The state of AI delivery and security in 2026: Implications for app infrastructures

Distributed inferencing has arrived

Data from the F5 2026 State of Application Strategy Report indicate that 78% of organizations are managing distributed inferencing services across an average of seven different AI models. This distributed AI landscape turns questions such as cost-aware model selection, routing, and input security into runtime concerns.

The systems and control points decision makers choose today will determine how well their organizations deliver, protect, and scale AI apps into the future. Download the ebook to learn more.

Explore key AI workload findings

More than 1,100 global IT decision makers shared their AI plans, concerns, and current state—revealing where AI and app strategies stand today. Key findings include:

AI appsDistributed inferencingMulti-model AIAgentic AI

AI is embedded in the app path

AI apps are no longer experimental. They’re integrated into production systems, decision loops, and operational workflows.

Control matters

Nearly eight in 10 organizations operate their own inferencing service, and the average number of services per organization is two. Because enterprises want control of their data and costs, public AI as a service is the least common model, chosen by only 36% of respondents.

Strategy drives model proliferation

Considerations ranging from cost optimization to API compatibility drive the use of multiple models. As with hybrid multicloud app deployment, the benefits outweigh the complexity of distributed, multi-model AI.

Who’s accessing your resources?

Nearly half (47%) of organizations plan to address the issues by implementing identity-aware infrastructures to interact with AI agents. Still, those agents are likely to reshape access control, identity governance, traffic management—possibly even the costs of user-based economic models—faster than enterprises may expect.

AI apps

AI is embedded in the app path

AI apps are no longer experimental. They’re integrated into production systems, decision loops, and operational workflows.

Distributed inferencing

Control matters

Nearly eight in 10 organizations operate their own inferencing service, and the average number of services per organization is two. Because enterprises want control of their data and costs, public AI as a service is the least common model, chosen by only 36% of respondents.

Multi-model AI

Strategy drives model proliferation

Considerations ranging from cost optimization to API compatibility drive the use of multiple models. As with hybrid multicloud app deployment, the benefits outweigh the complexity of distributed, multi-model AI.

Agentic AI

Who’s accessing your resources?

Nearly half (47%) of organizations plan to address the issues by implementing identity-aware infrastructures to interact with AI agents. Still, those agents are likely to reshape access control, identity governance, traffic management—possibly even the costs of user-based economic models—faster than enterprises may expect.

Download the ebook to explore the trends driving distributed, multi-model inferencing and understand how converged app services can help reduce complexity and increase visibility to ensure AI scalability and success.

Update your AI strategy

Treat AI like any other critical production system, with strong control planes and converged delivery and security services, including:

More insight from F5