined by the software itself, forming SaaS's core migration barrier.

But in the AI era, this premise has been shattered. AI agents enable business logic to migrate entirely from the software application layer to the agent layer. Enterprises no longer rely on software's built-in functional modules but instead entrust judgment, orchestration, and execution to AI.

Simply put, traditional SaaS is [humans adapting to software], while the AI era is [software adapting to humans]. When AI can understand business contexts and make autonomous decisions, traditional SaaS's business logic barriers collapse.

Third Blow: The per-seat pricing model is utterly severed by AI

Traditional SaaS's moat is built on three pillars: per-seat pricing, users adapting to complex UIs/UXs, and higher barriers from more closed functionality.

This model has allowed the SaaS industry to enjoy two decades of prosperity, with giants like Salesforce, Adobe, and ServiceNow achieving soaring market caps.

But Claude Cowork's emergence directly shatters these three pillars, ushering in the AaaS (AI-as-a-Service) era.

Pricing shifts from per-seat to output-based, interaction logic moves from complex UIs to zero-UI backend operations, and system value transforms from closed to open interfaces.

A company that once needed 100 Salesforce or Zendesk seats can now accomplish the same work—or better—with just 10 Claude agents.

Seat fees are the lifeblood of SaaS companies, and AI is precisely severing this revenue stream with a scalpel.

This impact (impact) will also create second-order effects. Software companies are cloud providers' largest customers. When SaaS firms are disrupted by AI, cloud giants like Microsoft Azure and Oracle OCI won't escape unscathed. Meanwhile, compute companies like Nvidia and AMD will see sustained pressure on profit margins due to soaring data center costs.

SaaS Won't Vanish—It Will Become AI Infrastructure

The real threat isn't AI functionality but AI-native architecture.

Many SaaS companies aren't sitting idle. Salesforce has Einstein, Adobe has Firefly, and Notion, Zoom, and HubSpot have all gone [All in AI].

But the market continues to vote with its feet for one reason: bolted-on AI ≠ AI-native.

SaaS + AI remains fundamentally about feature stacking. Currently, most SaaS AI paths involve adding an [intelligent button] to existing modules for summarization, generation, prediction, and recommendation. This may enhance the experience short-term but doesn't alter the software's core form.

It's like adding an [AI write formula] feature to Excel—but Excel remains Excel.

True AI-native systems operate without preset workflows, module emphasis, or user training on [how to use]. Users simply state their goals, not operation paths.

Sequoia partner Konstantine Buhler bluntly stated: [AI may not destroy SaaS but could accelerate enterprise consolidation, hardening Top companies (top companies') moats.]

The essence of this market plunge isn't [AI killing SaaS] but AI conducting a ruthless species selection in the software industry.

This software stock collapse doesn't signify the end of the SaaS industry but a ruthless correction.

It declares that the SaaS era of [piling on features, profiting from seat fees, and locking in customers with closed barriers] has Completely ended (completely ended).

In the past, SaaS sold feature collections; AI sells problem-solving capabilities.

AI replaces [human software operation], not software itself.

Core enterprise operations—customer management, order processing, compliance audits, financial settlements—will always require stable, deterministic, auditable software systems to handle data, rules, and transaction execution.

AI agents also need underlying software to provide data and functional support, creating core opportunities for SaaS companies.

In the future, leading SaaS companies' primary task won't be competing with AI but transforming their products into standardized capability modules for the intelligent agent era, becoming infrastructure called upon by AI.

Microsoft deeply integrates Copilot into Office and Azure, making Office AI's content creation foundation and Azure its compute infrastructure.

Salesforce launches AI sales agents that combine its CRM data with AI, enabling agents to make decisions based on real customer data.

Simply put, traditional SaaS was a standalone product, while AI-era SaaS is part of the AI ecosystem.

Gartner predicts that by the end of 2026, 40% of enterprise-grade SaaS will include outcome-based pricing elements.

This means per-seat pricing will fade into the past, and SaaS companies must establish new AI-compatible pricing systems: shifting from charging per user to charging per output, value, or result.

In the AI era, software features are easily imitated, but understanding industry rules, governing complex systems, and maintaining long-term client trust are irreplaceable core moats.

Conclusion:

Software won't vanish—it will only become more [busy], serving as the underlying foundation for AI agents, the core infrastructure for enterprise digitalization, and a vital carrier of commercial value in the intelligent agent era.

Currently, competition among AI giants has shifted from model parameter arms races to contend for (vying for) application layer control—specifically, who can dominate workflows.

SaaS was once the ultimate form of the cloud era, but in the AI era, it resembles more of a foundation. Foundations don't disappear, but the houses built on them will certainly change. And this transformation has only just begun.

Partial sources referenced: Silicon-Based Observation Pro: [Behind the $285 Billion Market Cap Evaporation: Software's 'Premium' No Longer Exists], NetEase Technology AGI: [Claude Spends Tens of Millions to 'Slam' OpenAI at the 'U.S. Super Bowl'; Altman Retorts: Hypocrite!], Geek Park: [A Minor Update from Claude Collapses Silicon Valley's Entire Software Industry], Industry Observer: [In 2026, the AI SaaS Tide Is Surging In]