Deutsche Bank Called the AI Correction—But the Story Is More Complex

The Correction Validates a Structural Threat, Not Just Sentiment

The article's headline correctly identifies the core mechanic behind the trillion-dollar selloff: investors had priced software stocks as if AI would be universally additive, when in reality it represents an existential substitution threat for many business models. The correction wasn't triggered by abstract fear—it was catalyzed by a concrete product release. On February 4, 2026, Anthropic released a new automation tool that pushed AI agents into workflows traditionally dominated by enterprise software, including legal review, compliance, research, and data analysis. This wasn't speculation anymore; it was a functional demonstration that AI could bypass, not augment, existing software infrastructure.

The market responded with precision: approximately $300 billion evaporated across SaaS, data, and software-heavy investment firms in a single day. Companies like Thomson Reuters, LegalZoom, PayPal, Expedia, Equifax, and Intuit experienced double-digit percentage crashes. Unity Software, Rapid7, and Braze have each lost more than half their market value since the start of the year. This wasn't indiscriminate panic—it was targeted repricing of companies whose revenue models depend on information resale, labor substitution, or process-heavy workflows that AI can now replicate instantly and cheaply.

The Business Model Under Attack: Subscription Economics in an Agent-First World

What the article underplays is the fundamental economics being challenged. The SaaS model thrived on predictable subscription revenue, low churn, and high recovery value—assumptions that are now unraveling. The core question isn't whether AI will improve productivity; it's what software businesses can realistically charge when intelligent systems can replicate outputs instantly and cheaply. Subscription-heavy models lose leverage when the value proposition shifts from "access to a tool" to "automated execution of the outcome."

Goldman Sachs projects that by 2030, more than 60 percent of software economics could flow through agentic systems rather than legacy SaaS seats. This represents a profit pool migration, not just margin compression. The value will accrue to AI agents and systems that own execution rather than interfaces. Companies like Salesforce, Workday, Thomson Reuters, SAP, and ServiceNow are now viewed by markets as having "terminal values at risk" because their moats were built on workflow complexity and data aggregation—precisely the areas where AI agents excel.

Wall Street is actively "de-rating" software stocks, meaning the market no longer inclines to pay a premium for long-term growth in companies like Adobe. This isn't just a multiple compression—it's a recognition that the revenue defensibility thesis has changed. Capital allocation over the last decade assumed predictable revenue streams; those assumptions are now being stress-tested in real time.

Why "Old Economy" Disruption May Be Overstated—For Now

Deutsche Bank's Jim Reid suggests the repricing in "old economy" sectors feels overdone, and he may be right in the near term. The article quotes Reid arguing that "even by the end of this year, we still won't have enough evidence to identify the structural winners and losers with confidence." This is a critical distinction: software companies face immediate displacement risk because AI can automate their core product, whereas industries like logistics, consulting, and IT services have more time to adapt because they involve physical constraints, regulatory friction, and human judgment that AI can't yet fully replace.

The speed of the correction reflects what Ed Yardeni calls AI's ability to "feed on itself"—AI writing AI code, accelerating obsolescence at "warp speed" for both hardware and software, particularly large language models. This creates a unique feedback loop: each new AI release doesn't just compete with existing software—it makes prior generations of AI tools obsolete, compressing product lifecycles in ways unprecedented in tech history.

Yardeni characterizes the market swing as moving from "AI-phoria to AI-phobia," with investors rapidly exiting AI-exposed software stocks and reallocating toward cyclical and value-oriented sectors. This rotation suggests the market is treating AI disruption as a sectoral risk, not a systemic productivity boom. The Nasdaq composite fell as chip stocks faced fresh selling pressure, indicating even hardware suppliers are being reassessed.

Geographic Concentration Amplifies European Exposure

The article doesn't mention Europe, but the correction has significant geographic implications. Europe's software sector is valued at roughly €300 billion and heavily concentrated in a handful of companies, with SAP alone representing approximately €200 billion in market capitalization. This concentration means European equity markets are disproportionately exposed to software repricing compared to the more diversified U.S. market. If SAP's terminal value comes under sustained pressure, it could ripple through European pension funds and institutional portfolios in ways that dwarf the U.S. impact on a relative basis.

What Investors Are Missing: The Asymmetry of Disruption

The article rightly quotes Jeremy Siegel arguing that markets are "asking the right questions" about capital expenditure payback periods and competitive moats. But what's missing is the asymmetry of disruption. Companies spending $200 billion on AI infrastructure—like hyperscalers and chipmakers—may indeed face scrutiny on returns, but they're building the rails. Software companies, by contrast, are the passengers whose tickets just got canceled.

The real insight from Goldman Sachs' projection is that 60 percent of software economics flowing through agentic systems doesn't mean 60 percent revenue loss for incumbents—it could mean 100 percent displacement for companies on the wrong side of the agent economy, and windfall gains for those who control the agent layer. This is a winner-take-most dynamic, not a gradual transition.

Deutsche Bank's Jim Reid noted a "clear shift in markets from AI euphoria towards more differentiation between companies, and growing concern about its disruption to existing business models." That differentiation is the headline story, but the subtext is that markets are now pricing AI as a zero-sum game within software, where one company's automation gain is another's revenue loss. The article's framing—that investors "banked that almost every tech company would come out a winner"—captures this perfectly. The correction is markets realizing that AI doesn't lift all boats; it floods some harbors while building new ports elsewhere.

Implications: Volatility as the New Baseline

Reid's prediction that "big sentiment swings will continue to be the order of the day" is likely the most actionable insight for investors. With insufficient evidence to identify structural winners and losers even by year-end, the market will oscillate between optimism (when AI enables new capabilities) and pessimism (when it obsoletes existing revenue streams). This creates a volatility regime distinct from typical tech cycles, because the feedback loop—AI improving AI—means the pace of change is endogenous to the technology itself, not dependent on external adoption curves.

The article positions this as a "reckoning" that "should have been expected," and in hindsight, that's true. But the timing and trigger matter: it wasn't gradual reassessment—it was a product launch that made abstract risk concrete. That's the pattern to watch going forward: each major AI capability release will likely trigger repricing events as markets update their assumptions about which business models remain defensible.