Extreme previews on-prem AI zones for enterprise networks

Extreme Networks has previewed Exchange zones, an on-premises artificial intelligence infrastructure for enterprise networking. The product is due for general availability in the first half of CY27.

The system is designed to let businesses run the same AI applications locally and in the cloud while controlling where data is processed and stored. Extreme is targeting organisations that want tighter oversight of data residency, less reliance on cloud links and greater use of hardware already installed across branch and edge sites.

Under the design outlined by Extreme, each Exchange zone would combine either an Extreme Universal Compute Platform or a qualifying customer-owned server with spare processing capacity from access points and switches. Workloads would then be placed automatically according to customer rules on latency, resilience and data residency.

Models would run locally by default, with cloud inference available only when policy allows. Any such cloud use would be logged, according to Extreme.

The announcement reflects a broader shift in enterprise AI from centralised cloud deployments to distributed computing closer to where data is generated. That trend has been shaped by regulation, the cost of moving large volumes of data and the need for systems to keep operating when network connectivity is disrupted.

Extreme is positioning the product around those pressures, particularly for regulated industries and businesses with dispersed operations that may need AI workloads to remain within a defined site and country boundary rather than flow to external cloud infrastructure.

Edge focus

The proposed architecture is also aimed at environments such as retail stores, warehouses and branch offices, where local decision-making can matter when links to central systems are weak or unavailable. In those settings, businesses have increasingly looked for ways to run AI tasks at the edge without building separate, dedicated infrastructure at each location.

Exchange zones would allow organisations to pool compute resources from networking equipment already in place, including spare capacity on deployed access points and switches. Extreme argues that capacity can be redirected to AI applications and plug-ins.

"We've spent decades building networks that reach every corner of the enterprise, and those networks hold a lot of idle compute. Exchange zones will put that compute to work, letting AI run locally on infrastructure customers already own. That delivers sovereignty, local inferencing, and better economics than sending everything to the cloud. Customers control their data, inference, and cost inside a boundary they define, with a record of every exception. No one is better positioned to bring AI to the edge than Extreme. We already run the networks where that work happens," said Nabil Bukhari, Chief Technology Officer and President of AI Platforms, Extreme Networks.

Extreme said its networking products generate large volumes of operational data that can be used not only for network management but also for areas such as security, logistics and business operations. Exchange zones are intended to keep that data on premises for AI use cases that, in Extreme's view, are harder to support in a cloud-only model.

Market context

The move comes as infrastructure providers compete to define where enterprise AI workloads should run and who controls the software layer through which customers access models and applications. Many suppliers have focused on public cloud services, while others are trying to build hybrid systems that can move workloads between customer sites and remote infrastructure.

For networking vendors, the case for edge AI rests partly on installed base. Access points, switches and related hardware are already spread across large enterprises, giving suppliers a route to offer local processing without requiring a full rebuild of branch technology estates.

Brandon Butler, IDC Senior Research Manager, Networking and Infrastructure Services, said the approach addressed a gap in current AI distribution models.

"As AI moves toward the distributed sites where enterprise data is generated, organisations need to run AI applications on-premises to meet data sovereignty requirements and control inferencing cost - yet many AI marketplaces still assume a cloud-only model. Exchange zones is an innovative approach that brings a marketplace of AI applications to customer premises, using spare capacity in existing network infrastructure to keep data local," said Butler.

Extreme said Exchange zones will sit within a broader enterprise software stack that includes its network infrastructure, Platform ONE management and security layer, AI tools and Agent ONE interface, as well as Exchange for AI applications. The company is seeking to tie those layers together around the idea that data originates in the network and can be processed closer to that source rather than being sent automatically to the cloud.

Whether customers adopt that model at scale is likely to depend on how much unused compute exists in current deployments, how easily workloads can be orchestrated across mixed hardware, and whether the operational savings outweigh the added complexity of managing AI across distributed sites.

Each zone would be bound to a single site in a single country, with records showing what remained local, what crossed the boundary and why, and which model was resident, according to Extreme.