AI infrastructure and security software indexes are up 22% this year, while workplace software indexes are down as much as 34%

US software stocks are splitting into clear winners and losers this year, and the divide runs along a line most investors haven't drawn yet, according to the Kobeissi Letter. Infrastructure and cybersecurity software indexes are up 22% year to date, while vertical (industry-specific) and general application software indexes are down 28% and 34%, pulling the broad US software index down 13% for the year.

AI hype isn't the real dividing line, the pricing model is

Kobeissi frames this simply as AI dividing the industry into winners and losers. Deeper reporting suggests the actual mechanism is narrower and more mechanical: software sold per user seat is the one losing ground, regardless of what industry it serves. If an AI agent can do the work five employees used to do, a company now needs to buy one license instead of five, and industry coverage has tied that dynamic to a slowdown in vertical SaaS revenue growth from about 14% in 2024 to roughly 12.2% by late 2025.

Why Software and Cybersecurity Stocks Are Ripping in 2026 (Forbes, 2026-02-06)

출처: forbes.com

Not every 'vertical' product is actually exposed

A separate AllianceBernstein research note complicates the industry-vs-industry framing further. It argues the real fault line isn't infrastructure versus applications at all, it's how replaceable the software is: products with low switching costs and thin proprietary data, mostly seat-priced interface tools, are easy for an AI agent to replace, while systems embedded deep in a business, even some vertical software running banking or retail operations, are harder to dislodge no matter the category label.

Software's Big Sell-Off: Structural Risk or Narrative Noise? (AllianceBernstein)

출처: alliancebernstein.com

두 읽기 모두 승자와 패자가 갈리고 있다는 점은 같지만, 그 경계선을 업종에 긋느냐 요금 구조에 긋느냐에서 갈린다.

So what's the thing to actually watch

The question comes down to one thing: is this split really about what a company sells, or about how it charges for it? If it's mostly an industry story, infrastructure and security software should keep outperforming almost regardless of pricing model. If it's mostly a pricing-model story, some vertical software that has already moved away from flat per-seat licensing toward usage-based or outcome-based pricing should hold up better than this week's index split suggests, and some 'infrastructure' names still selling by the seat should eventually feel the same pressure. Next quarter's earnings season, when more companies disclose how their pricing models are shifting, will start to answer which read is closer to right.