Does artificial intelligence really threaten the business model of software publishers?

Technology is no longer the automatic safe haven it had become. After several years of euphoria, fueled by the promise of artificial intelligence capable of transforming all sectors, the markets have recently undergone a brutal change of tone. AI, yesterday the undisputed engine of growth, has suddenly appeared as a potential threat to some software publishers' business models.

In just a few weeks, the financial world has gone from enthusiasm to distrust. The question is no longer simply "who will benefit most from AI?", but "who risks becoming its victim?". True to its logic of anticipation, the market has quickly made its decision, sometimes faster than economic reality.

A software sector caught in turmoil

The figures illustrate the magnitude of the shock. On Wall Street, software and IT service providers have lost approximately 25% since the end of October.

January saw the sector's worst monthly performance since October 2008, with a 19% drop, followed by a further decline in February. Former tech stars like Unity, Rapid7, and Braze have seen their market capitalization halved since the beginning of the year. Even heavyweights like Palantir, Salesforce, and ServiceNow have lost nearly 40%. In Europe, the contagion is affecting SAP, Dassault Systèmes, and Sage. One of the major triggers for this correction was Anthropic's January launch of new enterprise plugins for its AI assistant, Claude. These tools, capable of automating complex tasks, including in finance, acted as a wake-up call. If AI can directly perform functions currently handled by specialized software, what becomes of the subscription models that underpin the sector's profitability?

In just a few trading sessions, the logic has reversed. The focus is no longer on identifying the winners of the revolution, but on identifying potential losers. Sell first, analyze later.

From AI-phoria to AI-phobia

Some observers speak of a sudden shift from "AI-phoria" to "AI-phobia," a term popularized by Wall Street veteran Ed Yardeni. The phenomenon is classic: exuberance gives way to widespread mistrust.

According to Wedbush Securities, investors now seem to be embracing a near-apocalyptic scenario, as if companies were going to dismantle their legacy software infrastructures in a matter of quarters in favor of autonomous agents. The contrast is striking with the industrial dynamic: several American giants are planning over $650 billion in AI investments by 2026. The paradox is obvious. At the very moment when industry is accelerating, finance is penalizing some of the players supposedly at the heart of this transformation. A crisis of timing. From the perspective of a research center or an innovation department, the current sequence looks primarily like a crisis of timing. Markets are reasoning as if AI were supposed to reconfigure the economy in twelve months. However, the history of technology tells a different story: demonstrators are rapid, organizational transformations are slow. The idea of an "agentic AI" capable of becoming the primary interface for work is technically credible. But between technological possibility and economic shift, there is a long way to go. Data governance, regulatory compliance, cybersecurity, legal responsibility, integration with information systems, change management: these are all obstacles hindering immediate substitution. Nvidia CEO Jensen Huang recently described the idea of a straightforward replacement of software with AI as "the most illogical thing in the world." According to him, AI will enhance existing systems far more often than it will replace them. JPMorgan Chase echoes this sentiment, believing that markets are anticipating scenarios of extreme disruption unlikely to materialize in the short term. Anticipation and Overreaction: The market operates by projection. This is its strength and, sometimes, its weakness. When visibility on future flows becomes blurred, valuation models become strained. The risk premium increases, multiples contract, sometimes even before accounts deteriorate.

The indicators celebrated yesterday—recurring growth, high margins, subscription revenues—are now being re-examined in light of a theoretical risk of disruption. It is not so much the current results that are being penalized as the assumption of a less predictable future.

Added to this is a broader sector rotation. Goldman Sachs strategist Ben Snider speaks of a "long-term downside risk" and draws a parallel with established sectors that underestimated profound changes. Capital is shifting towards stocks considered more cyclical or more "anchored" in the real economy.

Integration rather than substitution

The narrative of a rapid disappearance of software has a blind spot: technological transformation often occurs through gradual integration rather than outright substitution.

In Europe, IT services groups like Atos, Capgemini, and Sopra Steria play a pivotal role.

Their model is based on the integration and security of information systems. AI is therefore not a threat but rather a structural opportunity to support the necessary digital transformation of businesses. Rather than weakening their business model, AI increases the complexity of digital systems, strengthens governance and security requirements, and expands integration needs. In sensitive sectors, data sovereignty, compliance, and resilience are priorities. AI does not replace critical systems from scratch; it integrates with them gradually. Contracts, regulatory constraints, and internal organizations cannot be reconfigured in a few quarters. Memory of cycles and Jevons' paradox: Technological history suggests caution. The arrival of the web, mobile, cloud, or SaaS did not erase established players overnight. Uses have overlapped, models have adapted, and competitive hierarchies have been redistributed over several years. The questions "if the agent does the work, what is the point of the software, what is the point of IT services?" are legitimate. But companies are not just buying an interface. They are buying support, security, compliance, service guarantees, and contractual liability. A look at Jevons' paradox also sheds light on the debate. In the 19th century, the economist William Stanley Jevons observed that an improvement in the efficiency of a resource could increase, rather than decrease, its total consumption. Applied to AI, this suggests that a decrease in usage costs and an increase in performance could expand the market for IT services instead of shrinking it. More efficient tools can stimulate demand, multiply use cases, and create new application layers and integration needs within systems. The market, however, oscillates between euphoria and panic. But the real economy moves at a different pace. Between technological demonstrations and the overhaul of business models, the long term often remains the deciding factor.