Broadcom has introduced new AI data centre and edge networking platforms designed to bridge cloud AI and telecom edge computing infrastructure.
Working with partners including Samsung and several telecom operators, Broadcom is launching a portfolio of silicon and software intended to create a cohesive fabric that spans hyperscale data centres to network edge locations.
The initiative addresses the growing gap between centralized AI model training and the demand for low-latency inference in enterprise and consumer settings. Broadcom aims to provide the hardware foundation that helps network operators monetise their fibre and 5G investments through distributed AI services.
Unifying a fractured infrastructure
Running AI workloads closer to end users—directly on the network—creates new revenue opportunities for operators. The main challenge has been operational: deploying and managing compute infrastructure across thousands of dispersed edge sites is complex. Toolchains, hardware, and management planes for data centre AI and edge AI are often separate and incompatible.
Broadcom’s approach tackles this fragmentation by delivering a suite of interoperable technologies rather than a single product. The goal is to enable a single logical AI workload to be partitioned so its components run where they are most efficient—for example, large-scale model training in the cloud while real-time inference or data pre-processing runs at a cell tower, factory floor, or enterprise campus.
The new platforms rely on several key technology upgrades:
- 50G PON: An upgrade to the Passive Optical Network standard that increases backhaul capacity beyond the current 10G baseline. This provides the bandwidth necessary to support thousands of AI-enabled edge devices.
- Wi-Fi 8: The next generation of wireless networking to deliver the throughput and low latency required for AI applications inside homes and enterprise environments.
- Fixed Wireless Access (FWA): Enhanced FWA solutions that offer fibre-like speeds over 5G, extending high-performance edge computing to locations where fibre deployment is impractical.
- Integrated AI accelerators: Purpose-built silicon for efficient AI inference, integrated into networking equipment to lower system cost, power consumption, and physical footprint at the edge.
These elements are intended to form a performance-matched chain: powerful AI accelerators at the edge are only effective if local wireless (Wi-Fi 8) and backhaul networks (50G PON) can support the required data throughput.
Vijay Nagarajan, VP of Marketing for Broadcom’s Wireless and Broadband Communications Division, said: “The true potential of the intelligent broadband edge requires a fundamentally new foundation for the smart home and the smart enterprise. By deploying NPUs across our Wi‑Fi 8 and broadband solutions, we empower service providers to secure user privacy, reduce network congestion, and deliver the multi‑gigabit, sub‑millisecond connectivity that enables the AI era.”
From disjointed components to a cohesive AI fabric
By creating a seamless continuum of networking and compute, enterprises can run AI-powered applications—such as factory-floor quality control or real-time retail analytics—without the latency penalty of communicating with a distant public cloud. For telecommunications providers, the architecture enables managed services like private 5G with on-site AI processing or content delivery networks that perform video transcoding at the edge.
Samsung’s involvement indicates these capabilities will be integrated across network infrastructure and end-user devices, supporting optimization from the network core down to handsets and customer premises equipment (CPE).
In this model, FWA and 50G PON provide the high-capacity transport while Wi‑Fi 8 and accelerator silicon deliver local processing where it’s needed. That combination is essential for AI agents that interact constantly with local sensors and data sources while leveraging larger models hosted in regional data centres.
Broadcom will compete with other silicon vendors such as NVIDIA and Marvell, which are also converging networking and compute. NVIDIA has promoted an “AI factory” concept that extends data centre capabilities outward with platforms aimed at the industrial edge, while Marvell has built a strong portfolio of custom silicon for 5G infrastructure and data processing units (DPUs). Broadcom’s advantage is its presence across the entire network data path—from data centre switches to home router chipsets—which could accelerate adoption of its integrated platform.
See also: Global operations depend on securing subsea cable networks
Learn more about AI and big data from industry experts. The AI & Big Data Expo events take place in Amsterdam, California, and London as part of TechEx, and are co-located with related technology conferences including IoT Tech Expo and Cyber Security & Cloud Expo.
Telecoms content is provided by TechForge Media. Explore other enterprise technology events and webinars with TechForge Media’s event listings.