Build the AI factory your network was made for
Service providers have the data centers, power, fiber, and enterprise relationships an AI factory needs. F5 supplies the delivery and security layer on carrier-grade Kubernetes.
Operators start with what every AI factory lacks
An AI factory needs power, buildings, fiber, presence, and customers who trust the operator with their data. Carriers have all of it, plus decades of running multi-tenant networks. What they lack is the layer that turns GPU racks into a service enterprises will buy: tenant isolation, inference routing, and border protection. F5 delivers it on the F5 Application Delivery and Security Platform.
Sell GPU capacity to enterprise tenants
Per-tenant isolation, ingress, egress, and firewall policy on shared GPU nodes.
Route inference where GPUs are free
Inference-aware routing on live serving signals, inside each site.
Reuse the edge you already run
The BIG-IP hardware and cloud-native functions at N6 protect AI services, too.
Built on carrier-grade Kubernetes
Developed with and proven at the largest US carriers for Kubernetes at scale.
F5 helps enterprises run AI in production
How F5 helps service providers monetize AI
GPU as a service (GPUaaS)
PRODUCT
Network-level tenant isolation and per-tenant policy at the cluster border.
PRODUCT
Gateway API implementation for application traffic inside tenant clusters.
PRODUCT
BIG-IP Next for Kubernetes deployed on NVIDIA BlueField DPUs
Tenant enforcement on the DPU, with host cores returned to sellable capacity in DPU mode.
PRODUCT
BIG-IP eBPF Observability for cloud-native applications
Real-time traffic visibility for tenant workloads on Kubernetes.
Inference at the network edge
PRODUCT
Cloud-native firewall, CGNAT, DNS, and policy for N6 traffic on the same platform.
PRODUCT
BIG-IP Next for Kubernetes deployed on NVIDIA BlueField DPUs
The same data path offloaded to the DPU at space- and power-constrained sites.
Protect AI services at scale
Sovereign AI for your market
PRODUCT
Per-tenant routing domains, egress policy, and per-tenant firewall logs.
Resources
Reference architectures
REFERENCE ARCHITECTURE
Explore the AI reference architecture
REFERENCE ARCHITECTURE
NVIDIA Cloud Partner and F5 BIG-IP Next for Kubernetes common networking
White papers
Frequently asked questions
Carriers already have what an AI factory needs: data centers, power, fiber, local presence, regulatory relationships, and decades of running multi-tenant networks at scale. An AI factory lets telecom operators sell GPU capacity, managed Kubernetes, and inference to the enterprises and agencies they already serve, often as a sovereign alternative to global providers, and place inference close to subscribers at the network edge.
An AI factory is less different than it looks from already-running Kubernetes platforms. An AI factory is a Kubernetes estate with GPUs, and it needs the same tenant isolation, ingress and egress control, firewalling, and observability as a 5G core or edge platform, plus inference-aware routing and controls for AI-specific risks. BIG-IP Next for Kubernetes provides the network layer for both, so the team that runs the network platform can run the AI platform.
Yes, AI services and network functions can run on the same edge site. F5 cloud-native network functions for firewalling, CGNAT, DNS, and policy run on Kubernetes at the edge, and BIG-IP Next for Kubernetes provides tenant isolation and inference routing on the same platform, on host CPUs or NVIDIA BlueField DPUs. Each stays in its own tenant boundary, so AI workloads and network functions share a site without sharing a blast radius.
No, NVIDIA BlueField DPUs are not necessarily needed. BIG-IP Next for Kubernetes runs on standard host CPUs. BlueField DPUs move the networking data path off the host and enforce policy outside the host operating system, which matters most at edge sites where space, power, and host cores are scarce. Performance results measured in DPU mode do not carry over automatically to host deployments.
F5 complements the accelerated computing, fabric, and model serving software in AI factory reference designs rather than replacing any part of them. BIG-IP Next for Kubernetes is validated in the NVIDIA Enterprise AI Factory validated design and documented in the NVIDIA Cloud Partner reference architecture for common networking, and F5 border and data services run on BIG-IP software, VELOS, and rSeries.