CIO Priorities in 2026: Why AI Stack Wins & Traditional SaaS Loses

The software market is sending a crystal-clear message about what matters to Chief Information Officers in 2026. And it's not what most SaaS companies expected.

The public markets tell the story better than any survey ever could. Of five major software sectors, only two are thriving. The other three are in freefall. The reason? The buying patterns are unmistakable: fund the AI stack, cut everything else.

This shift represents the most significant realignment in enterprise software preferences in a decade. Understanding these CIO priorities isn't just market trivia—it's essential intelligence for anyone building, investing in, or selling technology solutions.

Core Summary

• AI infrastructure & dev tools lead with +68.5% returns while business applications crash -36.2%, revealing CIO budget reallocation toward AI-ready tech
• Security spending surges to 11.6x EV/Sales as AI expands both attack surfaces and defense needs, making it the only sector besides infrastructure gaining investment
• Traditional seat-based SaaS is dying, with Salesforce cutting headcount by 44% because AI agents now handle customer interactions—a warning sign for every horizontal application vendor
• The market ignores growth rates when choosing winners—Infrastructure and AI mega-caps both grow ~21%, but one trades at premiums while the other is punished for relying on seat pricing
• AI adjacency is now the primary valuation driver, with token-path companies (those processing AI agent transactions) commanding premium multiples regardless of sector

CIO Budget Shift: The AI Stack Dominates Enterprise Spending

The divergence is stark and measurable. Over the past year, the public software market has split into clear winners and losers—and the split has nothing to do with revenue growth.

Infrastructure and Developer Tools have become the commanding heights of enterprise software, rising +68.5% over the past year. This category includes DigitalOcean (up +430%), Datadog, MongoDB, Cloudflare, Snowflake, and Confluent. These companies combine agent compute infrastructure, AI data pipelines, and developer tooling into a unified value proposition that CIOs desperately need.

The numbers tell the story: these companies are growing at 21.4% revenue growth while commanding a 10.0x EV/Sales multiple. That multiple—traditionally reserved only for the most exclusive software categories—now reflects CIO conviction that infrastructure underpinning AI is non-negotiable.

Compare this to Business Applications, which is experiencing absolute carnage at -36.2% over the past year. Salesforce, Workday, ServiceNow, and HubSpot anchor this 48-company bucket—which represents most of public SaaS. These companies are still growing at 12.5% annually, yet the market punishes them at a 3.4x EV/Sales multiple, less than one-third the valuation of infrastructure peers.

The irony is brutal: Business Applications grow slower than Infrastructure (12.5% vs. 21.4%), yet that's not why they're losing. They're losing because CIOs no longer believe in their fundamental business model.

Why Security Is the Second Pillar of CIO Spending

Security is the only other positive sector, rising +17.6% over the past year. But the story here is even more revealing than the infrastructure surge.

Security now commands the highest multiple in the entire software landscape at 11.6x EV/Sales—equal to infrastructure—and grows fastest at 24.1%. Palo Alto Networks, Fortinet, CrowdStrike, and Okta represent the category's core.

The reason is structural, not cyclical. AI is simultaneously expanding both the attack surface (more AI agents running, more inference endpoints, more data flowing through systems) and the defense surface (more sophisticated threats, more novel attack vectors, more endpoints to secure).

CIOs understand that AI deployment without security is essentially uninsurable. The market is pricing security not as a cost center, but as a strategic requirement for AI-enabled operations. Every percentage point of compute added to AI infrastructure requires corresponding investment in the perimeter defense.

This explains why security spending is holding its own even as business applications collapse. CIOs aren't cutting security—they're accelerating it, and the market rewards that conviction with premium valuations.

The Death of Seat-Based Pricing: What Marc Benioff Just Revealed

The clearest admission of what's happening came from an unexpected source: Salesforce's own CEO.

In September 2025, Marc Benioff told the Logan Bartlett Show that Salesforce had reduced headcount from 9,000 to 5,000—a 44% cut—because "I need less heads." The reason? Agentforce, Salesforce's AI agent platform, now handles approximately 50% of customer interactions.

This casual comment should terrify every horizontal SaaS vendor relying on seat-based pricing. Benioff is essentially admitting that AI agents are cannibalizing the core Salesforce unit economics. If Salesforce itself is replacing 44% of headcount with agents, what does that mean for every Salesforce customer running Service Cloud?

It means they're asking the same question: "Why are we paying per seat when agents can handle half the work?"

This isn't Benioff announcing layoffs. He's announcing that the fundamental business model of application software—charging per user—is becoming obsolete. When agents can substitute for human workers, the seat-priced layer becomes a liability, not an asset.

Communications and Collaboration sector data confirms this thesis. The entire category is down -6.6% over the past year on just 8.2% growth. But there's one exception: Twilio is up +62%.

Why Twilio? Because AI agents send messages. Twilio isn't seat-priced—it's transaction-priced. Every AI agent making outbound calls needs a phone number. Every AI agent responding to customer inquiries needs a communications channel. Twilio sits on what I call the "token path"—the infrastructure that AI agents depend on to operate.

Even inside a losing sector, the AI-adjacent, agent-enabled name wins.

The Valuation Paradox: Why Growth Doesn't Matter Anymore

Here's the puzzle that traditional software analysis can't explain: Infrastructure and AI Mega-cap Platforms both grow at nearly identical rates (~21%), yet the market rewards one and punishes the other.

AI and Mega-cap Platforms (Apple, Microsoft, Nvidia, Meta, Oracle, Palantir) are down -5.9% over the past year despite 21.5% growth. These are some of the most profitable, fastest-growing companies on Earth. Yet the market treats them like losers.

Why? Because the market is asking a different question than "Are you growing?" It's asking: "Are you on the token path?"

This distinction—first explored in depth in "So You Want to Sell Inference"—separates companies into two categories:

Payment processors: Companies that monetize AI infrastructure through transaction-based models. These sit on the token path because their revenue is tied directly to AI agent activity. Twilio processes agent communications. DigitalOcean processes agent compute. Cloudflare processes agent traffic. MongoDB processes agent queries.

Software businesses: Companies that still charge seats, subscriptions, or per-transaction for human-facing functionality, even if they're building AI capabilities on top.

The market is pricing the hypothesis that AI agents will drive volumetric growth in infrastructure, not necessarily in traditional application layers. A Datadog that charges for every trillion events logged by AI systems grows differently than a Salesforce that charges per seat to sales reps—even if both are technically growing.

CIOs are betting on the first model and divesting from the second.

What This Means for Enterprise Software in 2026

The 2026 CIO playbook is now legible:

First, fund the AI stack. This includes infrastructure (compute, storage, data pipelines), security (threat detection, perimeter defense), and agent dependencies (communication, routing, orchestration). These categories are receiving disproportionate budget allocation, and the market is rewarding that with premium multiples.

Second, reduce seat-based spending. Every dollar saved by replacing human workers with agents is a dollar no longer spent on the seat-priced application layer. This doesn't mean CIOs are abandoning Salesforce or Workday tomorrow—it means they're questioning why they need as many licenses, and they're finding agents increasingly viable.

Third, prioritize categories where AI increases demand. Security is the template here. AI agents running on your infrastructure don't reduce security spending—they increase it. Similarly, infrastructure spending scales with agent activity. These categories are defended by the fundamental growth dynamics of AI itself.

Fourth, question traditional multiples. A company growing 21% at a 3.4x multiple is not a bargain if the market believes its business model is under threat from substitution. Conversely, a company with identical growth at 10.0x multiple might be fairly valued if it sits on infrastructure the market believes is essential.

The message is unambiguous. CIOs have made their decision about where AI is going and how to fund it. The public market is simply pricing that decision.

Conclusion

The software market's 2026 verdict is clear: AI infrastructure and security will command CIO budgets, while traditional seat-based application software will continue to lose ground. This isn't speculation—it's observable in real-time across every major software sector.

For technology builders, the lesson is hard but clear: build for the token path, not the seat path. For CIOs, the opportunity is equally clear: fund infrastructure and security while rationalizing seat-based spending. For investors, the dispersion between infrastructure multiples and business application multiples reveals where the market believes the future value resides.

The CIO's priorities aren't hidden anymore. They're on display in every stock chart, every earnings call, and every budget reallocation happening across the enterprise right now.

---

Original source: The CIO's Choices are Clear in 2026

powered by osmu.app