Deploying private 5G networks in locations with limited space often forces a compromise between performance and physical footprint.
Although many industrial strategies call for private 5G to enable automation and ensure data sovereignty, fitting data centre–class hardware into retail branches or small manufacturing sites frequently stalls rollouts before they start.
The obstacle is not just connectivity but the infrastructure overhead needed to support it. Branch offices and remote industrial sites rarely have the square footage or power capacity for traditional server racks. Operators must therefore minimise power consumption to control cooling costs while delivering the low latency required by modern applications.
Ji-Yun Seol, Executive VP and Head of Product Strategy, Networks Business at Samsung, explained: “If you are involved in deploying and operating private networks for enterprises, you might be familiar with this gap between promise and practicality—a challenge often encountered by system integrators, managed service providers, and operators.”
To close that gap, enterprises are shifting from proprietary hardware stacks toward virtualised architectures that can run alongside existing compute resources.
Consolidating the stack for private 5G networks
Samsung has addressed these density constraints with a “Network in a Server” (NIS) design. The approach consolidates private network functions—including the mobile core, Radio Access Network (RAN), transport, and AI agents—onto a single platform.
This architecture replaces specialised telecom hardware with a software-driven model running on commercial off-the-shelf (COTS) servers. By virtualising containerised network functions, the system removes the need for separate physical appliances and uses standard compute resources to handle network operations and edge AI services concurrently.
Moving to general-purpose hardware changes the economics of deployment: fewer physical devices reduce shipping and logistics, rack space requirements, and overall power consumption.
For enterprises, this consolidation means lower operating expenses and a simpler support model with a single point of contact. Samsung leverages its experience virtualising macro cell components to deliver this enterprise-grade implementation.
Hardware and ecosystem integration
The performance of a single-server private 5G network depends on the underlying silicon. Samsung worked with AMD, Supermicro, and Wind River to build the NIS ecosystem.
The platform uses an AMD EPYC 8000 server CPU. This marks the commercial debut of Samsung’s virtualised portfolio on AMD processors. The server supports GPUs alongside the CPU, enabling it to offload demanding vRAN processing while keeping capacity available for AI and real-time analytics.
Integrating AI capabilities directly inside the network server enables local data processing. That on-premises approach is critical for environments requiring immediate responsiveness because it reduces latency by avoiding round trips to a central cloud. Local retention of data also helps organisations with strict on-premises security or regulatory requirements.
Samsung validated the private 5G system against industrial scenarios that demand millisecond-level network responses. These use cases rely on instant analysis where latency can create operational risk:
- Video analysis: The server processes CCTV feeds on the private network to flag safety incidents—such as identifying missing hard hats on construction sites, locating lost items in secure areas, or detecting early signs of fire.
- Integrated Sensing and Communication (ISAC): This capability uses radio signals for sensing while maintaining connectivity. Practical applications include parking-space detection, drone tracking, and vehicle monitoring to support road safety.
- Connectivity for emerging devices: The solution supports devices like AR glasses and XR headsets, which overlay contextual information in a user’s field of view—useful for stadium spectators viewing player stats or meeting participants seeing attendee details.
Not your regular connectivity
As AI adoption grows in industrial environments, these services are becoming central to next-generation operations. Running AI and other applications on the same infrastructure used for connectivity gives enterprises new paths to monetise beyond traditional access fees.
Merging network and compute resources enables private 5G deployment in locations that were previously impractical due to space or power restrictions. By reducing physical complexity, organisations can concentrate on application value rather than on maintaining supporting hardware.
Combining RAN and AI compute on a single server shifts the ROI calculation for private 5G, lowering the deployment friction caused by space and power constraints in edge locations such as retail branches or small factories. That said, assessing whether a single-server architecture provides sufficient redundancy for mission-critical operations is essential—some scenarios may still favour multi-node deployments for resilience.
See also: 6G may arrive as sensing infrastructure, not just connectivity
Want to learn about the IoT from industry leaders? Consider the IoT Tech Expo, held in Amsterdam, California, and London. The event offers a broad programme focused on IoT and is part of TechEx, co-located with events covering AI & Big Data and Cyber Security.
Telecoms is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars through TechForge’s events listings.