F5 Networks, Inc. - Experts & Thought Leaders

F5, the pioneer in delivering and securing every app and API, announces that Telenor AI Factory has selected F5 BIG-IP Next for Kubernetes as an intelligent AI service plane for its NVIDIA-powered sovereign AI infrastructure. Telenor AI Factory is designed to help Norwegian and European organisations access secure, high-performance AI services while maintaining control over where data is processed, stored, and accessed. Sovereign AI requires more than national GPU capacity. It requires an operational AI service plane that can govern where data flows, how tenants are isolated, and how policy is enforced in real time. Commercial AI services Telenor AI Factory addresses growing European sovereignty requirements by combining Telenor’s national infrastructure and connectivity with NVIDIA-powered AI infrastructure and F5’s intelligent traffic management, security, and governance capabilities. With F5 BIG-IP Next for Kubernetes, Telenor can apply policy, visibility, tenant isolation, and traffic control directly in the AI service path. In the future, Telenor also plans to leverage F5 technology to expand the service plane with token-aware governance and more advanced inference traffic optimisation as commercial AI services scale. Managing AI consumption F5 BIG-IP Next for Kubernetes on NVIDIA BlueField DPU provides the following capabilities for Telenor AI Factory: Secure multi-tenancy and network isolation enable customers to share sovereign GPU infrastructure while mitigating the challenge of tenant interference, policy drift, and data exposure. Policy-based traffic control aligns ingress, egress, API interactions, and operational telemetry with sovereignty and compliance requirements. Real-time observability and telemetry provide visibility for operations, auditability, and compliance reporting. Planned token-aware governance and circuit breaking help manage AI consumption, enforce usage policies, and support entitlement, reporting, and billing workflows. Planned inference-aware traffic optimisation improves utilisation and maintains consistent performance across dynamic AI inference workloads. Highly regulated industries Sovereign AI is defined not only by where infrastructure is located, but also by how AI traffic, customer data, telemetry, and policy are controlled. Telenor AI Factory is designed to help Norwegian and European organisations meet these requirements while supporting the needs of highly regulated industries. F5 BIG-IP Next for Kubernetes provides policy enforcement in the data path, enabling Telenor to apply consistent controls across ingress, egress, API interactions, observability, and AI inference traffic. Inference-aware traffic management Telenor plans additional AI-specific capabilities as commercial services scale, including token-aware usage visibility and policy controls, circuit breaking, and more advanced inference-aware traffic management. These capabilities from F5 are intended to support greater control over AI consumption as well as entitlement, reporting, and billing workflows. By running F5’s traffic-management data plane on NVIDIA BlueField-3, selected networking and security functions can be accelerated and isolated from host workloads. This preserves host CPU capacity for AI applications, strengthens the enforcement boundary between infrastructure and tenants, and helps reduce data-path bottlenecks as commercial AI services scale. Deeper token-aware governance “At Telenor AI Factory, our purpose is to give Norwegian and European enterprises access to trusted, secure, and high-performance AI services. Sovereignty has to extend beyond where infrastructure sits to how AI services are operated and governed. F5 gives us a policy and traffic-control layer we can build on as we scale services for customers,” said Kaaren Hilsen, CEO of Telenor AI Factory. “As AI factories move from infrastructure projects to commercial service platforms, service providers need a way to govern AI traffic, tenants, telemetry, and policy in real time. With BIG-IP Next for Kubernetes, Telenor is establishing an AI service plane that can grow with its sovereign AI services, including a path to deeper token-aware governance and DPU acceleration as requirements evolve,” said Kunal Anand, Chief Product Officer at F5. “Together, NVIDIA-powered AI infrastructure and F5 BIG-IP Next for Kubernetes on BlueField-3 help service providers such as Telenor deliver high-performance AI services while maintaining the governance, control, and security required for sovereign AI,” said Ash Bhalgat, Sr. Director, AI Ecosystem and Marketing, at NVIDIA.

F5, the global pioneer in delivering and securing every app and API, announces enhanced cloud-native network functions and new observability capabilities that help service providers modernise critical network services as they move from virtualised infrastructure to Kubernetes and cloud-native architectures. Service providers are under pressure to support growing 5G, edge, IT, and emerging AI workloads, boost efficiency, and maintain the performance, security, and operational control required for carrier-grade services. Moving network functions to Kubernetes lets operators use infrastructure more flexibly but also introduces additional requirements for traffic management, security, lifecycle operations, and visibility across increasingly distributed environments. F5 now accelerates this transition. Increasingly distributed environments “Service providers should not have to choose between cloud-native agility and the performance, security, and control required for critical network infrastructure,” said Kunal Anand, Chief Product Officer at F5. “BIG-IP Cloud-Native Edition brings proven F5 BIG-IP capabilities into Kubernetes, giving network and cloud operations teams a more consistent way to deploy, secure, observe, and operate 5G, edge, and AI infrastructure at scale.” Independent testing by The Tolly Group of selected BIG-IP Cloud-Native Network Functions within BIG-IP Cloud-Native Edition demonstrated substantial infrastructure-efficiency gains in Kubernetes-based DNS and CGNAT workloads compared with virtualised network function deployments. In a DNS scenario, F5 processed 1.9x more queries per second while using 91.5% less CPU. In a CGNAT scenario, F5 delivered comparable throughput while using 79.7% less CPU and 98.5% less memory. Infrastructure-efficiency gains F5 BIG-IP Cloud-Native Edition extends capabilities from the F5 Application Delivery and Security Platform within Kubernetes-based service provider environments, bringing together complementary capabilities across cloud-native networking, application delivery, security, and observability. BIG-IP Cloud-Native Network Functions extend core BIG-IP network and security services into Kubernetes-native environments, including DNS, CGNAT, network security, subscriber policy enforcement, and traffic management. BIG-IP Next for Kubernetes operates at the Kubernetes network layer, providing networking, application delivery, security, and manageability. BIG-IP eBPF Observability provides kernel-level application and network visibility for troubleshooting distributed environments. Kernel-level application Enhanced F5 BIG-IP Cloud-Native Network Functions provide Kubernetes-native traffic management and security capabilities designed for service provider environments. Expanded traffic management: Broad Layer 4–7 support, including ingress and egress networking and support for TCP, UDP, HTTP/2, Diameter, SCTP, GTP, and SIP, helps operators move a wider range of network services into Kubernetes rather than limiting modernisation to HTTP-based applications. Subscriber-aware policy enforcement: Operators can dynamically apply quality-of-service (QoS), traffic-shaping, and content-filtering policies based on subscriber and service requirements, supporting differentiated services while maintaining consistent policy enforcement. Cloud-native lifecycle management: Kubernetes-native deployment and management workflows let teams automate common configuration and lifecycle tasks as network functions move from VM-based to containerised infrastructure. Kubernetes-native deployments By moving these services into Kubernetes-native deployments, operators can make more flexible use of compute and memory resources, more efficiently using infrastructure as network traffic and service demands grow. Network, security, application, and platform teams often rely on different tools when troubleshooting the same problem. New F5 BIG-IP eBPF Observability provides a consistent source of application and network telemetry across distributed Kubernetes clusters, enabling teams to examine app behaviour, container activity, and network flows together when investigating performance or connectivity issues. Real-time, low-overhead telemetry: Operations teams gain granular network and app insights across Kubernetes environments without requiring application-by-application instrumentation. Scalable visibility: Kubernetes-native deployment models and a resource-efficient architecture support visibility across distributed environments, helping teams investigate issues without adding another heavy monitoring layer. Shared operational context: BIG-IP eBPF Observability provides correlated application and network context to determine whether an issue originates in the application, container, or network to reduce investigation time. Virtualised network function deployments The Tolly Group tested BIG-IP Cloud-Native Network Functions against virtualised network function deployments in the same Kubernetes environment across two essential service provider workloads. DNS: F5 processed 1.9x more queries per second with 91.5% less CPU demand and 48% lower average latency than the tested VNF configuration. CGNAT: F5 delivered comparable throughput with 79.7% less CPU demand, 98.5% less memory consumption, and 82% lower p95 latency than the tested VNF configuration. These results demonstrate how Kubernetes-native network functions can help operators increase workload density and reduce infrastructure requirements for the workloads tested. Critical networking and security services Service providers have relied on BIG-IP across physical and virtual infrastructure for critical networking and security services. As those environments increasingly incorporate Kubernetes, F5 is extending those capabilities into the architectures operators are adopting while providing a path to maintain familiar operational models and policies. BIG-IP Cloud-Native Edition gives service providers a broader cloud-native solution across traffic management, networking, security, and observability. It also creates opportunities for operators to adopt additional F5 capabilities as their infrastructure evolves across 5G core, edge, IT, and newer workload environments. BIG-IP Cloud-Native Network Functions, BIG-IP Next for Kubernetes, and BIG-IP eBPF Observability address different layers of that transition while helping customers maintain familiar networking, security, and operational models. BIG-IP Cloud-Native Edition and the capabilities highlighted in this announcement are available now for service provider environments.

F5, the global pioneer in delivering and securing every app and API, introduces significant enhancements to the F5 AI Gateway and integrated the solution into the F5 AI Security Platform. The enhanced F5 AI Gateway seamlessly enforces policies on every AI request, giving enterprises a unified control plane to govern how AI models, agents, and tools are accessed and used, while optimising the economics of AI at scale. Enterprises have moved past AI experimentation, but governance has not kept pace. AI traffic still moves through a patchwork of standalone proxies and monitoring tools that were never built for AI, with no guardrails in between. Unified control plane According to F5’s 2026 State of Application Strategy Report, 77% of organisations say inference, rather than model training or tuning, is now their dominant AI activity, and organisations are managing an average of seven AI models. As inference scales, tokenomics is becoming an increasingly important consideration, with every model request carrying implications for cost, performance, and security. Yet the piecemeal approach to standalone tools to secure AI traffic leaves enterprises without consistent control over how models and agents are accessed and used. Customers accelerate innovation "We're watching enterprises race to deploy AI while struggling to control it," said Kunal Anand, Chief Product Officer at F5. "Every AI request carries economic, security, and governance implications, yet most organisations are relying on fragmented tools that address only part of the problem. The result is rising costs, increased risk, and operational complexity. F5 AI Gateway, integrated into the F5 AI Security Platform, provides a single control point for managing AI across models, clouds, agents, and applications. We believe every enterprise will need an intelligent control layer for AI. F5 is building that foundation, helping customers accelerate innovation while maintaining visibility, security, and control." F5 AI Gateway brings three critical functions together in a single integrated solution: Model Gateway for model access and cost optimisation MCP Gateway for agent-to-tool governance AI Guardrails for prompt and response protection Single integrated solution Budgets, model routing policies and agent access controls are set once centrally and enforced across distributed environments wherever the models, agents and AI apps run, providing a single operations pane for AI platform ops, AI Security ops and finance teams. Rapid adoption of AI is fueling the need for AI gateways. According to a recent Gartner® report, "AI gateways have emerged as a critical part of AI infrastructure, as enterprises need tools to support safe, efficient and controlled access to AI models and MCP servers. Adoption of AI gateways will continue to accelerate among large enterprises." Fine-grained access controls Organisations that cannot see which teams, models, and providers are consuming tokens cannot assess the value and costs of using AI. F5 AI Gateway puts tokenomics under control: the Model Gateway function attributes every token by provider, model, team, and user, per-team budgets enforce limits as spend occurs rather than after the invoice arrives; and automated optimisation — smart routing and model tiering, semantic caching, and GPU-aware load balancing — routes each request to the right model at the right cost. The solution is designed to reduce token spend by up to 60 percent, with no application changes. As agents multiply, so do the MCP servers they call, usually with no central registry, no per-tool authorisation, and no record of what agents are doing. The MCP Gateway function of F5 AI Gateway provides fine-grained access controls that limit agents to the resources they are explicitly authorised to use, including APIs, data sources, and RAG systems. Redacting sensitive data A complete audit trail captures what was accessed, when, by which agent, and on whose behalf. MCP server registry gives teams a single source of truth for approved MCP servers, thereby removing friction for developers who would otherwise find it challenging to discover which servers and tools exist. Personal data, health records, and intellectual property move through AI applications every day, and that data usually reaches external models uninspected while injection and jailbreak attempts go undetected. F5 AI Guardrails inspect every prompt and response, redacting sensitive data before it reaches the model, blocking injection and jailbreak attempts, and failing closed when a request cannot be evaluated. Full audit trails, SIEM export, data residency controls, and alignment with SOC 2, ISO, and HIPAA frameworks give regulated industries the evidence they need before putting AI into production. Hybrid multicloud environments F5 introduced the F5 AI Security Platform earlier this year to give teams continuous visibility, governance, and protection across enterprise AI applications, models, agents, and the APIs connecting them. Four integrated pillars — AI governance, AI usage control, AI security testing, and AI runtime protection — plus an overarching observability layer, create a persistent security lifecycle rather than a one-time compliance exercise. F5 AI Gateway is where those pillars meet live traffic, putting policy into force at the point of interaction and controlling what AI can access, what it can expose, and, crucially, what it can cost the business. F5 AI Gateway is deployable across SaaS, hybrid SaaS, and hybrid multicloud environments, with air-gapped support planned for regulated and sovereign use cases.