Zscaler closed fiscal 2026 with quarterly revenue up 25% to $898.2 million, while annual recurring revenue reached $3.77 billion. Management attributes this growth to demand extending beyond employee access into workloads, branches, and AI systems. The company now describes AI security as both a threat-management problem and an infrastructure access problem, covering agents and models alongside conventional workloads. This shift moves security spending closer to the traffic between machines, services, and increasingly autonomous software.
The Shift from Seats to Connections
Seat-based cybersecurity scales roughly with employee populations, but workload security scales with applications, services, cloud instances, branches, APIs, and machine identities. AI agents add another layer because they can communicate with applications and data without a human directly initiating every transaction. Zscaler frames its Zero Trust architecture around connections between users, workloads, branches, and agents rather than giving those entities broad network access. At Zenith Live 2026, the vendor introduced the ZAgent Framework to automate administrative tasks like configuration and troubleshooting using natural language prompts. The platform also launched a Zero Trust Gateway for Google Cloud Platform, matching its existing support for Amazon Web Services to standardize security policies across different cloud providers.
The pricing structure remains opaque, with no public list prices available for the core platform tiers. Independent analyses converge on per-user ranges of roughly $140 to $390 per user per year for combined ZIA and ZPA, depending on the tier. However, these figures exclude significant add-ons that drive total cost. Buyers should expect steep step-ups between Business, Transformation, and Unlimited tiers, with advanced DLP, browser isolation, and privileged access often gated to higher levels. For context, a blended price of $15 per user per month at 5,000 users becomes $900,000 annually, making seven-figure contracts normal for large enterprises.
| Plan | Cost Model | Included AI/Workload Features |
|---|---|---|
| Essentials Platform | Custom quote | Zero Trust for Workloads (1 GB/user/month) |
| Zscaler Platform | Custom quote | Full SWG, Private Access, Inline Data Security |
| Add-On Modules | Custom per module | GenAI Security, DSPM, Endpoint DLP |
Operational Costs and Hidden Escalation
Multiple independent third-party analyses warn of a 30-100% cost escalation from common add-ons such as ZDX, workload licensing, and Data Protection. These modules are frequently required to secure the complex environments where AI workloads operate. The platform requires a genuine need to map tier-to-feature detail carefully before committing, since high-value capabilities are not universally included in entry-level subscriptions. Implementation costs also factor into the total expenditure, with enterprise deployments averaging $180,000 to $340,000 in professional services fees for traffic discovery and policy migration.
Zscaler’s financials show that full-year revenue rose 25% to $3.35 billion, yet the acquisition of Red Canary contributed $141 million of annual recurring revenue. Strip that out, and annual recurring revenue growth falls to 20%, while net new annual recurring revenue growth drops to 17%. This distinction matters because it suggests some growth comes from consolidation rather than pure expansion. Enterprises may consolidate existing security budgets rather than create entirely new ones for AI protection. How much incremental spending this creates remains harder to establish from one quarter alone.
The operational burden shifts from hardware management to software configuration, eliminating the need for on-premises firewalls or VPN concentrators. Siemens deployed the platform across 360,000 employees in 192 countries, while NOV reported saving millions after moving 27,000 employees off legacy VPNs. Careem cut its legacy firewall and VPN costs by approximately 55%, demonstrating the potential for savings through consolidation. Yet, the lack of a private backbone means performance relies on internet peering, which can expose latency-sensitive AI workflows to variability. Does the convenience of a single-vendor stack outweigh the risk of relying on public internet paths for critical machine-speed traffic?
