Hewlett-Packard Enterprise (HPE) today revealed it will expose the routers it recently gained via the acquisition of Juniper Networks to third-party artificial intelligence (AI) agents that IT teams have adopted to configure and deploy IT infrastructure and applications.

At the same time, HPE is also adding an 800-port line of Juniper PTX12000 routers capable of scaling to 345.6 terrabits-per-second (Tbps) in an 8-slot PTX12008 offering and up to 518.4 Tbps in a 12-slot PTX12012 version.

Additionally, there is a 2u router, dubbed the Juniper PTX10002 router line, that can be configured to provide 14.4 or 28.8 Tbps.

Finally, HPE is also rolling out a pair of HPE ProLiant Gen12 Telco Edge Compute servers that can fit in 1u and 2u racks to increase the amount of compute capacity that can be packed into a data center environment.

Julius Francis, senior director of product marketing for the HPE Routing Infrastructure Solutions business, said the Juniper Routing Director will make it possible for various domain-specific AI agents capable of managing IT infrastructure to be extended to Juniper routers by invoking a Model Context Protocol (MCP) server.

That capability will drive further convergence of roles and responsibilities across IT teams as it becomes simpler to unify the management of multiple domains, he added.

Based on the Juniper Express 5 ASIC processors, HPE is making a case for routers that can be deployed at the network edge that are more power-efficient than rival offerings, said Francis. That’s critical in an era where more latency-sensitive artificial intelligence (AI) applications are being deployed closer to the network edge, noted Francis.

It’s not clear to what degree AI workloads will become more distributed in the years ahead, but any effort to reduce AI application latency will require more workloads to be deployed closer to the network edge. The overall goal is to reduce the need to rely on cloud services to run AI inference engines

According to the Omida research, enterprises can expect the share of AI traffic on their networks to grow by about 5× in the next 36 months. Today, that traffic averages about 3% of total network traffic, which is expected to rise to 15% over the next three years and eventually eclipse traditional network traffic by 2031.

As AI applications become more pervasive, it’s now only a matter of time before internal IT teams assume more responsibility for deploying them. While data science teams will continue to build and train AI models, the inference engines that are relied on to enable AI to run in production environments will increasingly become one more type of workload managed by internal IT teams. The challenge, of course, is that those inference engines will be distributed everywhere from the network edge to the cloud.

Ultimately, each organization will need to determine at what pace to deploy those AI agents for themselves, but as more of them are deployed, the need for increased amounts of network bandwidth will become apparent. As such, IT teams should start planning for network upgrades now that are all but inevitable.