Software Sector Splits Into Winners and Losers as AI Redraws the Investment Map

The Great Software Industry Fracture

The US software industry is experiencing one of the largest internal performance gaps ever recorded, according to Evercore ISI analyst Kirk Materne. He describes a sector now cleanly split into two opposing groups, driven by investor perceptions of how each company will monetize artificial intelligence.

Data compiled by Evercore shows that an index of infrastructure software—the platforms and security tools enterprises need to run AI at scale—is up 51% year-to-date as of mid-July 2026. In contrast, application software has slumped 40%, and vertical software for specific industries has lost 27%. The divergence is so wide that Materne sees it as a sign the market no longer treats software as a uniform asset class.

The winners are concentrated in infrastructure, security and data platforms. Datadog has surged about 52% this year, while Palo Alto Networks and CrowdStrike have jumped roughly 43% and 41%, respectively. Cadence Design Systems has also posted gains. On the losing side, application software giants like Salesforce, Adobe, Workday and Atlassian have each shed about a third of their value.

Materne argues that investor capital is flooding into companies that directly benefit from the buildout of AI infrastructure, while those that sell traditional business applications are under deep suspicion that autonomous AI agents will eventually perform many of the tasks those applications currently charge for—a threat that is weighing on valuations regardless of current financial health.

Why Infrastructure Thrives as Applications Crumble

The Infrastructure-Application Divide

At the core of the split is a simple question: does a company provide the picks and shovels for the AI gold rush, or does it sell software whose functions could be replaced by AI? Evercore’s index breakdown makes this stark. Infrastructure and security names—Datadog for observability, Palo Alto Networks and CrowdStrike for cybersecurity—are seen as essential building blocks. They benefit from surging enterprise demand for safe, scalable AI deployment.

By contrast, application software firms like Salesforce and Workday, despite strong suites of customer-relationship and human-resources tools, face the narrative that generative AI agents could automate workflow tasks, reducing the need for licensed seats. This fear has caused a dramatic derating, with valuations contracting even when quarterly numbers have been solid.

Why the Winners Are Still Winning

Datadog’s strength has been specifically tied to its disciplined approach to AI investments, according to Monness, Crespi, Hardt & Co. analyst Brian White. The company is viewed as avoiding the massive, open-ended capex that worries investors about some larger tech firms. J.P. Morgan’s Joseph Gallo cited Palo Alto Networks’ strong visibility in its sales pipeline for the next nine months, reinforcing the idea that cybersecurity spend tied to AI transitions remains a durable growth driver.

Materne’s take is unambiguous: “There is little reason for investors to turn away from infrastructure and security as safe havens in the AI storm right now.” This confidence has driven sharp multiple expansion in these names.

The Genuine Threat of AI Agents

The sell-off in application software is not pure sentiment. The risk that AI agents could autonomously complete tasks—from drafting contracts to managing payroll—represents a structural challenge to traditional software licensing models. Even if the shift takes several years, the market is already pricing in the possibility, creating a potentially rational discount for companies like Atlassian and Adobe, whose tools might be partially automated away.

However, Materne does not think it is time to write off all application stocks. He singles out Oracle, Microsoft and Salesforce as his top picks for the second half of 2026, betting that these large, well-positioned firms can accelerate growth by integrating AI themselves. He tempers expectations, saying any re-rating is likely to materialize only after upcoming quarterly reports.

Concentration Risk Lurks in the Winners’ Camp

One paradox of the split is extreme concentration risk. Capital has piled into a narrow group of infrastructure and security stocks, inflating valuations to levels that could reverse quickly if AI investment cools. Even a moderate slowdown in enterprise AI spending—perhaps triggered by a macro shock or a period of digestion after the current buildout—could trigger sharp corrections in names like Datadog or Palo Alto Networks. The gap between the two camps could narrow not because application stocks recover, but because the winners give back gains.

What the Split Means for Software Investors

• Monitor AI infrastructure spending trends: The rally in Datadog, Palo Alto Networks and CrowdStrike hinges on continued enterprise AI investment. Any sign of a capex slowdown—from comments by large cloud providers or data-center suppliers—could quickly hit these high-multiple stocks.
• Watch for application software catalysts: Materne’s picks (Oracle, Microsoft, Salesforce) are scheduled to report quarterly results in the late summer/fall of 2026. If any of these companies show concrete AI-driven revenue growth or new product adoption that counters the agent-disruption narrative, their beaten-down shares could see substantial re-ratings.
• Assess AI-agent risk on a company-by-company basis: Not all application software faces the same level of agent disruption. Workflow-centric firms like Atlassian and document-heavy players like Adobe may be more exposed than companies offering deeply integrated platforms that can embed AI as a feature rather than being replaced by it.
• Beware of crowded trades: The extreme concentration in infrastructure and security names means that even strong fundamentals may not protect against a rotation if sentiment shifts. Position sizing and awareness of liquidity risk are especially important in this environment.