Telenor AI Factory has selected F5 BIG-IP Next for Kubernetes for its Nvidia-powered sovereign AI infrastructure, F5 said in a 21 October 2026 announcement issued in Seattle. The platform is intended to control traffic and separate customers using Telenor’s shared AI services in Norway and Europe.
Key points
- Telenor selected F5’s Kubernetes platform to manage traffic and isolate customers on shared GPU infrastructure.
- F5 says its traffic-management functions run on Nvidia BlueField-3, away from host workloads.
- Token-based usage controls and more detailed management of inference traffic remain planned capabilities.
Telenor’s shared GPU infrastructure
Telenor AI Factory combines Telenor’s national infrastructure and connectivity with Nvidia-powered AI systems, according to F5’s announcement. It is designed for Norwegian and European organisations that want AI services while retaining control over where their data is processed, stored and accessed. The selection adds F5’s traffic-management and security software to that infrastructure.
F5 says BIG-IP Next for Kubernetes can place tenant isolation, policy enforcement and traffic control in the path taken by AI service requests. The company says network isolation is intended to let different customers use the same sovereign GPU infrastructure while reducing the risk of one tenant interfering with another. F5 also identifies inconsistent application of policy and potential data exposure as problems the arrangement is designed to address.
That distinction matters to the service Telenor is building. Customers would share computing infrastructure, while the controls governing their traffic and data would operate separately, according to F5’s description. The announcement concerns Telenor’s choice of platform and the capabilities F5 expects it to provide.
F5 policies cover APIs and telemetry
F5 says the platform applies policies to ingress and egress traffic — data entering and leaving the service — as well as API interactions and operational telemetry. Those controls are intended to align traffic handling with customers’ sovereignty and compliance requirements. The company also says the platform provides real-time operational visibility for audits and compliance reporting.
The scope extends beyond access to the GPU systems. According to F5, its software can apply controls across the requests made to AI services and the information generated to monitor their operation. That gives Telenor a proposed way to administer shared infrastructure for organisations with different rules on data location and access.
F5 describes the arrangement as a Kubernetes-native control point for AI traffic. Kubernetes manages applications across computing systems. In Telenor’s case, F5 says its platform would govern traffic associated with those applications, including AI inference requests, rather than supply the GPU capacity on which they run.
BlueField-3 runs F5’s traffic functions
F5 says its traffic-management functions run on Nvidia BlueField-3, a data processing unit that handles selected networking and security work separately from the host computer, MarketScreener reported. F5 says the arrangement preserves host processor capacity for AI applications and can reduce bottlenecks in the path taken by service traffic.
F5 also says separating those functions from host workloads strengthens the boundary between customers and the underlying infrastructure. That is a different task from supplying more GPUs: the BlueField-3 component is intended to handle the networking and security work involved when customers use the available capacity.
Telenor plans token-based controls
F5 describes several AI-specific functions as plans tied to the expansion of Telenor’s commercial services. They include visibility into token usage, policies based on that usage and circuit breaking, which can interrupt requests when a control is triggered. F5 says these functions are intended to support customer entitlements, reporting and billing workflows.
F5 also describes more detailed management of inference traffic as a planned capability. The company says it is intended to improve use of the infrastructure and maintain performance as AI workloads change. In its account, token-aware controls would govern consumption, while inference-aware traffic management would govern how requests move through Telenor’s service.
